USF Libraries
USF Digital Collections

Gender differences in lung cancer treatment and survival

MISSING IMAGE

Material Information

Title:
Gender differences in lung cancer treatment and survival
Physical Description:
Book
Language:
English
Creator:
Kowski, Margaret Anne
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Chemotherapy
Gender Specific Survival
Lung Cancer
Lung Malignancy
Radiation Therapy
Surgery
Dissertations, Academic -- Epidemiology Biostatistics -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The objectives of this research were to test treatment and survival differences between women and men with lung cancer as there is minimal investigation in the literature. Three research questions were developed with statistical testing for gender differences based on similar cancer type, stage, treatment assignment and survival. Data for 44,863 primary lung cancer cases were collected from eight U.S. state-based cancer registries to investigate the research questions. The lung cancer incidence data included the morphological cell-types of adenocarcinoma (AC); squamous cell carcinoma (SCC); large cell carcinoma (LCC) and small cell carcinoma (SCC). Stage, grade, treatment type, as well as, individual characteristics such as gender, age at diagnosis, marital status at diagnosis and race were other variables obtained to be included in the statistical models. Reporting the overall effect for lung cancer gender specific treatment differences or survival has not been demonstrated in the literature to the author's knowledge. By convention, main effects and interaction effects are reported in the literature; without including an evaluation the overall effect of a variable on the outcome, possible misinterpretations could be made. For example, utilizing the Cox's Proportional Hazards model when the interaction effect of gender and treatment type received was examined, females were at an increased risk for death by as much 29% as compared to males (HR = 1.18, 95% CI 1.09 - 1.29). But when the gender effect on survival was assessed, there was an increase in females survivorship as compared to males by as much as 28% (HR = 0.80, 95% CI 0.72 - 0.97 ). In conclusion, by using a unique statistical approach, statistically significant Odds Ratios and Hazard Ratios were demonstrated for the research data set when the overall interaction effect on the outcome was examined. Recommendations to health care practitioners include adhering to current guidelines, e.g. American Medical Association, for lung cancer treatments. Standard treatment protocols were not always followed for early stage disease, e.g. females versus males with stage I lung cancer were 1.71 times more likely to receive chemotherapy in combination with radiation therapy versus a standard first treatment course of surgery (OR = 1.71, 95% CI 1.06 - 2.78). Also, depending on the lung cancer morphology and lung cancer treatment, females as compared to males could exhibit an increase in survivorship by as much as 28%. To improve the results of medical care decisions for lung cancer, clinicians may find the information presented in this study useful and encourage further research on which treatment increases survival for both men and women.
Thesis:
Disseration (Ph.D.)--University of South Florida, 2011.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Margaret Anne Kowski.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 349 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
usfldc doi - E14-SFE0004926
usfldc handle - e14.4926
System ID:
SFS0028174:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2011 flu ob 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004926
035
(OCoLC)
040
FHM
c FHM
049
FHMM
090
XX9999 (Online)
1 100
Kowski, Margaret Anne.
0 245
Gender differences in lung cancer treatment and survival
h [electronic resource] /
by Margaret Anne Kowski.
260
[Tampa, Fla] :
b University of South Florida,
2011.
500
Title from PDF of title page.
Document formatted into pages; contains 349 pages.
Includes vita.
502
Disseration
(Ph.D.)--University of South Florida, 2011.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: The objectives of this research were to test treatment and survival differences between women and men with lung cancer as there is minimal investigation in the literature. Three research questions were developed with statistical testing for gender differences based on similar cancer type, stage, treatment assignment and survival. Data for 44,863 primary lung cancer cases were collected from eight U.S. state-based cancer registries to investigate the research questions. The lung cancer incidence data included the morphological cell-types of adenocarcinoma (AC); squamous cell carcinoma (SCC); large cell carcinoma (LCC) and small cell carcinoma (SCC). Stage, grade, treatment type, as well as, individual characteristics such as gender, age at diagnosis, marital status at diagnosis and race were other variables obtained to be included in the statistical models. Reporting the overall effect for lung cancer gender specific treatment differences or survival has not been demonstrated in the literature to the author's knowledge. By convention, main effects and interaction effects are reported in the literature; without including an evaluation the overall effect of a variable on the outcome, possible misinterpretations could be made. For example, utilizing the Cox's Proportional Hazards model when the interaction effect of gender and treatment type received was examined, females were at an increased risk for death by as much 29% as compared to males (HR = 1.18, 95% CI 1.09 1.29). But when the gender effect on survival was assessed, there was an increase in females survivorship as compared to males by as much as 28% (HR = 0.80, 95% CI 0.72 0.97 ). In conclusion, by using a unique statistical approach, statistically significant Odds Ratios and Hazard Ratios were demonstrated for the research data set when the overall interaction effect on the outcome was examined. Recommendations to health care practitioners include adhering to current guidelines, e.g. American Medical Association, for lung cancer treatments. Standard treatment protocols were not always followed for early stage disease, e.g. females versus males with stage I lung cancer were 1.71 times more likely to receive chemotherapy in combination with radiation therapy versus a standard first treatment course of surgery (OR = 1.71, 95% CI 1.06 2.78). Also, depending on the lung cancer morphology and lung cancer treatment, females as compared to males could exhibit an increase in survivorship by as much as 28%. To improve the results of medical care decisions for lung cancer, clinicians may find the information presented in this study useful and encourage further research on which treatment increases survival for both men and women.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor:
Mason Stockwell, Thomas Heather J. G..
Advisor:
Mason Stockwell, Thomas Heather J. G..
653
Chemotherapy
Gender Specific Survival
Lung Cancer
Lung Malignancy
Radiation Therapy
Surgery
690
Dissertations, Academic
z USF
x Epidemiology Biostatistics
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.4926



PAGE 1

Gender Differences in Lung Cancer Treatment and Survival by Margaret Anne Kowski A dissertation s ubmitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Epidemiology and Biostatistics College of Public Health University of South Florida Co Major Professor: Thomas J. Mason, Ph.D Co Major Professor: Heather G. Stockwell, Sc.D Getachew Dagne, Ph.D. Tatyana Zhukov, Ph.D. Date of Approval: April 11 2011 Keywords: Gender Specific Survival, L ung Malignancy Chemo therapy, Radiation Therapy, Surgery Copyright 2011 Margaret Anne Kowski

PAGE 2

i TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ............ v ii LIST OF FIGURES ................................ ................................ ................................ ......... x ii i LIST OF ABBREVIATION S ................................ ................................ .......................... xi v ABSTRACT ................................ ................................ ................................ ....................... x v CHAPTER I : INTRODUCTION ................................ ................................ ......................... 1 Background ................................ ................................ ................................ .............. 1 Research Questions ................................ ................................ ................................ .. 5 CHAPTER II: LITERATURE REVIEW ................................ ................................ ............ 6 Overview of the Lungs ................................ ................................ ............................. 6 Anatomy and Physiology ................................ ................................ ............. 6 The Disease of Interest: Lung Cancer ................................ .......................... 8 Exposures of Interest: Gender and Lung Cancer Treatment Modality .................. 11 Epidemiol og y ................................ ................................ ............................. 14 Impact on Healthcare Resources ................................ ................................ 22 Origin ................................ ................................ ................................ ......... 24 Clinica l Si gns and Symptoms of Lung Cancer ................................ .......... 25 Procedures for Diagnosing Lung Cancer ................................ ................... 27 Screening ................................ ................................ ........................ 30 Pathology/Histology ................................ ................................ .................. 33 Staging/Extent of Disease ................................ ................................ .......... 36

PAGE 3

ii Lung Cancer Prognosis ................................ ................................ .............. 41 Lung Cancer Survival and R isk Factors ................................ ................................ 43 Gender ................................ ................................ ................................ ........ 43 Tobacco ................................ ................................ ................................ ..... 5 0 Race and Ethnicity ................................ ................................ ..................... 52 G enetics ................................ ................................ ................................ ...... 53 Family History ................................ ................................ ............... 55 Genetics and the Environment ................................ ....................... 56 Geographic Variation ................................ ................................ ................. 57 Alcohol ................................ ................................ ................................ ....... 59 Diet and Micronutrients ................................ ................................ ............. 60 Obesity and Body Mass Index ................................ ................................ ... 63 Occupation ................................ ................................ ................................ 66 Hormones ................................ ................................ ................................ ... 67 Socioeconomic Status ................................ ................................ ................ 68 Environmen t ................................ ................................ ............................... 6 9 Disea ses Associated with Lung Cancer ................................ ..................... 70 Treatments for Lung Cancer ................................ ................................ .................. 71 Confined to t he Lungs ................................ ................................ ................ 71 Local Spread ................................ ................................ .............................. 72 Distant Spread ................................ ................................ ............................ 72

PAGE 4

iii Lung Cancer Relapse ................................ ................................ ................. 73 Complications of Lung Cancer ................................ ................................ .. 73 Lung Cancer Treatment Modalities ................................ ................................ ....... 75 Radiation Therapy ................................ ................................ ...................... 75 Chemotherapy ................................ ................................ ............................ 77 Surgery ................................ ................................ ................................ ....... 81 Combination Therapy ................................ ................................ ................ 83 Em ergent Modalities ................................ ................................ .................. 84 Conclusions and Assessment of the Literature ................................ .......... 86 CHAPTER III : PROCEDURES AND METHODS ................................ .......................... 89 Introduction ................................ ................................ ................................ ............ 89 Aims/Hypothesis ................................ ................................ ................................ .... 89 Aim 1 ................................ ................................ ................................ ......... 89 Hypothesis 1 ................................ ................................ ............................... 9 0 Aim 2 ................................ ................................ ................................ ......... 90 Hypothesis 2 ................................ ................................ ............................... 90 Aim 3 ................................ ................................ ................................ ......... 90 Hypothesis 3 ................................ ................................ ............................... 90 Participant Description and Case Ident ification ................................ .................... 90 Variables of Interest (In clusion and Exclusion Inclusion Criteria) ..................... 102 In clu sion Criteria ................................ ................................ ..................... 102

PAGE 5

iv Exclusion Criteria ................................ ................................ .................... 104 Variable Identification and Coding ................................ ................................ ...... 108 Epidemiologic Research Design ................................ ................................ .......... 132 Data Collection Methods ................................ ................................ ..................... 133 Statistical Procedures ................................ ................................ ........................... 13 6 Study Question One ................................ ................................ ............................. 1 39 Study Question Two ................................ ................................ ............................ 142 Study Question Three ................................ ................................ .......................... 142 Preliminary Statistical Analysis ................................ ................................ ........... 143 Summary ................................ ................................ ................................ .............. 145 CHAPTER IV : PRESENTATION AND ANALYSIS OF DATA ................................ .. 147 Intr oduction ................................ ................................ ................................ .......... 147 Population Characteristics ................................ ................................ ................... 151 Demographics ................................ ................................ ................................ ...... 151 Testing the Hypothe ses ................................ ................................ ........................ 166 Hypothesis I ................................ ................................ ................................ ......... 166 Introduction ................................ ................................ .............................. 166 Potential Confounders, Multicollinearity and Interaction ....................... 168 Multinomial Logistic Regression ................................ ............................. 172 Multinomial Logistic Regression Main Effects ................................ .... 17 8 Multinomial Logistic Regression Interaction ................................ ....... 181

PAGE 6

v Multinomial Logi stic Regression Model Assessment ............................. 18 8 R andom Effect ................................ ................................ ......................... 19 1 Overall Interaction Effect on Treatment Received ................................ .. 193 Hypothesis I C onclusion ................................ ................................ .......... 2 15 Hypothesis II ................................ ....................... Introd uction and Survival Analysis ................................ ......................... 216 Hypothesis II Conclusion ................................ ................................ ......... 2 24 Hypothesis I II ................................ ................................ ................................ ....... 2 2 4 Introduction ................................ ................................ .............................. 2 2 4 CPHM Interaction Terms ................................ ................................ ......... 230 Residuals ................................ ................................ ................................ .. 247 Overall Interaction Effect on Survival ................................ ..................... 2 5 0 Hypothesis III Conclusion ................................ ................................ ....... 2 70 CHAPTER V : DISCUSSION ................................ ................................ .......................... 272 Introduction ................................ ................................ ................................ .......... 272 Assessment of the Major Findings ................................ ................................ ....... 273 Hypothesis I ................................ ................................ ................................ ......... 274 Hypothesis II ................................ ................................ ................................ ........ 275 Hypo thesis III ................................ ................................ ................................ ....... 276 Comparison and Consistency of Key Findings ................................ .................... 278 Comparison and Inconsistency of Key Findings ................................ ................. 279

PAGE 7

vi Study Limitations ................................ ................................ ................................ 282 Study Strengths ................................ ................................ ................................ .... 284 Public Health Importance ................................ ................................ .................... 289 Future Directions ................................ ................................ ................................ 291 REFERENCES ................................ ................................ ................................ ................ 293 APPENDICES ................................ ................................ ................................ ................. 310 A ppendix I : State D emographics ................................ ................................ ......... 3 11 A ppendix I I : Lu ng Cancer Distribution Tables ................................ ................... 319 A ppendix I II : Chemotherapy Agents .......... 3 2 5 A ppendix I V : Calculation of the Overall Interaction E ffect ............ 32 6 ABOUT THE A UTHOR ................................ ................................ ................... END PAGE

PAGE 8

vii LIST OF TABLES Table 1: Molecular Bi omarker for Lung Cancer (LC) ................................ ....................... 32 Table 2: AJCC TNM Staging System for Lung Tumors ................................ .................. 40 Table 3: Incidence and Mortality Rates ................................ ................................ ............. 5 8 Table 4: Lung Cancer and Food Intake Cohort Studies ................................ ..................... 63 Table 5: Lung Cancer Treatment Recommendations ................................ ........................ 80 Table 6: Selection Crite ria for State/State Cancer Registries ................................ ............ 92 Table 7: NAACCR Criteria for Gold/Silver Certification ................................ ................. 97 Table 8: Annual NAACCR Certification Designation ................................ ...................... 98 Table 9: Annual NAACCR Region and Certification ................................ ....................... 99 Table 10: Final NAACCR Eight State Cancer Registries ................................ ............... 1 0 1 Table 11: NAACCR Variable Code and Description Table 12: NAACCR Code and Description of Race ................................ ....................... 109 Table 13: NAACCR Code and Description of Spanish/Hispanic Origin ........................ 111 Table 14: NAACCR Code and Description of Laterality ................................ ................ 113 Table 15: NAA CCR Code and Description of LC Morphology ................................ ..... 11 4 Table 16: NAACCR Code and Description of LC Behavior ................................ ........... 115 Table 17: NAACCR Code and Description for Grade ................................ .................... 117 Table 18: NAACCR Code and Description Diagnostic Confirmation ............................ 119 Table 19: NAACCR Code and Descripti on for Reporting Source Type ......................... 120 Table 20: NAACCR Code and Description for Class of C ase ................................ ........ 121

PAGE 9

viii Table 21: NAACCR Code and Description for Payor at Diagnosis Table 22: NAACCR Code and Description SEER Summary Stage 1 977 ....................... 123 Table 23: NAACCR Code and Description SEER Summary Stage 2000 ....................... 124 Table 24: NAACCR Code and Description of Surgical Primary Site ............................. 125 Table 25: NAACCR Code and Description of Radiation T reatment .............................. 126 Table 26: NAACCR Code and Description for Chemotherapy ................................ ....... 127 Table 27: Derived AJCC S tage Group ................................ ................................ ............. 12 8 Table 28: NAACCR Code and Description of Follow Up Sources ................................ 131 Table 29: NAACCR Code and Description of Autopsy ................................ .................. 131 Table 30: State Cancer Registry Contact Information ................................ ..................... 134 Table 31: Final Data Lung Cancer Set Variables ................................ ............................ 149 Table 32: Classifi cation of Variables for Hypothesis Testing ................................ ......... 150 Table 33: State Cancer Registries versus Gender ................................ ........................... 152 Tab le 34: Lung Cancer Distr ibution ................................ ................................ ................ 153 Table 35: Lung Cancer Distribution ................................ ................................ ................ 155 Table 36 a : Lung Cancer Treatment Group and State ................................ ..................... 157 Table 3 6 b : Total Population for the Eight States ................................ ........................... 158 T able 37: Lung Cancer Distribution Treatment Group vs. Gender .............................. 160 Table 38: Lung Cancer Distribution Treatment Group vs. Stage ................................ 161 Table 39: Lung Cancer Distribution Treatment Group vs. Grade ................................ 163 Table 40: Lung Cancer Distri bution Treatment Group vs. Morphology ...................... 164

PAGE 10

ix Table 41: Lung Cancer Distribution Treatmen t Group versus Race ............................ 319 Table 42: Lung Cancer Distribution Treatment vs. Marital Status at Diagnosis ......... 321 Table 43: Lung Cancer Distribution Treatment vs. Age Group at Diagnosis ............... 323 T able 44 : Predictor Variable and Explanatory/Independent Variables ........................... 167 Table 45: Multicollinearity Assessment vi a Logistic R egression ................................ .... 171 Table 46: Type 3 Analysis of Effects ................................ ................................ .............. 174 Table 47 a : Main Effect of Morpho logy ................................ ................................ .......... 1 80 Table 47 b : Main E ffect of Race ................................ ................................ ..................... 18 1 Table 48: Gender and Stage Interaction Terms ................................ ............................... 18 3 Table 49: Gender and Marital Status Interaction Terms ................................ .................. 18 4 Table 50: Stage and Age Group at Diagnosis Interaction Terms ................................ .... 18 6 Table 51: Stage and Grade at Diagnosis Interaction Terms ................................ ............ 1 8 8 Table 52: Ra ndom Effect of State ................................ ................................ .................... 1 9 2 Table 53: Type III Analysis Main Effects and Interaction Terms ................................ ... 1 9 3 Table 54: Overall Variable Effect on L ung C ancer (LC) Treatment Received ............... 1 9 5 Table 55: Interaction Effect of Gender on LC Treatmen t Received ................................ 1 97 Table 56 a : Interaction Effect of Stage on LC Treatment Received ............................... 200 Table 56 b 1 : Interaction Effect of Stage on LC Treatment Received ............................ 202 Table 5b b 2 : Interaction Effect of Stage on LC Treatment Received ............................ 20 3 Table 56 b 3 : Interaction Effe ct of Stage : on LC Treatment Received ........................... 204 Table 56 c 1 : Interaction Effect of Stage I : on LC Treatment Received ......................... 206

PAGE 11

x Table 56 c 2 : Interaction Effect of Stage II on LC Treatment Received ........................ 207 Table 56 c 3 : Interaction Effect of Stage III on LC Treatment Rece ived ....................... 208 Table 56 d : Interaction Effect of Marital Status on LC Treatment Received .................. 209 Table 56 e 1 : Interaction Effect of Grade I on LC Treatment Received ......................... 210 Table 56 e 2 : Interaction Effect of G rade II on LC Treatment Received ........................ 211 Table 56 e 3 : Interact ion Effect of Grade III on LC Treatment Received ...................... 211 Table 57 a : Interaction Effect of Age Groups 4 and 5 on LC Treatment Received ........ 213 Table 57 b : Interaction Effect of Age Groups 6 and 7 on LC Treatment Received ........ 214 Table 58: Lung Cancer Survival ................................ ................................ ...................... 2 17 Tabl e 59: Survival Data for Lung Cancer Cases ................................ ............................. 2 19 Table 59 a : Extracted Life Table Survival Parameter Results ................................ ......... 2 2 0 Table 60: Gender Survival Estimates (in months) ................................ ........................... 2 21 Table 61: Life Tables Test of Equality over Strata ................................ ....................... 227 T a ble 61 a : The Cox Proportional Haza rds Model (CPHM 1 ) ................................ ......... 2 29 Table 62: Hazard Ratios and 95% Confidence Intervals ................................ ................. 2 32 Table 63: Hazard Ratios and 95% Confidence Intervals ................................ ................. 2 36 Table 64: Hazard Ratios and 95% Confidence Intervals ................................ ................. 240 Table 65: Hazard Ratios and 95% Confidence Intervals ................................ ................. 2 41 Table 66: Hazard Ratios and 95% Confidence Intervals ................................ ................. 2 46 Table 67: Hazard Ratios and 95% Confidence Intervals ................................ ................. 247 Table 68: Overall Effect on Survival ................................ ................................ ............... 252

PAGE 12

xi Table 69 a : Overall Effect of Gender on Survival ................................ ........................... 255 Table 69 b: Overall Effect of Gender on Survival ................................ ........................... 2 55 T able 69 c: Overall Effect of Morphology on Survival ................................ ................... 256 Tabl e 69 d: Overall Effect of Morphology on Survival ................................ .................. 257 Tabl e 69 e: Overall Effect of Morphology on Survival ................................ ................... 257 Tabl e 69 f: Overall Effect of Grade on Survival ................................ ............................. 258 Tabl e 69 g: Overall Effect of Grad e on Survival ................................ ............................. 259 Tabl e 69 h: Overall Effect of Stage on Survival ................................ ............................. 261 Tabl e 69 i: Overall Effect of Stage on Survival ................................ .............................. 262 Ta ble 69 j: Overall Effect of Stage on Survival ................................ .............................. 263 Tabl e 69 k: Overall Effect of Age Group on Survival ................................ .................... 264 Tabl e 69 l: Overall Effect of Age Group on Survival ................................ ..................... 265 Tabl e 69 m: Overall Effect of Race on Survival ................................ ............................. 266 Tabl e 69 n: Overall Effect of Treatment Type on Survival ................................ ............. 268 Table 69 o: Overall Effect of Treatment Type on Survival ................................ ............. 268 Table 69 p: Overall Effect of Treatment Type on Survi val ................................ ............. 268 Table 69 q: Overall Effect of Treatment Type on Survival ................................ ............. 269 Table 69 r: Overall Effect of Treatment Type on Survival ................................ ............. 269 Table 70: Geographic Area: Florida ................................ ................................ ............... 311 Table 71: Geographic Area: Idaho ................................ ................................ .................. 312 Table 72: Geographic Area: Indiana ................................ ................................ ............... 313

PAGE 13

xii Table 73 : Geographic Area: Massachusetts ................................ ................................ .... 314 Table 74: Geographic Area: Nebraska ................................ ................................ ............ 315 Table 75: Geographic Area: Oregon ................................ ................................ ............... 316 Table 76: Geographic Area: Rhode Island ................................ ................................ ...... 317 Table 77: Geographic Area: South Carolina ................................ ................................ ... 318 Table 78 : Chemotherapy Agents for Lung Cancer ................................ .......................... 325

PAGE 14

xiii LIST OF FIGURES Fi gure 1 : The Respiratory System ................................ ................................ ..................... 8 Figure 2 : 2006 Estimated US Cancer Cases ................................ ................................ .... 20 F igure 3 : 2006 Estimated US Cancer Deaths ................................ ................................ .. 20 Figure 4 : US Women Cancer Death Rates ................................ ................................ ....... 2 1 Figure 5 : US Men Cancer Death Rates ................................ ................................ ............. 2 1 Figure 6 : Structure of Morphology Code ................................ ................................ ......... 34 Fi gure 7 : Structure of a Complete ICD O Code ................................ ............................... 3 4 Figure 8: ICD O 3 Site (Lung) Codes ................................ ................................ ............. 35 F igure 9 : Lung Anatomy with ICD O 2/3 Codes ................................ ............................ 3 5 Figure 10: State Selection Process ................................ ................................ .................... 9 3 Figure 1 1: Residual Analysis: Treatment Groups I, II, III ................................ ............... 189 Figure 1 2: Residua l Analysis: Treatment Groups IV, V ................................ .................. 190 Figure 13: Residual Analysis: Treatment Groups V I V II ................................ ............... 191 Figure 1 4 : Life Table Method ................................ ................................ .......................... 218 Figure 1 5 : Cumulative Hazard Function (CHF) ................................ .............................. 222 Figure 16 : Transformation of the CHF ................................ ................................ ............ 223 Figure 1 7 : Residual Testing of the L ung Cancer Distribution ................................ ......... 249 Figure 1 8 : Residual Testing of the Lung Cancer Distribution ................................ ......... 250

PAGE 15

xiv LIST OF ABBREVIATIONS A CR American Cancer Society ATBC Alpha Tocopherol, Beta Carotene Cancer Prevention Study BMI Body Mass Index CARET beta Carotene and Re tinol Efficiency Trial CI Confidence Interval CO 2 Carbon Dioxide DHEW Department of Health, Education and Welfare ICD 9 International Classification of Disease: 9 th Edition ICD 10 International Classification of Disease: 10 th Edition LC Lung Ca ncer NTLDRI National Tuberculosis and Lung Dise ases Research Institute OR Odds Ratio PHS Physicians' Health Study SAS Statistical Analysis Software SEER Surveillance, Epidemiology, and End Result Program SES Socioeconomic Status

PAGE 16

xv ABSTRACT Th e objective s of this research w ere to test treatment and survival differences between women and men with lung cancer as there is minimal investigation in the literature. Three research questions were developed with statistical testing for gender differenc es based on similar cancer type, stage, treatment assignment and survival. D ata for 44,863 primary lung cancer cases were collected from eight U.S. state based cancer registries to investigate the research questions The lung cancer incidence data includ ed the morphological cell types of adenocarcinoma (AC) ; squamous cell carcinoma (SCC) ; large cell carcinoma (LCC) and small cell carcinoma (SCC). Stage, grade, treatment type, as well as, individual characteristics such as gender, age at diagnosis, marita l status at diagnosis and race were other variables obtained to be included in the statistical models Reporting the overall effect for lung cancer gender specific treatment differences or survival has not been demonstrated in the literature to the autho By convention, main effects and interaction effects are reported in the literature; without including an evaluation the overall effect of a variable on the outcome, possible misinterpretations could be made. For example, s Proportional Hazards model when the interaction effect of gender and treatment type received was examined, females were at an increased risk for death by as much 29% as compared to males (HR = 1.18, 95% CI 1.09 1.29). But when the gender effect on sur vival was assessed th e re was a n increase in females survivorship as compared to males by as much as 28 % (HR =

PAGE 17

xvi 0.80, 95% CI 0.72 0.97 ). In conclusion, by using a unique statistical approach, s tatistically significant Odds Ratios and Hazard Ratios w er e demonstrated for the research data set w hen the overall interaction effect on the outcome was examined Recommendations to health care practitio ners include adhering to current guidelines e.g. American Medical Association for lung cancer treatments S tandard treatment protocols were not always followed for early stage disease e.g. females versus males with stage I lung cancer were 1.71 times more likely to receive chemotherapy in combination with radiation therapy versus a standard first treatment cou rse of surgery (OR = 1.71, 95% CI 1.06 2.78) Also, d epending on the lung cancer morphology and lung cancer treatment, female s as compared to males could exhibit an increase in survivorship by as much as 28 %. To improve the results of medical care deci sions for lung cancer, clinicians may find the information presented in this study useful and encourage further research on which treatment increases survival for both men and women.

PAGE 18

1 CHAPTER I: INTRODUCTION Background There are many histological types o f lung cancer and finding an optimum treatment regimen is a challenge. Lung cancer typically is classified into two major divisions: small cell lung cancer (SCLC or oat cell carcinoma) and non small cell lung cancer (NSCLC) 2 5 SCLC accounts for approximately 20% of all the lung cancer cases, whereas about 80% of all lung cancer cases are NS CLC. There are many types of NSCLC but the three major histological classifications are adenocarcinoma, large cell carcinoma, and squamous cell carcinoma 6, 7 The treatment modalities for small cell lung cancer versus non small lung cancer are differen t due to the biological response of the particular cancer cell type to various treatment regimens 8 11 The medical interventions for each histological type can include any combination of treatment modalities such as surgery, radiation therapy, and/or chemotherapy. Adding to the complexity of lung cancer is that the incidence, prevalence, and survival rates are also dissimilar for the specific histological type 1, 2, 12, 13 One prognostic factor treatment/ modality exploration is the relationship between lung cancer treatment(s) and gender. There is limited research rega rding if the treatment modality, e.g. radiation therapy, s urgery, chemotherapy received is dependent upon being a woman wi th lung cancer as compared to a ma n with lung cancer. This is of particular interest because of all the various types of cancers and treatments available lung cancer has

PAGE 19

2 become the leading cause of death for women as there has been a 600% increase for women with lung cancer from 1930 to 1997 44 Any effect which gender exerts in the decision regarding which lung cancer treatment modality decided upon must be disentangled from other prognos tic factors. The study question(s) of this research attempted to enumerate the risk of being a woman with lung cancer and type of treatment received compared to a man with lung cancer and the type of treatment modality that he receives. An assessment was made to determine if a statistically significant association between gender and treatment modality exists. Another aspect of gender differences that was investigat ed included the impact on survivorship between women and men with lung cancer. Stratified analysis was based on the histological type, stage, grade, gender and the treatment modality or treatment modalities received in an attempt to investigate treatment effects on survival. Much of the scientific literature on lung cancer research does not ad dress survival and the relationship gender has to play due to the effects of specific histological lung cancer types, stage or progression of LC, and grade on gender and survival. The purpose of this research study is to provide a quantitative assessment of the outcome (survival) for women as compared to men based on the particular treatment received for lung cancer 14 17 Minorities will also be included in the subject selection; it is important not to exclude minorities as they can provide valuable epidemiologic information. In an attempt to facilitate minority research, United States government agencies, e.g. the National Cancer Institute (NCI), now mandate the inclusion of women, children, and

PAGE 20

3 minorities if government fun ding is provided for the study 18 20 For the purposes of this research, minorities are being included since the treatment modality selected for the treat ment of lung cancer may be dependent upon race as well as gender 21 In other words, race may or may not play a role in the treatment modality utilized for lung cancer. Although this research is primarily focused on what treatment modalities are utilized for men as compared to women the imp act of gender on lung cancer survival will also be considered There are several reasons why is gender important as risk factor for lung cancer First, according to the 2001 report by the Surgeon General 44 female lung cancer mortality increasing by 600% 44 Secondly, the causal pathways of lung cancer development are blurred for women 8, 13 For example, t he causes of lung cancer among women seemingly different from men, are still not resolved 3, 8, 13 One possible answer to this question is m uch of the current knowledge and treat ment patterns for lung cancer are based on research primarily done on men. Previously, the association between being a woman and the risk of lung cancer was considered neg ligible as report ed in the 1964 Report of the Advisory Committee to the Surgeon Gene ral DHEW Publication N umber PHS 1103 23 But as behavior and other temporal changes, such as cigarette smoking have occurred over the past several decades for women, lung cancer incidence and increased mortality rates of lung cancer 24, 76, 77 Women hist orically have been excluded from clinical trials or if included, the data was not analyzed 22 If women are at a greater risk for lung cancer than men at the same level of smoking one result of women being excluded historically from

PAGE 21

4 research studies is even there is no evidence in the literature to support this; results have been conflicting and limited 22 A 1964 report of the Surgeon General on W omen and S moking 23 did not reach any conclusions concerning what role gender difference may play in the development of lung cancer. The 1964 report did conclude that although smoking was risk factor for lung cancer in men, smoking was not a risk factor for lung cancer in women as there was not enough scientific evidence to establish causality 24 What was not known at the time of the 1964 report was the temporal effect due to when women started smoking on a large scale and the development of lung cancer (lag time of approximately twenty years). Hypotheses have been developed based on possible physiological response s to carcinogens and hormonal related differences in women as compared to men but inconsistent results in the literature remains 8, 15, 17, 25 27 Lung cancer is the leading cause of death of all cancers in both men and women in the United States 28 The overall lifetime risk in women is 1 in 17 and for men 1 in 13 for lung cancer development 27, 29, 30 Lung cancer has an extremely low 5 year survival rate of 15%. The primary cause of lung canc er is due to smoking cigarettes ; smoking is estimated as being a causal factor in 80 90 % of all lung cancer cases 14, 15 Some of the l i terature reports the susceptibility for lung cancer in women is different when compared to men with women by demonstrating an increased risk for lung cancer 8, 14, 15, 31, 32 Another source for concern for women is second hand smoke; of the indi viduals that die from that exposure, 65 % are women 33 This could possibly indicate hormonal differences may make women more

PAGE 22

5 susceptible to smoke. Sex differences in survival and susceptibility have been linked to estrogen as a lung cancer risk factor 229 Research Que stions Given what present day research has and has not found concerning the treatment of lung cancer and the role gender has played in the selection of the treatment modality, the following research questions will be addressed in this dissertation: Questi on One : Do men and women with the same histological type and stage/grade of lung cancer receive the same treatment modality? Question Two : Are there differences in survival between men and women regardless of the treatment modality received? Question Three : Do men and women with the same histological type, stage/grade of lung cancer, and same treatment modality, have comparable survival? As stated in the abstract and in the background, the study or research question(s) will focus on the association between the treatment received by women with lung cancer as compared to that received by men. The study will investigate the overall survival patterns based on the treatment that a woman with lung cancer receives versus a man.

PAGE 23

6 C HAPTER TWO: LITERATURE REVIEW O verview of the Lungs Anatomy and Physiology The lungs, part of the respiratory system, are coned shaped, sponge like, and highly elastic organs located in the chest. The functions of the respiratory system and in particular, the lungs, include gas exchang e, moisturizing the inhaled air, stabilizing the temperature of all air to body temperature, and filtering harmful substances 34, 35 As shown in Figure 5, the respiratory system includes the nasal cavity, the windpipe or trachea, and two lungs. The upper tract of the respiratory system includes the mouth or oral cavity, the nasal cavity, and the trachea 36 The lower tract of the respiratory system consists of lungs, bronchi, and alveoli. Inspiration and expiration are the two phases of respiration or breathing. During each phase of respiration, the volume or dimens ions of the chest cavity is changed, i.e. increased lung volume (inspiration) or decreased lung volume (expiration) 36, 37 Air entering into the body via the nose or mouth, contains approximately 21% oxygen with no carbon dioxide. The air is drawn into the trachea and bronchi, and then enters the lungs through the left or right bronchi. Air entering into the main branch of the bronchi will travel int o smaller bronchi which further divide into smaller, complex tubes called the bronchioles 36, 37 Mucus is secreted by the inner lining of the larger bronchial tubes. One of the purposes of the secretion is to filter or trap dirt from the air. In a continuous, sweeping process, the mucus is expelled from the lungs by cilia; cilia are

PAGE 24

7 similar to hair or brush like structures 38 Coughing is another method by which mucus is removed from the lungs. The final or most distal ends of the bronchioles are connected to small air sacs called alveoli. The exchange of gases occurs in the alveoli. T he alveoli are very thinly walled, balloon like structures that expand upon inspiration and relax or deflate upon expiration 37, 38 Each alveolus is surrounded by small blood vessels called capillaries. When the concentration of dissolved oxygen is greater in the alveoli than in the capillaries, oxygen diffuses across the alveoli walls into the blo od plasma contained in the capillaries. An increased concentration of CO 2 in the blood results in carbon dioxide diffusing from the capillaries into the alveoli. At the time air is exhaled, it contains approximately 16% oxygen and 4.5% carbon dioxide 37, 38 As previously described, the exchange of oxygen and carbon dioxide occurs in the lungs. Ea ch lung is identified by the apex, lobes, and base. The left lung is comprised of 2 lobe or sections; typically weighing 625 grams 34, 39 The right lung has three lobes, approximately 567 grams. The left lung is smaller than th e right to accommodate the heart and other structures in the mediastinum. The lungs have a surface area approximately equal to the size of a tennis court and while at rest, the entire body blood supply or blood volume, five liters, passes through the lung s each minute 38

PAGE 25

8 Figure 1: The Respiratory System Source : webschoolsoluti ons.com/patts/systems/lungs.htm The Disease of Interest: Lung Cancer Any obstruction of air flow through th e bronchial tree or at the alveoli can cause serious functional limitations or even death 37, 38 Besides the various diseases which can obstruct airflow and affect the cellular respiration the lung s can become cancerous 10 Physiological changes in the lung tissue where the lung becomes cancerous can be defined as an uncontrolled cell growth in the lung forming clumps of tissue referred to as maligna nt tumors 37 Exposure to carcinogens, such as those present in tobacco smoke, immediately causes changes to the tissue l ining the bronchi of the lungs (the bronchial mucous membrane ) the more cumulative damage to the lung tissue, the greater the probability a tumor will develop 9, 10 The non small cell lung cancers (NSCLC) are g rouped together because their prognosis and management is roughly identical 2, 9, 54

PAGE 26

9 There are 3 major subtypes of NSCLC: squamous, large cell, and adenocarcinoma 1, 2, 55 Squamous cell carcinoma starts in the larger breathing tubes but grows slower, this means that the size of these tumors vary when the diagnosis is made. Adenocarcinoma ( the slower growing type forms alveolar cell cancer ) starts near the gas exchanging surface of the lung 56 It is less closely associated with smoking. Large cell carcinoma is a fast growing form that grows near the surface of the lung 4, 57 It is primarily a diagnosis of exclusion, and when more investigation is done, it is usually reclassifi ed to squamous cell carcinoma or adenocarcinoma 56 Small cell carcinoma (SCLC, also called "oat cell carcinoma") is the less common form of lung cancer. Approximately 20% of all prim ary lung cancer diagnosed are small cell lung cancer and account for 30,000 to 35,000 cases per year in the United States 13, 28 Small cell LC tends to start in the larger breathing tubes and progresses rapidly becoming quite large 6, 10, 58, 59 SCLC is more sensitive to chemotherapy, but carries a worse prognosis and is often metastatic at presentation 2, 3, 33 This type of lung cancer is strongly associated with smoking 4 Exposure to carcinogens, such as those present in tobacco smoke, immediately causes cumulative changes to the tissue lining the bronchi of the lungs (the bronchial mucous membrane ) and the more tissue that gets damaged, the greater the probability a tumor will develop 4, 37 Squamous cell carcinoma usually is diagnosed after the disease has spread 1, 5, 12, 13 The overall prognosis for all non small cell lung cancers is poor, with a five year survival rate of about 15% 11, 13, 60 The survival rate is higher (close to 50%) when the cancer is detected and treated early 13 Survival rates after surgery vary 7, 43, 54, 61 63 For those with stage I

PAGE 27

10 disease, the five year survival rate is about 47% 13, 64 For those with stage III disease, the five year s urvival rate is 8% 2, 13, 33, 64 Even when surgery and other therapies are initially successful, there is a high risk of the cancer reoccurring 4, 27, 32, 65 This reflects the fact that squamous cell carcinoma is rarely restricted to just one area. Squamous cell car cinoma readily spreads to other parts of the body 4, 30, 66 Cancer is a multistep progression of changes or phases that occur in the genes 43, 52, 67 71 The genotypic changes are characterized by the loss of normal cellular differentiation and an alteration in tissue morphology due to an increase of unr epaired DNA damage and the formation of abnormal genomic variants 10 Lung cancer can result from an exposure of a susceptible host to carcinogenic agents; these exposures cause progressive changes in the cell from metaplasia, to atypia and dysplasia, then dev eloping into a carcinoma in situ and invasive cancer 72 The changes that occur on the cellular level are variable from individual to individual, and not all neoplasms follow the same progress 4 Me taplasia, the first phase of cancer development, is the transformation of a mature differentiated cell type into a different mature differentiated cell type 4 This transformation is in response to an injury or insult at a cellular level which can make the tissues more susceptible to a malignan t alteration. Atypia is defined as an abnormality associated with a precancerous process. An atypical cell (atypia) can also be an indication of an infection or irritation 4, 37 Atypia can be cau sed by a chronic irritation and this has been shown increases the probability of premalignant lesions 9 Dysplasia is typically an irreversible co ndition or change in the cell that is a precursor of invasive

PAGE 28

11 epithelial tumors. There levels or grades of dysplasia and high grade dysplasia can be difficult to distinguish from carcinoma in situ during histologic examination 4, 37 Exposures of Interest: Gender and Lung Cancer Treatment Modality There is limited research regarding the survival of women with lung cancer and the treatment received compared to the survival of men with lung cancer an d the treatment men receive 12, 40 45 Presently, there are no quantitative results that show whether there is a statistically significant difference regarding survival due to a particular treatment for women as compared to men having the same histological type and stage of lung cancer 46 48 The goal of this research is to investigate the exposures, gender and treatment modality and their effect on the outcome, survival. Several research questions must be answered in order to e valuate the relationship between these variables. Belani et.al., 2007, in the article emerging trends in clinical research emerging findings in the scientific literature reveal gender specific differences in cancer prognosis 41 T he authors expressed an urgent need to increase research and funding to improve lung cancer care, women in particular 41 Ringer et al. (2005) in the article "Influence of sex on lung cancer histology, stage, and survival in a Midwestern United States Tumor Registry." identified differences between me n and women with regard to lung cancer type, stage at diagnosis, and survival in a single hospital system cancer registry The study design was a r etrospective c ohort with a t arget population based on case i nformation from a lung

PAGE 29

12 cancer tumor registry at a single hospital system composed of 2 independent hospitals i n the Midwestern United States 27 This database included all patients from 1996 to 2002 with known lung cancer or abnormal findings on chest radiography or computed tomography (N=2618). Patient s with adenocarcinoma or squamous cell, small cell, or large cell carcinoma were included in the study. A total of 1216 men and 997 women were included in the study. The authors found n o significant difference in age between sexes at diagnosis 27 Women w ere significantly more likely to have adenocarcinoma or small cell carcinoma but less likely to have squamous cell carcinoma compared with men. There were no significant differences between sexes in the incidence of large cell carcinoma. No significant d ifferences were found between men and women in terms of cancer stage at diagnosis 27 There were significant differences in survival between the histologic types at years 3, 4, and 5. Only patients with stage I disease showed a difference between sexes an d only for years 2, 3, 4, and 5. This study did not investigate the impact of treatment modality on survival, gender, histological type and stage of lung cancer. Women were found to have a decreased survival with late stage lung cancer as compared to men 2 7 but there was no expansion of the results based on the type of treatment received for women and men. 8 gender differences in survival were examined. The authors 8 cited several articles that reported on cardiovascu lar disease and the survival advantage for men as compared to women research attempted to

PAGE 30

13 identify gender disparities in lung cancer survival. To test the hypothesis of a gender difference in lung cancer survival, a retrospective coh ort study of 104 women and 104 men was conducted. Women were found to have a higher incidence of small cell lung cancer (25% versus 12% as compared to men); whereas men had a greater percentage of squamous cell carcinoma (51% versus 38% as compared to wom en) 8 The authors noted there were no statistically significant survival differences between women and men but women were found to live, on average, 6 months longer th en men (mean survival women = 24 months, mean survival men = 18 months). Ouellette, et. al. reported when s tratified ana lysis based on the stage of lung cancer (Stage I, II, IIIa, IIIb, and IV) was assessed these two groups with a coefficient according to stage, there was a survival advantage in 8 The authors reported that this may be contributed to an intrinsic factor, e.g. hormones. Ouellette, et. al. concluded the overall survival between men and women was not statistically significant but that there was a significant survival difference between men and women with lung cancer when stratified on stage. The question about gender differences and lung cancer survival has not been resolved in the literature as conflicting results still exist 40, 41, 49 52 A recently published article investigating gender differences and survival by Wisnivesky and Halm, 2007, Differences in Lung Cancer Survival: Do Tumors Behave Differently in Elderly men 53 The study was based on SEER data collected from men and women diagnosed

PAGE 31

14 between 1991 and 1999 (N = 18,967) with stage I and stage II non smal l cell lung cancer. It was shown that for early stages of lung cancer, that women have a better overall and relative survival as compared to men (p = 0.001). The authors noted women as compared to men had a greater probability of being diagnosed with ade nocarcinoma, tended to be diagnosed at an earlier age, and when the disease had not metastized (localized) 53 Epidemiology Epidemiology is utilized to monitoring the consequences of an intervention and is used in the development of hypo theses for risk factors 73, 74 Epidemiological methods are used to study lung cancer for the identification of the disease frequency, determinants, and distribution of lung cancer in human populations 73, 74 For example, t here has bee n an increase of 6 00% in mortality for women with lung cancer since 193 0 28, 40, 60 and without monitoring or the identification of the disease frequency in epidemiological terms this epidemic rise in lung cancer mortality 4 4 24, 75 77 may not have been identified Alberg et. al, (2005) reported that in the 20 th century of the United States the lung cancer 40 T he rates of lung cancer in women were shown to have a differential increase in lung cancer incidence and mortality over time as compared to men 40 L ung cancer rates have peaked for men but the rates for women are still increasing i n many regions of the world 5, 16, 30, 65, 78, 79 While the gap between lung cancer gender differences is narrowing, the differences for in incidence an d mortality rates are declining 45, 46, 66, 80, 81 According to

PAGE 32

15 the International Agency for Research on Cancer (IARC ), rates of all lung cancer types among women and adenocarcinoma of the lung in men continue to rise in many Western countries 5 Worldwide, lung c ancer is the 10th leading cause of death and is the leading cause o f death for all types of cancer 5 The 5 yea r relative survival remains low; approximately 10% in Europe In Developing Countries, the incidence of smoking related lung cancer is rising rapidly 5, 30, 82 Countr ies such as China are expected to see a marked increase in lung cancer cases as smoking is exceedingly common 5, 78 Devesa and Bray 2005, reported recent total lung cancer incidence rates among males varied by 4 fold, from 83.6 among U.S. Blacks to 21.1 in Sweden 30 Rates in the Nordic co untries, which varied by 2 fold from a high in Denmark to a low in Sweden, still were generally lower than in other parts of Europe, where the incidence rate was highest in the Netherlands 30 Lung cancer r ates in Italy, Slovenia and France were higher as compa red to U.S. Whites or Canad ian LC incidence. The authors also noted that among females, recent incidence rates varied by almost 8 fold, with the highest among U.S. Blacks (35.8) and the lowest in Spain (4.6) 30 The ranking of rates among females paralleled th at in males, with the exception of Switzerland. Lung cancer r ates everywhere were higher among males than females 30 Male to female rate ratios varied from less than 2 in Iceland, U.S. Whites, Canada, Denmark and Sweden to more than 6 in Slovenia, Italy, and F rance and more than 10 in Spain 13, 30 Henschke et. al. (2006) reported that the US c ancer rates for men and women in their research show ed a dose (pack years) incidence (lung cancer) threshold as there was a biological gradient associated with increased pack

PAGE 33

16 ye ars with an increased risk of lung cancer 174 In the United States, t he American Cancer Society estimated that there were 92,305 new cases of lung cancer in men and 79,544 new cases among women in 2006 2 The majority of cancer deaths among women and men are attributed to lung cancer 2 According to the American Cancer Society, a pproximately 60% of newly diagnosed lung cancer cases die within the first year of diagnosis 2 The 5 year relative survival rate is approximately 15% in the U nited S tates. The prevalence rates of smoking as reflected in the National H ealth Interviews, Current Population Survey, notes that smoking attributable cancer mortality for males is approximately 90% and 78% for females 84 86 Current literature about smoking habits (age when started smoking, number of cigarettes daily, duration frequency of inhalation, use of dark tobacco, and non filter cigarettes) 87 89 notes that a smoker is twenty two times more likely to die from lung cancer than a nonsmoker 86 the book Epidemiology of Lung Cancer : Academic Press; 1998, a study from the Saskatchewan Cancer Foundation (a population based cancer registry) was referenced by Thomas J. Mason. He noted endogenous and exogenous factors may contribute to the development of primary lung cancer in women 83 Endogenous factors can be produced or can be synthesized within an organ in the body ; exogenous factors are agents or factors from outside the body (cigarette smoke) 37 Zang and Wynder conducted a hospit al based prospective, case control study on data collected from 1995 through 1995 that included 21,057 males as controls and 14,448

PAGE 34

17 female controls that were originally diagnosed with non smoking related diseases 81 The authors found that at the same level of lifelong exposure to cigarette smoke, women had a 1.5 times greater risk of developing lung cancer as compared to men 90 There was a statistically significant difference in the incidence of adenocarcinoma; females were at a higher risk of developing adenoc arcinoma versus males independent of tar yield per cigarette 90 Zang and Wynder noted a sta tistically significant difference between squ amous cell carcinoma for women as compared to men dependent upon the level of to tal tar per cigarette (> 6 kg). W omen were found to develop primary lung cancer at earlier age as compared to men, yet women smoked fewer cigarettes for a shorter time than me n 81 Smoking patterns have changed over the past thirty years and the change in the dominant histologic lung cancer classifi cations, possible differences between gender emerges 83 Lung cancer has a multivariable etiology and there are specific risks associated with the type of lung cancer 3, 91 95 These secular trends can provide a clue to the understanding of lung cancer and future research for the impact on diagnosis, treatment, and outcome 48 Other studies that identify patterns of risk by the histologic types include an article by Devesa, et al., 2005, utilizing data from the International Agency for Research on Cancer (IARC) databases 30 Morphology specific incidence data noted that the rates of all lung cancer types are increasing for women and adenocarcinoma is rising for men 30 This trend continues even with the decrease in prevalence of smokin g and the use of filtered and low tar cigarettes 13 These finding are

PAGE 35

18 consistent with current literature as the secular trends in histologic type with the annual rise in the incidence of adenocarcinoma 10, 96 98 Govindan et al., 2006 in the article anging Epidemiology o f Small Cell Lung Cancer in t he United States o ver t he Last 30 Years: Analysis o f t he Surveillance, Epidemiologic, a nd End Results Database that the proportion of women with SCLC increased from 28% in 1973 to 50% in 2002 99 When SCLC was compared to a ll lung cancer histologic types there was a decreased of SCLC from 17% in 1986 to 13 % in 2002 99 The authors also noted that although there was an overall decrease in small cell carcinoma, survival had not improved. Stockwell, et al., 1990, found t he histological type of lung cancer var ied by age, sex and the use of cigarettes ; this was based on observations from a population based cancer registry in Florida 96 A dose threshold for the amount of cigarettes smoked and the risk of lung cancer was not statistically significant. The author s noted that adenocarcinomas were more frequent in the younger aged population (< 60 years of age) for both genders. Men who smoked had a higher risk for squamous cell carcinoma whereas females very more at risk for small cell carcinoma 96 Adenocarcinoma was the most frequently encountered histological type for women who were nonsmokers 96 As there are differences in the incidence of histologic lung cancer types based on smoking patterns, the rates of incidence and mortality for lung cancer differ accor ding to regional areas across the United States 4 Geographic mapping of lung National Cancer Institute in Atlanta, Georgia 83 This novel approach allowed for the

PAGE 36

19 identif ication of regional differences in lung cancer rates; with this information Public Health resources were directed to areas with increased rates for purposes of prevention and monitoring of trends. There are differences in smoking attributable risk between males (>90%) and females (<80%) although ratio between male smoking rates and female smoking rate approach unity 83 These homogeneous and heterogeneous patterns of lung cancer etiology require further identificat ion of factors other than smoking as to quantify the risk for the four major histologic types of lung cancer. Currently, lung cancer incidence in women is approximately equal to that of men as reported by the American Cancer Society (see Figure 6 below) 2 Lung cancer in the United States is the most common cause of cancer death in women; today the mortalit y rate is more than two times what it was 25 years ago 2 In Figure 7 below, the estimated number of U.S. l ung cancer deaths is given for 2006. Cancer of the lung and bronchus is the most common and most fatal cancer in men (31%), followed by prostate cancer (10%), and colon & rectum cancer (10%) 2 The major killer of women from a cancer specific cause is lung cancer (27%), breast cancer (15%), and colon & rectum (10%) are the leading sites of cancer death 2 the death rate due to lung ca ncer steadily increased with a dramatic rise in mortality rates in 1965 (see Figure 8) 2 4, 79, 100, 101 Lung cancer mortality rates in women have reached a plateau since 1998 102 As shown in the Figure 9 below, the death rate from l ung cancer appears t o have peaked in 1990 for men 2 T he age adjusted lung cancer death rate in

PAGE 37

20 men has been decreasing since 1990. P rior to 1990 the major increase in cancer death rates for men was attributable to lung cancer 2 When comparing the mortality rates between men and women (Figure 8 and Figure 9), the temporal effect of gender specific smoking patterns associated with the increase in lung cancer mortality is clearly demonstrated 2, 102 Figure 2: 2006 Estimated US Cancer Cases Figure 3: 2006 Estimated US Cancer Deaths

PAGE 38

21 Figure 4: US Women Cancer Death Rates Figure 5: US Men Cancer Death Rates There are many other aspects of the association of the risk factor (gender) to the outcome of lung cancer. These aspects of a wom can include but are not limited to: smoking patterns (cigarette type, depth of inhalation, number of pack years), gender, occupation, dietary factors, nutrition, hormonal factors, a ir pollution obesity, and radia tion effects 45, 49, 62, 67, 81, 103 110 Incidence and prevalence

PAGE 39

22 rates are used in the evaluation of the overall disease (lung cancer) trends. These statistics are available on government sponsored data bases su ch as the National Cancer Registry SEER registries, state cancer registries, e.g. members of the North American Association of Central Cancer Registries, Inc. or other state registries, such as the Washington State Department of Health Occupational Mortali ty Data Base 1, 13 Impact on Health C are Resources has direct and i ndirect costs associated with lung cancer diagnosis and treatment. Some of the direct costs are medical care which include hospitalization, doctors visits, home health care, hospice care, and treatment modalities such as radiation therapy, surgery, and ch emotherapy 10 Direct non medical costs associated with lung cancer can consist of additional costs incurred due to changes necessary in the living conditions of the patie nt. Other considerations are the indirect costs which can be difficult to grasp the scope of exactly what can be involved with patient care 111 113 These costs such as time spent seeking med ical attention, time lost from work (lost productivity), or job replacement costs cannot be directly measured in some instances 111 113 The costs associated with lung cancer care are enormous according to the National Heart Lung & Blood Institute (NHLBI). I n 2003, there were 1.5 million deaths representing 47% of all deaths in the United States 111, 113, 114 These deaths were as result mainly of three disease processes; lung cardiovascular, and blood diagnosis. By 2006 these three diseases are expected to

PAGE 40

23 excee d $560 billion of medical costs 115 Lung cancer costs in 2004, shows medical expenditures as approximately 10 billion annually, accordi ng to the Centers for Medicare and Medicaid Services (CMS) 115 Lung cancer represents over 13% of the total cancer care costs for 2004. The non medical total or personal care exceeded 250 billion for the same time period. Lung cancer is one cancer which is more expensive to diagnose and treat because of the histologic types 64, 116 118 Many successfully treated cancer types have an early detection program or screening program for the early diagnosis of c ancer 4, 117, 119 Unfortunately, lung cancer does not have an effective screening tool and typically is not diagnosed until it has spread outside the diseased organ 64, 116 118 As the majority of lung cancer is diagnosed at later stages, the associated healthcare costs and resources required are increased 111, 113 In 2007, it is estimated that the total healthcare cost (HCC) for lung cancer will be 21 billion. According to CMS, lung cancer care and treatment accounts for 10% of the total US healthcare costs 115 The United States F ederal Office of the Actuary estimates that by 2016 every 20 cents of every dollar will go towards health care by 2016 115, 120 The annual forecast by a division of the Centers for Medicare and Medicaid Services (CMS) predicts a 10 year increase o f approximately 2 trillion dollars ( 2.1 trillion to $4 trillion ) for spending on health care in the United States 115, 120 This represents an ever increase amount of healthcare resources going to the detection and treatment of lung cancer. Using the estimated 4 trillion which is expected in 2016 with the total HCC being 400 billion, the lung cancer port ion of 13.3% would make the projected lung cancer

PAGE 41

24 portion over 53 billion 115 If the same relationship as seen in medical costs, estima ted non medical costs could exceed over 1 trillion. The projected lung cancer incidence and mortality rates are expected to increase as 77 million baby boomers will move into their rch to retirement of those who were heavy tobacco use will be responsible for even higher costs after 2016 115, 120 Origin The site of origin of lung cancer refers to the type of tissue from which the cancer cells develops 9, 121 Lung cancer is categorized by site of origin into hilar and peripheral types; as the structures where the disease originates are different 37 The majority of early cancers in the hilar region are squamous cell types, whereas many early stage lung cancers in the peripheral areas are adenoca rcinoma 121 Adenocarcinoma originates in glandular tissue; whereas a carcinoma originates in the tissu e that lines the organs and tubes of the lungs called epithelial tissue 122 NSCLC adenocarcinoma and large cell carcinoma typically are located in the peripheral of the lungs and can present as solitary nodules or masses 37 Squamous cell carcinoma and small cell carcinoma are normally found in the central portion of the lungs and can be misdiagnosed as a collapsed lu ng (atelectasis) or pneumonia (an inflammation of the lungs) 37 Small cell carcinoma is normally located in the mainstem which are a component of the bronchial epithelium 35

PAGE 42

25 The tissue layers are comprised of cells th at are similar in their structure and perform common functions. The intercellular material, e.g. RNA, DNA, are contained within the cells; genetic material is found within the intercellular material for that cell type. As a human embryo develops, three p rimary germ layers provide the basis for body tissue and organ formation 35 The three germ layers are the ectoderm, the endoderm, and the mesoderm. The e ctoderm and endoderm layers are considered epithelial tissue. The epithelial tissue from the endoderm lines the respiratory tract (the lungs and the air passageways from the pharynx to the lungs). Clinical Signs and Symptoms of Lung Cancer Early dete ction of cancer is credited with an increased survival; unfortunately for lung cancer, there is no early detection program that has clinically proven long term success 4, 117, 118, 123 One impact due to the lack of early detection for lung cancer is that lung cancer has become one of the most lethal of all cancers; mortality rates for lung cancer have surpassed colorectal, breast and prostate cancer combine d 4 The main difficulty in the diagnosis of lung cancer is that the majority of lung cancer cases do not have symptoms (asymptomatic) until the disease has progressed to an advanced stage 4, 64, 119, 124 It is estimated by the American Cancer Society that only 15% of all lung cancer cases are diagnosed in the early stage, i.e. Stage I 4 The average five year survival rate for lung cancer patients is 15%, this low survival rate is consistent with the current lack of an early diagnosis program 4, 64, 119, 124

PAGE 43

26 Common clinical lung cancer symptoms include a new or persistent cough, hemoptysis or blood in the sputum, chest pain, wheezing, hoarseness, shortness of breath, and repeated respiratory infections, e.g. bronchitis, pneumonia 4 The symptoms of lung cancer can vary a ccording to the tumor type and the extent of the disease or metastases. Recent articles in the literature have identified another area of concern in the diagnosis of lung cancer 4, 9, 125 128 Lung cancer diagnosis is currently done based on the symptomatic criteria outlined in textbooks that were written ten to twenty years ago 125 As physicians may not be aware of the changing patterns of lung cancer, this may add to the difficultly of a diagnosis, let alone an early diagnosis 125 Collins, et. al., 2007, noted that there have been epidemiologic changes or differences in the lung cancer patient population 4 Some of the current differences include an increased number of females with lung can cer; the most frequently encountered histological lung cancer type for males and females has changed to adenocarcinoma, and temporal differences in the age of diagnosis 4, 10 These epidemiologic differences may decrease the identification of specific symptomatic patterns in lung cancer cases which in turn could negatively impact the rate of early diagnosis 4 Approximately thirty to forty percent of lung cancer cases that are diagnosed hav e symptoms of metastatic disease 28 ; some of the most common organs that the cancer spreads to are the liver, the brain, the bones, spinal cord, and the adrenal glands. The symptoms of metastatic disease include bone pa in, personality changes, confusion, elevated alkaline phosphatase level, seizures, weakness, weight loss, nausea, vomiting,

PAGE 44

27 and palpable lymphadenopathy 4 There are several clinical manifestations of the s keletal and endocrine systems due to lung cancer spread. Some of the endocrine manifestations Common clinical symptoms of the skeletal system consist of digital clubbing and hypertrophic pulmonary osteoarthrop athy; these symptoms occurs in approximately ten percent of lung cancer cases 4 Clinical presentation and radiographic results are first steps in the differentiation or diagnosis of lung cancer type, i.e. small cell lung cancer (SCLC) or non small call lung cancer (NSCLC) 9 Small cell lung cancer is recognized by lymphadenopathy or the swelling of the lymph nodes and tumor invasion of the mediastinum 9 A characteristic of small cell lung cancer is the tumor or mass is seen in the hilum in a pproximately 78% of the cases. Patients with small cell lung cancer can present with paraneoplastic syndromes. Paraneoplastic syndromes are a collection of clinical signs and symptoms resulting from the byproducts of the tumor interrupting normal biologi cal function 4 Some of the syndromes resulting from small cell lung cancer include Lambert Eaton syndrome (muscle weakness), inappropriate antidiuretic hormone, and ectopic adrenocorticotrophic hormone pr oduction 9 Procedures for Diagnosing Lung Cancer Early detection and treatment is credited with increased survival for early stage lung cancer 13, 28 Early stage lung cancer is defined ICD 9 code as Stage I; w hich is less than 3 cm and with no evidence that the disease has may has spread outside the lung There are

PAGE 45

28 several early detection technologies providing diagnosis of lung cancer. These include are Computerized Tomography (CT), Magnetic Resonance Imagin g (MRI) and Positron Emission Tomography (PET) Scans 113, 129 Early treatment choices for lung cancer include chemotherapy, surgery, Radiation Therapy and combined modalities 4, 48, 130 In Radiation Therapy, there have been advancements in computer techn ology allowing the scanning results of CT, MRI and PET to be merged into a three dimensional (3 D) treatment planning system for more precise Radiation Therapy treatments 72 The technological treatment advancement of Intensity Modulated Radiation Therapy can focus the radiation beam into very specific treatment fields that are created or simulated on the treatment planning system by using CT scan and/ or by merging of CT and PET scans 72 Literature has shown that staging of cancer patients has vastly improved with the aid of PET 131 133 In many instances, the patient t reatment plan has been changed drastically. Cancer patient cases thought to be primary and hence the patient would have received a very aggressive treatment become palliative with less cost, physically, emotionally and economically, to the patient and comm unity at large 131 133 The gold standard for diagnosing lung cancer is with a tissue diagnosis. There are several diagnostic methods in order to obtain a tissue which include 1) sputum cytology, 2) a thoracentesis, 3) excisional biopsy of an accessible node, 4) flexible broncho scopy with or without transbronchial needle aspiration, 5) transthoracic needle aspiration, 6) video assisted thoracoscopy, and 6) thoracotomy 4 In order to select the most appropriate test or procedure th e physician, e.g. pulmonologist, interventional radiologist,

PAGE 46

29 or thoracic surgeon, must make a determination of which lung cancer type is suspected. Patients with suspected early stage non small cell lung cancer who are surgical candidates, commonly have a surgical procedure known as a thoracotomy 4, 9, 54, 59, 125, 134, 135 A patient can be staged as well as having a tissue diagnosis from this procedure 4, 9 Sputum cytology involves the collection of at least three samples of sputum; it is noninvasive, but if the results are negative, further testing may be required 4, 9, 135 This technique is recommended if the patient has hemoptysis; sputum cytology is ind icated for centrally located tumors in the chest cavity. The specificity for this test is 99% and the sensitivity for central tumors is 71%, peripheral tumors are less than 50% 4, 125, 127 If the pat ient has pleural effusion (fluid between the lung and the chest cavity); a thoracentesis can be performed. Sampling of the fluid can give an indication of the presence of lung cancer. The sensitivity for this procedure is 80% with a specificity of less t han 90%. In the case of palpable lymphadenopathy, a biopsy of an accessible node can be a method to obtain a tissue sample. Sputum cytology, flexible bronchoscopy, and transthoracic needle aspiration are procedures employed when the stage an d the cancer t ype are unclear. Flexible bronchoscopy involves passing a scope along the bronchial tract and taking samples of tissue via bron chial washings and/or biopsies. The sensitivity of this procedure or ability to correctly detect the presence of the disease is 88%. Computerized tomography (CT) of fluoroscopic guidance can be utilized while placing the catheter into The sensitivity and specificity of this test depends upon where the tumor is located or where the tissue sample is taken. Th e sensitivity for diagnosis of

PAGE 47

30 centrally located tumors utilizing flexible bronchoscopy is 88% and the specificity is 90%. The sensitivity for peripheral tumors drops to 60 to 70% with this technique. The procedure of choice for peripheral tumors (sensit ivity of 90%) under CT or fluoroscopic It is indicated in nonsurgical candidates with peripheral tumors when the transbronchial needle aspiration is inconclusive. One drawback or c omplication of this procedure is a pneumothorax (collapsed lung) in 25 to 30% of the patients undergoing the procedure 4, 9 Video ass isted thoracoscopy is a more recent procedure 4 and is used for small peripheral tumors less than 2 centimeters in diameter, pleural effusion or pleural tumors. The major advantage of a video assisted tho racoscopy is it can prevent an unnecessary surgical procedure, i.e. the thoracotomy. Lastly, the surgical procedure recommended for the treatment and the diagnosis of early stage non small cell lung cancer is a thoracotomy in cases with a clearly resectab le tumor 4, 9, 126, 136 Screening There are many histological types of lung cancer and finding a single biomarker or screening tool is a challenge. Several biomarkers are being evaluated as screening or predictors for lung cancer. An effective screening program for lung cancer is important for early detection of the disease which could increase s urvival. One of the research projects at the Moffitt Cancer Research Center, has the objective of finding a biomarker that will be used to develop an early screening and detection program for those people at risk for lung

PAGE 48

3 1 cancer 137, 138 Present research includes microscopic examination of sputum sample staining patterns. One of the possible biomarkers is monoclonal antibodies (Mabs). The pattern and stain intensity of the Mabs and varying cell characteristics are being investigated as a possible screening tool. The understanding of tumor biology has increased with the recognition of genetic and protein markers which precede malignancy 137 Mutations of particular genes c ontribute to the process of epithelial carcinogenesis. These mutations modify the control of abnormal cell growth. Heterogeneous nuclear ribonucleoprotein (hnRNP) has been linked as a marker with sputum cytology for early detection of lung carcinogenesis Data demonstrate that hnRNP is expressed in most lung cancer cases before any morphologic abnormality Other biological markers that are found in lung tumors include: tumor suppressor genes ( p53, Rb, p16, p21 ) p roto oncogenes (K ras, c myc, c erB 1 an d 2, HGF, HER 2), Telomerase (hTERT), hypermethylation and growth factors (GRP/BN, TGF b, FDGF, PTHrP, IGF I and II), apoptosis and angiogenesis (Bcl 2, VEGF), and gene amplification (HER 2) 137 These molecular markers can be importa nt markers for pulmonary carcinogenesis, used as an early diagnosis tools, and can be determinants in prognosis of a lung cancer treatment regimen. As shown in the Table 1 (Chart 2) below from Duarte, et. al, 2005, several biological markers are found wit h greater frequency with respect to the tumor type 139 The molecular marker Rb is found 30% in NSCLC but approximately 100% of the time it is detected in SCLC.

PAGE 49

32 Table 1: Molecular Biomarker for Lung Cancer (LC) Present ly, the National Cancer Institute is conducting a large scale clinical trial known as the Prostate, Lung, Colorectal, and Ovarian Screening Trial (PLCO) 140, 141 The objective of this study is to determine the efficacy of screening tools utilized during the trial and evaluate the death or mortality rate associated with that particular cancer under study 142 The major disadvantage with this trial for the early detection of lung cancer is the screening tool, a conventional chest x ray, does not detect lung neoplasms at early stages 140 Another resea rch study known as the National Lung Screening Trial (NLST) is sponsored by the National Cancer Institute for women and men and women at risk for lung cancer. This particular trial is comparing s piral CT scan s and conventional chest x rays and making a de termination which is a more effective screening tool in an effort to reduce death due to lung cancer. S piral CT scan s are effective in the visualization of lung nodules that cannot be seen in conventional chest x rays; this does creates moral and ethical issue as a spiral CT is proven to detect lung cancer in the early stages as compared to a chest x ray 143 Presently, the literature does not show that a spiral CT scan or a conventional chest x ray has not been demonstrated in the literature to

PAGE 50

33 reduce the risk of lung cancer mortality 9, 135, 140, 144 Pathology/Histology The re are two major categories of lung cancer; small cell carcinoma and non small cell carcinoma. NSCLC includes adenocarcinoma, squamous cell carcinoma and large cell undifferentiated carcinoma 3, 40 Each histological type has its own medical intervention that can include any combination of surgery, radiation therapy, and/or chemotherapy 7, 51, 67 The in cidence and survival rates are also different for the lung cancer type. There are many causal factors for lung cancer such as lifestyle, occupational risks, and environmental factors 10 The primary cause of lung cancer is smoking 9, 13, 145 151 Lung cancer remains a major public health problem lacking an early prevention and intervention program 4, 9, 125, 136 As far as Public Health consequences, lung cancer is the most lethal of all cancers as it has the highest mortality rate both among men and women 2, 13 with t of 13 percent. The four major types of lung canc er are adenocarcinoma, squamous cell carcinoma, large cell, carcinoma, and small cell carcinoma. The four major lung cancer types comprise 95% of a lung cancer cases 72 Adenocarcinoma is a malignant neoplasm; it originates in the epithelial cells of glandular tissue and forms glandular structures. This particular cancer is very commonly found in the periphery and accounts for 30 40% of all lung cancer types. Squamous cell carcinoma accounts for 20 30% of lung tumors and the origin is usually hilar 7, 152 95% of all small cell lung cancer is attributed

PAGE 51

34 to smoking. SCLC metastasizes early and has a five year survival rate of less than 15 %. Large cell carcinomas account for 10 15% of all lung neoplasms and are comprised of undifferentiate d or immature cells Large cell is the most aggressive of the NSCLC as they difficult to diagnose due to undifferentiated nature. These are commonly located in the central portion of the lung. The ICD O code for lung cancer pathology is classified by the morphology code and the topography code as shown below in Figure 6. Figure 6: Structure of a Morphology Code Source : http://training.seer.cancer.gov The topography code identifies the site and the sub site for the disease of interest. The complete ICD O code contains ten digits with the first four digits being the topography and the last six digits the morphology identifiers. Figure 7 is an example of the coding scheme for a squamous cell lung cancer. Diagnostic term: Poorly differentiated squamous cell carcinoma, upper lobe of lung C34.1 M 8070/33 Figure 7: Structure of a Complete ICD O Code Source: http://training.seer.cancer.gov

PAGE 52

35 The complete site specific topography code for lung cancer is shown in Figure 8 from SEER. Figure 9 displays the anatomy of the lungs with the associated ICD O code. The codes range from C34.0 as found in the main bronchus to C34.9 NOS (not otherwise specif ied). ICD O TERM C34.0 Main bronchus C34.1 Upper lobe, lung C34.2 Middle lobe, lung (RIGHT LUNG ONLY) C34.3 Lower lobe, lung C34.8 Overlapping lesion of lung C34.9 Lung, NOS C33.9 Trachea, NOS Prior to the Second Edition of ICD O, trachea and lung had the same ICD O code. With the advent of ICD O 2, trachea has a separate code (C33.9) from lung (C34._). The ICD O four digit subsites of C33.9 through C34.9 are considered part of a single primary site. Since lung is a paired organ, laterality mu st be coded. Figure 8: ICD O 3 Site (Lung) Codes Source: http://training.seer.cancer.gov/ss_module03_lung/unit03_sec01_icdo_codes.html Figure 9: Lung Anatomy with ICD O 2/3 Codes Source: http://training.seer.cancer.gov/ ss_module03_lung/unit03_sec01_icdo_codes.html

PAGE 53

36 Staging/Extent of Disease There are several purposes why it is necessary to stage a cancer case. First, the medical professional must assess the extent of the disease adequately 1, 4, 40 Correct ascertainment of the extent of the disease is crucial as the appropriate treatment regimen must be determined 10 A curative or palliative approach in disease management will be based on the stage of lung cancer the patient has. Secondly, staging can one of the Other indicators of prognosis incl ude tumor histology, grade of disease, and patient demographics, e.g. age, gender, race, socioeconomic status, and martial status 55, 139, 153, 154 Finally, with a comprehensive and standardized staging protocol, the exchange of information between the scientific communities can be accomplished. Staging or coding data are use for research and general health care information. The extent of the disease can be classified by the use of a number or coding system with increasing values representative of increasing disease severity. The anatomic coding system allows for analysis of similar cases with comparable characteristics based on disease extent 11 Cancer cases are described by the site of origin called the primary site and how far the ca ncer has spread from the primary site. Other essential variables include tumor size, the number of tumors (multiplicity), the depth of the tumor invasion, regional or distant tissue extension, regional lymph node involvement, and distant metastases 13 Coding information began on an international level in 1893 for mortality data.

PAGE 54

37 staging a disease in 1929. One of the first descriptions of the extent or the stage of the disease was for carcinoma of the cervix 13 After World War II, the World Health Organization established guidelines for the classification of disease. In 1948 the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD), manual was published; the coding scheme was used to code and tabulate morbidity and mortality data. The American Cancer Society 2 in 1951, developed the first code manual for the classification of tumor morphology. The tumor codes were comprised of the first two numbers being t he indicator of the tumor type with a third number representing the behavior of the neoplasm 2, 29 WHO adopted a cod ing system based on the ACS in 1956 2, 29 The International Agency for Research on Cancer (IARC) was commissioned by the World Health Organization to help develop a world wide classification scheme for oncology. The first edition of this manual was in 1976 called the International Classification of Disease for Oncology (ICD O) 2, 29 There have been several updates and revision for the classification scheme but morphology code uses standardized three and four character categories 1, 2, 29 In the United States, the Surveillance, Epidemiology, and End Results (SEER) Program of the Nati onal Cancer Institute routinely gathers data on cancer statistics from designated population based cancer registries 11, 13, 60 SEER has developed a two stage classification system, extent of the disease and a summary stage. Summary staging is based in how the cancer advances or grows and there are five main categories 11 The Commission on Cancer (CoC) of the American College of Surgeons uses the American

PAGE 55

38 Joint Committee on Cancer (AJCC) staging system. An earlier version of the AJCC classification scheme was first introduced in 1958 by the Union Internationale Contre le Cancer (UICC) 29 In 1959, the American Joint Committee on Cancer Staging and End Resu lts Reporting (AJC) was organized and adopted the UICC coding system 29 The AJC changed their name to The American Joint Committee on Cancer in 1980. This system classifies the tumor in terms of the primary tumor (T), the region al lymph nodes (N), and distant metastasis 155 T is the indication for size and the invasiveness of the primary tumor. It (T) describes the size of the tumor in millimeters or centim eters with the extension of the disease into the adjacent tissue, e.g. mucosa, submucosa, muscularis, subserosa, serosa. T0 indicates there is no tumor; T1 indicates carcinoma in situ and limited to surface cells, and T1 4 reflects increasing tumor size a nd disease extension 1, 11 The N component is indicative of nodal involvement or no lymph node involvement. An increasing numerical value represents increasing disease extension into the lymph nodes. M is used in the identification of metastatic disease and distant lymph node involvement 1, 11 Another numeric system commonly in cancer registries used to describe or classify the extent of the disease is Stage 0 through Stage IV 1, 11 Stage 0 limits the disease to the surface and is also known as cancer in situ, Stage I confines the cancer growth to the tissue of origin and gives evidence of cancer growth, Stage II describes the cancer as limited local spread, Stage III is extensive local and regional spread, and Stage IV is used to classify distant metastasis 29

PAGE 56

39 The ICD O coding system uses a morphology code based on the histology (cell type), behavior code, and grade 1, 11 The behavior indicate s if the tumor is malignant, benign, in situ, or if the diagnosis is uncertain. Grading is determined by microscopic examination of the tumor cells 1, 11 The cells are classified Grade I through Grade IV; Grade I, the cells are slight ly abnormal and well differentiated, Grade II cells are moderately differentiated and the cells appear more abnormal, Grade III cells are very abnormal and poorly differentiated, and Grade IV are undifferentiated and immature. Immature or primitive cells are undifferentiated and highly abnormal in appearance. If a cell is well differentiated, it can appear like a mature or specialized cell. Table 2 below summarizes the AJCC staging system originally based on clinical data or surgical f indings 11

PAGE 57

40 Table 2: AJCC TNM Staging System for Lung Tumors American Joint Committee on Cancer (AJCC) TNM Staging System for Lung Tumors Tx Primary tumor cannot be assessed, or tumor proven by the presence of malignant cells in sputum or bronchial washin gs but not visualized by imaging or bronchoscopy T0 No evidence of primary tumor Tis Carcinoma in situ T1 Tumor 3 cm or less in greatest dimension, surrounded by lung or visceral pleura, without bronchoscopic evidence of invasion more proximal than the bronchus T2 Tumor with any of the following features of size or extent: More than 3 cm in greatest dimension Involves main bronchus, 2 cm or more distal to the carina Invades the visceral pleura Associated with atelec tasis or obstructive pneumonitis that extends to the hilar region but does not involve the entire lung T3 Tumor of any size that directly invades any of the following: chest wall (including superior sulcus tumors), diaphragm, mediastinal pleura, parietal pericardium; or tumor in the main bronchus less than 2 cm distal to the carina, but without involvement of the carina; or associated atelectasis or obstructive pneumonitis of the entire lung T4 Tumor of any size that invades any of the following: mediasti num, heart, great vessels, trachea, esophagus, vertebral body, carina; or separate tumor nodules in the same lobe; or tumor with a bronchial pleural effusions associated with lung cancer are due to tumor. However, in a few patients, multiple cytopathologic examinations of pleural are negative for tumor. In these cases, fluid is not bloody and is not an exudate. Such patients may be further evaluated by videothoracoscopy (VATS) and direct pleural biopsies. When these elements and clinical judgment dictate that the effusion is not related to the tumor, the effusion should be excluded a staging element and the patient should be staged T1, T2, or T3. §M1 includes separate tumor nodule(s) in a different lobe (ipsilateral or contralateral). NX: Regional lymph nodes cannot be assessed N0: No regional lymph node metastas is N1: Metastasis to ipsilateral peribronchial and/or ipsilateral hilar lymph nodes, and intrapulmonary nodes N2: Metastasis to ipsilateral mediastinal and/or subcarinal lymph node(s) N3: Metastasis to contralateral mediastinal, contralateral hilar, ipsil ateral or contralateral scalene, or supraclavicular lymph node(s) Stage 0: Carcinoma in situ Stage IA: T1, N0, M0 Stage IB: T2, N0, M0 Stage IIA: T1, N1, M0 Stage IIB: T2, N1, M0, T3, N0, M0 Stage IIIA: T1, N2, M0, T2, N2, M0, T3, N1, M0, T3, N2, M0 Stage IIIB: T4, N0, M0, T4, N1, M0, T4, N2, M0, T1, N3, M0, T2, N3, M0, T3, N3, M0, T4, N3, M0 Stage IV: Any T, any N, M1 Source: American Joint Committee on Cancer (AJCC)

PAGE 58

41 Lung Cancer Prognosis There are c ertain factors that affect prognosis, i.e. quality of l ife, chance of recovery, survival, for a disease. In particular, the prognostic factor of interest for this research is to expand the scientific knowledge concerning gender differences in women and men with lung cancer and survival. The literature has ci ted several prognostic factors that affect lung cancer survival 49, 55, 139, 154 The prognosis can be dependent upo n 1) the stage of lung cancer (the size of the tumor, whether the cancer is confined to the lungs or has spread to other places in the body, i.e. metastized), 2) the histologic type of lung cancer, 3) if there are respiratory symptoms, e.g. coughing, diffi culty breathing, and 4) being 4 Early stage disease (Stage I, Stage II, resectable Stage III) prognostic factors most critical to decreased survival have been shown to include large tumor size and presence of lymph node metastasis, male gender, age greater than 60 years, and having a wedge resection versus a lobectomy or pnumonectomy 55, 61, 156 In advanced stage lung cancer, poor performance status, weight loss, male gender, elevated serum l actate dehydrogenase, and liver/bone metastasis are key prognostic factor for poor survival 17, 61, 157, 158 Clinical research has identified more than 150 risk or prognostic factors accord ing to Blanchon, et. al., 2006 153 research in cluded age, gender, socioeconomic status, possibility of occupational origin of the cancer, stage of cancer at time of diagnosis, smoking history (pack years, duration, discontinuation, date of discontinuation), geographic location, histology, stage, vita l

PAGE 59

42 status, treatment modality, and performance status at time of diagnosis 153 Performance status was based on the classification as given by the Eastern Cooperative Oncology Group 158, 159 based on a 0 to 4 scale. The final univariate model included age, sex, performan ce status, histologic types, and stage of disease. Patients younger than 50 years had a greater probability of survival as compared to those greater than 70 years. Men had a decreased survival (Hazard Ratio (HR) = 1.17; 95% CI: 1.05 1.31) versus women. The HR for risk of death increased with increasing stage of disease (Stage IV HR = 3.53; 95% CI 3.05 4.09) and performance status (Performance Status 4 HR = 4.97; 95% CI: 3.83 6.43) 153 These five predictor variables served as the most important prognostic factors Other medical research has reported similar individual characteristics ( prognostic factors) such as gender, sex, stage of disease, performance status, molecular biologic markers, marital status, smoking status, and psychosocial factors as predictors for lung cancer survival 61, 156, 157, 160, 161 ; similar to Blanchon et al. ( 2006 ) at the National Cancer Center, et. al., 2006, reported the importance of having a database with available prognostic factors for lung cancer survival 160 The authors stressed the epidemiologic, psychosocial, and molecular biology data as to improve lung cancer patient outcome by increase treatment efficiency. In this particula r study, biologic material was collected and several questioners at baseline and subsequent follow up intervals 160

PAGE 60

43 Epidemiologic studies have show n that there is a causal relationship between smoking and lung cancer 149, 162, 163 Smoking patterns and status serves as a prognostic factor in survival after the diagnosis of lung cancer but the majority of heavy smokers do not develop lung cancer 139 Duarte, et. al. 2005 has suggested that genetic factors affect Molecular changes in lung cancer may serve as an indicator (prognostic factor) for survival. Several genetic factors have been investigated, but presently there is n o evidence that a single parameter has sufficient treatment efficiency 139, 153 Principle molecular markers prima rily found in lung cancer will be expanded upon in the section on Genetic Risk Factors in this chapter. Biologic or molecular markers can be important as prognostic variables but also in the identification of treatments targeting the cancer cell based on for an effective cancer cell kill. This is becoming increasingly important for the treatment of lung cancer and survival Lung Cancer Survival and Risk Factors Gender Why is gender important as risk factor for lung cancer surviva l ? Although lung cancer mortality has reached epidemic levels for women (an increase of 6 00 % since the the causal pathways are blurred for women 40 The exact etiology of a suscepti bility (reported in the literature as different from men ) to lung cancer still not resolved 49 One potential reason for the current deficit of knowledge for gender

PAGE 61

44 difference s in the etiology and subsequent susceptibility is that the treatment patterns for lung cancer are based on research done on men. Previously, the association between being a woman and the risk of lung cancer was considered negligible by the medical community as referenced by published reports by the US Surgeo n General 24, 76 But as behavior and other temporal changes, such as cigarette smoking have occurred over the past several decades, the inc idence of lung cancer has increased with a resultant increase in mortality 2, 29 Women historically have been exclud ed from clinical trails or if included, the data was not analyzed 18 As one result of this disparity of being excluded from research studies, women may be at a greater risk of lun g cancer than men at the same level of smoking but the evidence in the literature has be conflicting and is limited 22 A report of the Surgeon General 24, 76 on women and smoking did not reach any conclusions concerning what role gender difference may play in t he development of lung cancer (US Department of Health, and Human Services, 2001). Hypotheses have been developed based on possible response to carcinogens and hormonal related differences in women as compared to men (Ryberg et al., 1994) but conflict in the literature remains 27, 40, 45, 50, 107, 164 Lung cancer incidence and mortality rates are higher in men as compared to women 3, 10 The fact that women have a lower prevalence of smoking may account for this difference. The primary cause of lung cancer in women and men is due to smoking tobacco products, in particular cigarettes 23, 75, 165, 166 In 2006, it is estimated lung cancer will account for 30 p ercent of all cancer deaths in the United States. Among men, lung

PAGE 62

45 cancer incidence and mortality have been declining since the early 1980s and 1990s The peak death rate in men in the 1990s coincided with a lag period of approximately 25 years after the h ighest per capita c igarette consumption. Women started smoking approximately 20 years after men, lung cancer incidence rates did not begin to fall in women until 1998. For the first time in 1995, mortality rates in women have stabilized, after increasing for several decades There has been an increase of 6 00% in mortality for women with lung cancer since 193 0 28, 40, 60 Lung cancer has overtaken breast cancer as the number one cause of cancer related deaths of women in the United States. In the next several articles cited in this chapter, epidemiological studies were designed to quantify the differences in lung cancer risk between men versus women 8 15 17 27 167 In the paper, "Lung cancer in women compared with men: stage, treatment, and survival" Ouellette et al., 1998, conducted a retrospective cohort study, t o test the hypot hesis of a survival difference among men and women with lung cancer 8 The t arget population consisted of 104 women and 104 men with incident cases of lung cancer diagn osed between March 1998 and June 1990 at a university hospital in Montreal, Canada The authors concluded t here was no difference in mean age of lung cancer diagnosis for female s ( 60.97 + 10.89 years of age) and for male patients ( 61.49 + 10.29 years of age ) There was a statistically significant difference in the distribution of the different histologic types of lung cancer between men and women ( p = 0.028) When Ouellette et. al. stratified on lung cancer stage, t he stage of the disease positively influe nced survival between the women and men 8 After adjustment, w omen appeared

PAGE 63

46 to live 12 months longer than men at any stage and a statistically significant survival adva ntage in women was found ( p = 0.02) 167 The authors (Ouellette et. al. ) found that w omen did received less surgical interventions than men; although not statistically signific ant 8 Twenty four women received chemotherapy compared with fourteen men, although this was not found to be statistically significant according to the authors A limit ation in this study could be due to the small number of lung cancer cases contributing to the non significant finding (decreased power) Ouellette et al., 1998 reported m en and women received similar treatments for their disease in this study. This diffe rs from studies on coronary artery disease in which it was thought that physicians may pursue less aggressive management in women as noted by Steingart et al. 1991 8 Unlike Ouellette, et. al., 1998, A itakov et al. 1998, found that more men in Russia underwent surgery with a ratio of men to women of 7.4:1.0 167 Ouellette et al. did not fin d such a disparity; the ratio was 1.17:1.0 men to women, and noted that is probably reflect ed the tendency to offer similar treatments to both sexes in the Western world. A population based study by Radzikowska et al., (2002), investigate d demographic fa ctors (gender, age, and smoking) and factors connected with the disease (histology, performance status, stage, treatment and survival) for lung cancer patients The t arget population was comprised of members of community based cancer registr ies Approxim ately 20,561 lung cancer cases from all parts of Poland f rom 1995 to 1998, were registered with the National Tuberculosis and Lung Dise ases Research Institute

PAGE 64

47 (NTLDRI ) as well as being registered with the Polish Cancer Register. From this population, 2, 875 women and 17,686 men were selected 15 It was determined that w omen developed lung cancer at a younger age, were more likely to be lifetime non smokers, consumed fewer cigarettes per day and smoked for a short period of time 15 The authors commented that all those factors suggested that women are more susceptible to the carcinogenic compounds of cigarette smoke and environmental noxious conditions that possibly damaged the genetic distribution f or the population 15 Women were found to be more likely to have adenocarcinoma and SCLC as compare d to men. Squamous cancer was the predominant type of lung cancer among men, and less than ten percent of men had adenocarcinoma. Different patterns of histological types of lung cancer were observed in Poland as compared to the USA, China or Denmark, wh ere over representation of adenocarcinoma has been noted The distribution of main histological types of lung cancer in Poland was similar to that described in Finland and Scotland The most i mportant prognostic factors for lung cancer patients were perf ormance status, clinical stage of the disease and surgical treatment The authors did not evaluate different treatment moda lities and the effect on survival. Radzikowska et al., found that females with lung cancer had a survival benefit compared with male s, taking into account age, histology, performance status, extension of the disease and treatment. This overall survival advantage of women was described first in data based o n Danish Register information Although Radzikowska, et al. found an increased survival based on gender the other researchers found the opposite. Kirsh

PAGE 65

48 the Lung in W omen that the survival among women in a younger age group was significantly lower than for both groups of women in the older age group (p = 0.0335) and men in the younger age group (p = 0.0033). This was believed to be due to the higher incidence of both Stage III disease and adenocarcinoma in younger women. Visbal et al. (2004) "Gender differences in non small cell lun g cancer survival: an analysis of 4,618 patients diagnosed between 1997 and 2002 evaluate d the magnitude of the gender effect on non small cell lung cancer (NSCLC) survival across disease stage, tumor histology, and therapies 17 The t arget population of 4,618 newly diagnosed NSCLC were patients at the Mayo Clinic in Olmsted County, Minnesota between 1997 and 2002 There were 2,724 men (59%) and 1,894 women (41%), with a median age at diagnosis of 68 years in men and 66 in women (p < 0.01). Women were d iagnosed on average two years younger than men 17 A s compared to men, women began smoking at a later age, smoked less cigarettes per day and fewer years. More men smoked and were heavier smokers than women 17 A denocarcinoma was the common subtype in both genders; 59.5% of the women and 48.2% of men The difference between women and men with adenocarcinoma was significant with a p value of < 0.001. For the other histological types (squamous, unclassified NSCLC, large cell, adenosquamous) of lung cancer the difference was not significant between men and women with. The estimated relative survival in men was 51% (95% CI: 49%, 53%) at one year and 15% (95% CI: 12%, 17%) at five years The estimated one year relative survival in women was 60% (95% CI: 58%, 62%) and 19% (95% CI: 16%, 22%) at five years 17 Men were at a

PAGE 66

49 significantly increased risk of mortality compared to women following a diagnosis of NSCLC (adjusted relative risk: 1.20, 95% CI: 1.11, 1 .30), particularly for patients with stage III/IV di sease or adenocarcinoma. Male gender was found an independent unfavorable prognostic indicator for NSCLC survival 17 This particular study by Visbal et. al. was of interest to this research as Visbal studied treatment types, stage and histology of the disease; as noted previously reports on stage, grade, histology, and treatment are limited Some of the weaknesses of this study included the lack of interaction terms in the model as without evaluating interaction any significant effects could be masked. 27 attempted to exp and the current knowledge of gender differences in men and women with lung cancer and survival (N= 2618). Squamous cell carcinoma was the predominant histologic type of lung cancer for men; women had an increased likelihood of having adenocarcinoma or sma ll cell lung cancer. There was no statistically significant difference for men with large call carcinoma versus women with large call carcinoma. The stage of disease at diagnosis for men as compared to women was not significant 27 Differences in survival were demonstrated between the lung cancer types during the cutpoints of 3, 4, and 5 years. A very pronounced survival difference existed between men and women for stage IV squamous cell carcinoma (274 mean days for men versus 153 mean days for women, p = 0.005). Women diagnosed with squamous cell carcinoma stage II again had

PAGE 67

50 decreased survival versus men (636 mean days for men, 379 mean days for women, p = 0.043). This study demonstrated the important effect histologic type and stage at diagnosis plays in overall survival when comparing gender differences 27 It has been hypothesized that women are more susceptible to tobacco products as compared to men. Tobacco The number one cause of lung cancer, approximately 80% of lung cancer cases can be attributed to smoking tobacco products, and the subsequent decrease in survival for lung cancer cases is due to tobacco smoke 10, 168 The etiology of lung cancer is multi causal with a complex pathway of development that includes carcinogen exposure, metabolism, and genetics 48 Tobacco smoke is recognized as the chief risk factor for lung cancer 25, 67, 92, 169 It has been estimated by the International Agency for Research on Cancer (IARC) to contain at least 80 known mutagens and carcinogens, e.g. polycyclic a nit ro amines, and aromatic amines. Yach and Wipfli stress the importance of tobacco control efforts as tobacco kills five million people annually with an estimated increase of 100% (10 million) by mid 2020 170 Today the population attributable risk percent (PARP) for men is approximate ly 90% and for women the PARP is approximately 80% 10 The effects of tobacco smoking were The increase in lung cancer rates were first attributed to factors other than tobacco smoke such as atmospheric pollution 170 As the rise in lung cancer ra tes and the increase in

PAGE 68

51 mortality rates for lung cancer, clinical evidence and case series were reported in the risk of lung cancer. In 1938, Raymond Pearl, M.D. from Johns Hopkins University reported that smokers do not live as long as nonsmokers 170 There were two classic epidemiologic studies that demonstrated a strong association between the risk of lung cancer and smoking by 1) Sir Richard Doll and 2) Sir Bradford Hill 10 Sir Richard Doll conducted a case control study in 1947 and compared hospitalized patients with and without lung cancer. He collected information on smoking history and found a greater than 20 percent increased risk for lung cancer. Zang and Wynder found that women were at an increased risk (1.2 to 1.7 times) for lung cancer than men independent of tobacco smoking level 171 with an associated decrease in survival. Passive smoke, also known as second hand smoke or environmental tobacco smoke (ETS) has been shown to increase the risk of lung cancer 10, 172 Stockwell et al. (1992) demonstrated an increased risk, OR = 2.4; 95%CI = 1.1 5.3, for women with 40 or more smoke years of household (ETS) exposure as a n adult. The authors also found a statistically significant association with women exposed to ETS during childhood or during adolescence for 22 smoke years with an OR = 2.4; 95%CI = 1.1 5.4 173 Investigators for the International Early Lung Cancer Action Program found that women have an increase susceptibility to lung cancer as compared to men, OR = 1.9; 95%CI = 1.5 2.5 yet women had a decrease hazard rat io for survival versus men, HR = 0.48; 95%CI= 0.25 0.89 174 The study population was comprised of 1202 men and women

PAGE 69

52 from New York City undergoing a baseline screening at Weill Medical College of Cornell University 174 Race and Ethnicity Worldwide statistics for lung cancer obtained from the International Agency for Res earch on Cancer, IARC, and GLOBOCAN 2002 demonstrate that lung cancer incidence and mortality rates are less in women than men 5 The decreased rates for women are not dependent upon ethnicity and race as reported by IARC. Some of the variat ion in the incidence rates may be attributed to the quality of the data collected by the cancer registries and diagnostic methods employed. Gadgeel and Kalemkerian (2005) reported that race is not a biologically relevant parameter but racial differences in lung cancer characteristics and outcomes have been reported 21 Blacks consistently have higher rates of lung cancer incidence as compared to whites. African Americans in the United States have the greatest incidence of lung cancer (8.5% risk of lung cancer diagnosis) ; they also hav e the highest lung cancer mortality rates (7.6% risk of death form lung cancer) 10 Racial differences in smoking habit sand SES, have been associated with an inc reased risk of lung cancer 21 ung C ancer: F ocus on Asian Americans and Pacific Islanders, American Indians and Alaska Natives, and Hispanics and Latinos racial and ethnic groups 175 The authors noted racial and ethnic differences in smoking habits,

PAGE 70

53 presentation, stage at diagnosis, metabolism of nicotine, treatment received, and outcomes will impact lung cancer survival and health care over th e next several decades in the United States 175 The identification of ethnic groups is an important aspect of lung cancer research for possible gender differences based on ethnicity and lung cancer survival. The Multiethnic/ Minority Cohort study was established to study diet and cancer in the United States 176 The data can be and is utilized to evaluate lung cancer patterns of incidence and mortality and the effect gender, race, and ethnicity play. The cohort consisted of 215, 251 living in California (primarily in Los Angeles County) and Hawaii with the cohort consisting of 16.3% African American, 22.0% Latino, 26.4% Japanese American, 6.5% Native H awaiian, 22.9% White, and 5.8% of other ancestry. African American had the highest rate of smoking, 28.5%, followed by Native Hawaiian at 20.1%. The lowest groups of smokers were Japanese Americans, 15.5%, and whites, 15.9%. Both females and male African American and Native Hawaiian females had the highest prevalence of smoking as compared to male and female Japanese Americans and Latin as Lung cancer incidence was 54.0% lower for Japanese American men (p value < 0.001) and 71.0% lower for Latina women ( p value < 0.001) as compared to African Americans 176 Genetics Prior to 1970, scientific evidence about the etiology of lung cancer was unavailable. A pathway of understanding t he complex route of lung cancer development

PAGE 71

54 with the consequent effects of gender differences and survival became accessible with advances in molecular genetics 83 Cancer is a complex process which involves an ini tial damage of the genetic material (DNA, RNA) of a cell; this leads to a mutation or change in the chromosomes 19, 67, 69, 71, 177 180 The mutations can be inherited (germ cells) or may become cancerous (malignant and uncontrolled cell growth). Although identification of biologic materials provides evidence o f genetic susceptibility; inter individual variation can modify the effects of carcinogenic exposures and the resultant effects. For an example, the majority of long term smokers will not develop lung cancer. A predictive model for lung cancer genetic su sceptibility among smokers was developed by Bach et al. (2003) 92, 117, 118 Utilizing data from the CARET trial, 18 172 individuals, the statistical model predicted only a quarter of the lung cancer cases based on genetics predisposition; individual variation of metabolis m, DNA repair, cell cycle, inflammation and microenvironment would be a possible explanation for the inability to calculate an accurate and precise model of behavior 92 Several genetic and epigenetic mutations of tumor supp ressor genes have been observed in lung cancer 48 Thomas et al. (2006) noted in 95% of small cell lung cancers and 40 to 70% of NSCLC, the most frequently encountered genetic alteration was p53. The p53 affects the biological pathway in G1 and G2 cell cycle regulation; p53 stops the cell division that occurs when there is damage to the DNA 48 The p53 mutation leads the formation of DNA adducts (abnormal piece of DNA bonded to a cancer causing agent) as

PAGE 72

55 a response to the effects of smoki ng; this mutation inhibits the normal cell repair and is involved in carcinogenesis. Women have been found to have increased DNA adducts per pack years versus men. K ras (a proto oncogene) is another biomarker and forms DNA adducts when damaged; women are three times more likely than men to have the K ras mutation; this mutation is associated with adenocarcinoma. Family History Another risk factor impacting lung cancer survival can be associated with family history or familial aggregation 10, 172 Familial aggregation can serve as a surrogate (indirect measure) for lung cancer etiology and resultant survival rates bas ed on genetic predisposition 181 Etzel et al. (2003) examined risks for smoking related cancers for relatives of lung cancer patients 169 Siblings were found to have a statistically significant association, RR = 1.85; p value = 0 .003, of lung cancer risks as well as an increased risk for smoking related cancers, RR = 1.29; p value = 0.01 169 When stratification on age of disease onset was done, there was no association between familial aggregation and lung cancer risk for ages les s than 55 years. The authors did find evidence of familial aggregation, RR = 1.71; p value < 0.001, for lung cancer risk between relatives of late onset cases of lung cancer. Schwartz et al. (1999) found evidence that common susceptibility genes may incr ease the risk for lung cancer among relatives of nonsmoking lung cancer cases 182 The study population was obtained from families of nonsmoking lung cancer cases (257 populati on based) and nonsmoking controls (277 population

PAGE 73

56 based), residing in metropolitan Detroit The first degree nonsmoking lung cancer cases relatives were (OR = 1.5 ; 95% CI = 1.02 2.27) at increased risk for cancer of the digestive system after adjustment f or each relative's gender race, age and smoking status. T here was an elevated OR of 1.12 for an increased risk for lung cancer for first degree relatives but the findings were not statistically significant, i.e. 95% CI = 0.65 1.93 182 Genetics and the Environment A powerful design to disentangle the interplay between genetics and environmental influences in the studies of human disease incidence, mortality, and survival can be accomplished by the use of twin studies 10, 172 Twin studies serve to ological makeup and environmental influences. Identical twins (monozygotic (MZ)) develop from fission of single fertilized egg and have inherited identical genetic material; fraternal (dizygotic (DZ)) twins derive from two distinct fertilized eggs meaning they have the same genetic makeup comparable to siblings 10, 172 Genetic effects would be determined significant if there was concordance for cancer among MZ twins as compared to DZ twins (on average share 50% of their separated genes). Environmental factors would be the determining factor for increased lung cancer risk if the concordance was similar for both types of twins 183 Lichtenstein et al, ( 2000) combined data from three different national twin and cancer registries (44,788 pairs of twins from Swedish, Danish, and Finnish twin registries ) 183 The

PAGE 74

57 authors found that there were stat istically significant risks associated with colon and breast cancer, lung cancer was not. This implies the inherited genetic factors do not make an individual susceptible to lung cancer and survival but environmental factors play the major role 183 Braun et al. (1995) concluded that genetic susceptibility had influence on lung cancer mortality 184 in men. There was an excess risk of lung cancer mortality for dizygotic twin pairs (DZ SMR = 2.2; 95%CI = 1. 3 3.4) but the risk was not statistically significant for monozygotic twin pairs (MZ SMR = 2.1; 95%CI = 1.0 3.7). This suggests a predisposition for lung cancer in males. Geographic Variation Survival rates based on lung cancer incidence and mortality c luster in geographic regions that have a high prevalence of smoking 10 Devesa et al. (1999) examined data from the IARC cancer registries of morphology specific lung cancer. Squamous cell carcinoma had declined by 30% in North America 30 Rates in the Nordic countries, which varied by 2 fold from a high in Denmark to a low in Sweden, still were generally lower than in other parts of Europe, where the rate was highest in the Netherlands 30 The l ung cancer incidence rates among males varied by 4 fol d: 83.6 among U.S. Blacks to 21.1 in Sweden A mong females, recent rates varied by almost 8 fold, with the highest among U.S. Blacks (35.8) and the lowest in Spain (4.6) Incide nce rates among females paralleled that in males, with the exception of Switzerland. Rates everywhere were

PAGE 75

58 higher among males than females. Male to female rate ratios varied from less than 2 in Iceland, U.S. Whites, Canada, Denmark and Sweden to more tha n 6 in Slovenia, Italy, and France and more than 10 in Spain Developing countries demonstrate a higher ratio of lung cancer incidence and mortality for men versus women As shown below in Table 3, developing countries, have a higher ratio of male and fe male lung cancer incidence and mortality rates Table 3: Incidence and Mortality Rates Incidence and Mortality Rates, Crude and Age Standardized (World) rates, per 100,000 Country/Region Incidence Mortality Cases Crude Rate ASR(W) Deaths Crude Rate AS R(W) World 1352132 43.5 47.6 1178918 37.8 41.5 More developed regions 676681 114.7 72 584979 99.2 61.2 Less developed regions 672221 26.7 35.3 591162 23.4 31.2 Females World 386891 12.6 12.1 330786 10.7 10.3 More developed regions 194731 31.7 1 7.1 161472 26.3 13.6 Less developed regions 191192 7.8 9.4 168481 6.8 8.3 Males World 965241 30.9 35.5 848132 27.1 31.2 More developed regions 481950 83 54.9 423507 72.9 47.6 Less developed regions 481029 18.9 25.9 422681 16.6 22.9 Source: GLOB OCAN 2002, IARC In the United States, Kentucky had the highest incidence of lung cancer, 40 per 100,000, and lung cancer mortality, 115 per 100,000 for the years 1997 2001 for males

PAGE 76

59 10 Utah has the lowest incidence and mortality rates for lung cancer for males and females based on data from the American Cancer Society. Alcohol Another risk factor associated gender differences and lung cancer survival is alcoho l 10 The causal relationship between lung cancer and alcohol is complicated and remains controversial 10 Confounding by smoking is a major consideration and tobacco smoking commonly exists in setting where alcohol is consumed 185 Nine studies were cancer; they found only five of the studies adjusted for smoking 185 This meta analysis concluded that methodological issues explain the elevated risks in the stud ies of alcohol as misclassification errors based of smoking status were common in the nine studies. Prescott et al. (1999) found a protective association between lung cancer risk and wine drinking 186 T his was based on the results of three prospective cohort studies in Denmark. Men who consumed more than 13 glasses of wine per week had an RR of 0.78; 95%CI 0.63 0.97 compared to nondrinkers of wine. statistically significant for beer drinkers (RR = 1.36; 95%CI 1.02 (more than 41 drinks per week RR = 1.57; 95%CI 1.06 2.33). The study made the determination that the type of alcohol consumed impacted the association between lung cancer and alcohol after adjustment for smoking status 186

PAGE 77

60 Diet and Micronut rients Scientific evidence from the literature exists pointing to dietary factors as protectors against lung cancer induction 83, 187, 188 This risk factor, diet, may prove to increase survival as it may serve as a protector against lung cancer for men and women. Some of the dietary factors inclu de fruits, vegetables, carotenoids, vitamin C, phenols, flavones, vitamin E, selenium, isothiocyanates, folate, fat, and alcohol 10 The article by Smith fter controlling for smoking, there was a sixteen to twenty three percent reduction in lung cancer risk (RR = 0.77; 95% CI 0.67 0.87; p value (test for trend) =0.001) for men and women that had an increased consum ption of vegetables and fruits versus study participants that had a limited intake of fruits and vegetables Table 4 (Table 1 from Smith Warner et al. ( 2003 )), lists the prospective cohort studies used in the pooled analysis used for the research. The American Institute for Cancer Research present ed a summary of seventeen case control and seven cohort studies and concluded evidence existed that with an increase d in take of fruits, vegetables, and in particular dark, leafy, green vegetables, lung cancer risk was decreased 189 Brennan et al., 2005 studied whether cruciferous vegetables were protecti ve against lung cancer in a case control study 190 C ruciferous vegetables contain isothiocyanates non nutrient compounds found to be effective inhibitors of tumorigenesis 10 The authors found that weekly consumption of cruciferous vegetables decreased lung cancer (OR = 0.78; 9 5% CI 0.64 0.96) as

PAGE 78

61 compared to men and women that consumed cruciferous vegetables less than monthly 190 One of the weaknesses of the study by Brennan et al. was there was no identification or analysis by gender and the possible gender effect of cruciferous vegetables consum ption on lung cancer Insufficient adjustment for smoking has been examined as a possible residual confounder 191 to explain the protective effects demonstrated by an increased intake of fruits and vegetables for a decreased lung cancer risk and subsequent increase in survival. The results o f the research done by Skuladottir et al., 2004, found that there was an inverse relationship between an increased intake of fruits, vegetables, and lung cancer risk even after the influences of smoking as a confounder analyzed via stratified analysis. Me n and women in the highest quartile of intake of fruits and vegetables demonstrated a thirty five percent lower risk of lung cancer as compared to individuals in the lowest quartile of fruits and vegetable intake 191 The lung cancer risk and fruits and vegetable intake association was decrea sed when stepwise adjustment for smoking status, duration of smoking, and the number of cigarettes smoked per day was done but the relationship remained statistically significant for study participants that had the highest intake of all plant food (fully a djusted rate ratio = 0.65; 95% CI 0.46 0.94) 191 Dietary carotenoids have been identified as the possible micronutrients in fruits and vegetables that may decrease lung cancer risk 84, 192 195 When this effect was tested utilizing c linical trials with high doses of carotenoids, in particular beta carotene, a reduct ion in lung cancer risk was not demonstrated 196, 1 97 One particular clinical trial,

PAGE 79

62 the Alpha Tocopherol Beta Carotene Cancer Prevention 198 Trial, was stopped early due to the unexpected result of a statistically significant increase of lung cancer after receiving beta carotene as compared to participants receiving the placebo 197, 198 Stram et al., 2002 suggested that biases introduced from the method of smoking assessment resulted in the failure of three prospective beta carotene clinical trials: CARET, ATBC, and the PHS 199 202 Gender specific l ung cancer survival may be influence b y a dietary factor, fat. There has been extensive research into the association between dietary fat and lung cancer risk 203 211 Alavanja et al. (2001), Goodman, et al. (1992), an d De Stefani et al. (1997) (study restricted to men) reported an elevated risk of lung cancer with an increased consumption of fat 204, 206, 212 A non statistically significant association was demonstrated by Swanson et al., 1997 210 for intake of red meats and increased lung cancer ri sk after adjusting for confounders. Conflicting results in the literature as cited in this chapter could be suggestive of inaccurate reporting and possible recall bias 10

PAGE 80

63 Table 4: Lung Cancer and Food Intake Cohort Studies Obesity and Body Mass Index (BMI) The major controversies concerning the exposure disease relationship between gender, obesity, BMI, lung cancer risk and lung cancer survival exist in the literature. There are conflicting opinions in the scientific community concerning th e association 40, 213 of low BMI and elevated lung cancer risk. In case control studies, the literature 26, 108, 214 219 cite increased risk of lung cancer. The scientific unit for BMI is 1 kg/m 2 220 The three levels of exposure are: low BMI (> 25 kg/m 2 ), n ormal or the reference group BMI ( > 21.9 kg/m 2 to <25 kg/m 2 ), and the high BMI group (> 25 kg/m 2 ). As noted in a report from the January February 2005 FDA Consumer Report, the average or median BMI has increased from 25 kg/m 2 in 1960 to 28 kg/m 2 in 2002 fo r the general US population. BMI can serve as a proxy

PAGE 81

64 measure for overweight and obesity. It is widely cited in the literature 220 th at a high BMI (> 25 kg/m 2 ) is associated with an increased risk to hypertension, diabetes mellitus, ischemic heart disease and in particular most cancers. A contradiction exists with lung cancer and low BMI (> 25 kg/m 2 ) 40, 218, 219 where there is an inverse relationship. As pointed out previously, the evidence based on BMI and cancer is blurred due to the progression of lung cancer and that effect on mortality 106, 220 Weight loss may occur (which affects BMI) due to the cancer process prior to the disease being diagnosed adding to the difficulty asso ciated with making a definitive causal inference in the BMI/Lung Cancer association 5, 106, 108, 220, 221 Some of the limitat ions of previous obesity, BMI, and lung cancer risk study designs may have served to mask a true association. Kanashiki 218 mentioned in his article that the previou s case control studies investigating the relationship between low BMI and lung cancer were based on participants with symptomatic lung cancer. This may have caused a misinterpretation of the relationship between the exposure and the disease because weight loss may be a sign or clinical symptom of the disease, lung cancer. Kanashiki 218 further reported that the re is a statistically significant association between low BMI and lung cancer for men ; women did not demonstrate a statistically significant association between low BMI and lung cancer (increased lung cancer risk) Previous cohort studies 108, 219, 222 such as Kabat and Wynder used self reported body size during data collection. Usin g a method of self reported body size has been noted to be problematic in the literature 223 ; as overestimates of height and underestimates of weight

PAGE 82

65 can be reported with this method. As noted by Henley 213 another issue with all previous prospective studies is that the numb ers were not large enough to exclude those that were smokers or that have preexisting diseases that may reduce body weight. This may have resulted in a spurious association bet ween low BMI and lung cancer. The main effect of BMI and lung cancer has been s hown by Rauscher 26 to have an increased Odds Ratio of 1.33 (95% CI 1.13, 1.57) in matched analysis for men and women with h igh BMI and being a non smoker. This conflicts with other studies tha t associate a low BMI with an increase in the risk of cancer as compared to a normal BMI group. These conflicting results serve as an example for the need for the additional clarification that the proposed research study will provide. Biologic plausibili ty defined by Gordis 224 as a consistency of the epidemiologic findings with existing bio logic knowledge. Therefore, without biologic plausibility, interpreting the data or making a definitive statement about the association between the exposure and the disease becomes problematic. In the case of suspect biologic plausibility, Gordis 224 suggests that the requirements for the sample size and the significance of any differences t hat may be observed may have to be escalated, e.g. increase the sample size to decrease the variability in the sample 155 Hennekens 73 states that biologic plausibility is a causal criterion for an association and that a known biologically plausible mechanism enhances the cause and effect rel ationship. From the literature review in the section above, the disease process in lung cancer has been shown to influence BMI levels or visa versa. Changes in the association between the exposure and the disease, for example, could be a result of change s in

PAGE 83

66 physiology during the preclinical phase of the disease; which would provide a biologically plausible mechanism. changes BMI or how BMI influences lung cancer has not been clearly established. Occu pation Several occupational exposures are known carcinogens and have been classified by IARC (an international agency) and the Occupational Safety and Health Administration (OSHA) in conjunction with the National Institute of Safety and Health (a US based agency) 225 The list of substances considered by NIOSH include arsenic and inorganic arsenic compounds, dintrotoluenes, beryllium and beryllium compounds, cadmium compounds, nickel compounds, and crystalline forms of silica 225 Diesel exhaust, coal tar pitch volatiles, coke oven emissions, and environmental tobacco smoke are other substances of variable chemical composition and are considered carcinogens by NIOSH. Other occupational risk factors (agents) include radon, vinyl chloride, polycyclic aromatic compounds, asbestos, and bischoloromethyl ether 10 Epidemiologic studies estimate a range of attributable risk percent associated with lung cancer and occupational exposure of 9% to 15% 10 Hessing and Hartung explored the excessive rates of respiratory c ancers for European underground metal miners in 1879 226 Radford and Renard (1984) examined the increased dose response relationship for radiatio n and lung cancer 227 The expected death rate for nonsmoking miners wi th lung cancer was 1.8 but the observed mortality rate for nonsmoking miners due to lung cancer was 18 227

PAGE 84

67 Deposits of uranium and radium and the subsequent by products of radioactive decay (radon) were determined to be the causative agent for the development of lung cancer for miners. Other occupational investigational st udies included Doll (1952) who demonstrated an increased risk of lung cancer for gas workers and Morgan (1992) reported that mortality from lung cancer had a standardized mortality ratio SMR of 1.65 226 Hinds et al. (1985) determined risk factors for lung cancer based on excessive relative risks for a number of occupational groups exposed to coal and tar pitch, diesel fuel and exhaust, arsenic, c hromium, asbestos, nickel, and beryllium 228 Hormones Sex differences in susceptibility and survival have been attributed to estrogen as a lung cancer risk factor and prognostic factor 229 Gender specific estrogen receptor (ER ) expression may offer a biologi cally plausible influence in female lung carcinogenesis 230 Schwartz et al. (2005) conduc ted a study of lung cancer tissue samples from two population based, case control studies (214 women and 64 men) 229 Normal lung tissue was obtained for comparison from subject during autopsy that did not have lung cancer. The association between the ER receptor status, subject characteristics, and survival were analyzed. The lung tissue was tested for the presence of nuclear estrogen receptor (ER) alpha and ER beta with immunohistochemistry. Lung tissue sample for tumor and normal tissue did not sta in positive for ER Nuclear ER receptors were found in 61% of the lung tumor samples (70% of the men and 58% of the

PAGE 85

68 women) and in 20% of normal tissue. Females were less likely to have positive ER tumors than males (OR = 0.54; 95%CI = 0.27 1.08). When the analysis was stratified on histologic type, women with adenocarcinoma were less likely to have positive ER tumors than males ((OR = 0.40; 95%CI = 0.18 0.89). Han et al. (2005) found that the gender specific estrogen receptor (Er ) may offer a plausible explanation that inter individual difference in Er expression (present in the lung) impact carcinogen metabolism and mutation. The research was based on genome studies of genes (CYP1A1, CYP1B1) key in carcinogen metabolism; those genes were the most responsive to cig arette smoke extract (CSE) in normal bronchial epithelial (NHBE) cells 230 Socioeconomic Status The risk of lung cancer and socioeconomic status patterns can be dependent upon In Canada, Mao et al. (2001) reported males with a lower socioeconomic status had an increased risk of lung cancer as compared to individ uals at a higher SES level 231 Females did not show an association between lung cancer risk and SES after adjustment for occupation, education le vel, income, and social class was made 10 Singh et al. (2002) reported on changing US area socioeconomic patterns for lung cancer mortality for the years 1950 through 1998 232 Temporal changes in the distribution of lung cancer mortality were shown for women in all age groups with a 7 times increased risk between 1950 and 1998 with a n overall higher mortality of women

PAGE 86

69 with lung cancer in higher socioeconomic groups. Lung cancer mortality for men (25 64 years) was 56% (95%CI = 49% 64%) higher in the lowest socioeconomic groups 232 The authors concluded that lung cancer mortality risk based on socioeconomics reversed for males from 1950 to 1998 with women demon strating an increased risk for lung cancer mortality independent of socioeconomic status 232 Environment Environmental factors play a distinct role in lung cancer etiology and survival patterns. Passive or environmental tobacco smoke (ETS) and occupational exposures are risk factors for lung cancer and causal associations have b een established. Veneis et al. (2007) investigated ETS and traffic related air pollution 233 Attributable risk percents for the proportion of lung cancer cases of never smokers and former smokers were 16 to 24%. The authors concluded that a reduction in air pollution levels, as measured by nitrogen dioxide (NO 2 ) levels less than 30 g/m 3 would prevent 5 to 7 % of all lung cancer cases 233 Indoor air pollutants have been studied as risk factors for lung c ancer in developed and developing countries 234 In a recent article by Ramanakumar et al. (1997) the risk of lung cancer and residential heating and cooking fuels was assessed for a North American population. The odds ratio for women as compared to men exposed to both traditional cooking and heating sources was 2.5; 95% CI = 1.5 3.6. Oriental women have been shown to be a t increased risk for lung cancer, in particular adenocarcinoma, which is attributed to prolonged and concentrated exposures

PAGE 87

70 to cooking and heating sources 172, 234 Diseases Associated with Lung Cancer the risk of lung cancer incidence and mortality 10, 235 Cigarette smoking and chronic respiratory diseases play a key role is carci nogenesis due to a continued cycle of injury and repair. Schabath et al. (2005) compared the medical histories of 1,375 health controls and 1,553 lung cancer cases in a case control study (1995 through 2003) with a focus on respiratory diseases (asthma, e mphysema, bronchitis, hay fever, pneumonia, and TB) Two biologically relevant b iomarkers for lung cancer, polymorphic genes ( matrix metalloproteinase 1 and myeloperoxid ase) were also assessed. Those with emphysema had an elevated for the risk of lung ca ncer (OR = 2.87; 95%CI = 2.20 3.76). positive for the adverse genotype had a significantly higher risk of lung cancer; OR metalloproteinase 1 + = 4.98; 95%CI = 2.94 8.44) and OR myeloperoxidase + = 4.23; 95%CI = 1.84 9.73 235 A previous history of hay fever was found to be protective with an OR of 0.32; 95%CI = 0.21 0.50. Alavanja et al. (1992) examined preexisting lung disease in nonsmoking women and the risk of lung cancer. The OR = 1.7 for the risk of adenocarcinoma and previous lung disease; the overall OR for all lung cancer typ es was 1.8. were significant for lung cancer risk; OR asthma = 2.7 and OR pneumonia = 1.5. The OR for e mphysema was 2.6 and tuberculosis OR = 2.0 for former smokers. The authors found

PAGE 88

71 an attributable risk percent (16%) among wom en that were nonsmokers and had a previous history of emphysema, asthma, pneumonia, and tuberculosis 235 Further investigation is warranted as the biologic role of respiratory diseases in lung cancer etiology is unclear as the evidence is not consistent. Treatments for Lung Cancer How lung cancer spreads throughout the body can be classified into three categories: intrathoracic (local), lymphatic (regional), and hemat ogenous (distant). The sequence of the cancer growth is sporadic and does not necessarily follow any particular order 72 Small cell carcinoma (oat cell) has the greatest probability of distant spread as compared to non small cell lung cancer; adenocarcinoma of the three NSCLC ty pes has the highest incidence of distant spread or metastasis 2, 4, 40 Dependin g upon the diagnosis of the lung cancer histologic type, stage, grade and the health of the patient, a clinical decision is made by the physician how the treatment will proceed. These treatment decisions are based on years of empirical data, research, and clinical trials; the standards of care established by the medical community are overseen by several organizations such as the American Medical Association, the American College of Surgeons, the National Cancer Institute, and the American College of Radiol ogy 33 Confined to the Lungs NSCLC confined to the lung is considered early stage disease and is treated with a

PAGE 89

72 surgical resection 72 Postoperative radiation therapy with or without chemotherapy is recommended for the treatment of microscopic disease 135 o morbidities are other considerations if the patient is a surgical candidate. Radiation Therapy is then the treatment of choice if the patient cannot tolerate surgery. Early stage small cell lung cancer is considered limited disease spread and the prima ry treatment is chemotherapy plus concurrent radiation therapy. There has been current interest in surgical procedures for limited stage SCLC; according to Anraku and Waddell (2006) when the disease is confined, surgery improves the local control and incr eases survival 54 The authors also note that continued research via clinical trials is warranted to confirm long term results. Local Spread Once the tumor has spread beyond the hemithorax and there is mediastinal lymph node metastasis, the treatment options include chemothe rapy and radiation therapy; at this stage surgery is contraindicated. Surgery could be an option in cases of limited mediastinal lymph node involvement in combination with chemotherapy or radiation therapy 72 Prior to the advent of CT scans and the ability to detect mediastinal lymph node metastasis, it has been estimated in the literature that 30% of all NSCLC patients would have received a surgical resection unnecessarily by current medical standards 72 Distant Spread Extrathoracic, distant, or hematogenous spread involves the growth of lung cancer into multiple organs. Treatment of extensive disease for non small cell lung cancer and

PAGE 90

73 small cell lung cancer is chemotherapy alone 135 The anticancer agents treat the disease systemically or throughou t the body. The future direction for treatments includes targeting therapies, i.e. the treatment of cancer cells treated with bio agents that attack the cells at the molecular level 135 Lung Cancer Relapse In the case a lung cancer relapse, the stage, grade, histologic type, and patient condition once again determines next steps in treatment options. Angeletti, et. al. (1995), noted that surgery is warranted in the case of an early stage lung cancer relapse or a second locally confined primary in the lung 236 Al though long term survival and local control has not been validated, the authors suggest this as a viable optio n to increase patient survival. Another approach suggested by Johnson, et. al. (1990) suggested a combination of cyclophosphamide doxorubicin, a nd vincristine or etoposide or both vincristine and etoposide for SCLC. C omplications of Lung Cancer The lungs are highly vascular and are supplied by a system of lymphatic glands so the greatest complication of lung cancer is the spread of cancer throug h different tissue and organs of the body (metastasis). Another major complication of lung cancer is the reappearance of the disease in the form of another primary neoplasm or the development of a secondary tumor 72 According to the American Cancer Society, lung cancer relapse

PAGE 91

74 commonly occurs within two years even in the event of a positive treatment course 2 Complications are associated with each treatment regimen; in particular chemotherapy agents have been shown to decrease survival resulting in the early termination of clinical trials 237 Venuta, et al., 2006, found in a retrospective study based on one hundred and thirty nine patients (100 males and 39 females), preoperative functional param eters, type of operation, associated disorders, staging, induction regimen (chemotherapy alone or combined with radiation therapy), all added to the complication rate for surgical procedures 238 After multivariate analysis for morbidity and mortal ity and controlling for age and lung functional volume; the results were not statistically significant. The complications for radiation therapy lung cancer treatments include decreased ventilation function, hemoptysis, and local relapse. Historically, th e major disadvantage to the use of radiation therapy was the amount normal tissue that had to be irradiated thereby reducing lung function in a lung already compromised by cancer. Advances in radiation therapy treatment modalities, e.g. Intensity Modulate d Ration Therapy (IMRT), Image Guided Radiation Therapy (IGRT), Respiratory Gated Radiation Therapy (RGRT), all allow for techniques to minimize motion and to increase the precision of the target or tumor coverage 239 241 Underberg et. al., 2005 found a 50% reduction in the primary tumor volume irradiated with the new treatment technique s. This means that the newer radiation therapy techniques can offer a consistently smaller irradiation volume so decreased toxicity can be expected; any longer term effect on survival would have to be investigated 242

PAGE 92

75 Lung Cancer Treatment Modalities The type of lung cancer treatment modality is dependent upon several factors that include the histological type of lung cancer, the size of the tumor, the location, extent or degree of regional s pread of the disease, and the general condition of the patient. Many of the treatment modalities, e.g. radiation therapy, surgery, chemotherapy, photodynamic therapy, can be used separately or combined to treat the cancer 239 241 As there are several treatment options, there are at the very least two main goals that are anticipated upon c ompletion of the treatments. First, there is an expectation that if a complete cure is not achieved, the progression of the tumor has been confined. This should have an overall effect of increased survival. Secondly, there is an expectation that the sym ptoms of the disease will be diminished for the patient, thereby imp roving quality of life issues. In the next several sections in this chapter, current and emergent technologies will be expanded upon. Radiation Therapy Radiation therapy or radiotherapy utilizes high energy rays to kill cancer cells 4, 47 The radiation is delivered to a very specific region of the diseased lung, with the goal of a minimal radiation dose given to normal tissue. Prior to a surgical procedure, radiation can be given to de crease the tumor size and destroy peripheral, microscopic disease surrounding the tumor 4 239 241 Radiation therapy can also be used for the relief of lung cancer symptoms, such as shortness of breath. There are several radiation therapy

PAGE 93

76 treatment options for lung cancer; external beam therapy and radioisotope therapy (brachytherapy) 6 External beam is the application of a radiation beam from an external source, e.g. linear accelerator, cyclotron, to the affect site of cancer. Recent reviewed the newer technologies to reduce radiation damage of the normal tissue, thereby preserving lung function 72 One of the newest and most costly (1.2 billion US dollars) radiation therapy treatment involves treating with a proton beam to reduce radiation induced morbidities. Protons deposit their energy maximally at a specific distance from the patient surface 72 This maximum deposition occurs at the Bragg Peak of the proton; the amount of radiation deposited is dependent upon the energy of the proton (a charged ionizing particle) and the medium or material of interaction. There is minimal interaction with normal tissue and decreased lung function toxicity. Endobronchial brachytherapy involves the application of a radioactive material to the affect site in the lung; a commonly used radioisotope is Iridium 1 92 72 Normally the treatment is done for a more circumscribed area of the lung (less than 10 centimeters in length) as compared to external beam rad iation. The application of the radioactive source can be done by two methods; either a high dose rate application (taking several minutes) or by the low dose method (can take several hours) 72 For the high dose rate (HDR) method, the radioactive source is introduced into the lung tumor via a flexible catheter. This catheter is place prior to the HDR during a bronchoscopic procedure. The treatment g oal is to minimize any normal tissue damage and deliver a therapeutic dose (3 to 10

PAGE 94

77 Gray per fraction) 72 The advantage of this technique is an incr eased local control of the tumor progression and it is less invasive than a surgical procedure with the associa ted postoperative morbidities. The main disadvantages to this technique is fistula formation and fatal hemoptysis 243 Chemotherapy Chemotherapy is another lung cancer treatment option, as shown in Table 5, which uses drugs to obliterate cancer cells The purpose of chemotherapy is to kill the cancerous cell or interr upt the cancer cell cycle, thereby preventing the growth of the tumor. One of the harmful effects of chemotherapy agents is the drug(s) destroy normal cells in the process of destroying the cancer cells. The destruction of normal cells/tissue can cause v arious side effects that include nausea, vomiting, diarrhea, increased susceptibility to infections, and in some instances, death 9 The various c hemotherapy drugs are administered either by infusion, orally, or as a combination of both during a treatment. NSCLC and SCLC can both be treated with chemotherapy agents; the optimum or most effective drugs for the treatment of lung cancer are platinum ba sed drugs, cisplatin and carboplatin 13 5 Other non platinum based chemotherapy agents include docetaxel, paclitaxel, gemcitabine, and irinotecan; these can be used in conjunction with cisplatin and carboplatin during the treatment regimen. For further information, Appendix I II Table 75 cont ains a listing with chemotherapy drugs typically used in the treatment of lung cancer, the type of agent, and major side or adverse effects.

PAGE 95

78 Typically, chemotherapy is used as an adjunct or secondary treatment to surgery or radiation therapy, dependent up on the histological type and stage of the disease. In the case of small cell lung cancer, chemotherapy is the treatment of choice because the procession of the disease is typically widespread throughout the body when the diagnosis of lung cancer is made 4, 135 Small cell lung ca ncer accounts for approximately 20 to 25% of incident lung cancer cases diagnosed 13 Of all SCLC diagnosed, only 40% of the cases have the disease limited to the chest cavity (thorax). The regimen of choice is chemotherapy plus radiation therapy. In some instances, prophylactic irradiation of the brain is given to tr eat micrometastasis; which is the early spr ead of the cancer to the brain. The five year survival rate for limited stage SCLC is 15 25% The median survival for patients that do not receive chemotherapy in combination with radiation therapy is 6 to 12 weeks 237 If the SCLC has developed into an extensive stage, the recommendation is to use chemotherapy alone; these cases account for 60% of all the newly diagnosed SCLC 237 When a case of SCLC is diagnosed at this stage, the disease has normally progressed or metastized to the brain, liver, bone, and/or bone marrow. According to Carney in his New England J chemotherapy drugs combinations and varying treatment regimens, the most significant improvement in survival was on average two months 244 At advanced stages of lung cancer, clinical trials are still being evaluated for efficacy 54, 59, 245 249 Gerold Bepler of the H. Lee Moffitt Cancer Center and Research Institute, noted

PAGE 96

79 that there is a renewed interest in t he development of new strategies for chemotherapy to increase survival for NSCLC patients 250, 251 This has c ome about due to advances in genetic research and the identification of genes on the chromosomes that have been identified as prognostic factors in the treatment of lung cancer. The level of the genetic makeup can determine which specific chemotherapy drug or combination of chemotherapy drugs will be most effective to treat the cancer 250 An increased level of two genes, RRM1 and ERCC1, has been shown to decrease the effectiveness of chemotherapeutic drugs 251 In the clinical trial, the International Adjuvant Lung Cancer Tr ial, Olaussen et al. (2006), demonstrated that patients with ERCC1 negative tumors had increased survival (adjusted hazard ratio = 0.65, 95% CI 0.50 0.86; p value = 0.002); whereas patients with ERCC1 positive tumors (adjusted hazard ratio = 1.14; 95% CI 0.84 1.55; p value = 0.40) 252 Table 5 gives a summary of the recommendations found in Collins, et al 2007 4 and describes the primary or first choice of treatment for a particular lung cancer type, i.e. Non Small Cell Lung Cancer and Small Cell Lung Cancer, an d stage of the disease. The secondary treatment modality recommendations and the associated five year survival rates are also listed in Table 5.

PAGE 97

80 Table 5: Lung Cancer Treatment Recommendations Stage Primary Treatment Modality Secondary Treatment M odality Survival Rate Non Small Cell Lung Cancer (NSCLC) I Surgery (Resection) Chemotherapy 5 Yr survival rate, > 60 70% II Surgery (Resection) Chemotherapy with or without Radiation Therapy 5 Yr survival rate, > 40 50% III A (resectable) Pre operative chemotherapy followed by surgical resection (preferable) or surgical resection Chemotherapy with or without Radiation Therapy 5 Yr survival rate, 15 30% III A (unresectable) Chemotherapy plus concurrent radiotherapy (preferable) or chemothera py followed by radiotherapy None 5 Yr survival rate, 10 20% III B (pleural effusion) or IV Chemotherapy with 2 agents for 3 or 4 cycles (preferable) Surgical resection of solitary brain metastasis and surgical resection of primary (T1) lesion None None Median survival, 8 10 mo 1 Yr survival rate, 30 35% 2 Yr survival rate, 10 15% 5 Yr survival rate 10 15% Small Cell Lung Cancer (SCLC ) Limited Disease Chemotherapy and concurrent Radiation Therapy None 5 Yr survival rate, 15 25% Extensiv e Disease Chemotherapy None 5 Yr survival rate, < 5% All chemotherapy regimens include either cisplatin or carboplatin. A complete list of clinical trials is available at http://www.cancer.gov and up to date appr oaches to the treatment of non small cell and small cell lung cancer are available from the National Comprehensive Cancer Network at http://www.nccn.org. a small but significant survival advantage with cisplatin based adjuvant therapy. Physicians should strongly consider such therapy for appropriate patients. ial therapy. Source: Adapted from Collins, et. al., 2007

PAGE 98

81 Surgery Surgery removes lung cancer or the tumor during an operation. The operation or surgical procedure is based on the histology or type of lung cancer and where the tumor is located. A surgi cal procedure is not recommended for small cell lung cancer, unless the disease is very circumscribed and is not too advanced at the time of diagnosis 4 Surgery is recommended by the American College of Ch est Physicians and the American Joint Commission on Cancer 4, 9, 253 for NSCLC. According to Rivera, et. al., (2003), Stage I and Stage II have the optimum prognosis, with a average five year survival rate of 6 0% to 70% for Stage I and a greater than 40 50% for Stage II 9 Other considerations for a surgical procedure for the excision of a lung tumor i nclude the location of the tumor, i.e. the tumor may be too close to the heart or the trachea, if the patient is a surgical candidate and can withstand the surgical intervention, and the stage or extent of the disease 135 If the disease is too extensive throughout the body, a surgical proced ure to remove the lung tumor is not recommended. Radiation Therapy and/or chemotherapy may be other options if surgery is not indicated 4, 9, 135 Different types of surgical procedures include 1) wedge resec tion, 2) segmentectomy, 3) lobectomy, 4) lymph node removal, and 5) pnumonectomy. A wedge resection consists of the removal of part of the lung; it is used when the tumor is confined to a particular location in the lung. When the tumor is removed, a port ion of the healthy or non diseased lung is also removed. This is done in an attempt to eliminate any microscopic disease around t he periphery of the tumor bed. A segmentectomy is the

PAGE 99

82 procedure where a greater margin is taken around the tumor bed. A lobe ctomy entails the removal of one of the five lobes of the lung; the right lung has three lobes and the left lung has two lobes. This type of surgical resection involves the surgeon making an Once the incis ion has been made the surgeon then spreads the ribs apart; this makes the lung tumor accessible for removal. A video assisted lobectomy (VATS) employs a small video camera (a thorascope) inserted into the chest cavity; the images received form the thorasc ope guides the surgeon to the operative area in the chest. This particular procedure minimizes the bleeding and complications of the surgery; the patient stay is reduced from five to seven days to two to three days. A lymphectomy or lymph node removal ca n be accomplished during the surgical procedure. The lymph nodes are examined by the pathologist for signs of disease; if the nodes were positive, this would be an indication of disease spread. In the case of positive nodes, the surgeon may opt not to re move any of the tissue as the stage of the disease has progressed and surgery is not an indication 4, 54, 59, 253 A pnumonectomy is the comple te removal of the entire lung. This procedure can A pnumonectomy is recomm ended for centrally located tumors and tumors that involve more then one lobe. The removal of the lung is warranted when the tumor has spread throughout the lung but has not metastized to other parts of the body. The complications from surgery may includ e internal bleeding, infection, lymohocytopenia (low white

PAGE 100

83 count), and possible reoccurrence of the disease. Combination Therapy Treatment modalities are combined in an effort to increase the length of survival for lung cancer patients. The standard of practice for small cell lung carcinoma with limited disease (no evidence of spread) is chemotherapy drugs concurrent with radiation therapy 4, 135 According to Anraku and Waddell (2006), surgery can be warranted under certain conditions for early stage SCLC. Chemotherapy with a combination of surgery for patients with T 1 2 N 0 SCLC may enhance local control but the authors also note that clinical trials are needed to validate these results 54 Surgery and chemotherapy can be offered to patients that have a mixed tumor type (SCLC and NSCLC components ) as the anticancer agents are less effective against NSCLC in the limited or early stages of the disease 54 Stage I and Stage II NSCLC use the combination of surgical resection, radiation therapy, and/or chemotherapy. Surgery is the primary treatment with radiation and/or chemotherapy as the adjuvant therapy. Other treatment regimens include surgery with or without preoperative chemotherapy for resectable Stage IIIA with the adjuvant therapy of chemotherapy with or without radiation therapy. The five year survival rate i s 15 to 30% for this course of therapy 4, 9 The recommendation for Stage IIIA unresectable NSCLC is chemotherapy with concurrent radia tion therapy or radiation therapy treatments after the course of chemotherapy treatments. The five year survival rate declines for this combination of modalities to 10 to 20% 4, 9, 135 Patients with Stage II IB

PAGE 101

84 (pleural effusion) or IV are given chemotherapy, resection of a primary T1 tumor and primary brain metastasis. Emergent Modalities The optimum treatment for lung cancer is based on stage and grade of the disease 4, 144, 243, 249, 253, 254 There are several recently developed endoscopic modalities for the treatment of early stage lung cancer. 243 Although the newer treatment techniques offer a less invasive technique, decreased perioperative morbidity and reduced cost as compared to conventional modalities, the techniques and methods require validation 4, 243 Endoscopic therapies such as photodynamic therapy, brachytherapy, neodymium yttrium aluminum garnet laser, electrocautery, and cryotherapy; offer an alternative in the treatment of early sta ge lung cancer. These modalities can be applied during a procedure known as a fluorescence bronchoscopy. Pathological changes in the appearance of normal lung tissue can be detected during bronchoscopy utilizing fluorescence. Normal tissue fluoresces (e mission of light) at varying energy levels when compared to cancerous tissue; this difference in energy levels is seen by the human eye as differences in color 243 The main d isadvantage to this detection method is the high false positive rate; inflammatory processes or trauma can cause changes in the light patterns that could be perceived as cancerous 243, 255 Lam, et al., 1993, reported a sensitivity of 72.5% and a specificity of 94% in the detection of advanced dys plasia 243, 255 When comparing fluorescenc e bronchoscopy to conventional bronchoscopy, the conventional technique

PAGE 102

85 had a sensitivity of 48.4% 255 A multicenter trial comparing the two techniques found fluorescence bronchoscopy detection rate of invasive lung carcinoma of 95% as compared to conventional bronchoscopy with a 65% detection rate 256 Photodynamic therapy (PDT), as a treatment for lung cancer, has been used in conjunction with fluorescence bronchoscopy 243 PDT involves targeting the lung tumor cells with a photo sensitizing agent and an application laser light to the affected area during a bron choscopy. The laser light, typically 630 nanometers, activates the chemical sensitizer producing a photochemical reaction at the cellular level. This results in the destruction of the tumor cells by an oxidative process 243 Lung cancer treatments can be done by a neodymium yttrium aluminum garnet (Nd: YAG) laser. This treatment has been done for palliation purposes but the literature does not support a significant contributio n in the treatment of early stage lung cancer 243 Gerasin et al. (1990), did report success with this technique for early stage lung cancer in the contralateral lung when a lo bectomy is contraindicated 257 Scientific advancements in the fields of chemotherapy, surgery, and radiation therapy involving tumor developme nt (carcinogenesis) and lung cancer progression have been possible with the discovery of genetic materials that are involved in the disease process 4, 9 There has been recent evidence that a drug, epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor, has been effective in the treatment of bronchialveolar cell carcinoma (a NSCLC type). This particular cancer type is comm on in women and in non smokers. The drug causes the shutdown of the epidermal growth factor receptor

PAGE 103

86 protein, thereby preventing the development of the cancer. Conclusions and Assessment of the Literature The purpose of this research is to test the hypoth esis of a survival difference in women with lung cancer as compared to men, dependent upon treatment modality (surgery, chemotherapy, radiation, or combination), histological type, and stage of disease. An extensive literature search and review was prefor med to investigate the variables that could be involved in the associated lung cancer treatment received and how There is limited research regarding the risk of being a woman with lun g cancer and the treatment received as compared to men. As described in the literature review, there are no studies combining the specific treatment modality received or combinations of treatment received gender, stage, grade, morphology and demographic factors with respect to survivorship. Visbal et.al. (2004) noted the Relative Risks between males and females and survival adjusting for stage, histological, and treatment type. Although there were significant results for stage (Stage IIIB RR = 1.22, 95% CI = 1.02, 1.46 and Stage IV RR = 1.29, 95% CI = 1.15, 1.44), there was no significant results for treatment type received. Also the authors did not mention if there were any combined treatments of chemotherapy, radiation, or surgery 17 In the article by Radzikowska et. al. (2002), (n = 11,479) a multivariate survival analysis was performed based on a ge (categorized into two groups; group I < 50 years old, group II > 50 years old) gender performance status, clinical

PAGE 104

87 stage of lung cancer (four stages), h istology (adenocarcinoma, squamous, SCLC, and other) and two treatment types (surgery and non surgical) 15 In the Cox Proportional of lung cancer, performance status and stage of the disease. no confidence intervals; any truly significant results could not be evaluated. The reported (squamous cell RR = 1.09, SCLC RR = 1.42, Other Histological Type RR = 1.46) wi th the exception of the reference group (RR = 1.0) There were significant p values all values < 0.004 with the exception of the histological type for squamous cell (p value = 0.29). Lung cancer was six times more frequent in males versus females and wome n with lung cancer were younger and smoked less than males with lung cancer. T here are many histological types of lung cancer, finding an optimum treatment regimen that will increase survival for a particular lung cancer type is challenging. Each histolo gical type has its own medical intervention that can include any combination of surgery, radiation therapy, and/or chemotherapy. Gender specific incidence and survival rates were shown to be different for the lung cancer types. In the article by Thomas e t. al. 48 placing women at a greater risk were reported to i nclude molecular variables such as different metabolism of tobacco related carcinogens, possible association with human papilloma virus ( HPV ) infection, and that women have less DNA

PAGE 105

88 repair capacity (DRC) ; the authors also noted that women had better surviv al outcomes stage for stage than men 48 Presently, there are no published quantitative results that show whether there is a statistically significant difference regarding survival due to a particular treatment for women as compared to men having the same histological type grade and stage of lung cancer. The statistical methods performed in the articles cited in the literature review did not evaluate interaction effects of the independent variables; this is a major limitation as any gender specific diffe rences based on any moderating variables could not be evaluated. Performing this research pr ovide s the scientific evidence to answer the question concerning gender, stage, grade, morphology, age, martial status and race and their impact on survival. F indin gs of treatment differences by major histological types are presented in this dissertation. In conclusion, the goal of this research was to provide a statistical and biologically plausible model demonstrating gender differences in lung cancer survival exis t based on the treatment received.

PAGE 106

89 CHAPTER III: PROCEDURES AND METHODS Introduction The purpose of the research presented in this dissertation was to determine if gender differences exist for a treatment (s) utilized for lung cancer and if that treat ment (s) received impacts gender specific survival. The epidemiologic study design was based on a historical cohort of lung cancer cases drawn from population based state wide cancer registries Each cancer registry were members of the North American Assoc iation of Central Cancer Registries, NAACCR 258 The participants were men and women with newly diagnosed/incident primary lung cancer diagnosed between January 1, 2000 and December 31, 2004. All lung cancer ca ses selected were pathologically confirmed and classified on the four major histological types of the disease i.e. adenocarcinoma, squamous cell carcinoma, large cell carcinoma and small cell carcinoma Aims/Hypothesis Aim 1 : The first aim was to deter mine if men and women with the same histologic type, stage, and grade of lung cancer receive d the same treatment type. Any effect, such as any interaction that the covariates ( histologic type, stage, and grade) may exert on the relationship between gender and treatment received must be evaluated. It If the lung cancer treatment is gender dependent, this may impact gender specific survivorship.

PAGE 107

90 Hypothesis 1 : Women wi th the same histological type, stage and grade of lung cancer will receive the same treatment modality as compared to men with the same histological type, stage and grade of lung cancer. Aim 2 : The second aim is to evaluate the overall relationship betwe en survival and gender for the lung cancer cases. The goal is to obtain an assessment of the overall survivorship by gender Hypothesis 2 : There is a statistically significant difference in survival in women with lung cancer as compared to men with lung cancer. Aim 3 : The third aim of this study is to expand the investigation of treatment modality differences and gender specific survival. The goal is to determine if men and women with lung cancer grouped or stratified by treatment modality, histologic t ype, stage, and grade exhibit or demonstrate gender specific survivorship. Hypothesis 3 : Wo men with the same histological type, stage, grade of lung cancer, and the same treatment modality differ significantly in survival as compared to men with the same histological type, stage, and grade of lung cancer, and the same treatment modality. Participant Description and Case Identification The study participants in this research are primary lung cancer cases drawn from population based state wide cancer regi stries. Data on cancer cases and cancer deaths are collected, managed, and analyzed by a system of state based cancer registries 1 The

PAGE 108

91 majority of state cancer registries are members of the National Progra m of Cancer Registries 259 ; the NPCR was established in 1992 by the Cancer Registries Amendment Act 259 The NPCR is administered by the Centers for Disease Contro l and Prevention (CDC) and is under the Division of Cancer Prevention and Control 259 Prior to the establishment of the National Program of Cancer Registries, there were ten states with no cancer registry; today ther e are forty five states with cancer registries to include the Virgin Islands, the Republic of Palau, Puerto Rico, and the District of Columbia 258 Another international cancer registry organization that certifi es NPCR is the North American Association of Central Cancer Registries 258 NAACCR was established in 1987; it represents state cancer registries and professional organizations such as the Americ an College of Surgeons, the American Cancer Society, and the Public Health Agency of Canada 258 ncer registries; this ensures the data collection methods used by each state registry are complete, accurate, and done on a timely basis. The each state registry that is a member of NAACCR submits cancer case data obtained from medical facilities, e.g. ho spitals, surgical centers, laboratories, outpatient facilities, physician offices, and radiation therapy centers, to the central cancer registry. Cancer information is collected or abstracted in a standardized format into highly specified field arrangemen ts 258 The standardization of field information allows for the intercomparison of data within the state, with other states, and on a national level 258, 259

PAGE 109

92 One of the first steps in the selection of the primary lung cancer cases for this research study was to identify the eligible cancer registries. The criteria for the state/state cancer registries selected had to be established. Each selection parameter or crite rion was critical; the ultimate goal being selection criteria that could be utilized to generate a dataset free from bias. The criteria for selection of a state/state cancer registry (Table 6) for this research were as follows: Table 6: Criteria for State /State Cancer Registry Selection 1. The registry must exist in a state in the United States of America, 2. The registry must be population based, 3. Each registry must be selected from the four US regions as defined by the US Census Bureau, 4. The individual lung c ancer cases must be randomly selected, as each lung cancer case in the state cancer registry must have an equal chance of being included or excluded in the registry, 5. 6. The data must include primar y lung cancer cases diagnosed between 1 1 2000 and 12 31 2004, a five year time frame, 7. The state registry must be a member of NAACCR, 8. Each state registry must meet the criteria of the Standards for Cancer Registries: Standards for Completeness, Quality, 258 and 9. The state cancer registry must have achieved NAACCR certification (a minimum of 3 years gold certification and maximum of 2 years silver certification see Table 6. 10. Each state must be randomly selected as not to introduce selection bias, 11. The data must be accessible and retrievable to the researcher conducting the study.

PAGE 110

93 T he NAACCR guidelines for certification are measurable standards and each year state registrie silver designation (if the goals were achieved) by NAACCR. The selection of states for the research study began with a process as outlined in Figure 10. Figure 10: State Selection Process 37 USA States (50 13 SEER States) 50 USA States 29 USA NAACCR States (Reporting Non Reporting States) 14 USA States (Meet Certification Guidelines) 8 USA States (Randomly Selected)

PAGE 111

94 bias If one state is selected over another state it must be purely by chance thereby eliminating any possible influences or bias. In other words, each state cancer registry must have an equally likelihood of being selected; one state may be inherently different and this difference would tend to diminish with the having all the states randomly selected. Another source of bias may be introduced by the investigator if that investigator selected a particular state over another state due to personal or unscientific reasons the random process would be invalid . The first text box of Figure 10 represents individual lung cancer cases fr om the fifty US states which serve as the population from which the final research data set (Criterion 1) was drawn. The research protocol was developed as to include lung cancer cases selected from population based cancer registries (Criterion 2). The f ifty states are sub divided into geographic regions by the US Census Bureau (Criterion 3) The Bureau identifies four major US regions (South, Midwest, West, and the Northeast) which correspond to regional populations from which the lung cancer cases will be selected As discussed in the literature, p opulation characteristics can differ with geographic location 155, 260 262 and it is critical that the individuals selected are randomly selected lung cancer cases as not to introduce bias (Criterion 4) Of the 50 states, 13 states were excluded (text box 2 of Figure 6) leaving 37 NAACCR states. The states excluded were members of SEER and those SEER states were: 1 Connecticut, 2 Hawaii, 3 Iowa, 4 Louisiana, 5 New Jersey, 6 New

PAGE 112

95 Mexico, 7 Utah, 8 Georgia (multi country areas of Atlanta & rural Georgia), 9 Michigan (Detroit), 10 California (San Francisco Oakland, San Jose Monterey, Los Angeles county, remaining counties of California), 11 Washington (Seattle Puget Sound), 12 Arizona (American Indians), and 13 Alaska (Alaskan Natives). SEER is an active registry system and those states under SEER do not meet selection criterion 5. This is important as there are two main types of cancer registries, passive and active, that differ in the method data are collected. An active cancer registry collect s the data from the medical facilities whereas a passive registry has the data sent to the registry from the medical facilities (reporting facilities) that are part of the state wide system. An example of an active state cancer registry would be a member of SEER such as the Kansas Cancer Registry. NAACCR cancer registries are passive; some of the state cancer registries are listed in In this research, passive registries were only selected (Criterion 5). Having only passive registries served the following standardization of the states selected; this helped to minimize selection bias by only selecting states that have a similar reporting mechanism and reporting criteria. The thirty seven states were evalu ated (text box 3 Figure 10.) for their NAACCR status. Eight of the thirty seven states were not members of NAACCR for a portion of the years under study (2000 2004), thereby excluding them from participation in this research; 29 US NAACCR states remained (Criteria 6 and 7). Selection Criterion 6, restricting the time period under study (a 5 year range), will help to reduce any temporal differences associated with lung cancer treatments. A temporal bias may be introduced

PAGE 113

96 when studying extended time period s or ranges. Treatment modalities have drastically changed over the past 20 years, and trying to determine treatment effects over that 20 year time period would be more difficult to ascertain. Survival differences during this 5 year range are of particul ar concern; the treatment modality utilized to treat the lung cancer case should be more consistent and this could impact survivorship. Therefore, the date/time range established for lung cancer case data inclusion starts January 1, 2000 and ends December 31, 2004. The reporting system used for the cancer registries of interest for this research is based o n Criteria and Standards for Eligibility of NAACCR Registry Certification and CINA Combined Rates The criteria and the standards are displayed below in Table 7. Using this criterion, fifteen (15) states, as shown in Table 8., were excluded and fourteen (14) NAACCR states remained from the 29 states (text box 4 of Figure 10.). The selection criteria for state inclusion were based o n the grading scales established by the central cancer registry as outlined in Table 7 below ( NAACCR Criteria and Standards for Gold/Silver Certification ). NAACCR c ertifie s for igh Quality Incidence Data ncer registry is contained in the North American Association of Central Cancer Registries, Inc. Standards for Cancer R egistries Volume III Standards for Completeness, Quality, Analysis, and Management of Data For example, the criteria and standards ar e used to evaluate characteristic variables of the tumor such as tumor morphology (histology and behavior), stage, grade, and the method of diagnostic confirmation. The standards are used to assess

PAGE 114

97 the quality of the information in the individual state ca ncer registry, completeness of the data reported, and timeliness of reporting 258 In summary, a total of fifteen (15) of the twenty nine (29) states did not meet the selection criteria as outlined in Criterion 9; thereby excluding those states from participation (see Table 7). Table 7: NAACCR Criteria and Standards for Gold/Silver Certification Source: Standards for NAACCR Cancer Registries: Standards for Completeness, Quality, Analysis, and Mana gement of Data Note: DCO = Death Certificate Only Table 8 lists the twenty nine NAACCR states from the four US Census Bureau defined regions and their certification status over the study 5 year time period (2000 2004).

PAGE 115

98 Table 8: Annual NAACCR Certifi cation Designation Five Years of Certification States (29) by Region (4) REGION STATE 2000 2001 2002 2003 2004 Annual Incidence Average Annual Cases WEST COLORADO gold silver gold gold gold 54.9 2011 IDAHO gold gold gold gold gold 55.2 683 MONTANA silver silver silver gold gold 66.5 663 NEVADA gold gold gold gold gold 76.9 1622 OREGON gold gold gold gold gold 69 2489 SOUTH ALABAMA silver silver gold silver gold 75.2 3569 ARKANSAS silver silver silver 81.1 2404 DELAW ARE silver gold gold 76.6 649 FLORIDA gold gold gold gold gold 74.6 15838 NORTH CAROLINA gold silver silver silver gold 68.5 5611 OKLAHOMA gold gold gold 83.9 3068 SOUTH CAROLINA silver silver gold gold gold 74.7 3111 TEXAS gold gol d 68 12162 WEST VIRGINIA gold gold gold gold gold 87.7 1915 MIDWEST ILLINOIS gold gold gold gold gold 72.8 8836 INDIANA silver gold gold silver gold 80.6 4931 KANSAS gold gold gold gold 66.5 1841 MINNESOTA gold gold gold gold 58.8 2843 MISSOURI gold silver gold gold gold 78.6 4731 NEBRASKA gold gold gold gold gold 62.4 1118 OHIO silver silver silver silver 74.6 8993 SOUTH DAKOTA gold silver 58.7 486 WISCONSIN gold gold gold gold 65.9 3700 NORTHEAST MAINE gold g old gold gold 79.1 1183 MASSACHUSETTS gold gold gold gold gold 70.5 4764 NEW HAMPSHIRE gold gold gold silver gold 67.7 860 PENNSYLVANIA gold gold gold gold 70.4 10292 RHODE ISLAND gold gold gold gold gold 73.8 860 VERMONT silver gold 63.9 418

PAGE 116

99 Of the 29 states, 15 states were excluded as the state had to meet at a minimum three years of gold certification and a maximum of 2 years of the silver certification as defined the NAACCR criteria (see Table 7.) for the five year time frame under s tudy (Criterion 9). Table 9 lists the final 14 states; the minimum number of states in a region (the South and the Northeast) was three with a maximum number of states of four in the West and Midwest. Table 9: Annual NAACCR Certification Designation by R egion and State (14) REGION STATE 2000 2001 2002 2003 2004 Annual Incidence Average Annual Cases WEST COLORADO gold silver gold gold gold 54.9 2011 IDAHO gold gold gold gold gold 55.2 683 NEVADA gold gold gold gold gold 76.9 1622 OREGO N gold gold gold gold gold 69 2489 SOUTH FLORIDA gold gold gold gold gold 74.6 15838 SOUTH CAROLINA silver silver gold gold gold 74.7 3111 WEST VIRGINIA gold gold gold gold gold 87.7 1915 MIDWEST ILLINOIS gold gold gold gold gold 72.8 8836 INDI ANA silver gold gold silver gold 80.6 4931 MISSOURI gold silver gold gold gold 78.6 4731 NEBRASKA gold gold gold gold gold 62.4 1118 NORTHEAST MASSACHUSETTS gold gold gold gold gold 70.5 4764 NEW HAMPSHIRE gold gold gold silver gold 67.7 860 RHO DE ISLAND gold gold gold gold gold 73.8 860 The fourteen states were distributed from the four US geographic regions from each region the intent was to randomly select two states (Criterion 10). As stated previously, t he states were selected at random as not to introduce bias. The reasons for

PAGE 117

100 selecting two states from each region were: 1. At least two states were needed from each region to measure or account for variability within the regions. 2. Including all fourteen states was not feasible due to limited resources, e.g. cost considerations, manpower, and time constraints. In the selection of the states from the four US geographic regions, a random sample selection was made utilizing a SAS program; these samples represent the population from which they we re drawn (all US primary lung cancer cases) To account for any variability of the population within each region, more than one state for the region had to be selected. At a minimum, at least two states from each region must be selected in order to accou nt for any variability within the region As part of the selection criteria the data must be accessible (Criterion 11); not all states consent ed to having the data distributed to an outside individual. Logistical issue s, such as data unavailability w ould prevent the selection of a state registry. An additional logistical issue that could be encountered could be although a particular state registry may meet NAACCR requirement for certification, the state reporting system may not report a variable needed for the research under study. The final step was to call each of the state registries and request the procedure the particular state registry utilized for a data request (Criterion 11). West Virginia was hesitant to participate due to concerns for the p

PAGE 118

101 that was submitted to the University of South Florida (the application contained confidential information about the author and committee members); the registry would not accept the official IRB approval letter. Also, West Virginia does not/did not report the date is reported by year and does not include the day and month. As West Virginia did not meet selection Criterion 11, South Carolina was selected to participate; the resultant 8 states selected are given in Table 10. Additionally, after fourteen mon ths of requesting data from the Missouri Cancer Registry with no data forthcoming and in the interest of completing this research, a decision was made to randomly select a different state from the Midwest region, Nebraska was selected. Table 10: Final NAAC CR Eight State Cancer Registries REGION STATE 2000 2001 2002 2003 2004 Annual Incidence Average Annual Cases WEST OREGON gold gold gold gold gold 69 2489 IDAHO gold gold gold gold gold 55.2 683 SOUTH FLORIDA gold gold gold gold gold 74.6 15838 SOUTH CAROLINA silver silver gold gold gold 74.7 3111 INDIANA silver gold gold silver gold 80.6 4931 MIDWEST NEBRASKA gold silver gold gold gold 78.6 4731 MASSACHUSETTS gold gold gold gold gold 70.5 4764 NORTHEAST RHODE ISLAND gold gold gold gold gold 73.8 860 In summary, the selection criterion for this research utilized a process to minimize selection bias. For example, Criterion 4, the state registry must use passive reporting

PAGE 119

102 methods as outlined by NAACCR and not use other methods for r eporting as in SEER State registries. Possible geographic differences in the population under study are addressed with selection criterion 3. It is critical to get a fair comparison of lung cancer cases; selecting cases just from the Northeast could intr oduce bias possibly invalidating several study results. The 10 th item for selecting a state (registry) is the selection cannot be done by the researcher in a biased manner; the selection must be made by a random assignment. The last criterion, Item 11, ( selection of the state based cancer registry) is that the data must be available for acquisition. If the data cannot be acquired from a state registry by the author that state cancer registry will be excluded from selection. The seventh criterion is that the state registry must be a member of NAACCR. The NAACCR has standardized guidelines for abstracting data. Data can only be compared if the methods and information collected are consistent and complete. Deviations from a standardized format can intro duce error into the study results affecting internal validity and external validity. Variables of Interest (I nclusion and Exclusion Criteria) Inclusion C riteria Primary lung cancer cases from state population based cancer registries that are a member of the North American Association of Central Cancer Registries were identified. A primary site classification (NAACCR Code 400) is made by the state cancer registry based on site of tumor origin and specified in the medical record. In this research the

PAGE 120

103 prima ry site code is specified by ode 36 for lung. This classification is made in accordance with ICD O coding schemes. Cases that were diagnosed between January 1, 2000 and December 31, 2004 were selected. An extensive inclusion criteria are listed in th is Chapter ( Three ) criteria include a known date of diagnosis, that the case must be confirmed from a tissue or cell sample and in terms of the case assessment : the primary lung cancer case w as classified as analytic 1, 258 In order to perform survival analysis, it is critical to have an origin or beginning (date of diagnosis will serve as the origin), an observed time range, and an endp oint/conclusion for the study or a valid analysis cannot be completed Also, for this research, a lung cancer case must be diagnostically confirmed by means of a positive histology from the tumor tissue or a positive cytology (cells examined microscopical ly) as not to bias any subsequent results with the addition of histologically unconfirmed primary lung cancer cases in the data set T he codes include 1 2, 4, 5 the codes describe the methods of a diagnostic technique with a lung tissue/cell sample of the tumor. An analytic lung cancer case classification code denotes that part of the diagnosis and/or treatment of the lung cancer case was performed at the reporting (cancer registry) facility. An analytic case can further be defined under the NAACCR classification Class of Case (NAACCR Code 610) describes the criterion for inclusion as an analytic case with the codes 0 2 as shown in T able 3.12. Non analytic cases (codes 3 through 9) are cases that can have a greater chance of error

PAGE 121

104 as the information is abstracted from case information not directly associated with the reporting facility. The data from non analytic lung cancer cases c an be subject to bias when the information is provided by a patient (recall bias) that had the diagnosis and treatment at a facility different from the reporting facility. The decision was made to include only analytic cases for this research. Exclusion C riteria As only primary lung cancer cases are included in this study, secondary or recurrent lung cancer case will be excluded. The variables of interest such as gender include codes or categories that are not of interest in this research; those data ar e excluded These gender codes include 3= hermaphrodite, 4 = transsexual, or 9 = unknown/not stated 258, 263 Other variable codes of interest that have a code category not known or not stated (code = 9) must be identified during exploratory analysis. The number of missing values (codes for the sex of an individual cancer case) can possibly impact the research results. These values can provide information and knowing the exact number of individuals not represented in the final data set is important as it can decrease the va lidity of the results (increases the uncertainty) if the number of missing values is large. Some of the other variables that include the not know or not stated category (code = 9) are primary site, histology, stage, grade, treatment type (chemotherapy, su rgery, radiation therapy), tobacco use, marital status, and vital status. It was important to address the missing values and record the number for each category as to assess the

PAGE 122

105 impact those missing values on the study results. In the next (Chapter Three) each of the research patient parameters or variables of interest will be described The central cancer registry information/data variables that were requested by the investigator for each state cancer registr y are listed in Table 11.

PAGE 123

106 Table 11: NAACCR Variable Code and Description NAACCR Code Description of Variable NAACCR Code Description of Variable 1 20 Patient ID Number 24 480 Morphology Coding System Original 2 150 Mari tal Status at DX 25 490 Diagnostic Confirmation 3 160 Race 1 26 500 Type of Reporting Source 4 161 Race 2 27 560 Sequence Number Hospital 5 162 Race 3 28 580 Date of 1st Contact 6 163 Race 4 29 610 Class of Case 7 164 Race 5 30 630 Primary Pay or at DX 8 190 Spanish/Hispanic Origin 31 759 SEER Summary Stage 2000 9 220 Sex 32 760 SEER Summary Stage 1977 10 230 Age at Diagnosis 33 1200 Record Date of First Surgery 11 240 Birth Date 34 1210 Record Date of First Radiation 12 250 Birthpla ce 35 1220 Record Date of First Chemotherapy 13 340 Tobacco History 36 1290 Record Summary Surgical Primary Site 14 390 Date of Diagnosis 37 1360 Record Summary Radiation 15 400 Primary Site 38 1390 Record Summary Chemotherapy 16 410 Lateralit y 39 1750 Date of Last Contact 17 419 Morphology Type and Behavior ICD O 2 40 1760 Vital Status 18 420 Histology (92 00) ICD O 2 41 1910 Cause of Death 19 430 Behavior (92 00) ICD O 2 42 1930 Autopsy 20 521 Morphology Type and Behavior ICD O 3 43 1940 Place of Death 21 522 Histology (92 00) ICD O 3 44 2090 Date Case Completed 22 523 Behavior (92 00) ICD O 3 45 2110 Date Case Exported 23 440 Grade 46 3000 Derived AJCC Stage Summary Some of the research variables (independent or explanatory and dependent or

PAGE 124

107 outcome) include the individual lung cancer case identifier assigned by the cancer registry (this number was de identified to the researcher of this study), the cancer registry identification number, the date of first contact (the date may be representative of a ray, or laboratory test), the treatment modality received (radiation, surgery, chemotherapy), and the type of reporting source, e.g. hospital, outpatient facility. It should be noted that the d ate of lu ng cancer d iagnosis NAACCR recommends that the best approximation for the date of diagnosis should be used versus coding the date as unknown (9) 258 Therefore i n the situation of the year of diagnosis being kn own but no other information on the month or day is given, the general abstracting instruction is to use June 15 for the year indicated. If the year and month are available but not the day, the 15 th of the month is entered. The other p atient demographic variables of interest include gender, race, marital status at time of diagnosis, primary insurance payer at diagnosis, a birthplace Geocode, and birth d ate. Table 11 includes NAACCR variable names and the associated NAACCR item number to that variable. T able 11 provides the complete list of patient or individual lung cancer case information that was intended to be utilized in this study. Tumor information or data are collected to describe the cancer case; this descriptive information includes the tumor h istology, tumor type and stage, date of diagnosis, and how the diagnosis was made. The individual lung cancer case data that identifies dates that can be used for an origin or beginning of the study time period and stop or end date are required for surviv al analysis. All variables such as the date of diagnosis, vital status (alive or dead), date of last contact or date of

PAGE 125

108 death, were used to answer the research questions as it pertains to survival in conjunction with gender, treatment type, and/or tumor d escriptor variables. Variable Identification and Coding Patient Identification (ID) Number NAACCR Designated Item Number = 20 The identification number is a unique NAACCR number assigned to a particular individual (patient). The ID number serves sever al functions and purposes as a unique identifier for a particular individual. It is the recommendation of NAACCR, that a previously assigned patient ID number is never reused or reissued if a patient file is deleted. This number will follow the patient a ssigned a unique Patient Identification (ID) Number Different state cancer registries may report tumor information on the same patient to NAACCR or commonly referred to as the central registry. In this instance, the central registry will identify that in dividual, verify any duplicate records, and then assign a unique p atient ID number exclusive to that patient. This number assignment serves to follow the individual patient throughout his/her cancer history regardless of any subsequent tumors that are re ported for the patient. Marital Status at Diagnosis (DX) NAACCR Designated Item Number = 150 diagnosis date is recorded. The martial status can be different depending upon the t umor being reported, as an individual may have different tumor sites. This variable is important as the incidence and survival has been shown to vary by marital status with

PAGE 126

109 particular cancer types 264, 265 The codes used by the central registry for marital status are 1) Single (never married), 2) Married (including common law), 3) Separa ted, 4) Divorced, 5) Widowed, 9) Unknown. Race 1 NAACCR Designated Item Number = 160 The Race coding used by NAACCR is taken from the 2000 US Census Bureau definitions for race, see Table 12 below. Table 12: NAACCR Code and Description of Race Code NAAC CR Designation Code NAACCR Designation 01 White 20 Micronesian, NOS 02 Black 21 Chamorran 03 American Indian, Aleutian, or Eskimo (includes all indigenous populations of the Western hemisphere) 22 Guamanian, NOS 04 Chinese 25 Polynesian, NOS 05 Ja panese 26 Tahitian 06 Filipino 27 Samoan 07 Hawaiian 28 Tongan 08 Korean 30 Melanesian, NOS 09 Asian Indian, Pakistani 31 Fiji Islander 10 Vietnamese 32 New Guinean 11 Laotian 96 Other Asian, including Asian, NOS and Oriental, NOS 12 Hmong 97 Pacif ic Islander, NOS 13 Kampuchean 98 Other 14 Thai 99 Unknown 15 Micronesian, NOS Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Race 2 NAACCR Desi gnated Item Number = 161

PAGE 127

110 When an individual is multiracial, Race 2 through Race 5 is coded and prior to 2000 Race 2 through 5 was blank. The acceptable coding for each race designation is shown in Table 12 under Race 1. Race 3 NAACCR Designated Item Numbe r = 162 Race 4 NAACCR Designated Item Number = 163 Race 5 NAACCR Designated Item Number = 164 Spanish/Hispanic Origin NAACCR Designated Item Number = 190 Spanish or Hispanic origin does not use the same code as Race 1 through Race 5. Origin as defined country of birth of the person the United States. The NAACCR Coding Manual states that p eople who identify their origin as Span ish, Hispanic, or Latino may be of any race 266 This parti cular item variable was used in an attempt by the US Census Bureau to increase the reporting accuracy of the data.

PAGE 128

111 Table 13: NAACCR Code and Description of Spanish/Hispanic Origin Code Description of Spanish/Hispanic Origin 0 Non Spanish; non H ispanic 1 Mexican (includes Chicano) 2 Puerto Rican 3 Cuban 4 South or Central American (except Brazil) 5 Other specified Spanish/Hispanic origin (includes European; excludes Dominican Republic) 6 Spanish, NOS 7 Hispanic, NOS 8 Latino, NOS 9 'unkn own whether Spanish or not' should be used Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Sex NAACCR Designated Item Number = 220 This variable is one of the key variables under study for gender differences research. Under the coding scheme of NAACCR, there are 5 codes for this classification : 1) 1 = Male 2) 2 = Female 3) 3 = Other (Hermaphrodite) 4) 4 = Transsexual and 5) 9 = Not Stated or Unknown. A ge at Diagnosis NAACCR Designated Item Number = 230 scheme is shown in actual years of age, e.g. a 57 year old would be coded as 057. Other examples of age coding are given as 000 for less than 1 year old, 001 for 1 year old, but less than 2 years, 002 represents 2 years of age, 101 for 101 years, 120 for 120 years old, and 999 for an unknown age. Birth Date

PAGE 129

112 NAACCR Designated Item Number = 240 The NAACCR format is given as MMDDCCYY where MM is the month (01 12), DD the day (01 31) and CCYY, the year. The birth date is coded in an 8 character format as either a valid date in the NAACCR format or 99999999 (8 characters) if unknown. The North American Association of Cen tral Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary recommends that if the birth year is unknown, t he birth date can be calculated from the age of diagnosis and the year of diagnosis. The coding for the mont h and the day would be 9999 as unknown but the calculated year would be used. The NAACCR Standards further state an estimated birth date is better than an unknown value. Birthplace NAACCR Designated Item Number = 250 The coding of the birthplace of an in dividual is found in the SEER Program Code Manual Appendix B. This variable is of interest as variations in disease patterns, genetic and socioeconomic characteristics have been demonstrated in the literature varying on place of birth 82, 261, 267 Tobacco History NAACCR Designated Item Number =340 NAACCR does not have a designated code for tobacco use or tob acco history. Coding schemes for tobacco use varies across state cancer registries. Date of Diagnosis NAACCR Designated Item Number = 390

PAGE 130

113 The date of diagnosis is the initial date the primary lung tumor was identified. The coding of the date is done wit Primary Site NAACCR Designated Item Number = 400 The primary tumor site for this particular research is 36, lung. The coding used by NAACCR is designated by the International Classification of Di sease Oncology. Laterality NAACCR Designated Item Number = 410 Laterality is used for paired organs and describes which lung (left or right) has been diagnosed with the primary tumor (Table 14). Table 14: NAACCR Code and Description of Laterality Code Description of Laterality 0 Not a paired site 1 Right: origin of primary 2 Left: origin of primary 3 Only one side involved, right or left origin unspecified 4 Bilateral involvement, lateral origin unknown; stated to be single primary; including both ovaries 9 Paired site, but no information concerning laterality, midline tumor Source: The North American Association of Central Cancer R egistries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Morphology Type and Be havior ICD O 2 NAACCR Designated Item Number = 420 The morphology code is representative of the cell type and the biological activity the tumor presents. The ICD O coding system uses a morphology code based on the histology (cell type), behavior code, an d grade; the codes are given in Table 15. The first

PAGE 131

114 four digits of the morphology code denote the cell histology, the fifth digit is the behavior and the last digit is the grade. The ICD O 2/3 terms include 1) C34.0 for the main bronchus, 2) C34.1 for the u pper lobe of the lung 3) C34.2 for the m iddle lung lobe ( this can be for the right lung only) 4) C34.3 l ower l obe, lung 5) C34.8 an o verlapping lesion of lung and 6) C34.9 Lung, NOS The histologic and behavior codes vary as a function of the lung ca ncer type. Table 15: NAACCR Code and Description of LC Morphology Lung Cancer Type ICD O Morphology Codes Lung Cancer Type ICD O Morphology Codes Small C ell Lung C ancers 80413 80423 80433 80443, 80453 Large C ell C arcinoma 80123 Squamous or E p idermoid 807_3 Adenosquamous C arcinoma 85603 Adenocarcinoma 814_3 Non S mall C ell C arcinoma 80463 Bronchi alveolar 82503 Source : Florida Cancer Data System November 2003 Monthly Memo Histology (92 00) ICD O 2 NAACCR Designated Item Number = 420 Ther e are three parts of the coding scheme for the morphology code and the tumor morphology code. A complete explanation of the history classification scheme was given in Cha pter Two under the Pathology/Histology section. As shown in Table 15, the first four digits are representative of the coding used for lung cancer cases. Each lung cancer

PAGE 132

115 type is classified and coded by standardized methods, as discussed in Chapter Two. Be havior (92 00) ICD O 2 NAACCR Designated Item Number = 430 The behavior of a tumor is the way or the mode of tumor growth or progression within the human body. The physician, normally a pathologist, observes the tumor behavior and classifies the growth p attern. It would be important to select or include Table 16: NAACCR Code and Description of LC Behavior Code Description of Behavior /0 Benign /1 Uncertain whether benign or malignant, borderline malig nancy, low malignant potential, uncertain malignant potential /2 Carcinoma in situ, intraepithelial, non infiltrating, non invasive /3 Malignant, primary site /6 Malignant, metastatic or secondary site /9 Uncertain whether primary or metastatic site S ource: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Morphology Type and Behavior ICD O 3 NAACCR Designated Item Number = 521 Coding for the lung cancer type or morphology essentially did not change from ICD O 2 to ICD O 3; for other tumor types and disease classifications, the ICD O coding did change. As part of the quality assurance procedure, ICD O 2 lung cancer cases will be compared to the ICD O 3 cases, to ensure that the data are consistent. Another method to identify any errors or errors in duplication will be to utilize the NAACCR There should be

PAGE 133

116 consistency between Morphology Type and Behav ior ICD O 2, Morphology Type and Behavior ICD O and Morphology Coding System Original. Histology (92 00) ICD O 3 NAACCR Designated Item Number = 522 ICD O 3 Histology or cell/tumor type designation is the same coding scheme as Histology ICD O 2. Beh avior (92 00) ICD O 3 NAACCR Designated Item Number = 522 The ICD O 3 Behavior or cell growth pattern designation is the same coding scheme as described in Behavior IDC O 2. Grade NAACCR Designated Item Number = 440 The code for grade describes the cel ls of the tumor. Grades I through IV (codes 1 4 as shown in Table 17) are utilized and the other grades listed in that table below are not applicable to this research.

PAGE 134

117 Table 17: NAACCR Code and Description for Grade Code Description of Grade Equivalent Term* 1 Grade I Grade I; grade 1; Well differentiated; Differentiated, NOS 2 Grade II Grade II; grade 2; Moderately differentiated; Moderately well differentiated; Intermediate differentiation; Low grade; Partially well di fferentiated; Relatively well differentiated; Generally well differentiated; Fairly well differentiated; Intermediate differentiation; Grade I of 3 category system; Grade I II; Trabecular 3 Grade III Grade III, grade 3; Poorly differentiated; Ded ifferentiated; Medium grade; Moderately undifferentiated; Relatively undifferentiated; Relatively poorly differentiated; Grade II of 3 category system; Grade II III 4 Grade IV Grade IV; grade 4; Undifferentiated; Anaplastic; High grade; Grade III of 3 c ategory system; Grade III III 5 T cell 6 B cell 7 Null cell 8 NK (natural killer) cell 9 Grade/differentiation unknown, not stated, or not applicable Cell type not determined, not stated or not applicable; No grade/differentiation in the primar y site even if a grade is given for a metastatic site. Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 *Source: Florida Cancer Data System Data Acquisit ion Manual 2006 Morphology Coding System Original NAACCR Designated Item Number = 480 The morphology coding system originally used will be utilized as a second check to ensure data quality in the reporting and data received from the cancer registries.

PAGE 135

118 Diagnostic Confirmation NAACCR Designated Item Number = 490 Diagnostic confirmation is essential for verification of a particular tumor type. This confirmation can be made utilizing various methods such as an examination by a pathologist or cytologis t of tissue s or cells via the microscope. For this research, certain criteria for the variables of interest must be used when evaluating a particular tumor type, in particular, in the study of lung cancer. One criterion to know is the specific scientific method used to diagnosis the tumor type. Biological confirmation of the cancer type is the gold standard in the medical community. In the NAACCR coding scheme, other diagnostic confirmation methods other than biologic confirmation such as direct visuali zation of the tumor are included and are shown in Table 18. The codes for the reporting method for confirming a particular tumor type in this research will only include biologically confirmed methods: 1) p ositive histology 2) p ositive cytology, no positi ve histology 4) p ositive microscopic confirmation, method not specified and 5) Positive laboratory test/marker study as shown in Table 18.

PAGE 136

119 Table 18: NAACCR Code and Description Diagnostic Confirmation Code Description of Diagnostic Confirm ation 1 Positive histology 2 Positive cytology, no positive histology 4 Positive microscopic confirmation, method not specified 5 Positive laboratory test/marker study 6 Direct visualization without microscopic confirmation 7 Radiography and other i maging techniques without microscopic confirmation 8 Clinical diagnosis only (other than 5, 6, or 7) 9 Unknown whether or not microscopically confirmed Source: The North American Association of Central Cancer Registries Standards for Cancer Registries V olume II Data Standards and Data Dictionary 2007 Type of Reporting Source NAACCR Designated Item Number = 500 The tumor information is contained in many different records at different facilities. As the data are abstracted or collected in a standardize d manner, it is necessary to identify where the information was obtained from. As an example, information from laboratory reports identified from the medical record in a medical oncology center would have the reporting center coded as 2 see Table 19. I t is well documented in the literature 57, 229 that death certificates many have incomplete information and may not It would be impor tant to the investigator to be aware that reporting s ource with a code of 7 as to address any discrepancies during the analysis.

PAGE 137

120 Table 19: NAACCR Code and Description for Reporting Source Type Code Description of Type of Reporting Source 1 Hospital inpatient; Managed health plans with comprehensive, unified medical records 2 Radiation Treatment Centers or Medical Oncology Centers (hospital affiliated or independent) 3 Laboratory only (hospital affiliated or independent) 4 Physician's office/pr ivate medical practitioner (LMD) 5 Nursing/convalescent home/hospice 6 Autopsy only 7 Death certificate only 8 Other hospital outpatient units/surgery centers Source: The North American Association of Central Cancer Registries Standards for Cance r Registries Volume II Data Standards and Data Dictionary 2007 Sequence Number NAACCR Designated Item Number = 560 This code is used by the cancer registry to identify primary, secondary, or multiple lung tumors. As this research is focused on primary l ung cancer cases, this designation is 00 and codes other than the 00 will be identified. Date of Admission or First (Adm/1 st ) Contact NAACCR Designated Item Number = 580 This variable designates the date the first time a case was contacted either as an ou tpatient or inpatient. variable, Birth Date (240). The date may be representative of an outpatient procedure, an x ray, or pathology report associated with the diagnosis of the tumor. Clas s of Case NAACCR Designated Item Number = 610 Class of case describes the location of the reporting facility where the diagnosis

PAGE 138

121 was made and the codes for Class of Case are described in Table 20. Analytic cases are those that are diagnosed at the report ing facility and include the codes 0 through 2. The other codes include nonanalytic cases that are identified at the reporting facility but were diagnosed and treated at a different facility. These cases also include those diagnosed at autopsy. Table 20 : NAACCR Code and Description for Class of Case Codes Description of Class of Case Analytic Cases 0 Diagnosis at the reporting facility and the entire first course of treatment was performed elsewhere or the decision not to treat was made at another fa cility. 1 Diagnosis at the reporting facility, and all or part of the first course of treatment was performed at the reporting facility. 2 Diagnosis elsewhere, and all or part of the first course of treatment was performed at the reporting facility. Non analytic C ases 3 Diagnosis and the entire first course of treatment were performed elsewhere. Presents at your facility with recurrence or persistent disease. 4 Diagnosis and/or first course of treatment were performed at the reporting facility prior to the reference date of the registry. 5 Diagnosed at autopsy 6 Diagnosis and the entire first course of treatment were completed by the staff with admitting privileges at the repo rting facility. 7 Pathology report only. Patient does not enter the reporting facility at any time for diagnosis or treatment. This category excludes tumors diagnosed at autopsy. 8 Diagnosis was established by death certificate only. Used by central regi stries only. 9 Unknown. Sufficient detail for determining Class of Case is not stated in patient record. Used by central registries only. Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data S tandards and Data Dictionary 2007

PAGE 139

122 Primary Payor at Diagnosis NAACCR Designated Item Number = 630 Information about the insurance carrier at the time of diagnosis can be an important variable. Possible treatment disparities between minority groups with lung cancer and the association between the types of insurance coverage could be of interest. Table 21 lists the codes that are associated with the particular payor at the time of lung cancer diagnosis. Table 21: NAACCR Code and Description for Payor at Diagnosis Code Description of Primary Payor at Diagnosis 01 Not insured 02 Not insured, self pay 10 Insurance, NOS 20 Private Insurance: Managed care, HMO, or PPO 21 Private Insurance: Fee for Service 31 Medicaid 35 Medicaid Administered through a Managed Care plan 60 Medicare/Medicare, NOS 61 Medicare with supplement, NOS 62 Medicare Administered through a Managed Care plan 63 Medicare with private supplement 64 Medicare with Medicaid eligibility 65 TRICARE 66 Military 67 Veterans Affa irs 68 Indian/Public Health Service 99 Insurance status unknown Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007

PAGE 140

123 SEER Summary Stage 1977 NAACCR Des ignated Item Number = 760 The coding scheme for SEER Summary Stage 1977 (Table 22) is used at the time of initial primary tumor diagnosis. Stage was a variable that was historically monitored for time trends. Table 22: NAACCR Code and Description SEER Summary Stage 1977 Codes Description of SEER Summary Stage 1977 0 In situ 1 Localized 2 Regional, direct extension only 3 Regional, regional lymph nodes only 4 Regional, direct extension and regional lymph nodes 5 Regional, NOS 7 Distant 8 Not app licable 9 Unstaged Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 SEER Summary Stage 2000 NAACCR Designated Item Number = 2000 SEER Summary Stage 2000 at initial diagnosis is a variable that includes the description of the reportable tumor. Table 23 exhibits the site specific single digit coding scheme explicit to the tumor location.

PAGE 141

124 Table 23: NAACCR Code and Description SEER Summary Stage 2000 Codes Description of SEER Summary Stage 2000 0 In situ 1 Localized 2 Regional, direct extension only 3 Regional, regional lymph nodes only 4 Regional, direct extension and regional lymph nodes 5 Regional, NOS 7 Distant 8 Not applicable 9 Unstage d Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Record (RX) Date of First Surgery NAACCR Designated Item Number = 1200 The date of the first surgery for the primary tumor is coded with the NAACCR format given as MMDDCCYY where MM is the month (01 12), DD the day (01 31) and CCYY, the year. The surgical date is coded in an 8 character format as either a valid date in the NAACCR format or 99999999 (8 characters) if it is unknown if any surgical procedure was performed. If there was no surgical procedure performed or if the individual was an autopsy only case, the code would be 00000000 Record (RX) Date of First Radiation NAACCR Designated Item N umber = 1210 This is the date that the treatment modality, radiation therapy began at any their treatment. The coding for this variable is the NAACCR format M MDDCCYY where MM is the month (01 12), DD the day (01 31) and CCYY is the year. Other

PAGE 142

125 variations include 00000000 when radiation therapy is not administered; autopsy only case 88888888 if radiation therapy was scheduled as part of the first course o f therapy, but was not started at the time and 99999999 if the date was unknown it was unknown whether any radiation therapy was administered; or if the case was only identified by death certificate Record (RX) Date of First Chemotherapy NAACCR Designate d Item Number = 1220 This designation is the date that chemotherapy was first started. The format used The other codes admissible include 00000000 when chemotherapy is not administered or in the case of an autops y Record (RX) Summary of Surgery for Primary Site NAACCR Designated Item Number = 1290 The summary of the surgery performed for the primary tumor site is given below in Table 24. As the disease of interest for this research is lung cancer, all surgical sites will be specific to regions of the lung. Table 24: NAACCR Code and Description of Surgical Primary Site Code Description of Record (RX) Summary of Surgery for Primary Site 00 None 10 19 Site specific code; tumor destruction 20 80 Site specific codes; resection 90 Surgery, NOS 98 Site specific codes; special 99 Unknown Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007

PAGE 143

126 Record (RX) Summary of R adiation NAACCR Designated Item Number = 1360 explanation the type of radiation treatment the lung cancer case received. Table 25: NAACCR Code and Description of Radiation Treatme nt Code Description of Radiation Treatment 0 None 1 Beam radiation 2 Radioactive implants 3 Radioisotopes 4 Combination of 1 with 2 or 3 5 R adiation, NOS method or source not specified 6 Currently allowable for historic cases only; see Note below 7 8 Radiation recommended, unknown if administered* 9 Unknown if radiation administered Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards a nd Data Dictionary 2007 : In the SEER program, a code 2 for other radiation was used between 1973 and 1987. When the radiation codes were expanded to add codes '2' radioactive implants and '3' radioisotopes, all cases with a code '2' and diagnosed in 1973 1987 were converted to a Record (RX) Summary of Chemotherapy NAACCR Designated Item Number = 1390 The chemotherapy codes used for the NAACCR Designated (Item Number = 1390) Record Summary of Chemother apy are listed in Table 26. The code is specified when a chemotherapy agent/drug is received or not administered to an individual case as part of the first treatment for lung cancer. Also a code is given to identify when it is unknown if the lung cancer case received chemotherapy, i.e. codes 88 and 99.

PAGE 144

127 Table 26: NAACCR Code and Description for Chemotherapy Code Description of Chemotherapy Treatment 00 None, chemotherapy was not part of the planned first course of therapy. 01 Chemotherapy, NOS 02 Chem otherapy, single agent. 03 Chemotherapy, multiple agents. 82 C hemotherapy was not recommended nor administered because it was contraindicated due to patient risk factors, i.e., comorbid conditions, advanced age 85 Chemotherapy was not administered beca use the patient died prior to planned or recommended therapy. 86 physician, but was not administered as part of first course therapy. No reason was stated in the patient record. 87 Ch 88 Chemotherapy was recomme nded, but it is unknown if it was administered. 99 It is unknown whether a chemotherapeutic agent(s) was recommended or administered because it is not stated in patient record; death certificate only cases. Source: The North American Association of Centr al Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Derived AJCC Stage Group NAACCR Designated Item Number =3000 This variable Derived AJCC Stage Group, encompasses all stage designations from the AJC C Sixth Edition TNM stage, SEER Summary Stage 1977, and SEER Summary Stage 2000 and complies the different coding into this one item number, 3000. The coding designation, shown in Table 27 came into effect as a result of a joint task force so a common, un iform set of rules and coding will be available. Representatives from SEER, ACoS, CDC, NAACCR, NCRA, and AJCC collaborated on the coding designation to standardize the grouping of disease stage.

PAGE 145

128 Table 27 : Derived AJCC Stage Group AJCC Code Display String Comments 00 0 Stage 0 01 0a Stage 0a 02 0is Stage 0is 10 I Stage I 11 INOS Stage I NOS 12 IA Stage IA 13 IA1 Stage IA1 14 IA2 Stage IA2 15 IB Stage IB 16 IB1 Stage IB1 17 IB2 Stage IB2 18 IC Stage IC 19 IS Stage IS 23 ISA Stage ISA (lymphom a only) 24 ISB Stage ISB (lymphoma only) 20 IEA Stage IEA (lymphoma only) 21 IEB Stage IEB (lymphoma only) 22 IE Stage IE (lymphoma only) 30 II Stage II 31 IINOS Stage II NOS 32 IIA Stage IIA 33 IIB Stage IIB 34 IIC Stage IIC 35 IIEA Stage IIEA ( lymphoma only) 36 IIEB Stage IIEB (lymphoma only) 37 IIE Stage IIE (lymphoma only) 38 IISA Stage IISA (lymphoma only) 39 IISB Stage IISB (lymphoma only) 40 IIS Stage IIS (lymphoma only) 41 IIESA Stage IIESA (lymphoma only) 42 IIESB Stage IIESB (lymp homa only) 43 IIES Stage IIES (lymphoma only) 50 III Stage III 51 IIINOS Stage III NOS 52 IIIA Stage IIIA 53 IIIB Stage IIIB 54 IIIC Stage IIIC 55 IIIEA Stage IIIEA (lymphoma only)

PAGE 146

129 56 IIIEB Stage IIIEB (lymphoma only) 57 IIIE Stage IIIE (lymphoma only) AJCC Code Display String Comments 58 IIISA Stage IIISA (lymphoma only) 59 IIISB Stage IIISB (lymphoma only) 60 IIIS Stage IIIS (lymphoma only) 61 IIIESA Stage IIIESA (lymphoma only) 62 IIIESB Stage IIIESB (lymphoma only) 63 IIIES Stage IIIES ( lymphoma only) 70 IV Stage IV 71 IVNOS Stage IV NOS 72 IVA Stage IVA 73 IVB Stage IVB 74 IVC Stage IVC 88 NA Not applicable 90 OCCULT Stage Occult 99 UNK Stage Unknown Source: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Date Case Report Received NAACCR Designated Item Number = 2111 and received by the central cancer re gistry 258 In the event of multiple reports on the same individual and one date is needed, the protocol is to use the first date the record was received. This variable can be used to evaluate t he reporting timeliness of the cancer registries. This variable may also be used to measure how long the individual cancer registry takes to submit the data when the date the report (2111) is received is compared to the date of first contact (580).

PAGE 147

130 Dat e of Last Contact NAACCR Designated Item Number =1750 The D ate of L ast C ontact is the last date of known contact but also can represent the date of death. The date is obtained from either an active or a passive follow up. The sources include the state r egistries date of last contact or death (passive) or from an active state SEER registry or the National Death Index. The date coding follows the NAACCR format of MMDDCCYY, where MM is the month (01 12), DD the day (01 31) and CCYY is the year. The ma in purpose of this variable is to record the date of last contact or date of death. Vital Status NAACCR Designated Item Number =1760 The vital status of the individual as given by the NAACCR is 0 for dead, 1 for alive, 4 for dead. A code of 0 is obtained from states that report based on the guidelines of the Commission on Cancer (passive registry) and a code of 4 is based on information obtained from a SEER state (active registry). Follow Up Source NAACCR Designated Item Number =1790 The source is given for the most currently recorded information for an individual Any discrepancies in the record can be reviewed and cross checked with other variable coding. Table 28 lists the sources or the contributors to the follow up data. It includes information re ported from the Department of Motor Vehicles, death certificate information, patient or physician reporting, Medicare/Medicaid files, and if the data is unknown, not stated in the patient record.

PAGE 148

131 Table 28: NAACCR Code and Description of Follow Up Source s Code Description of Follow Up Source 0 Reported hospitalization 1 Readmission 2 Physician 3 Patient 4 Department of Motor Vehicles 5 Medicare/Medicaid file 7 Death certificate 8 Other 9 Unknown, not stated in patient record Source: The North Ame rican Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Autopsy NAACCR Designated Item Number =1930 This designation is the indicator if an autopsy was performed or not. This info rmation could be use to verify the correctness of the coding for other variables. For example, the coding for this NAACCR designated item number, 1930, could be compared to a patient status code for a particular individual to check for agreement. A case could not be alive if an autopsy was performed. Table 29: NAACCR Code and Description of Autopsy Code Description of Autopsy 0 Not applicable; patient alive 1 Autopsy performed 2 No autopsy performed 9 Patient expired, unknown if autopsy performed S ource: The North American Association of Central Cancer Registries Standards for Cancer Registries Volume II Data Standards and Data Dictionary 2007 Place of Death NAACCR Designated Item Number = 1940

PAGE 149

132 This variable is useful to correlate to the date of la st contact. If the patient is coded as alive and a place of death is documented, further investigation is warranted. Date Case Completed NAACCR Designated Item Number = 2090 The date can be used to assess the quality and timeliness of reporting for the d ata. Date Case Exported NAACCR Designated Item Number = 2110 The date can be used to assess the quality and timeliness of reporting for the data. Derived AJCC Stage Summary NAACCR Designated Item Number = 3000 This variable is used to compare stage inf ormation. Epidemiologic Research Design The epidemiologic study design for this research was based on a historical cohort of primary lung cancer case. Female lung cancer cases were compared to male lung cancer cases and this comparison between genders i nclude d the histological type, stage, and grade of lung cancer and the treatment received (chemotherapy, radiation therapy, surgery, or combination) as variables of interest. Some of the strengths of this study design are multiple effects of the exposure w ere assessed simultaneously. Historically, a weakness of a retrospective or historical study design is it can be prone to bias due to recall or information bias. This particular limitation or weakness was minimized as the data were collected by a standar dized, controlled method utilizing trained cancer registry abstractors ; information was provided by medical records and non analytic cases were

PAGE 150

133 excluded, e.g. information provided by the patient or members of the family As the majority of the states are mandated by law to report cancer case information; non compliance is minimal and monitored by the NAACCR; therefore the case information is assumed to be complete. Data Collection Methods Prior to the collection of any lung cancer case information, approv al from the University of South Florida Internal Review Board Data (IRB) was sought by the principle investigator, PI. The chief concern of the IRB was that the case information could not be used to identify any one particular individual. The research da ta acquired for primary lung cancer cases for this study are cases from state cancer registries that are members of NAACCR. Initially, t here were two main approaches to collecting the lung cancer case information. The first method was to contact the eigh t state s selected (Table 10). As stated previously, West Virginia was randomly selected but the cancer registry was unable to comply with the data research request. Table 30 outlines the contact process followed for the cancer registri es and the contact i nformation. The data set acquisitions was done in accordance to protocol and procedures outlined by each state based cancer registry.

PAGE 151

134 Table 30: State Cancer Registry Contact Information State Contact Information Requirements for the Release of Data Oregon Catherine Riddell (great resource) Research Analyst Oregon State Cancer Registry Phone: (971) 673 1113 FAX: (971) 673 0996 catherine.a.riddell@state.or.us Requested Lung Cancer Data and sent the following three files: 1. The PDF file of the approval letter from the IRB at the University of South Florida concerning my dissertation research. 2. The variables of interests are outlined in "NAACCR VARIABLE" (the data will be utilized as part of my dissertation/research project). 3. The eight state cancer registries selected to participate in my dissertation research project that includes Oregon. Idaho Cancer Data Registry of Idaho 615 N. 7th Street P.O. Box 1278 Boise, Idaho 83701 http://www.idcancer.org/generalinfo.html The Cancer Data Registry of Idaho has a release requirement and form that must be submitted prior to the release of any data. Florida Florida Cancer Data System ht tp://fcds.med.miami.edu/ The FCDS has a release requirement and form that must be submitted prior to the release of any data. The data request forms are located on the FCDS website at: http://fcds.med.miami.edu/inc/datarequest.shtml South Carolina S.C. Department of Health & Environmental Control S.C. Central Cancer Registry 810 Dutch Square Blvd., Ste. 220 Columbia, SC 2921 0 The South Carolina Central Cancer Registry has a release requirement and form that must be submitted prior to the releas e of any data. Telephone # (803) 731 1419 Fax # (803) 731 1455 Indiana Indiana State Department of Health Epidemiology Resource Center 2 North Meridian, 5K Indianapolis, IN 46204 317 233 7807 317 234 2812 FAX The Indiana State Department of Health has a release requirement and form that must be submitted prior to the release of any data. Massachusetts Massachusetts Cancer Registry 250 Washington Street, 6th Floor Boston, MA 02108 Phone: (617) 624 5642 Fax: (617) 624 5695 Annie McMillan ( great resource and extremely helpful) The Massachusetts Cancer Registry has a release requirement and form that must be submitted prior to the release of any data. Confidential Data Officer Privacy and Data Access Office Massachusetts Department of Public Health 250 Washington Street, 2nd Floor Boston, MA 02108 4619 TEL: (617) 624 5229 FAX: (617) 624 5234 Nebraska Nebraska Cancer Registry Nebraska Comprehensive Cancer Control Program Janis Singleton (very nice) IRB approval process DHHS, Division of P ublic Health 301 Centennial Mall South Lincoln NE 68509 Rhode Island 3 Capitol Hill Providence, RI 02908 (401)222 1172 Fax: 222 3551 http://www.health.ri.gov/disease/cancer/regis try.p hp The Rhode Island Cancer Registry has a release requirement and form that must be submitted prior to the release of any data. Main Contact (extremely helpful) : John P. Fulton, PhD, RI Department of Health: 401 277 1394 x115

PAGE 152

135 The other or second method of data case acquisition was attempted through the central cancer registry NAACCR, for the states of interest (Table 9). To answer the research questions, data from the eight states was required. It was considered a viable option that collecting the da ta from a centralized data bank, like NAACCR, would streamline the data collection process. The second option was pursued via multiple data requests made directly to NAACCR. The complete approval / disapproval process for data release took over a year and after multiple requests and multiple re submissions for the lung cancer data NAACCR determined they would provide the information to the researcher. Within 1 month of receiving the NAACCR approval letter, the investigator was again contacted by NAACCR an d was told more approvals by another NAACCR committee were required and the approval was withdrawn Ultimately, after several more months, a letter was received by the investigator and was told by NAACCR the data would not be provide d to the researcher ; m ultiple reasons were given During this process of awaiting the second NAACCR approval, the investigator was contacted by telephone by one of the NAACCR committee members ; that particular committee member said that the dissertation research did not serve any scientific merit NAACCR can be contacted via the URL http://www.naaccr.org/ Any other future attempts to use the NAACCR data base were deemed unproductive therefore method one of contacting the state s direct ly was utilized for the primary lung cancer case data collection When the data sets were acquired from each state cancer registry the data set was assessed for completeness of information i.e. that the variables of interest were included

PAGE 153

136 Also the cate gories for the variables of interest were evaluated for coding as the format would have to be similar as outlined by the standard NAACCR protocols. The comparison of results from state based registries for lung cancer cases w ould not be possible without t his standardization. When each state cancer r egistry was contacted, e ach state representative was particularly interested in how the data are used for the specified or requested research project and that patient confidentiality would not be compromised Additionally, it was requested that the database not shared with anyone other than the researchers identified on the data request form (the researchers must attest to this ; in some instances the state data request form had to be notarized), and any confide ntial patient information inadvertently discovered must be kept confidential. Statistical Procedures Prior to any statistical procedures, complete assessment of the study design methods was completed. The methods outlined in the selection criteria for c ases identification and state cancer registry selections were based on epidemiological principles so ultimately valid assessments could be made after utilizing the most appropriate statistical procedures. In other words, data sets that contain inherent fl aws due to bias would never result in invalid conclusions regardless of the statistical methods applied. For example, r andomization during the selection of the regional state based c ancer registry was part of the initial selection process. Any bias that may have been introduced by the selection of a state in theory was minimized by that randomization

PAGE 154

137 process. In the previous sections, the importance of the data selection or variables of interest (inclusion and exclusion criteria) was discussed. Collectin g data from cancer registries that utilize a standardized format 258 was critical. Data not collected in such a manner could be subject to bias resulting in erroneous results. Each lung cancer c ase acquired from each state cancer registry was assessed for completeness of the data. In other words, all the variables of interest should be included in the case information and the categories for the variables of interest should be coded in a similar manner as outlined by the standard NAACCR protocols. The comparison of results from state based registries for lung cancer cases would not be possible without the standardization of the variables of interest. The selection of the most appropriate statisti cal model that best represented and accounted for the behavior of the data was critical in the evaluation of the three research questions. The statistical procedures applicable to each research question are discussed and reviewed in the following section; these procedures are utilized so that the research questions were answered appropriately. Each of the eight state based primary l ung cancer case data was concatenated into one data set ; this data set was used to answer the research questions. The lung c ancer case information from each state is representative of the lung cancer cases that state (Idaho, Oregon, Florida, South Carolina, Indiana, Massachusetts, Nebraska, and Rhode Island) as each case within the state cancer registry has as likely a chance t o be included in the state registry as another The ability to identify which state the individual lung cancer case originated from in the merged data

PAGE 155

138 set will be done by means of a generated variable (code). This coding enable d the researcher to not onl y look at the aggregate population statistics but also by state and region One of the first statistical procedures performed on the data sets was exploratory analysis. These descriptive procedure s enabled the investigator to identify any differences, as well as the similarities of the patient population under study. This was accomplished by the PROC FREQ statistical procedure in SAS and univariate analysis by investigating the resultant means of the continuous variable age ; results of the testing are gi ven in Chapter Four For example, it was useful to examine the difference between men and women, age of diagnosis, and state/region. Several studies recently published, suggest that the age at diagnosis of lung cancer is less for women versus men 40 15 and the opportunity to compare the results of this data set analysis served to increase the validity of this research. Any differences in the lung cancer case population were de termin ed for the number of men and women in each particular variable category such as morphology (histology and behavior) group, treatment group, or the age at tumor diagnosis categories. The results of the SAS PROC FREQ statistical procedure were examined for the other research variables of stage grade, marital status at the time of diagnosis, race, vital status and state. The statistical procedure s were used to count the fre quencies of the data variables and calculate d p ercentages In summary, all categorical variables are displayed in Chapter Four tables and were also classifi ed according to gender (male and female). The continuous variable of age at diagnosis was evaluated by the SAS PROC

PAGE 156

13 9 UNIVARIATE procedure and described in terms of the mean, medians and ranges compar ing to m ales and females Age was categorized into five age groups and those categorical classifications were used in the subsequent analysis to answer the three research questions. Study Question One Null Hypothesis: Females receive the same treatment as men regardless of the histological type, stage and gr ade of lung cancer. The outcome variable is treatment, the exposure is gender (a main effect) and the variables classified as other main effect variables include stage, grade, and histological type; the demographic independent variables included age grou p at diagnosis, marital status at diagnosis and race. The statistical model used to examine the relationship between the dependent and independent variable, gender will be the multinomial logistic regression model (MLRM) The multinomial logistic regres sion model facilitates the examination of the categorical outcome variable (Treatment Group) an d the relationship between the independent variables. In this research the outcome variable is multinomial or polychotomous and is coded on a nominal level. Each treatment type or treatment group (txgrp) or combination of treatments is categorized on a nominal scale meaning the levels (scale) do not represent a better or worse category. These nominal outcome levels and the independent variables were modeled o r fitted to a multinomial logistic regression model. There are eight treatment types (outcome variable levels) 1) radiation, 2) surgery, 3) chemotherapy, 4) radiation combined with surgery, 5) radiation combined with

PAGE 157

140 chemotherapy, 6) surgery combined wi th chemotherapy, 7) radiation combined with chemotherapy and surgery, and 8) no treatment received that were analyzed to answer question one. A representative equation for the full model is listed below and is defined as: Pr ( y i =1) = G ( 0 + 1x i1 + 2x i2 k xik ) Where i s the cumulative distribution functi on for a logistic variable and upon transformation is referred to a s the logit model. This model makes the assumption the chance ( odds ) of an outcome given a response level ( in this case, a particular treatment modality received) are constant regardless of which level (treatment type) selected. For n ominal outcome logistic models with k + 1 possible l evels for the outcome variable the logistic model can be extended to a mul tinomial model called a generalized or baseline category logit model, and is shown below: l n (Pr(Y = i| x ) ) / ( Pr(Y = k +1| x )) = i + i x i k Where the 1, k are the k intercept parameters and the k are the k vectors of th e slope parameters; these models are designated as a special case of the discrete choice or conditional logit model. The coefficients resulting from multinomial logistic regression upon exponentiation are commonly referred to as an Odds Ratio. As gender wa s one of the main interests of this research, the effect that gender had

PAGE 158

141 on the selection of the treatment received classifie d gender as the primary variable for this particular research question (Question I) Each covariate, gender, stage, grade, morphology, age gro up, race, and marital status had different coding or levels within the this chapter. The statistical results including the Odds Ratios and 95% Confide nce Intervals generated by the MLRM included the main effects, interaction effects, and the overall gender effect on the outcome, i.e. treatment type received, are included in the overall assessment to answer Question One. In summary, the multin omial logi stic regression model utilizing categorical variable s was used as the statistical model to test Hypothesis I in order to answer question one. Additionally, a random effect model utilizing the SAS PROC GLIMMIX procedure with a link function to the genera lized logit model was included to evaluate any effect state had the outcome as compa red to another state. The identification of each state in the data set was important with respect to the study of any random effect s introduced by a state on the relations hip between the outcome (lung cancer treatment) and the independent variables. A random effect model was useful in the identification of one state that behaved differently (in the statistical sense) or having variability as compared to another state. As Florida is known as a retirement state population has an overrepresentation of individuals with a greater probability of cancer incidence. Statistics based on the treatment of Florida lung cancer female and male cases when compared to

PAGE 159

142 another state such as Nebraska (different population base) may lead to variability in the relationship of lung cancer treatment and the independent variables dependent upon state. Study Question Two The statistical procedures to address the second study question ( I s there a statistically significant difference in survival in women with lung cancer as compared to men with lung cancer regardless of the treatment modality received?) i nclude d t he Kaplan Meier and the Life Table method s for overall lung cancer survival analysis between men and women The log rank statistical test was utilized to test for survival differences between women and men by examining for any statistical signifi cance. Survival was defined in this study as the time (in months ) from the diagnosis of lung cancer to death or to the date of last contact when the individual was reported as alive a cutoff date of 12 31 2004 was used to censor individuals that had a d ate of last contact greater than 12 31 2004 Study Question Three omen with the same histological type, stage/grade of lung cancer, and the same treatment modality differ significantly in survival as compared to men w ith the same histological type, stage/grade of lung cancer, utilized the Cox Proportional Hazards model The Cox Proportional Hazards model estimated the relative risk or hazard ratio for death for women as compared to men This model was used to address gender differences in overall survival while ad justing for the primary main effect s, demographic main effects and interaction term

PAGE 160

143 moderation The effect of gender on survival was examined by determining the estimated rel ative risk (hazards ratio) of death for women as compared to men by adjusting by stage, histology, grade, treatment type race, marital status, and age group as well as interaction terms with the a djusted Cox Proportional Hazards model The proportionalit y assumption of for each variable was tested by evaluating the graphs of the survival function and noting that the distance between the levels or strata of a variable did not change (increase or decrease) over time or cr oss. In each case, the proportionality assumption held with the exception of the variable, treatment groups. In this case there was crossover between two of the treatment groups suggestive of limitations in the analysis. Residual analysis was completed f or the final model, there were some trends demonstrated in the Martingale Residuals over time but the majority of the residuals were varying about zero as expected d emonstrating no trends. Preliminary Statistical Analysis Initially, a study was conducte d by the investigator to determine if there were statistically significant differences between females versus males and the treatment received prior to the development of this dissertation. The data set was drawn from the Florida Cancer Data System 1 in which Commercial File 4505 had 2,393,853 cancer cases from the years 1981 through 2003 The lung cancer cases (n = 13 9 926) were categorized by the International Classification of Diseases Oncology (ICD O) and included the four major histological lung cancer types: adenocarcinoma, squamous cell

PAGE 161

144 carcinoma, large cell carcinoma, and small cell carcinoma. Other FCDS gender categories : 3 = Other (Hermaphrodite) 4) 4 = Transsexual and 5) 9 = No t Stated or Unknown, were exclude d from the analysis as those particular gender categories did not contribute to the research question The major treatment modalities (chemotherapy, surgery, hormone use, and radiation therapy) for FCDS lung cancer cases w ere included; other treatment modalities were excluded as this research focused on the major treatment modalities used to treat cancer. The major races/ethnic groups were selected based on the overall FCD S statistics; white and African American were sele cted as the two racial/ethnic groups. Inclusion criteria for smoking status consisted of never smoking, past history of smoking, and presently smoking as referenced to the date of lung cancer diagnosis. The mean age for FCDS males (n = 88 248) was 68.96 years of age and for FCDS females (n = 51,678) 68.66 years of age was calculated with the SAS PROC UNIVARIATE program Variable f requencies classified by gender were determined wi th the SAS PROC FREQ procedure. The majority of the FCDS women were marrie d (14.2%), had a history of smoking (90.43%), and were white (n = 49,227 (95.26%)). Adjusted Odds Ratios were derived from the logistic regression model utilizing SAS Institute Inc., Cary, NC, USA Version 9.1 software. Based on a the statistics generate d by a logistic regression model, t here were statistically significant differences between gender and the treatment modality after adjusting for race, age and tobacco use. FCDS females had a decrease odds of receiving radiation therapy (OR = 0.939 ( 95%CI = 0.919, 0.961)) and surgery (OR = 0.940 ( 95%CI = 0.915, 0.966)) as compared to FCDS

PAGE 162

145 men. Additionally, there was decrease in the odds of having radiation therapy as a treatment modality for lung cancer for white FCDS females as compared to African America n females (OR = 0.806 ( 95%CI = 0.771, 0.841)). FCDS African American females had a decrease odds of receiving surgery as a treatment modality (OR = 0.605 ( 95%CI = 0.569, 0.643)) and chemotherapy (OR = 0.790 ( 95%CI = 0.751, 0.831)) for lung cancer as compa red FCDS white women. FCDS females had a greater probability or risk of adenocarcinoma and small cell carcinoma as compared to FCDS males Some of the limitation s of initial FCDS study were that the treatment groups were not stratified to examine a combination of receiving more than one treatment type nor were interaction terms considered in the relationship between treatment and gender. Summary Initially, this research was based on a preliminary investigation of primary lung cancer cases for the Florida Cancer Data System 1 that studied if the treatment modality selected to treat a lung cancer case was based on gender. This research expanded the concept of lung cancer treatme nt differences based on gender to include the determin ation of survival differences in women as compared to men dependent upon the treatment modality received The initial study objective was to investigate differences in major treatment modalities by gen der, all four major histological lung cancer types combined by gender, and the major histological lung cancer types by gender. The relationship between treatment modalities and other variables (race, smoking status, vital

PAGE 163

146 status, and marital status) was evaluated. In particular, race was investigated to determine if any disparities between race and treatment type existed for FCDS women. This research expands on the preliminary findings and the patient population by including other NAACCR associated can cer registry lung cancer data in an attempt to determine if t he particular treatment modality used to treat a woman with lung cancer affect s her survival as compared to a man As statistically significant differences in the association between gender and treatment have been demonstrated previously in the preliminary findings of this research it was important to address these findings durin g the next phase of gender differences in lung cancer survival research. This study is a first step in the determinat ion of survival in women with lung cancer and differences in treatment patterns as compared to men utilizing the data from state registries that are members of NAACCR. Results from this newly combined database will be in an attempt to quantify the extent of a gender specific treatment effect and the impact of this effect on survival. Another novel statistical approach in the study of gender differences in treatment selection and gender specific survival is the addition of interaction terms in the analysis Also with the inclusion of interaction terms the calculation of an overall gender effect of the treatment outcome and on survival could be possible. In the literature reviewed and cited throughout this dissertation, this approach has not been demonstrat ed. This approach adds another dimension in the study of gender differences for lung cancer treatments and survival as statistically significant results were demonstrated.

PAGE 164

147 CHAPTER IV: PRESENTATION AND ANALYSIS OF DATA Introduction This chapter present s the study findings. The study population consisted of lung cancer cases drawn from state based passive cancer registries in the United States. The lung cancer cases that were selected from each state cancer registry were intended to be r epresentative of all the lung cancer cases for that particular state For each state, the lung cancer case had equally as likely a chance of being included or excluded from the cancer registry. The study individuals were selected from state cancer registry lung cancer ca ses diagnosed during a five year time period, 1 1 2000 through 12 31 2004. The time or date of diagnosis served a dual purpose as that date was also used to specify the origin or start date for subsequent Survival Analysis. As previously stated in Chapte r Three, the eight state cancer registries with the lung cancer case s were randomly selected from NAACCR US state cancer registries in four geographic regions. The reason for selecting cancer registries from four different geographic regions in the United States was reduce or eliminate any biases, e.g. selection, treatment, that may have been introduced by selecting cases from only one geographic region. The overall intent was to account for any differences in the population characteristics. Forty six va riables for each lung cancer case were requested from the eight NAACCR cancer registries. Each state reviewed the requested information and provided data that was consistent with their Internal Review Board (IRB) protocol, policies, and procedures. W hile some of the individual states did not provide information on all 46

PAGE 165

148 variables requested, the data provided by each state, did allow for a complete assessment so that the three research questions proposed in this study could be answered Many of the study variables requested were intended to be utilized in a quality of data assessment. For example, evaluating the number of autopsies reported and comparing that frequency with Vital Status (alive versus dead) could be used to check the integrity of the data. As some data were either incomplete or unavailable to the researcher, a quality assessment or test could not be completed. Additionally, in the original request for specific variables, a number of states would not provide the variable information that t hey (the state registry) determined could possibly compromise the confidentially of a particular lung cancer case. Some of the state cancer registries made the determination of the variables or variables that were needed to answer the three research questi ons and provided only that information. providing the date of death, the South Carolina Cance Survival time was calculated as the number of months from date of diagnosis to date of death or censure time (12/31/2004) Of the original 46 variables reques ted from the cancer registries, eleven variables were chosen to answer the three research questions. The list of variables in Table 31 include gender, stage of disease, grade of lung cancer, morphology (histology and behavior), treatment group, age at dia gnosis, age group at the time of diagnosis, race,

PAGE 166

149 marital status at diagnosis, state of the cancer registry, vital status, and survival time ( number of months from the date of diagnosis to date of death or censure time (12/31/2004) ) Four variables of the original forty six variables were selected as primary variables and are listed below in the Table 31 The primary independent variables are gender morphology, stage, and grade and all four are included in the analysis to answer research questions one an d three. When Hypothesis II for question two was tested g ender was used as the primary independent variable. Table 31 Final Data Lung Cancer Set Variables Description of Variables Gender Morphology (Type and Behavior) Stage Grade Marital Statu s at Diagnosis Race Age at Diagnosis Group Vital Status Survival Time* Treatment Group State Survival time in months : from date of diagnosis to date of death or censure time (12/31/2004) From the originally requested 46 variables, several variab les were intended to be used as quality indicators be evaluated as possible confounders and to test for interaction effects Table 31 secondary variables include race, marita l status at the time of diagnosis, and age group at the time of diagnosis. Treatment Group was used as a response variable for testing Hypothesis I and as an independent variable for testing Hypothesis III. Table 32

PAGE 167

150 provides clarification and gives a des cription of the independent and outcome variables used to answer each of the research questions via hypothesis testing. Table 32: Classification of Variables for Hypothesis Testing Independent Variables (Predictor) Dependent Variables (Response) Hypot hesis I Gender, Stage, Grade, Morphology, Race, Marital Status, Age Group, and State* Treatment Group Hypothesis II Gender Survival Time, Vital Status Hypothesis III Gender, Stage, Grade, Morphology, Race, Marital Status, Age Group, Treatment Group, Survival Time, Vital Status *State was used in a separate model when testing for any random effect, i.e. any effect that state could have on the relationship between the outcome and the independent variables. In conclusion, eleven variables w ere utilized for the final analysis in order to answer the three research questions via hypothesis testing The list of variables include gender, stage of disease, grade of lung cancer, morphology (histology and behavior), treatment group, age at diagnosi s, race, marital status at diagnosis, state of the cancer registry, vital status, and survival time ( number of months from date of diagnosis to date of death or censure time (12/31/2004) )

PAGE 168

151 Population Characteristics Demographics The demographic character istics of the population under study from each state are given in Tables 33 through 36. There were a total of 44, 863 primary lung cancer cases included in the analysis after the study selection criteria for inclusion and exclusion were met and as outline d in Chapter Three. Briefly, the combined primary and secondary variable data set consisted of lung cancer cases that excluded individuals that were not diagnostically confirmed lung cancer, e.g. a diagnosis made by a cell/tissue sample or that were not a nalytic. For a lung cancer case to be considered analytic, one of three criterion must be met: (1) the di agnosis at the reporting facility and the entire first course of treatment was performed elsewhere or the decision not to treat was made at another fa cility (2) the d iagnosis at the reporting facility, and all or part of the first course of treatment was performed at the reporting facility and (3) the d iagnosis was made elsewhere, and all or part of the first course of treatment was performed at the r eporting facility. Also, for each individual lung cancer, any missing or NOS (not otherwise specified) values for the primary and secondary variables were excluded. As shown in Table 33 below, Florida provided the major contribution of lung cancer cases at 24,602 (55.5% of all females, 54.9% of all males) with the overall data set minimum for lung cancer cases from Idaho (2.0% of all females, 2.0% of all males). As expected, for all states, there were a higher percentage of males with lung cancer as comp ared to females with lung cancer. Overall, the data set has 19,994 females (44.6%

PAGE 169

152 of the total lung cancer cases) and 24,869 males (55.4% of the total lung cancer cases) shown in Table 36. Table 33: State Cancer Registries versus Gender Lung Cancer Dis tribution from the Eight State Cancer Registries Females Males State Cancer Registry Frequency Percent Frequency Percentage Total Florida 11089 55.5 13513 54.9 24602 Idaho 400 2.0 496 2.0 896 Indiana 2333 11.7 3107 12.5 5440 Massachusetts 2823 14.1 2992 12.0 5815 Nebraska 702 3.5 1018 3.5 1720 Oregon 735 3.7 824 3.3 1559 Rhode Island 459 2.3 666 2.7 1125 South Carolina 1453 7.3 2253 9.1 3706 Total 19994 100.0 24869 100.0 44863 The demographic characteristics for the lung cancer cases ( ge nder, vital status, race, age group, and marital status at diagnosis ) are listed in Table 24. The ages for the combined data set (primary lung cancer cases diagnosed between 1/1/2000 12/31/04) ranged from 40 89 years old. The mean age for the data se t (N = 44,863) was 67.9 years, SD + 10.2; for females (N f emale = 19,994) the mean age was 67.9 years, SD + 10.4 and for males (N m ale = 24,869), the mean equaled 68 years, SD + 10.0. For hypothesis testing, the fied into age groups (categorical variables ) ge Group at D strata or intervals were generated and are described in Table 24. The Age at Diagnosis (in years) Group 7 ( > 70 < 80 ) had greatest frequency of lung cancer cases with 16,404 (36. 6 %), followed by Group 6 ( > 60

PAGE 170

153 < 70 ) with 13,536 (30.2 %), Group 5 ( > 50 < 60 ) 7,179 (16.0 %), and the minimum number in an age group was the > 40 < 50 age interval, Group 4 with 2, 352 (5.2 %). Table 34: Lung Cancer Distribution Gender, Vital Statu s, Race, Age Group, and Marital Status at Diagnosis Frequency Percent Variable Gender Female 19994 44.6 Male 24869 55.4 Total 44863 100 Vital Status Dead 31869 71.0 Alive 12994 29.0 Total 44863 100 Race White 41458 92.4 Black 3042 6.8 Other 363 0.8 Total 44863 100 Age Group at Diagnosis > 40 < 50 yrs 2352 5.2 > 50 < 60 yrs 7179 16.0 > 60 < 70 yrs 13536 30.2 > 70 < 80 yrs 16404 36.6 > 80 < 90 yrs 5392 12.0 Total 44863 100 Marital Status at Diagnosis Single 4427 9.9 Married 26759 59.6 Separated 367 0.8 Divorced 4920 11.0 Widowed 8390 18.7 Total 44863 100 ge Group at D iagnosis Those age range

PAGE 171

154 groups not listed in Table 34 were > 0 <10 years old > 1 0 < 2 0 years old > 2 0 < 3 0 years old > 9 0 < 10 0 years old and > 10 0 years old. T he decision was made to limit the number of age groups based on the following: f irst, there were limited numbers of lung cancer cases that were younger than 40 and older than 90 The cumulative percent was less than 1% for the lung cancer data set for those lung cancer cases less than forty years of age and for those cases greater than 90 years old. An analysis and subsequent results would be subject error due to the limited sa mple size (decreased power or lack of ability Secondly, the population for the extremely young and extremely old, as referenced to lung cancer, is different and would not contribute to the relevance of the lung cancer cases selected for this research. In summary, the decision was made to exclude these age range groups. Seventy one percent of the lung cancer cases (31,869) were classified under the s alive. The study set, under Race, consisted mainly of White lung cancer cases (41,458 ( 92.4 %) ) with 3,042 (6.8 %) Black Table 34 also displays marital status at the time of lung cancer diagnosis. Appro ximately 60 percent of all the lung cancer cases (26,759, 59.6%) were classified as married at the time of diagnosis. The next classification with the greatest frequency was windowed (8,390, 18.7%) followed by divorced (4,920, 11.0%) and single (4,427, 9. 9%) with the minimum number classified as separated of 367 (0.8%). The primary research variables (main effect) listed in Table 35 includes stage,

PAGE 172

155 grade, and morphology. Morphology coding as previously stated in Chapter Three includes coding for the hi stological type of lung cancer combined with the behavior code of the disease. All the primary lung cancer cases in this data set have a behavior code of 3, meaning all lung cancer cases in this data set were classified as malignant. Table 35: Lung Cancer Distribution Stage, Grade, and Morphology Frequency Percent Variable Stage I 12028 26.8 II 4107 9.2 III 10359 23.1 IV 18369 40.9 Total 44863 100 Grade I 3153 7.0 II 12715 28.3 III 22417 50.0 IV 6578 14.7 Total 44863 100 Morphol ogy Adenocarcinoma 16139 36.0 Squamous 13425 29.9 Large Cell 8473 18.9 Small Cell 6826 15.2 Total 44863 100 Stage IV lung cancer accounts for 40.9 % of the total four stage classification scheme with the minimum number of cases found with Stage II at 9.2 %. Adenocarcinoma was the major morphological type with 16,139 cases (36.0 %), squamous cell had the second highest frequency with 13, 425 (29.0 %), followed by large

PAGE 173

156 cell carcinoma with 8,473 (18.9 %) cases, and lastly, small cell carcinoma hav ing 6,826 cases made up 15.2 % of the total four different stages of the lung cancer data base (N = 44,863). The grade of lung cancer (Table 35) consists of four classifications, most commonly found was Grade III (22,417, 50.0%); the Grade II lung cancer s consisted of 12,715 (28.3%) cases, Grade IV (6,578, 14.7%), and Grade I had the minimum number of lung cancer cases of 3,153 (7.0%). One of the last table s of demographic data Table 36 a consist s of the frequency and percent for state each cancer regi stry and t he treatment groups. Of the eight states listed, Florida was the major contributor of the lung cancer cases as expected due to a greater number of residents see Table 3 6 b for the 2000 2004 annual estimated population. Additional demographics for each s tate are provided in Appendix I in Tables 70 through Table 77 There are eight treatment classifications in Table 36 a w hich include a single treatment modality (Radiation Therapy (I), Chemotherapy (II), or Surgery (III)) treatment group, combinations of tre atment modalities received (Radiation and Surgery (IV), Radiation and Chemotherapy (V), Surgery and Chemotherapy (VI), or Radiation combined with Surgery and Chemotherapy (VII) and the last classification consisted of lung cancer cases that received no tre atment (Treatment Group VIII).

PAGE 174

157 Table 36 a : Lung Cancer Treatment Group and State Lung Cancer Distribution (Frequency and Percent) Frequency Percent Variable State Florida 24602 54.8 Idaho 896 2.0 Indiana 5440 12.1 Massachusetts 5815 13 .0 Nebraska 1720 3.8 Oregon 1559 3.5 Rhode Island 1125 2.5 South Carolina 3706 8.3 Total 44863 100 Treatment Group Radiation 4351 9.7 Chemotherapy 6472 14.4 Surgery 12728 28.4 Radiation + Surgery 1063 2.4 Radiation + Chemotherapy 7955 17. 7 Surgery + Chemotherapy 1249 2.8 Radiation + Surgery + Chemotherapy 1348 3.0 No Treatment 9697 21.6 Total 44863 100

PAGE 175

158 Table 3 6 b : Total Population for the Eight States* Males Fem ales Total N % n % N % State Florida 7,797,715 48.8 8,184,663 51.2 15,982,378 40.0 0 Idaho 648,660 50.1 645,293 49.9 1,293,953 3.24 Indiana 2,982,474 49.0 3,098,011 51.0 6,080,485 15.24 Massachusetts 3,058,816 48.2 3,290,281 51.8 6,349,097 15.91 Nebraska 843,351 49.3 867,912 50.7 1,711,263 4.29 Oregon 1 ,696,550 49.6 1,724,849 50.4 3,421,399 8.58 Rhode Island 503,635 48.0 544,684 52.0 1,048,319 2.63 South Carolina 1,948,929 48.6 2,063,083 51.4 4,012,012 10.06 Total 19,480,130 100 20,418,776 100 39,898,906 100 *Source: U.S. Census Bureau, Census 2000, and used as most current source of population statisitics for estimate purposes only Although this research was focused primarily on specific treatment modalities, a proportion of lung cancer cases received no treatment (no radiation, chemotherapy, and/or surgery) were classified as Treatment Group VIII (Table 36 a ). This classification allowed for the investigation of lung cancer cases that received no treatment by comparing the no treatment group to the other levels of the treatment groups. Treatment Gro up VIII had the second largest number of lung cancer cases as shown in Table 36 a with 9,697 (21.6%) subjects One concern regarding the utilizing this treatment group (VIII) would be a possible bias being introduced from utilization of the no treatment re ceived group as the reference group during the statistical testing/analysis. For example, utilizing Treatment Group VIII as the reference group could introduce a differential classification bias. This bias would be resultant from using a set/group of lu ng cancer cases (VIII) that were different from all the other lung cancer cases

PAGE 176

159 (Treatment Group I VII) as they (VIII) never received any treatment i.e. radiation chemotherapy surgery radiation + chemotherapy radiation + surgery, chemotherapy + surgery, or radiation + chemotherapy + surgery. This could bias the null in any direction and any effect from the comparison of other treatment groups could be masked. Simply stated, the category of no treatment group used as a reference group cannot be lung cancer cases that are comprised from a different population. The population characteristics of the eight treatment groups w ere compared; there were no observable trends that suggested that Group VIII had any observable differences suggesting a dissimilar population mix. The following tables, Table 37 through Table 40, contain the assessment of the primary variable by the indi vidual treatment groups. Each treatment group was evaluated for any dissimilarity or variability in gender, morphological type, stage, and grade of lung cancer. Table 37 displays the lung cancer distribution between the Treatment Groups versus Gender. In Table 37 f or all treatment groups, the frequency of males with lung cancer is greater than females with lung cancer. When evaluating the eight treatment groups versus gender, the greatest number of males with lung cancer is found in Treatment Group III (S urgery only) with 6,718 cases. The minimum number of male lung cancer cases (612) was documented for Treatment Group IV (Radiation and Surgery). The maximum number of females (6010) was found in Treatment Group III (Surgery only) with the minimum number of female lung cancer cases (451) in Table 37 receiving a combination of Radiation and Surgery (Treatment Group IV).

PAGE 177

160 Table 37: Lung Cancer Distribution Treatment Group vs. Gender Treatment Group Gender Frequency Percent Radiation I Female 1772 4 0.7 Male 2579 59.3 Chemotherapy II Female 2946 45.5 Male 3526 54.5 Surgery III Female 6010 47.2 Male 6718 52.8 Radiation + Surgery IV Female 451 42.4 Male 612 57.6 Radiation + Chemotherapy V Female 3346 42.1 Male 4609 57.9 Surge ry + Chemotherapy VI Female 585 46.8 Male 664 53.2 Radiation + Surgery + Chemotherapy VII Female 608 45.1 Male 740 54.9 No Radiation, Surgery, and/or Chemotherapy VIII Female 4276 44.1 Male 5421 55.9 Note: During statistical testing sur gery was designated as the reference Treatment Group (VIII), the No Radiation Surgery, and/or Chemotherapy Group (No Treatment Group) was designated as Treatment Group III The next three tables compare the eight treatment groups with the primary varia bles of stage (Table 38), grade (Table 39) and morphology (Table 40).

PAGE 178

161 Table 38: Lung Cancer Distribution Treatment Group vs. Stage Treatment Group Stage Frequency Percent Radiation I 658 15.1 I II 278 6.4 III 820 18.9 IV 2595 59.6 Chemothera py I 337 5.2 II II 277 4.3 III 1224 18.9 IV 4634 71.6 Surgery I 8332 65.5 III II 1847 14.5 III 1972 15.5 IV 577 4.5 Radiation + Surgery I 195 18.3 IV II 200 18.8 III 482 45.3 IV 186 17.5 Radiation + Chemotherapy I 585 7.4 V II 551 6.9 III 2677 33.7 IV 4142 52.1 Surgery + Chemotherapy I 334 26.7 VI II 163 13.1 III 551 44.1 IV 201 16.1 Radiation +Surgery+ Chemotherapy I 119 8.8 VII II 160 11.9 III 816 60.5 IV 253 18.8 No Radiation, Surgery, I 1468 15. 1 and/or Chemotherapy II 631 6.5 VIII III 1817 18.7 IV 5781 59.6

PAGE 179

162 The purpose of Table 38 was to compare the eight different treatment groups with the four stages of lung cancer. There were four treatment groups having the greatest percent of Sta ge IV lung cancer, Group I Radiation (59.6%), Group II Chemotherapy (71.6%), Group V Radiation and Chemotherapy (52.1%), and Group VIII (59.6%). One treatment group, the surgical treatment group, Group III, had the greatest percent (65.5%) for Stage I. Stage three lung cancers had the highest percentage in Group IV (Radiation and Surgery) at 45.3%, Group VI Surgery and Chemotherapy (44.1%) and Treatment Group VIII which combine d all three treatment modalities: Radiation, Surgery, and Chemotherapy. S even of the eight treatment groups (Table 39) had the highest percentage of lung cancers considered Grade III as compared to the other three grades : Treatment Group I (58.7 % ), Treatment Group II (47.5 % ), Group IV (52.5 % ), Group V (53.3 % ), Group VI (49.6 % ), Gro up VII (59.4 % ), and Group VIII (55.5 % ). There was only one treatment group (those receiving surgery only, Treatment Group III ) in which the highest proportion (44.8%) of lung cancer cases were Grade II. In Table 40, Adenocarcinoma was the most frequent histological type of lung cancer for Treatment Groups III ( 46.2% ), IV ( 45.6% ), VI ( 46.3% ), and VIII ( 33.7% ). Squamous cell lung cancer was most common in Groups I (36.7%) and V (27.3%). Lastly, the most common histologic type of lung cancer for Group II Chemotherapy (38.5%) was Small Cell Lung Cancer (SCLC).

PAGE 180

163 Table 39: Lung Cancer Distribution Treatment Group vs. Grade Treatment Group Grade Frequency Percent Radiation I 244 5.6 I II 1171 26.9 III 2553 58.7 IV 383 8.8 Chemotherapy I 216 3.3 II II 926 14.3 III 3072 47.5 IV 2258 34.9 Surgery I 1627 12.8 III II 5697 44.8 III 5194 40.8 IV 210 1.7 Radiation + Surgery I 63 5.9 IV II 418 39.3 III 558 52.5 IV 24 2.3 Radiation + Chemotherapy I 280 3.5 V II 1496 18.8 III 4239 53.3 IV 1940 24.4 Surgery + Chemotherapy I 103 8.3 VI II 463 37.1 III 619 49.6 IV 64 5.1 Radiation +Surgery + Chemotherapy I 52 3.9 VII II 429 31.8 III 801 59.4 IV 66 4.9 No Radiation, Surgery, I 568 5.9 and/or Chemoth erapy II 2115 21.8 VIII III 5381 55.5 IV 1633 16.8

PAGE 181

164 Table 40: Lung Cancer Distribution Treatment Group vs. Morphology Treatment Group Morphology Frequency Percent Radiation Adenocarcinoma 1430 32.9 I Squamous 1596 36.7 Large Cell 980 22.5 Small Cell 345 7.9 Chemotherapy Adenocarcinoma 1747 27.0 II Squamous 1087 16.8 Large Cell 1145 17.7 Small Cell 2493 38.5 Surgery Adenocarcinoma 5880 46.2 III Squamous 4610 36.2 Large Cell 2142 16.8 Small Cell 96 0.8 Radiation + Surgery Adenocarcinoma 485 45.6 IV Squamous 426 40.1 Large Cell 144 13.6 Small Cell 8 0.8 Radiation + Chemotherapy Adenocarcinoma 2106 26.5 V Squamous 2171 27.3 Large Cell 1597 20.1 Small Cell 2081 26.2 Surgery + Chemotherapy Adenocarcinoma 602 4 8.2 VI Squamous 350 28.0 Large Cell 238 19.1 Small Cell 59 4.7 Radiation + Surgery + Adenocarcinoma 624 46.3 Chemotherapy Squamous 419 31.1 VII Large Cell 241 17.9 Small Cell 64 4.8 No Treatment Adenocarcinoma 3265 33.7 VIII Squamous 2766 28.5 Large Cell 1986 20.5 Small Cell 1680 17.3

PAGE 182

165 The tables and results of the analysis by treatment group versus the secondary variables are located in Appendix II Tables 41 through 43. The tables display a comparison of the treatment groups v ersus age race (Table 41), marital status group at diagnosis (Table 42), and age group at the time of lung cancer diagnosis (Table 43). There were no obvious differences in treatment groups versus and distribution of race, Table 41 in Appendix II The majo rity of lung cancer cases are White ranging from 91.3% of all lung cancer cases in Group V (Radiation and Chemotherapy) to 94% of all lung cancer cases in Group III (Surgery). the least amount of lung cance r cases for each treatment group with each Treatment Group having a minimum of approximately one percent within each treatment classification (I VIII). A ll treatment groups in Table 42 (Treatment Group vs. Marital Status at Diagnosis) have the greatest percentage of the lung cancer cases classified as married at the time of diagnosis ranging from 52.2 percent for Treatment Group VIII (no treatment) to maximum percentage of 69.9 percent for surgical and chemotherapy, Group VI. Lastly, each treatment group was evaluated for Age Group at Diagnosis. T he treatment groups, I VII did not display that Treatment Group VIII was any different or displayed any trends that would suggest that the patient population was not comparable to the other seven treatment gro ups It was determined that the lung cancer cases in Group VIII were just as likely to receive a treatment modality or combination of treatment modalities when comparing the treatment patterns for Groups I through VII. The decision was made to include G roup VIII for the analysis.

PAGE 183

166 Testing the Hypotheses Hypothesis I Women with the same histological type, stage and grade of lung cancer receive d the same treatment modality as compared to men with the same histological type, stage and grade of lung cancer. Introduction The first aim was to determine if men and women stratified by histologic type, stage, and grade of lung cancer received the same treatment type. The relationship between gender and the treatment modality received was evaluated including othe r independent variables and interaction terms. Interaction terms were included in the model to assess the role of moderating variables. A moderating variable can change the association (Odds Ratios) between the independent variable and the outcome variab le at different levels of that moderator. It was important to establish if different treatments, e.g. radiation therapy, chemotherapy, surgery, were received based on gender; this has not been addressed specifically in the literature. Also after determin ing if the type of lung cancer treatment received was gender dependent, further analysis or study of that impact on gender specific survivorship could be addressed in Hypotheses II and III. As stated in Chapter Three, the statistical model selected to exa mine the relationship between the outcome variable (treatment group) and the independent variables was the multinomial, polychotomous or polytomous logistic regression model. The multinomial as more than two

PAGE 184

167 outcomes; the research outcome variable, treatment group, was comprised of eight categories or eight different treatment selections. Table 44 lists the variables used during the testing of Hypothesis I; the variables were coded as catego rical variables and were classified on the nominal scale. Also shown in Table 44 are the reference groups, for example Group VIII (Surgery) was utilized as the reference group during statistical testing with the generalized logit model. Table 44: Outc ome Variable and Independent Variables Variables for Testing Hypothesis I Multinomial Logistic Regression Model (MLR 1 ) Outcome Variable* Independent Variables** Treatment Group Gender I. Radiation Therapy II. Chemotherapy III. No Treatment Assignment (No Radiatio n, Chemotherapy and/or Surgery) IV. Radiation + Surgery V. Radiation + Chemotherapy VI. Surgery + Chemotherapy VII. Radiation + Chemotherapy + Surgery VIII. Surgery Stage Grade Morphology Race Marital Status at Diagnosis Age Group at Diagnosis Outcome Variable Reference Group: Treatment Group = Surgery ** Independent Variable Reference: Gender = 2 (Male), Stage = IV, Grade = IV Morphology = 4 (SCLC), Race = 3 (Other, Non White), Marital Status = 5 (Widowed) and Age Group = 5 ( > 80 < 90 yrs) The outcome variable profile given previously in Table 36 had a total frequency of 44,863 lung cancer cases; there was one outcome variable, treatment group, with eight treatment groups. The total frequency of each treatment group of lung cancer cases were

PAGE 185

168 also given in Table 36 with the minimum number of cases in treatment group IV (Radiation and Surgery) with 1,063 cases and a maximum of 12,728 lung cancer cases in treatment group VIII (Surgery). Potential Confounders, Multicollinearity and Interaction s Thi s next section reviews three topics of interest: potential confounders, multicollinearity and interaction as each can impact the study by biasing the results due to the design and in the analysis phase. First, a potential source of error affecting the val idity of study can be confounding variables. Confounding can cause a distortion in the measure of association due to an unequal distribution of a determinant of the outcome 73, 224 Confounding is a problem of comparison, a problem that arises when impor tant extraneous factors are differentially distributed across groups being compared. A confounding variable is related to the outcome variable and the independent variable but not on the direct causal pathway between the outcome and independent variable o f interest 73, 224 The following methods were utilized to control for confounding in the design phase of the study. In the study d esign phase, two methods selected to reduce any confounding were 1) randomization in the selection of the states and 2) res triction: some of the restrictions included selecting only primary lung cancer cases and cases from NAACCR cancer registries. Smoking is a major confounding variable for lung cancer but smoking was not addressed as a variable in this study due to several reasons. There was difficulty in

PAGE 186

169 and collection of smoking history can be dissimilar and are listed below: I. Smoking was coded differently, e.g. the NAACCR standard cod e format was not followed in each state cancer registry. II. The data for smoking or smoking history were not collected in similar manner across all states, e.g. different start/inception dates to begin the collection of smoking information. III. No smoking history was collected or the information was not available from two state cancer registries under study, therefore a complete assessment could not be made. IV. Whether a person smokes or not is not a variable of interest in this research. In the opinion of the resea rcher, s moking or not smoking is not associated or rather will not determine if a lung cancer case receives a particular treatment modality, e.g. radiation, surgery, and or any combination of treatment modalities. Next, a second possible source of error could be in the case of collinearity or highly collinear values between two independent variables that could affect the relationship of either or both variables on the research outcome, i. e. the lung cancer treatment received. This is commonly referred to high level of intercorrelation between the independent variables, the effects of the

PAGE 187

170 independent variables may not be separated resulting in statistical inferences made about the data that could be unreliable. For the research data set, a similar method described by Hosmer and Lemeshow 1 for categorical variables was utilized to test for multicollinearity. First, a logistic regression model was run or generated with all seven independent variable: gender, stage, g rade, morphology, marital status, age group, and race. Then seven logistic regression models (variable subsets) were generated, i.e. each model dropped one of the independent variables that were originally included in the full model. The full model coeff icients and standard errors were examined and then compared to the coefficients and standard errors for the seven other logistic regression models (Model I all variables Model II stage excluded Model III grade excluded Model IV morphology excluded Mode l V gender excluded Model VI marital status excluded Model VII race excluded Model VIII age group excluded ). There were no appreciable differences between the standard errors and the coefficients An appreciable difference as noted by Hosmer and Lemesh ow could be a change in the beta coefficients or standard errors by an order of magnitude. For example, in Table 4 5 the data extracted from the full model, Model I and Model II (morphology removed from the other independent variables) does not display suc h a change From was deemed minimal; therefore the coefficients were assumed to be unbiased. 1 Pages 140 141

PAGE 188

171 Table 4 5 : Multicollinearity Assessment via Logistic Regression Compariso n of Coefficients and Standard Error Extracted from the Logistic Regression Models Parameter Coefficient Standard Error MODEL I (Full Model) Morphology 1 0.285 0.017 Morphology 2 0.007 0.018 Morphology 3 0.016 0.019 Grade 1 0. 221 0.029 Grade 2 0.198 0.020 Grade 3 0.189 0.016 Stage 1 1.364 0.018 Stage 2 0.381 0.023 Stage 3 0.321 0.016 Gender 1 0.037 0.009 Race 1 0.097 0.035 Race 2 0.103 0.039 Marital Status 1 0.061 0.030 Marital Status 2 0.162 0.023 M arital Status 3 0.071 0.077 Marital Status 4 0.021 0.029 Age Group 4 0.212 0.032 Age Group 5 0.162 0.020 Age Group 6 0.127 0.017 Age Group 7 0.013 0.016 Grade 1 0.324 0.027 Grade 2 0.301 0.017 Grade 3 0.121 0.014 Stage 1 1.363 0.0 18 Stage 2 0.379 0.023 MODEL II (Morphology Excluded) Stage 3 0.328 0.016 Gender 1 0.047 0.009 Race 1 0.091 0.035 Race 2 0.110 0.039 Marital Status 1 0.063 0.030 Marital Status 2 0.170 0.023 Marital Status 3 0.074 0.077 Marital Status 4 0.022 0.029 Age Group 4 0.232 0.032 Age Group 5 0.166 0.020 Age Group 6 0.121 0.017 Age Group 7 0.025 0.016 *Model I included the independent variables of gender, stage, grade, morphology, marital status, race, and age group ** Model II inc luded the independent variables of gender, stage, grade, morphology, marital status, race, and age group

PAGE 189

172 Another area of importance in the testing of the hypothesis was the evaluation of possible interaction or effect modification that might significantl y affect the relationship between the outcome and the independent variables. An effect modifier is a variable that changes the relationship between the independent variable and the outcome variable. A variable that acts as an effect modifier is contained in the interaction term and the outcome/independent variable relationship changes at different levels or strata 73, 74, 226 The interaction terms were evaluated for their impact on the outcome and any effect on the overall fit of the model equation. In the statistical testing and analysis stage, interaction or effect modification was addressed by evaluating the stratified multivariate analyses. In summary, stratification based on the independent variables, such as stage, grade and morphological lung ca ncer type was employed in the statistical methods and interaction terms were included to evaluate the effects of any possible moderating variable on the relationship between the independent variable and outcome variable at different levels of that moderati ng variable. Multinomial Logistic Regression Stepwise multinomial logistic regression (MLR) testing was used to select the variables (main effects and interaction terms) for the full model. The stepwise process consisted of a forward selection of covari ates with a backward elimination of variables that did not meet a specified significance level. The stepwise procedure incorporated (specified in the model statement) the four primary variables (gender, stage, grade, and

PAGE 190

173 morphology) throughout the process selection and included variables with second degree interaction terms in the stepwise model selection process. The criterion for inclusion into a model was a significance level of 0.05 and an elimination of variable/interaction terms when the significanc e level was greater than 0.05. In the stepwise multinomial logistic procedure, Type III Analysis of Effects showed the change in the model fit when an independent variable was dropped while keeping the other variables in the model. In all, there were el even different models generated. The resultant multinomial logistic model included seven main effects for the variables of gender, morphology, stage, grade, marital status, race, age group and five interaction terms of gender*morphology, gender*stage, sta ge*grade, gender*marital status and stage*age group Table 46 gives the results for the statistical test that was generated for the final model, i.e. Type III Analysis of Effects. The Type III test statistic is associated with the estimated coefficients in the model and represents an effect due to a particular variable, e.g. gender. The statistic for the Type III test is the amount of variation in the response when a particular variable, e.g. gender, is added to the model that already contains all the o ther variables. Also, the Type III test statistic is not depended upon the order that the independent variables (to include interaction terms) are specified in the model. In the full model, gender was not significant (p value >0.05) as a main effect but was significant in the interaction terms of gender and stage (gender*stage) and gender and marital status ( gender*marital status) Other statistically significant interaction terms included stage and grade ( stage*grade) and stage and age group (stage*age group). The

PAGE 191

174 interpretation of the exponentiated parameter estimates are presented as Odds Ratios with the associated 95% confidence intervals in the following paragraphs of this chapter Also as noted in Table 46, there are large Wald Chi Square values f or several main effect variables and interaction terms. Large Wald Chi Square statistics are an indication of the variability of the data contained in the model. For example, this result could be attributed to the multinomial nature of the output variabl e, treatment group. The variability of the parameter estimates in the model could be increased due to the fact there are eight different levels of the outcome variable, treatment group. Table 46: Type III Analysis of Effects Main Effect and Interaction Terms* Multinomial Logistic Regression Model (MLR 1 ) Effect DF Wald Chi Square Pr > ChiSq Gender 7 10.63 0.156 Morphology 21 879.88 <.0001 Gender*Morphology 21 29.43 0.104 Grade 21 132.55 <.0001 Stage 21 260.73 <.0001 Gender*Stage 21 45.31 0.002 S tage*Grade 63 178.92 <.0001 Marital Status 28 325.40 <.0001 Gender*Marital Status 28 52.94 0.003 Age Group 28 948.87 <.0001 Stage*Age Group 84 222.51 <.0001 Race 14 88.36 <.0001 Note: Age Group and Marital Status at defined on /at Date of Diagnosis indicates interaction term

PAGE 192

175 The Multinomial Logistic Regression (MLR) model (also known as the generalized or baseline category logit model) with treatment group as the outcome variable is represented by t he following equation (coefficients are not displayed): = Gender + Morphology + Gender*Morphology + Grade + Stage + Gender *Stage + Stage*Grade + Marital Status + Gender*Marital Status + Age Group + Stage*Age Group + Race There were seven categories of t he response variable (treatment group) for the multinomial logistic regression model. From the equation above, in the numerator Y represents the treatment type; when i = 1, the treatment group is radiation alone if i = 2 the treatment group is chemothera py alone i = 3 the treatment group is no treatment, i = 4, the treatment group is radiation + chemotherapy, i =5 the treatment group is surgery + radiation, i = 6 the treatment group is surgery + chemotherapy, and when i = 7 the treatment group is radiati on + chemotherapy + surgery. Also in the equation above, in the denominator Y is the reference treatment group for each of the treatment groups, 1 through 7. For the term Y (in the denominator) Y is the reference group and is the eighth treatment group (surgery). In the equation above, k + 1 is a numerical expression that specifies or is representative of the reference treatment group 8 (surgery); k = 7 therefore k +1 = Y =8 From Table 46, the following equation was derived from the full model (the int eraction term of Gender*Morphology was not included because it was not statistically significant and therefore would not affect the outcome) :

PAGE 193

176 Logit (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 gender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI + 10 morphologyI + 1 1 raceI + others The Wald statistic in Table 46 is a parameter/statistic that can be utilized to assess be interpreted as how well a mathematical equation estimates the behavior of the data. If a large variability of Wald Chi Square (a statistic derived from the parameter estimate and standard error) existed, this could be interpreted that the wrong model was selected to examine the data. A brief review will be presented concerning the issue of how the assessment of the model fit was evaluated prior to reviewing the MLR model results. Model fit or the assessment of the predicted results versus the truth (the actual results from the data) for a statistical test can be performed after the analysis because the researcher wishes to ensure that he/she are using the correct method to test or assess their data and that the resu lts are valid Prior to inferences being made for the fitted model, an assessment of the model fit via diagnostics for the multinomial logistic regression model was made. Residual testing is a common statistical approach in the evaluation of the error in the model equation comparing the predicted or estimated results with the data. Because residual analysis was not available in the SAS software for a multinomial or multiple outcome variable (treatment) levels ; Hosmer and Lemeshow 2 suggest assessing 2 David W. Hosmer a tion Section 4.5

PAGE 194

177 the f it via logistic regression models for each outcome (seven logistic regression models in total) The results of the residual analysis were compared and examined for any trends that would demonstrate a lack of model fit. This type of assessment was made fo r the seven possible treatment outcomes of 1) radiation alone, 2) chemotherapy alone, 3) no treatment, 4) radiation + chemotherapy, 5) radiation + surgery, 6) surgery + chemotherapy, and 7) radiation + chemotherapy + surgery) with surgery as the reference group for each treatment group Examples of the residual analysis results for three of the treatment outcome groups are given in Tables 52 through 54. No trends were demonstrated for the seven models meaning the use of the multinomial logistic regression model to the best of our knowledge was appropriate. In the next five sections, the statistical results obtained from the logistic regression models for the coefficient estimates, standard errors, the Odds Ratios and 95% confidence interval are evaluated The first section reviews the ORs /95% CIs for the statistically significant main effect variables morphology and race (Tables 47: a b ) Note these main effects are reported because they are not included in any statistically significant interaction t erms. If morphology or race were in an interaction term that was statistically significant, that result due to the interaction would be reported. As the outcome could change due to the interaction; the outcome could be misinterpreted if the main effect v ariables were the only variables considered. In the second section the MLRM ORs of the interaction terms (Tables 48 51 ) are evaluated. The third s ection compares and contrasts the results of the assessment of the model fit The fourth section

PAGE 195

178 present s the results of a random effect component for state in the multinomial logistic regression model. The results given in the fifth s ection are the statistics and interpretation of the overall variable effect on the treatment type received. Section 1: Mul tinomial Logistic Regression Main Effects When examining only main effect results from a statistical model, any type of effect modification between an independent variable and the outcome based on a moderating variable is not accounted for. When two vari ables interact in determining the chance of a particular outcome, it is inappropriate to just report the main effect as it will give misleading results. In this research the relationship between some of the main effect variables and the outcome variable c hanged when a moderating variable was present, as in the interaction term. The only time it is appropriate to report the main effect results is when that main effect variable gives statistically significant results and any interaction term containing that main effect variable is not statistically significant. Table 47 a consists of non intervals generated in the full model, MLR 1 for the main effect of morphology. The main effect of morphology is report ed because this variable is significant and the interaction term of gender and morphology (gender*morphology) was not statistically significant (the results for the interaction of gender and morphology term are not reported in the final analysis). When ev aluating morphology in Table 47 a, lung cancer cases with adenocarcinoma versus other lung cancer morphological types were at a statistically

PAGE 196

179 0.23) with the exception of adenocarci noma lung cancer cases receiving radiation therapy in combination with surgery (OR = 1.55, 95% CI 0.52, 2.65) as compared to receiving surgery after adjustment for gender, gender*morphology, grade, stage, gender*stage, stage*grade, marital status, gender*m arital status, age group, stage*age group, and race Comparing the Odds Ratios between adenocarcinoma and treatment type to lung cancer cases with squamous cell and treatment type, the same relationship was exhibited. There was an decreased risk that lun g cancer cases with squamous cell carcinoma as compared from 0.04 to 0.28) with the exception of radiation in combination with surgery (OR = 1.92, 95% CI 0.64, 3.01) as com pared to receiving surgery alone but this was not statistically significant. The last table for a main effect is Table 47 b, with race as the main effect with the outcome of treatment type. There was one statistically significant association between race and treatment type; no trends were exhibited for white and black lung cancer cases receiving a particular treatment versus receiving surgery. A possible limitation to this analysis of race was the reduced number of the reference group, other lung cancer cases (n = 363, 0.8%). There was an increased number of white (n = 41,458, 92.4%) and black lung (n = 3,042, 6.8%) cancer cases. This reduced number as the reference group could have introduced some bias into results (artificial inflation of the ORs). A s the ORs only demonstrate d one statistically significant result and as the variability o f the confidence

PAGE 197

180 intervals was minimal, the decision was made to keep this group (other) as the referent group. Table 47 a: Main Effect for Morphology Extracted From the Full Model (MLR 1 ) Odds Ratios and 95% Confidence Intervals by Treatment Group Treatment Type (Outcome) Main Effect Odds Ratio 95% LCI 95% UCI Morphology Radiation Adenocarcinoma 0.14 0.10 0.52 Chemotherapy Adenocarcinoma 0.03 0.02 0.37 No Tr eatment Adenocarcinoma 0.07 0.05 0.41 Radiation + Surgery Adenocarcinoma 1.55 0.52 2.65 Radiation + Chemotherapy Adenocarcinoma 0.04 0.03 0.38 Surgery + Chemotherapy Adenocarcinoma 0.23 0.13 0.80 Radiation +Surgery + Chemotherapy Adenocarcinoma 0.21 0. 12 0.75 Surgery Small Cell* 1 .00 Radiation Squamous 0.28 0.19 0.66 Chemotherapy Squamous 0.04 0.03 0.38 No Treatment Squamous 0.11 0.08 0.45 Radiation + Surgery Squamous 1.92 0.64 3.01 Radiation + Chemotherapy Squamous 0.08 0.06 0.42 Surgery + Chemotherapy Squamous 0.22 0.13 0.79 Radiation +Surgery + Chemotherapy Squamous 0.22 0.13 0.76 Surgery Small Cell* 1 .00 Radiation Large Cell 0.30 0.21 0.69 Chemotherapy Large Cell 0.06 0.04 0.41 No Treatment Large Cell 0.13 0.10 0.48 Radiation + Surgery Large Cell 1.64 0.55 2.75 Radiation + Chemotherapy Large Cell 0.10 0.07 0.44 Surgery + Chemotherapy Large Cell 0.34 0.19 0.92 Radiation +Surgery + Chemotherapy Large Cell 0.29 0.17 0.84 Surgery Small Cell* 1 .00 Signifies Referent or Re ference Group A djusted for gender gender*morphology, grade, stage, gender*stage, stage*grade, marital status, gender*marital status, age group, stage*age group, and race

PAGE 198

181 Table 47 b: Main Effect of Race Extracted From the Full Model (MLR 1 ) Odds Ratios a nd 95% Confidence Intervals by Treatment Group Treatment Type (Outcome) Main Effect Odds Ratio 95% LCI 95% UCI Race Radiation White 0.99 0.62 1.46 Chemotherapy White 0.67 0.45 1.08 No Treatment White 0.86 0.59 1.23 Radiation + Surgery White 0.88 0.43 1.58 Radiation + Chemotherapy White 0.83 0.56 1.23 Surgery + Chemotherapy White 1.31 0.62 2.05 Radiation +Surgery + Chemotherapy White 0.85 0.46 1.48 Surgery Other 1 Radiation Black 1.74 1.06 2.23 Chemotherapy Black 0.82 0.53 1.25 No Treatment Black 1.32 0.89 1.72 Radiation + Surgery Black 0.94 0.44 1.69 Radiation + Chemotherapy Black 1.22 0.80 1.63 Surgery + Chemotherapy Black 1.34 0.62 2.12 Radiation +Surgery + Chemotherapy Black 0.84 0.43 1.51 Surgery Other 1 Signi fies Referent or Reference Group Adjusted for gender, morphology, gender*morphology, stage, grade, gender*stage, stage*grade, marital status gender*marital status, age group and stage*age group. Section 2: Multinomial Logistic Regression Interaction Terms The next section lists the results that contain statistically significant interaction terms for the full model (Tables 48 intervals are given for the interaction term of gender and stage. In this research, the overall gender effect is reported later in this chapter which utilizes the results of the main effect of gender and the interaction term of gender*stage. Also the results from Table 48 sus Binomial Logistic Regression

PAGE 199

182 The Odds Ratios and 95% Confidence Intervals are given in Table 49 for the treatment groups and the interaction term of gender and marital status. The moderating variable of gender demonstrated the interaction eff ect varied according to the level of marital status in the interaction term (gender*marital status). Once again the statistics will not be discussed for Table 49 as this is covered in section five in portion of this chapter

PAGE 200

183 Table 48: Gender and Stage Interaction Terms Significant Terms Extracted From the Full Model (MLR 1 ) Odds Ratios and 95% Confidence Intervals by Treatment Group Treatment Type (Outcome) Interaction Term Odds Ratio 95% LCI 95% UCI Gender St age Radiation Female I 1.35 1.05 1.60 Radiation Female II 0.91 0.65 1.23 Radiation Female III 1.04 0.81 1.29 Surgery Male* IV* 1 Chemotherapy Female I 1.53 1.14 1.82 Chemotherapy Female I I 1.03 0.74 1.35 Chemotherapy Female III 1.29 1.02 1.5 2 Surgery Male* IV* 1 No Treatment Female I 1.12 0.91 1.33 No Treatment Female II 1.16 0.89 1.41 No Treatment Female III 1.14 0.91 1.36 Surgery Male* IV* 1 Radiation + Surgery Female I 0.95 0.61 1.39 Radiation + Surgery Female II 0.94 0.60 1.39 Radiation + Surgery Female III 1.07 0.72 1.47 Surgery Male* IV* 1 Radiation + Chemotherapy Female I 1.17 0.91 1.42 Radiation + Chemotherapy Female II 0.99 0.76 1.27 Radiation + Chemotherapy Female III 1.26 1.02 1.48 Surgery Male* IV* 1 Surgery + Chemotherapy Female I 0.99 0.67 1.38 Surgery + Chemotherapy Female II 0.84 0.53 1.30 Surgery + Chemotherapy Female III 1.25 0.86 1.63 Surgery Male* IV* 1 Radiation +Surgery + Chemotherapy Female I 0.62 0.38 1.11 Radiation +Surgery + Chemotherapy Female II 0.68 0.43 1.14 Radiation +Surgery + Chemotherapy Female III 1.27 0.90 1.61 Surgery Male* IV* 1 Note: = The reference group LCI = Lower Confidence Interval, UCI = Upper Confidence Interval ** Adjusted for morphology grade stage*grade marital status gender*marital status age group stage*age group and race

PAGE 201

184 Table 49: Gender and Marital Status Interaction Terms Significant Terms Extracted From the Full Model (MLR 1 ) Odds Ratios and 95% Confidence Intervals by Treatme nt Group Treatment Type (Outcome) Interaction Term Odds Ratio 95% LCI 95% UCI Gender Marital Status Radiation Female Single 0.73 0.53 1.05 Radiation Female Married 0.88 0.71 1.10 Radiation Female Separated 1.67 0.69 2.56 Radiation F emale Divorced 0.67 0.49 0.97 Surgery Male* Widowed* 1 Chemotherapy Female Single 1.12 0.82 1.44 Chemotherapy Female Married 1.10 0.89 1.32 Chemotherapy Female Separated 2.47 1.03 3.35 Chemotherapy Female Divorced 1.04 0.77 1.34 Surgery Male* W idowed* 1 No Treatment Female Single 0.76 0.58 1.02 No Treatment Female Married 0.94 0.78 1.12 No Treatment Female Separated 1.89 0.85 2.69 No Treatment Female Divorced 0.75 0.58 1.00 Surgery Male* Widowed* 1 Radiation + Surgery Female Singl e 0.75 0.42 1.33 Radiation + Surgery Female Married 0.70 0.46 1.12 Radiation + Surgery Female Separated 0.65 0.10 2.47 Radiation + Surgery Female Divorced 0.65 0.37 1.20 Surgery Male* Widowed* 1 Radiation + Chemotherapy Female Single 0.83 0.62 1. 14 Radiation + Chemotherapy Female Married 0.78 0.63 0.99 Radiation + Chemotherapy Female Separated 1.59 0.71 2.40 Radiation + Chemotherapy Female Divorced 0.74 0.56 1.02 Surgery Male* Widowed* 1 Surgery + Chemotherapy Female Single 1.50 0.83 2.0 8 Surgery + Chemotherapy Female Married 0.86 0.56 1.29 Surgery + Chemotherapy Female Separated 1.52 0.35 3.00 Surgery + Chemotherapy Female Divorced 0.69 0.40 1.25 Surgery Male* Widowed* 1 Radiation +Surgery + Chemotherapy Female Single 1.38 0.78 1.95 Radiation +Surgery + Chemotherapy Female Married 1.07 0.72 1.47 Radiation +Surgery + Chemotherapy Female Separated 2.27 0.57 3.65 Radiation +Surgery + Chemotherapy Female Divorced 1.42 0.84 1.93 Surgery Male* Widowed* 1 Note: = The refere nce group LCI = Lower Confidence Interval, UCI = Upper Confidence Interval ** A djusted for morphology grade stage stage*grade gender* stage, age group stage*age group and race

PAGE 202

185 Table 50 displays the results for the interaction term of stage and age group for treatment groups containing statistically significant ORs. The moderating variable of age group at the time of diagnosis demonstrated interaction between stage (independent variable) and the treatment group (outcome) based on the level of the mo derator Lung cancer cases with stage I lung cancer were 78% less likely to receiving radiation therapy treatments for Age Group V ( O R = 0.22, 95% CI 0.13 0.75) and 63% less likely for Age Group VI (O R = 0.37, 95% CI 0.24 0.79) as compared to receivin g surgery. This result was expected as radiation alone as a treatment for early stage lung cancer would not be the standard of care. For the other age groups with stage I lung cancer receiving radiation, the results were not statistically significant. C linically, it would be predicted that the od d s ratios for the youngest age group or Age Group 4 with early stage lung cancer would demonstrate a statistically significant decrease in the probability of receiving radiation alone but the results were not sta tistically significant (Table 50). For stage I lung cancer cases receiving radiation in combination with chemotherapy, there was a trend demonstrated that as age increased the ORs approached 1.00. Overall there was a decrease likelihood of being treate d with chemotherapy combined with radiation. For early stage disease the youngest age group was 85% less likely to receive chemotherapy combined with radiation with the oldest age group (age group VII ) being 32% less likely to receive radiation in combina tion with chemotherapy after controlling for gender, morphology, gender*morphology, grade, gender*stage, stage*grade, gender*marital status, marital status, and race As shown in Table 50,

PAGE 203

186 overall, the younger the early stage lung cancer case was, the le ss likely the lung cancer case had of being treated with radiation or radiation in combination with chemotherapy as compared to receiving surgery. Table 50: Stage and Age Group at Diagnosis Interaction Terms Significant Terms Extracted From the Full Mod el (MLR 1 ) Odds Ratios and 95% Confidence Intervals by Treatment Group Treatment Type (Outcome) Interaction Term Odds Ratio 95% LCI 95% UCI Stage Age Group Radiation I 4 0.08 0.03 1.14 Radiation I 5 0.22 0.13 0.75 Radiation I 6 0.37 0.24 0.7 9 Radiation I 7 0.69 0.47 1.08 Surgery IV* 8 1 Radiation III 4 0.31 0.14 1.08 Radiation III 5 0.48 0.29 0.97 Radiation III 6 0.71 0.46 1.13 Radiation III 7 0.87 0.58 1.28 Surgery IV* 8 1 Radiation + Chemotherapy I 4 0.15 0.07 0.92 Radi ation + Chemotherapy I 5 0.27 0.16 0.79 Radiation + Chemotherapy I 6 0.42 0.26 0.88 Radiation + Chemotherapy I 7 0.68 0.44 1.13 Surgery IV* 8 1 Note: = The reference group; LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Age G roups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs ) ** A djusted for gender, morphology, gender*morphology, grade, gender*stage, stage*grade, gender*marital status, marital status, and rac e Table 51 contains the results for the interaction terms of stage and grade. There were no statistically significant ORs for Stage II at any level of Grade (the moderating variable) and was not presented in Table 51. Stage I and Stage III with the mo derating variable of Grade that contained statistically significant results for a treatment group are listed. A fter adjustment for gender, morphology, gender*stage, stage* age group marital

PAGE 204

187 status, gender*marital status, age group, and race g rade II sta ge I lung cancer cases as compared to other lung cancer cases were 65% less likely to receive chemotherapy (OR = 0.35, 95% CI 0.20 0.93) This particular result was not unexpected as early stage disease, as the standard of care is not to receive chemoth erapy as the only treatment for lung cancer; grade moderated the relationship between stage and the treatment received. Stage III grade I lung cancer cases as compared to other lung cancer cases were 7 times more likely (OR = 7.03, 95% CI 1.50 8.58) to receive radiation in combination with surgery after adjustment for gender, morphology, gender*stage, stage* age group marital status, gender*marital status, age group, and race. The moderating effect of grade on the independent and the outcome relationshi p was highly significant for stage III grade I but this relationship was not significant for stage III with grade II or III. When considering just the ORs and not the confidence intervals this could be suggestive of a treatment difference base on grade as there was a trend of decreasing ORs with an increasing grade.

PAGE 205

188 T able 51: Stage and Grade at Diagnosis Interaction Terms Significant Terms Extracted From the Full Model (MLR 1 ) Odds Ratios and 95% Confidence Intervals by Treatment Group Treatme nt Type (Outcome) Interaction Term Odds Ratio 95% LCI 95% UCI Stage Grade Chemotherapy I I 0.69 0.35 1.37 Chemotherapy I II 0.35 0.20 0.93 Chemotherapy I III 0.35 0.20 0.90 Surgery* IV* IV* 1.00 Radiation + Chemotherapy I I 0.34 0 .17 1.01 Radiation + Chemotherapy I II 0.29 0.17 0.83 Radiation + Chemotherapy I III 0.30 0.18 0.82 Surgery* IV* IV* 1.00 Radiation + Surgery III I 7.03 1.50 8.58 Radiation + Surgery I II II 2.83 0.75 4.16 Radiation + Surgery I II III 2.45 0 .66 3.76 Surgery* IV* IV* 1.00 Note: = The reference group; LCI = Lower Confidence Interval, UCI = Upper Confidence Interval ** A djusted by gender, morphology, gender*morphology, gender*stage, gender*marital status, marital status, age group, stag e*age group, and race Section 3: Multinomial Logistic Regression Assessment of Fit The comparison of the residual analysis results for the multinomial logistic regression model s for the individual treatment are given next three tables ( Figures 11 13 ) The results are given for the treatment groups of radiation therapy, chemotherapy, no treatment, radiation + surgery, radiation + chemotherapy, surgery + chemotherapy, and radiation + chemotherapy + surgery. The Pearson residual is the residual divided b y the variance for a particular observation and is the individual contribution to the Pearson Chi Square statistic. The deviance residuals are a measure of the amount of deviance

PAGE 206

189 contributed by the individual observation. In each distribution, the residu als are centered about zero, do not demonstrate a distinctive trend, and are similar for each treatment outcome. From the observed data of the residual patterns, the determination was made that the multiple logistic regression model was appropriate as a model for the research lung cancer data set. Figure 11: Pearson and Deviance Residual Analysis Treatment Groups I, II, III with Treatment Group VIII (Surgery) as Reference

PAGE 207

190 Figure 12: Pearson and Deviance Residual Analysis Treatmen t Groups IV and V with Treatment Group VIII (Surgery) as Reference

PAGE 208

191 Figure 13: Pearson and Deviance Residual Analysis Treatment Groups VI and VII with Treatment Group VIII (Surgery) as Reference Section 4: The Random Effect Component The statistic al testing and analysis used in testing Hypothesis I also included a multinomial logistic regression model with a random effects component to investigate any random effect of state may have had on the model results. The method used in SAS was the Proc Gli mmix procedure that utilizes statistical modeling approach to account for of the model for the eight states in the cancer data set and estimates the variance of those

PAGE 209

192 a djustments separately for each level of the response variable, treatment group. In the Proc Glimmix procedure, the overall random effect of sta te was evaluated and in Table 5 2 the estimates of the variance are given. Because the variances do not demonst rate a wide range of variability, the random effect of eight states with respect to which treatment the lung cancer cases received meaning the heterogeneity (differences) was minimal. Table 5 2 : Random Effect of State Covariance Parameter Estimates: Inte rcept Method Subject Random Effect Group Treatment Group Estimate Standard Error State Radiation 0.340 0.175 State Chemotherapy 0.247 0.126 State No Treatment 0.094 0.050 State Radiation + Chemotherapy 0.091 0.060 State Radiation + Surgery 0. 331 0.168 State Surgery + Chemotherapy 0.101 0.061 State Radiation + Chemotherapy + Surgery 0.024 0.016 Note: Surgery is the reference treatment group ull multinomial logistic regression model (Table 46) with the 3 ) generated with a random effect of state; there were no significant differences in the p values of the independent and interaction terms. Therefore, evaluating the Type III tests

PAGE 210

193 for the model with and without the random effect, as the p values did not change, the overall conclusions drawn during the Hypothesis I testing could be that the random effect is not influenci ng the conclusions but the random effect should be account for/assessed. Table 5 3 : Type 3 Analysis of Effects Main Effects and Interaction Terms* Multinomial Logistic Regression Model (MLR 2 ) (Random Effect of State) Effect DF F Value Pr > F Gender 7 1.24 0.2739 Morphology 21 71.78 <.0001 Gender*Morphology 21 1.45 0.0851 Grade 21 21.00 <.0001 Stage 21 150.00 <.0001 Gender*Stage 21 2.16 0.0015 Stage*Grade 63 2.77 <.0001 Marital Status 28 12.76 <.0001 Gender*Marital Status 28 1.70 0.0115 Age Gr oup 28 39.88 <.0001 Stage*Age Group 84 2.67 <.0001 Race 14 5.40 <.0001 Note: Age Group and Marital Status at defined on /at Date of Diagnosis indicates interaction term Section 5: Overall Effect of Int eraction on the Outcome By convention, t he Odds Ratios (ORs) for the main effects and interaction terms are reported in the literature as a statistic used to evaluate the effect of the independent variables on the outcome of interest In this section, th e overall effect of interaction on the outcome of interest, treatment type received was also examined. In this research, the

PAGE 211

194 ORs for the interaction effect on the treatment received was also evaluated in order to perform a more complete assessment of the relationship between the overall effect variables with the outcome In determining the overall effect on the outcome, main effects and statistically significant interaction term variables that contained the variables of interest were included. In the equ ation below, the expression used to determine the overall interaction effect from the full model (variables listed in Table 46) is given as: Logit (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 gender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI + 10 morphologyI + 1 1 raceI + others O ver the next section and in Appendix IV the method use d to calculate the Odds Ratios and 95% Confidence Internals are given for the variable combinations of gender, stage, marital status, grade and age group T he summary of the overall effect variable combinations are given in Table 54. The statistically sign ificant results for the overall interaction effects are given in Tables 5 5 through 57 b.

PAGE 212

195 Table 54: Overall Variable Effect on Lung Cancer (LC) Treatment Received Overall Interaction Effect Variable Interaction Term Odds Ratios Gender Gender 1 *Stag e I OR = exp ( 1 + 6 ) Gender 1 *Marital Status 1 OR = exp ( 1 + 7 ) Stage Stage I *Gender 1 OR = exp ( 2 + 6 ) Stage 1 *Grade I OR = exp ( 2 + ) Stage I Age Group 1 OR = exp ( 2 + ) Marital Status Marital Status 1 *Gender 1 OR = exp ( 3 + 7 ) Grade Grade 1 *Stage I OR = exp ( 4 + 8 ). Age Group Age Group 1 *Stage I OR = exp ( 5 + 9 ) Note: Considering only G e nder 1 Stage I, Grade I, Marital Status 1 Age Group 1 in this example ; w here Gender 1 = female, Age Group 1 = Age Group IV and Marital Status 1 = Single As previously discussed, a multinomial logistic regression model was utilized having eight treatment groups (Y) S urgery was the reference treatment group; therefore there were seven possib le outcome categories or levels. Also, there were four categories for stage (Stage IV = reference) and five categories for marital status (Marital Status V (Widowed) = reference ). Odds Ratios were calculated with the 95% Confidence Intervals for one of the treatment groups I n the equations below the treatment group is radiation ( with surgery as the reference ) with gender = 1 for females and gender = 0 for males given for s tageI and m arital _s tatu s I The effect of gender on the probability of receiving radiation therapy as a treatment, given that the patient is at stageI, is determined as: Female : Logit (Y=Radiation|gender=1, stageI) = + 1 + 2 stageI + 3 marital_statusI + 4 *stageI + 5 mar ital_statusI Male : Logit (Y=Radiation|gender=0, stageI) = + 2 stageI + 3 marital_statusI By subtracting the Logit for males from Logit for females the following equation

PAGE 213

196 is given as : Logit (Y=Radiation |gender=1, stageI) = 1 + 4 *stageI + 5 marital_statusI Next at the variable stageI which is coded as 1 for stage I and 0 for stage IV (reference), the results are: Logit (Y=Radiation |gender=1, stageI=1) = 1 + 4 + 5 marital_statusI Logit (Y=Radiation |gender=1, stageI=0) = 1 + 5 marital_statusI Marital_statusI appears in both logits whether stageI is 1 or 0. In other words it does not matter whether the patient is single or married. This can also be stated as when the interaction between gender and stageI is exa mined marital_status I is fixed or controlled for. For estimating the effect of gender (female as compared to male) on the probability of receiving radiation treatment, given that the patient is at stage I (stage=1) and after adjusting for marital_status I the resultant equation for the Odds Ratio is given as : OR = exp ( 1 + 4 ). The Odds Ratio s were calculated by exponentiating the beta coefficients for gender (female) plus the beta coefficient for the interaction term of gender (female)*stage I The 95% lower (LCI) and upper (UCI) confidence intervals were calculated wi th the following equation: LCI, UCI = exp (( 1 + 4 ) + (1.96* [(var ( 1 ) +var ( 4 ) + 2covar ( 1 4 )] 0.5 )) For each of the six other possible outcomes (treatment group s ) remaining, each outcome (chemotherapy, no treatment, radiation + surgery, radiation + chemotherapy, surgery + chemotherapy, radiation + chemotherapy + surgery) would have twenty possible results (ORs and 95% CIs) based on the level of stage for a total of 21 possible

PAGE 214

197 outcomes As shown in Table 5 5 on the next page, females with stage I as compared with males with stage I are 1.71 times more likely to receive a combination of chemotherapy and radiation therapy versus receiving surgery alone (OR = 1.71, 95%CI 1.06 2.78) after adjustment. In other words for patients with stage 1 lung can cer, females are 1.71 times more likely to receive a combined chemotherapy and radiation therapy than males do. After adjustment, stage 3 females as compared to males with stage 3 were 1.85 times more likely to receive radiation therapy in combination wit h chemotherapy versus receiving surgery alone for the treatment of lung cancer (OR = 1.854, 95%CI 1.151 2.986). The results for females with stage II versus males with stage II lung cancer were not statistically significant. Also, none of the other six treatment types demonstrated statistically significant results. Table 5 5 : Interaction E ffect of Gender on LC Treatment Received Gender and Gender*Stage Treatment Group 5 (Radiation Therapy in Combination with Chemotherapy) Odds Ratio 95% LCI 95% UC I Females with Stage I (Treatment Group 5) 1.71 1.06 2.78 Females with Stage II (Treatment Group 5) 1.46 0.88 2.41 Females with Stage III (Treatment Group 5) 1.85 1.15 2.99 Reference for Gender = Males, Reference for Treatment Group = Surgery Reference for Stage = Stage IV

PAGE 215

198 Next, t he effect of gender on the probability of receiving radiation therapy as a treatment, given that the patient is at marital_status I, is determined as: Female : Logit (Y=Radiation|gender=1, marital_statusI) = + 1 + 2 stageI + 3 marital_statusI + 4 *stageI + 5 marital_statusI Male : Logit (Y=Radiation|gender=0, marital_statusI) = + 2 stageI + 3 marital_statusI By subtracting the Logit for males from Logit for females the following equation is given as : L ogit (Y=Radiation |gender=1, stageI) = 1 + 4 *stageI + 5 marital_statusI A t the variable marital_statusI which is coded as 1 for marital_statusI and 0 for marital_status I ( V = reference), the results are: Logit (Y=Radiation |gender=1, marital_status I =1) = 1 + 4 *stageI + 5 Logit (Y=Radiation |gender=1, marital_statusI =0) = 1 + 4 *stageI Thus, from above stageI appears in both logits whether marital_statusI is 1 or 0. That is, it does not matter whether the patient is stage 1 o r stage III. Another way of stating this fact is that when the interaction between gender and marital_statusI is looked at stageI is fixed or controlled for. For estimating the effect of gender (female as compared to male) on the probability of receivin g radiation treatment, given that the patient is at marital_statusI and after adjusting for stageI (stage=1) the resultant equation for the Odds Ratio is given as: OR = exp ( 1 + 5 ) For each of the six other possible outcomes (treatment groups) remaining, each outcome (chemotherapy, no treatment, radiation + surgery, radiation + chemotherapy, 5%

PAGE 216

199 CIs based on the level of marital status for a total of 28 possible outcomes The overall gender effect from gender and the interaction term of gender*marital status was also calculated for the treatment received. There were no statistically significa nt results. Overall Effect of Stage on the Treatment Received In the next three tables (Tables 56 a, 56 b, 56 c), the statistically significant ORs a, the interacti on between stage and gender is examined with grade and age group at the time of diagnosis being controlled for. As noted in the table, all the ORs demonstrate a decrease probability to receive a particular treatment for females as compared to males. Femal es as compared to males with stage 1 and stage 2 lung cancer are less likely to receive one of the seven treatment types. The ORs range from 0.008 to 0.137 for stage 1 and 0.023 to 0.929 for stage II lung cancer. After adjustment, females versus males wit h stage I lung cancer are 86.3% less likely to receive a combination of surgery and chemotherapy (OR = 0.137, 95% CI 0.103 0.782). For stage 2 lung cancer, females as compared to males are 73.1% less likely to receive surgery in combination with chemoth erapy (OR = 0.269, 95% CI 0.191 0.380). S tage 3 lung cancer females versus males after controlling for grade and age group, result in four treatment types (radiation, chemotherapy, no treatment and radiation in combination with chemotherapy) in which f emales can be up to 98% less likely to receive one of those four particular treatment s For example females versus males with stage 3 lung cancer are 91.6% less likely to

PAGE 217

200 receive radiation alone as their treatment for lung cancer (OR = 0.84, 95% CI 0.070 0.102) after adjustment. Also, contrary to the previous tables being presented with two significant figures, to see the variability between the statistics, the tables (Tables 56 a through Table 56 f 2 ) are presented to the third significant figure. Table 56 a: Interaction Effect of Stage on LC Treatment Received Stage and Stage*Gender Treatment Type (Outcome) Stage Gender OR 95% LCI 95% UCI Radiation I Female 0.019 0.016 0.023 Chemotherapy 0.008 0.006 0.009 No Treatment 0.019 0.016 0.022 R adiation + Surgery 0.074 0.054 0.103 Radiation + Chemotherapy 0.013 0.011 0.015 Surgery + Chemotherapy 0.137 0.103 0.182 Radiation +Surgery + Chemotherapy 0.032 0.022 0.046 Radiation II Female 0.029 0.022 0.037 Chemotherapy 0.023 0.018 0.02 9 No Treatment 0.037 0.030 0.044 Radiation + Surgery 0.329 0.235 0.461 Radiation + Chemotherapy 0.047 0.038 0.058 Surgery + Chemotherapy 0.269 0.191 0.380 Radiation +Surgery + Chemotherapy 0.189 0.133 0.267 Radiation III Female 0.084 0.07 0 0.102 Chemotherapy 0.092 0.077 0.109 No Treatment 0.092 0.078 0.108 Radiation + Chemotherapy 0.019 0.016 0.023 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Refe rence for Stage = Stage IV Reference for Gender = Males

PAGE 218

201 The statistically significant ORs and 95%CIs for Table 56 b 1, 56 b 2, and 56 b 3 are given for the effect of stage on the probability of receiving a particular lung cancer given the lung cancer case is at a specific grade. After adjustment, for stage 1 grade 1 lung cancer cases as compared to stage 4 grade 4 lung cancer cases, stage 1 grade 1 lung cancer cases are 98.9% less likely to receive chemotherapy alone (OR = 0.011, 95% CI 0.007 0.017) or 88.1% less likely to receive radiation in combination with surgery (OR = 0.119, 95% CI 0.050 0.286). For grade 2 stages 1 though 3 (Table 56 b 2), once eff 1 grade 2 and stage 2 grade 2, there are seven treatment types given as possible outcomes with four treatment types (radiation chemotherapy no treatment and radiation in combination with chemotherapy) for stage 3 grade 2. In Table 56 b 3, there are seven treatment type outcomes for stage 1 grade 3 and stage 2 grade 3 and five treatment b 1through 56 b 3 there is a decrease probability to receive the particular treatment type listed, in other words for effect of stage does not increase the probability of receiving a treatment type. Also as the stage of the lung cancer patient increases there is a trend of increasing ORs.

PAGE 219

202 Table 56 b 1: Overall Interaction Effect of Stage on LC Treatment Received Stage and Stage*Grade (Grade I ) Treatment Type (Outcome) Stage Grade OR 95% LCI 95% UCI Radiation I I 0.025 0.017 0.037 Chemotherapy 0.011 0.007 0.017 No Treatment 0.032 0.024 0.045 Radiation + Surgery 0.119 0.050 0.286 Radiation + Chemotherapy 0.012 0.008 0.020 Surgery + Chemotherapy 0.126 0.067 0.238 Radiation +Surgery + Chemotherapy 0.070 0.028 0.172 Radiation II I 0.047 0.027 0.0 83 Chemotherapy 0.028 0.015 0.053 No Treatment 0.050 0.032 0.076 Radiation + Chemotherapy 0.061 0.037 0.101 Surgery + Chemotherapy 0.312 0.146 0.668 Radiation +Surgery + Chemotherapy 0.150 0.044 0.503 Radiation III I 0.221 0.138 0.352 Ch emotherapy 0.129 0.077 0.215 No Treatment 0.184 0.123 0.274 Radiation + Chemotherapy 0.474 0.314 0.716 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for S tage = Stage IV, Reference for Grade = Grade IV

PAGE 220

203 Table 56 b 2: Interaction Effect of Stage on LC Treatment Received Stage and Stage* Grade (Grade II ) Treatment Type (Outcome) Stage Grade OR 95% LCI 95% UCI Radiation I II 0.016 0.013 0.019 Chemotherapy 0.005 0.004 0.007 No Treatment 0.018 0.015 0.021 Radiation + Surgery 0.081 0.057 0.116 Radiation + Chemotherapy 0.010 0.008 0.013 Surgery + Chemotherapy 0.095 0.068 0.133 Radiation +Surgery + Chemotherapy 0.038 0.025 0.058 Radiation II II 0.039 0.030 0.050 Chemotherapy 0.023 0.017 0.031 No Treatment 0.037 0.030 0.046 Radiation + Surgery 0.340 0.234 0.495 Radiation + Chemotherapy 0.051 0.040 0.065 Surgery + Chemotherapy 0.277 0.190 0.402 Radiation +Surgery + C hemotherapy 0.247 0.165 0.369 Radiation III II 0.086 0.069 0.106 Chemotherapy 0.065 0.052 0.082 No Treatment 0.077 0.064 0.094 Radiation + Chemotherapy 0.183 0.151 0.222 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confi dence Interval Reference for Treatment Group = Surgery Reference for Stage = Stage IV Reference for Grade = Grade IV

PAGE 221

204 Table 56 b 3: Interaction Effect of Stage on LC Treatment Received Stage and Stage* Grade (Grade III) Treatment Ty pe (Outcome) Stage G r ade OR 95% LCI 95% UCI Radiation I III 0.016 0.014 0.019 Chemotherapy 0.005 0.004 0.006 No Treatment 0.017 0.015 0.020 Radiation + Surgery 0.069 0.050 0.094 Radiation + Chemotherapy 0.010 0.009 0.012 Surgery + Chemotherap y 0.169 0.128 0.224 Radiation +Surgery + Chemotherapy 0.039 0.028 0.054 Radiation II III 0.023 0.018 0.028 Chemotherapy 0.019 0.015 0.024 No Treatment 0.030 0.025 0.036 Radiation + Surgery 0.315 0.233 0.426 Radiation + Chemotherapy 0.041 0.034 0.049 Surgery + Chemotherapy 0.298 0.215 0.414 Radiation +Surgery + Chemotherapy 0.237 0.178 0.316 Radiation III III 0.072 0.061 0.085 Chemotherapy 0.072 0.061 0.084 No Treatment 0.081 0.070 0.095 Radiation + Surgery 0.665 0.513 0.863 Radiation + Chemotherapy 0.162 0.139 0.188 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Stage = Stage IV, Reference for Grade = Grade IV

PAGE 222

205 Ne xt in Tables 56 c 1 through 56 c 3, the effect of stage on the probability of the treatment received given the patient is at age group I are given. For stage 1, there are seven possible outcomes with ORs ranging from 0.003 to 0.324 for age group 4, six ou tcomes for age group 5 (ORs range from 0.007 to 0.084), and 7 possible treatment outcomes for age group 6 and 7. Note that these results are based on comparing the reference group of surgery for the treatment type, age group 8 as the reference for the age group at the time of diagnosis and stage 4 as the reference group for stage. Table 56 c 2 gives the statistics for stage 2 and age groups 4 through 7. There are 5 treatment outcomes dependent upon the effect of stage (Stage 2 and age groups 4, 5, and 6) and four treatment groups for stage 2 in age group 7 with the ORs ranging from 0.038 to 0.080. Stage 2 age group 7 lung cancer cases have a 92.0% less likely probability to receive radiation therapy alone (OR = 0.080, 95% CI 0.35 = 0.182) versus receiving surgery after controlling for gender and grade. For stage 3 there are five possible treatment outcomes for age group 4 and 5 with age group 6 and 7 having four treatment types given with statistically significant ORs (Table 56 c 3). For age groups 6 an d 7, the treatment types are radiation, chemotherapy, no treatment and radiation in combination with chemotherapy. The effect of stage on the treatment type received was to decrease the probability of receiving the treatment given in the tables It is in teresting to report that the effect of stage does not increase the probability of receiving a particular treatment type given stage 1, 2 or 3

PAGE 223

206 Table 56 c 1: Interaction Effect of Stage on LC Treatment Received Stage and Stage*Age Group (Stage I ) Treatm ent Type (Outcome) Age Group Stage Odds Ratio 95% LCI 95% UCI Radiation 4 I 0.003 0.001 0.010 Chemotherapy 0.004 0.002 0.011 No Treatment 0.016 0.007 0.033 Radiation + Surgery 0.079 0.018 0.354 Radiation + Chemotherapy 0.010 0.004 0.022 Surge ry + Chemotherapy 0.324 0.107 0.979 Radiation + Chemotherapy + Surgery 0.073 0.021 0.258 Radiation 5 0.008 0.004 0.017 Chemotherapy 0.007 0.004 0.013 No Treatment 0.017 0.009 0.030 Radiation + Surgery 0.075 0.021 0.271 Radiation + Chemothe rapy 0.018 0.010 0.032 Radiation + Chemotherapy + Surgery 0.084 0.030 0.235 Radiation 6 0.014 0.007 0.026 Chemotherapy 0.013 0.007 0.022 No Treatment 0.024 0.014 0.040 Radiation + Surgery 0.129 0.038 0.443 Radiation + Chemotherapy 0.027 0.016 0.047 Surgery + Chemotherapy 0.308 0.132 0.716 Radiation + Chemotherapy + Surgery 0.104 0.039 0.279 Radiation 7 0.026 0.014 0.049 Chemotherapy 0.014 0.008 0.025 No Treatment 0.024 0.014 0.041 Radiation + Surgery 0.130 0.038 0.445 Ra diation + Chemotherapy 0.045 0.027 0.076 Surgery + Chemotherapy 0.315 0.133 0.744 Radiation + Chemotherapy + Surgery 0.137 0.052 0.365 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Reference for Stage = Stage IV Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs )

PAGE 224

207 Table 56 c 2: Interaction Effect of Stage on LC Treatment Received Stage and Stage*Age Group (Stage II ) Treatment Type (Outcome) Age Group Stage Odds Ratio 95% LCI 95% UCI Radiation 4 II 0.009 0.003 0.034 Chemotherapy 0.012 0.004 0.040 No Treatment 0.024 0.010 0.060 Radiation + Chemothera py 0.014 0.005 0.038 Radiation + Chemotherapy + Surgery 0.092 0.011 0.739 Radiation 5 0.025 0.010 0.062 Chemotherapy 0.020 0.008 0.050 No Treatment 0.026 0.012 0.055 Radiation + Chemotherapy 0.025 0.011 0.058 Radiation + Chemotherapy + Sur gery 0.106 0.015 0.746 Radiation 6 0.043 0.018 0.098 Chemotherapy 0.037 0.015 0.087 No Treatment 0.037 0.018 0.076 Radiation + Chemotherapy 0.039 0.017 0.086 Radiation + Chemotherapy + Surgery 0.131 0.019 0.906 Radiation 7 0.080 0.035 0.1 82 Chemotherapy 0.042 0.018 0.099 No Treatment 0.038 0.019 0.077 Radiation + Chemotherapy 0.064 0.029 0.141 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Reference for Stage = Stage IV Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs )

PAGE 225

208 Table 56 c 3: Interaction Effect of Stage on LC Treatment Rece ived Stage and Stage*Age Group (Stage III ) Treatment Type (Outcome) Age Group Stage Odds Ratio 95% LCI 95% UCI Radiation 4 III 0.013 0.004 0.043 Chemotherapy 0.034 0.012 0.098 No Treatment 0.061 0.027 0.139 Radiation + Chemotherapy 0.048 0.019 0 .119 Radiation + Chemotherapy + Surgery 0.168 0.032 0.874 Radiation 5 0.034 0.016 0.073 Chemotherapy 0.055 0.025 0.119 No Treatment 0.066 0.034 0.126 Radiation + Chemotherapy 0.087 0.043 0.177 Radiation + Chemotherapy + Surgery 0.193 0.044 0.849 Radiation 6 0.058 0.029 0.115 Chemotherapy 0.101 0.050 0.205 No Treatment 0.094 0.051 0.172 Radiation + Chemotherapy 0.135 0.070 0.261 Radiation 7 0.109 0.056 0.212 Chemotherapy 0.116 0.058 0.232 No Treatment 0.096 0.053 0.175 Ra diation + Chemotherapy 0.222 0.116 0.426 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Reference for Stage = Stage IV Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs ) Overall Effect of Marital Status on the Treatment Received For the effect of marital status on the probability of receiving a particular treatment g iven that the patient is female, Table 56 d demonstrates statistically significant ORs and 95% CIs for married, separated and divorced patients. Females versus males that were married were 48.2% less likely to receive radiation (OR = 0.518, 95%CI 0.388 0.690) or 45.3% less likely to receive no treatment (OR = 0.547, 95% CI 0.433 0.691). Married females versus married males were 2.144 more likely to receive surgery in combination with chemotherapy after adjustment. Also divorced females as compared

PAGE 226

209 to males that were divorced were 1.738 time more likely to receive surgery in combination with chemotherapy. There was a probability of receiving no treatment for females versus males that were separated by as much as 72.8% (OR = 0.493, 95% CI 0.274 0.887 ) Table 56 d: Interaction Effect of Marital Status on LC Treatment Received Marital Status and Marital Status *Gender Treatment Type (Outcome) Gender Marital Status Odds Ratio 95% LCI 95% UCI Radiation Female Married 0.518 0.388 0.690 No Treatment 0 .547 0.433 0.691 Surgery + Chemotherapy 2.144 1.335 3.444 No Treatment Female Separated 0.493 0.274 0.887 Surgery + Chemotherapy Female Divorced 1.738 1.012 2.985 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confi dence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Reference for Marital Status = Widowed Overall Effect of Grade on the Treatment Received I n examining the effect of grade on the probability of receiving a particular t reatment type, for stage 1, 2, and 3, there is a decreased probability in all instances listed in Table 56 e 1. For stages 1, 2 and 3, there are seven possible treatment outcomes for grade 1 As the severity of the grade increase, there are less treatme nt types given as outcomes that are statistically significant For grade 2 stage 1, there are two statistically significant ORs/95% CIs for the treatment types of chemotherapy and radiation in combination with chemotherapy. In Table 56 e 3, the effect of grade 3 on the probability

PAGE 227

210 of the treatment type received are given. For grade 3 stage 1 there 1 treatment type listed that is statistically significant (radiation in combination with chemotherapy) and three treatment types for stage 2 grade 3 and stage 3 grade 3. The ORs for grade 3 range from a minimum 0.385 to a maximum of 0.570. Table 56 e 1: Overall Interaction Effect of Grade on LC Treatment Received Grade and Stage*Grade (Grade I ) Treatment Type (Outcome) Stage Grade Odds Ratio 95% LCI 95% UCI Ra diation I I 0.445 0.267 0.741 Chemotherapy 0.254 0.157 0.411 No Treatment 0.526 0.372 0.744 Radiation + Surgery 0.320 0.113 0.901 Radiation + Chemotherapy 0.127 0.080 0.202 Surgery + Chemotherapy 0.331 0.179 0.613 Radiation + Chemotherapy + Surgery 0.300 0.118 0.761 Radiation II 0.297 0.168 0.528 Chemotherapy 0.129 0.076 0.218 No Treatment 0.299 0.191 0.469 Radiation + Surgery 0.224 0.065 0.765 Radiation + Chemotherapy 0.107 0.066 0.173 Surgery + Chemotherapy 0.247 0.108 0 .565 Radiation + Chemotherapy + Surgery 0.158 0.053 0.472 Radiation III 0.327 0.187 0.572 Chemotherapy 0.130 0.079 0.211 No Treatment 0.295 0.191 0.457 Radiation + Surgery 0.190 0.056 0.644 Radiation + Chemotherapy 0.112 0.071 0.178 Surge ry + Chemotherapy 0.434 0.193 0.973 Radiation + Chemotherapy + Surgery 0.159 0.055 0.462 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Stage = Stage IV Reference for Grade = Grade IV

PAGE 228

211 Table 56 e 2: Overall Interaction Effect of Grade on LC Treatment Received Grade and Stage*Grade (Grade II ) Treatment Type (Outcome) Stage Grade Odds Ratio 95% LCI 95% UCI Chemotherapy I II 0.380 0.211 0.682 Radiation + Chemotherapy 0.210 0.118 0.373 Radiation II 0.496 0.313 0.786 Chemotherapy 0.193 0.128 0.290 No Treatment 0.431 0.312 0.596 Radiation + Chemotherapy 0.177 0.125 0.251 Surgery + Chemotherapy 0.343 0.197 0.600 Radiation + Chemotherapy + Su rgery 0.322 0.145 0.716 Radiation III 0.545 0.331 0.897 Chemotherapy 0.194 0.127 0.296 No Treatment 0.426 0.293 0.618 Radiation + Chemotherapy 0.186 0.126 0.273 Radiation + Chemotherapy + Surgery 0.326 0.139 0.765 Note: OR = Odds Ratio, LC I = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Stage = Stage IV Reference for Grade = Grade IV Table 56 e 3 : Overall Interaction Effect of Grade on LC Treatment Received Grade a nd Stage*Grade (Grade III ) Treatment Type (Outcome) Stage Grade Odds Ratio 95% LCI 95% UCI Radiation + Chemotherapy I III 0.434 0.246 0.765 Chemotherapy II 0.401 0.255 0.632 Radiation + Chemotherapy 0.366 0.245 0.547 Surgery + Chemotherapy 0.325 0.168 0.627 Chemotherapy III 0.403 0.278 0.584 Radiation + Chemotherapy 0.385 0.277 0.533 Surgery + Chemotherapy 0.570 0.331 0.982 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Gro up = Surgery Reference for Stage = Stage IV Reference for Grade = Grade IV

PAGE 229

212 Overall Effect of Age Group on the Treatment Received In the next two tables, Tables 5 7 a and 57 b the effect of age group of the probability of receiving a particular trea tment type given the patient is at stageI are given. These tables display statistics that show not only a decrease d probability of receiving a specific treatment but also an increased probability to receive a particular treatment based on the overall inte raction effect of age group For age group 4 and 5, there is an increased probability to receive surgery + chemotherapy as well as radiation + chemotherapy + surgery for all stages (stages 1, 2, and 3). For stage 1 age group 4 patients, there is a 5.716 times increased in the risk to receive surgery in combination with chemotherapy and a 7.975 increase risk to receive radiation in combination with surgery plus chemotherapy. Of all age groups, the maximum risk (OR = 11.377, 95% CI 3.387 38.214) to rece ive a particular treatment (radiation + chemotherapy + surgery) is for age group 4 stage 3 lung cancer patients. There is a decrease probability to receive radiation or no treatment for age groups 4, 5, 6, and 7 for all three stages of lung cancer (Tables 57 a and 57 b ) after adjustment. The ORs range from 0.044 for age group 6 stage 1 to 0.382 for age group 4 stage 3. Age group seven is the only age group that lists ORs that a show a decreased risk to receive a given treatment type. This could be indic ative of that as age patient s are not treated as aggressively.

PAGE 230

213 Table 57 a : Interaction Effect of Age Group on LC Treatment Received Age Group and Stage*Age Group (Age Group 4 and Age Group 5) Treatment Type (Outcome) Stage Age Group OR 95% LCI 95% UCI Radiation I 4 0.085 0.035 0.210 No Treatment 0.278 0.191 0.406 Surgery + Chemotherapy 5.716 2.930 11.150 Radiation + Chemotherapy + Surgery 7.975 2.419 26.285 Radiation II 4 0.225 0.118 0.429 No Treatment 0.299 0.171 0.525 Rad iation + Surgery 3.260 1.136 9.356 Surgery + Chemotherapy 7.375 2.851 19.082 Radiation + Chemotherapy + Surgery 9.164 2.650 31.688 Radiation III 4 0.382 0.219 0.667 No Treatment 0.426 0.255 0.712 Radiation + Surgery 5.636 2.084 15.245 Radia tion + Chemotherapy 2.313 1.313 4.074 Surgery + Chemotherapy 5.429 2.212 13.328 Radiation + Chemotherapy + Surgery 11.377 3.387 38.214 Radiation I 5 0.057 0.020 0.161 Chemotherapy 0.340 0.138 0.837 No Treatment 0.211 0.112 0.397 Radiation + Chemotherapy 0.409 0.195 0.859 Surgery + Chemotherapy 3.110 1.140 8.488 Radiation + Chemotherapy + Surgery 4.624 1.194 17.914 Radiation II 5 0.150 0.103 0.217 Chemotherapy 0.542 0.339 0.867 No Treatment 0.226 0.179 0.286 Surgery + Chemoth erapy 4.013 2.243 7.181 Radiation + Chemotherapy + Surgery 5.314 1.841 15.343 Radiation III 5 0.254 0.173 0.371 No Treatment 0.322 0.233 0.445 Radiation + Surgery 4.125 1.822 9.338 Surgery + Chemotherapy 2.954 1.427 6.116 Radiation + Chemot herapy + Surgery 6.597 2.162 20.129 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Reference for Stage = Stage IV Age Groups: 4 = ( > 40 < 50 yrs), 5 = ( > 50 < 60 yrs), 6 = ( > 60 < 70 yrs), 7 = ( > 70 < 80 yrs), 8 = ( > 80 < 90 yrs)

PAGE 231

214 Table 57 b : Interaction Effect of Age Group on LC Treatment Received Age Group and Stage*Age Group (Age Group 6 and Age Group 7) Treatment Type (Outcome) S tage Age Group OR 95% LCI 95% UCI Radiation I 6 0.044 0.016 0.122 Chemotherapy 0.252 0.104 0.611 No Treatment 0.187 0.101 0.345 Radiation + Chemotherapy 0.250 0.121 0.516 Surgery + Chemotherapy 2.915 1.111 7.650 Radiation II 6 0.115 0.072 0.1 84 Chemotherapy 0.402 0.233 0.692 No Treatment 0.201 0.141 0.288 Radiation + Chemotherapy 0.457 0.292 0.714 Surgery + Chemotherapy 3.762 1.793 7.892 Radiation III 6 0.196 0.152 0.251 No Treatment 0.286 0.241 0.340 Radiation + Chemotherapy 0.705 0.520 0.954 Surgery + Chemotherapy 2.769 1.579 4.856 Radiation + Chemotherapy + Surgery 3.002 1.063 8.476 Radiation I 7 0.045 0.016 0.125 Chemotherapy 0.206 0.085 0.500 No Treatment 0.220 0.120 0.403 Radiation + Chemotherapy 0.161 0 .078 0.331 Radiation II 7 0.119 0.075 0.188 Chemotherapy 0.329 0.192 0.564 No Treatment 0.236 0.166 0.335 Radiation + Chemotherapy 0.294 0.189 0.457 Radiation III 7 0.201 0.145 0.280 Chemotherapy 0.608 0.395 0.938 No Treatment 0.336 0.258 0.438 Note: OR = Odds Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Reference for Stage = Stage IV Age Groups: 4 = ( > 40 < 50 yrs), 5 = ( > 50 < 60 yrs), 6 = ( > 60 < 70 yrs), 7 = ( > 70 < 80 yrs), 8 = ( > 80 < 90 yrs)

PAGE 232

215 Hypothesis I Conclusion In conclusion, in testing Hypothesis I, the null hypothesis of no differences in treatment outcomes between men and women, was rejected as there were statistical ly significant results when the overall gender effect was examined. A multinomial logistic regression model was used to test Hypothesis I for differences between men and women with the same histological type, stage, and grade of lung cancer and the treatm ent they received When considering the random effect that state may have made on the conclusions, the reported estimates of the variances demonstrated that there was minimal heterogeneity due to the states with regard to which treatment modality the lun g cancer cases received. Also when comparing the p values for the Type III tests for the model with and without the random effect, the overall conclusions drawn during Hypothesis I testing did not change. Although it could be said the random effect was n ot influencing the overall conclusions, the random effect should be assessed for the possible impact on the resultant analyses. Lastly, when the overall interaction effect of gender stage, grade, marital status and age groups on the treatment received wa s examined, a complete assessment of the outcome was given For example, u tilizing gender and the interaction terms of gender*stage and gender*marital status, statistically significant ORs and 95% CIs were demonstrated for an increased or risk to receive a particular treatment modality based on gender and a specific stage classification

PAGE 233

216 Hypothesis II There is a statistically significant difference in survival between women with lung cancer as compared to the survival of men with lung cancer. Introduction and Survival Analysis During hypothesis testing, the purpose for Hypothesis 2 was to examine if there was or was not an association between gender and survival without adjustment for the other research covariates in the lung cancer data set. To test Hypothesis II and assess the relationship between gender groups and survival, a non parametric survival method, the Life Table (Actuarial) method was utilized. The l ife table method is appropriate for large data sets with grouped data and t he observation times are subdivided into intervals of fixed length. Table 5 8 lists the results of the survival distribution function generated by the Life Table method. Although three test statistics (the log rank, the Wilcoxon, and 2Log (LR)) were gen erated for the survival distribution function, the log rank test statistic was selected as the standard reporting statistic. This particular statistic gives equal weight to every lung cancer case death time. The other two statistics adjust for difference s in the survival distribution function depending upon the time of the death in the time, i.e. the event (death) will be weighted. The log rank test statistic was significant with a p value less than 0.0001 which could be interpreted as a difference exist ed in survival between men and women.

PAGE 234

217 Table 5 8 : Lung Cancer Survival Survival Distribution Function Testing the Equality over Gender Life Table Method Test Chi Square DF Chi Square P Value Log Rank 213.70 1 <.0001 DF = Degrees of Fr eedom The graph of the survival function or cumulative survival, S(t) for Life Table method is shown below in Figure 1 4 As displayed in the graph, with no adjustment, females had an increased probability of survival and survived longer than males. The shape of the curves for the survival probability for males and females in Figure 1 4 were females and males did overlay, this would suggest that there was no diff erence with a 4 any crossing of the gender specific survival curves could indicate changes in the survival patterns between males and females or possible interaction. Also shown in Figure 1 4 after 10 months, the difference between the curves were approximately parallel over the 5 year time period; this suggested that the proportionality assumption was not violated meaning the hazard or risk of death did not change over time between the male an d female lung cancer cases.

PAGE 235

218 Figure 14: Life Table Method Survival Distribution Function Table 5 9 displays the s ummary of the number of lung cancer case that lived (no event = censored) and the number of lung cancer cases that died (even t occurred = u ncensored ). During survival statistical testing it was important to evaluate the number of cases that died and the number of cases that lived in the data set because if the number of lung cancer cases that lived were disproportional between the two groups (females and males) the resultant statistics could be biased and subsequent interpretations for Hypothesis II could be limited. Table 59 gives the total number of females and male lung cancer cases in the lung cancer data set; overall 33.3 1% of the female lung cancer

PAGE 236

219 cases did not die and 25.47% of the male lung cancer cases did not die over the study 5 year time period. In Table 5 9 although fewer males lived as compared to females (6333 m ales vs. 6661 females ) when examining Table 5 9 a, th e differences in the number of females and males that lived in time intervals (1 month intervals) was not disproportionate. Overall, the survival time consisted of 60 time intervals of one month, examples (Table 5 9 a) for the early time intervals and lat er time intervals are given for number of cases that died (d), the number of cases that survived (c) and the effective number of lung cancer cases at risk (n) in each time interval, I (one month) Therefore, the difference between the number of female and male lung cancer cases that lived (c) would not impact or limit the interpretation of the survival results from Hypothesis II testing. Table 5 9 : Survival Data for Lung Cancer Cases Lung Cancer Distribution Summary Total Total Percent Gender Total Died Lived Lived Female 19994 13333 6661 33.31 Male 24869 18536 6333 25.47 Total 44863 31869 12994 28.96

PAGE 237

220 Table 5 9 a : Extracted Life Table Survival Parameter Results Life Table Model for the Lung Cancer Data Set Time Interval (months) Lower, Upper Number Failed (Died) Number Censored (Lived) Effective Sample Size d c n Females 0 1 2665 1436 19276.0 1 2 1310 656 15565.0 2 3 1082 452 13701.0 3 4 967 379 12203.5 4 5 821 363 10865.5 56 57 7 12 1926.0 57 58 4 11 1907.5 58 59 4 10 1893.0 59 60 1265 619 1574.5 Males 0 1 3672 1278 24230.0 1 2 1877 641 19598.5 2 3 1617 459 17171.5 3 4 1445 375 15137.5 4 5 1215 360 13325.0 56 57 7 3 1971.5 57 58 5 18 1954.0 58 59 9 11 1934.5 59 60 1394 526 1657.0 Note: Originally there were 60 time intervals of 1 month In Table 60 the results for the quartile estimates and the 95% confidence intervals for the survival probabilities with the mean survival times are given for males and female s lung cancer cases. The confidence intervals are reported because each estimate of the survival probability contains random variation resulting in an inherent imprecision. th percentile is of the greatest o f interest as it represents the median survival time. The median survival time is defined as the survival time for a cumulative survival function of 0.5. The median survival time for the

PAGE 238

221 the lung cancer cases survived; this would only be true if there were no censored observations prior to that with a 95% confidence interval of 7.46 to 7.95 months. The median survival time for males was 6.30 months with a 95% confidence interval of 6.17 to 6.44 months; this was 1.39 months less than the female median survival time. Also shown in Table 60 the mean survival time for females is 19.83 months whereas the me an (or average) survival approximately 3.01 months longer than the males. Table 60 : Gender Survival Estimates (in months) Summary Statistics of the Lung Cancer Di stribution Quartile Percent Point Estimate 95% LCI 95% UCI Mean Standard Error Female 75 34.29 31.56 37.41 50 7.69 7.46 7.95 19.83 0.19 25 2.63 2.53 2.73 Male 75 19.70 18.85 20.46 50 6.31 6.17 6.44 16.37 0.15 25 2.23 2.14 2.30 L CI = Lower Confidence Interval U CI = Upper Confidence Interval Another test statistic generated with the Life Table Method is the cumulative hazard function, CHF. The cumulative hazard function corresponds to the total number of deaths over an inte rval of time. In Figure 1 5 the graph of the cumulative hazard function is representative for the overall study time of 60 months. The x axis for the

PAGE 239

222 overall time is annotated in ten month time intervals which are further subdivided into one month survi val time intervals as noted on the graph as a circle (o) for females and a plus (+) for males. The cumulative hazard function illustrates the probability for the outcome of interest, death and how that probability changes with time. The cumulative probab ility of death increased for both females and males over the time interval in Figure 1 5 with males having an increase in the probability of death as compared to females. Figure 15: Cumulative Hazard Function for the Life Table Model Female and Males Lung Cancer Cases

PAGE 240

223 When examining the cumulative hazard function plot a comprehensive representation of how the hazard is changing over time cannot be completely ascertained. Survival data are not normal and therefore to view the cumulative ha zard function transforming the data with a logarithmic function allows for a more complete examination of the hazard between males and females. As shown in Figure 1 6 after the 6 transformation of the data, the hazard is shown to remain constant between ma les and females over the time under study (see Figure 16 ). This was important to examine (constant hazard) to ensure the assumption of proportionality was not violated, i.e. the probability of death was constant over the study time period between males an d females. Figure 16: Transformation of the Cumulative Hazard Function Female and Males Lung Cancer Cases

PAGE 241

224 Hypothesis II Conclusion In conclusion, for Research Question Two, the null hypothesis was rejected as there were s tatistically signifi cant difference s in gender specific survival W omen lung cancer cases had an increased probability of survival (a survival advantage) versus men with lung cancer I t is reported in the literature that there is a distinct survival advantage for women with lung cancer as compared to men with lung cancer T his result of increased survival for women was verified during Hypothesis II testing and analyses; these results are consistent with published literature Hypothesis III Wo me n with the same histological type, stage, grade of lung cancer, and the same treatment modality differ significantly in survival as compared to men with the same histological type, stage, and grade of lung cancer, and the same treatment modality. Introduction The thir d aim in testing the hypothesis of this research study was to expand the investigation of lung cancer treatment differences for females versus males in order to answer the research question of whether differences in treatment assignment based on gender alt ered survival. For lung cancer cases, females have been shown to have a distinct survival advantage relative to males 8 15 17 The statistical modeling approaches that demonstrate this survival advantage for females do not account for any effects due to moderating variables on the relationship or association between the independent variable

PAGE 242

225 of gender and survival The intent was to determine under which conditions females demonstrated or did not demonstrate a survival advantage as compared to males by investigating differences in gender specific survival for men and women lung cancer cases grouped or stratified by treatment modality, histologic type, stage, grade and other research covariates and by expanding the modeling approach to include interaction terms. The research statistical approach to test Hypothesis 3 was to employ Survival to that are censored (the research study event is not observed) and are not normal (lack of normali ty) due to censoring. Under the conditions described in this research, the lung cancer data from the eight cancer registries were right censored. Right censoring is defined as the non observance of the study event, i.e. death, during a specified time ran ge. During the specified time frame (1 1 2000 through 12 31 2004) under study, any non observed event (death) would classify that individual lung cancer case as censored. As stated in Chapter Three, the model selected to examine the relationship between survival and the covariates was the Cox Proportional Hazards model. This particular model is categorized as semi parametric as the baseline hazard is not specified but other assumptions such as time invariant covariates across the study period are assumed Time invariant covariates imply that the ratio of the hazard for any two observations is similar across the period of study. Prior to hypothesis testing with the Cox Proportional Hazards statistical model, the first step to answer Research Question Thr

PAGE 243

226 Meier Survival the data set (N =44,863) preliminary analyses 3 This initial or preliminary testing was done to evaluate the survival function and shape of the survival curves for each covariate over survival time. The survival curves for the groups or strata of the individual covariates were utilized to examine the proportionality between the strata (groups) for each variable. When the groups or strata for the independent variable are proportional, the curves of th e survival function graphs between the strata appear approximately parallel (the lines of the graphs do not diverge or do not cross). After evaluating the survival function curves between gender vs. survival time, stage vs. survival time, grade vs. surviv al time, and morphology vs. survival time for the lung cancer cases diagnosed over the 5 year study period, it was determined that utilization of the Cox Proportional Hazard model was appropriate as the proportionality assumption held for gender, stage, gr ade, and morphology over survival time. There were some overlapping survival curves for independent variables of age group at diagnosis, marital status at diagnosis, race, and treatment group over time which could have been problematic as the assumption o f proportionality (constant hazard) could possibly be violated making the Cox Proportional Hazard model inappropriate to use with the lung cancer data set. The initial non proportionality concern was addressed later in this section with the variables (age group at diagnosis, marital status at diagnosis, race, 3 Marubini and Maria Grazia Valsecchi; section 3.3.2., page 54.

PAGE 244

227 and treatment group) being tested via residual analysis (results are shown in Figures 13 and 14). During the initial phase of univariate survival analysis, the statistics generated by Test of E quality over Strata for each independent categorical variable. The lifetest procedure generated log rank test statistic, with the p values (<0.0001) being significant for gender, stage, grade, morpho logy, race, marital status at diagnosis, age group at diagnosis, and treatment values are significant. A possible limi tation of the statistical analysis for this data set for the highly significant p values may not just be a result of a true null hypothesis (no difference) but rather may be attributed to the extremely large sample size (N = 44,863) influencing the statist ics. Table 61: Life Table Test of Equality over Strata Parameter Test Chi Square Pr > Chi Square Gender Log Rank 206.90 < 0.0001 Stage Log Rank 50 2.70 < 0.0001 Grade Log Rank 844.80 < 0.0001 Morphology Log Rank 417.80 < 0.0001 The next pha se in testing Hypothesis 3 was assessing the research question by fitting the data to the Cox Proportional Hazard model which is a semi parametric

PAGE 245

228 the model build or selec tion of the covariates and interaction terms to be included in the and backward elimination of the variable or interaction term if the significance level of 0.05 was n ot met. Four variables (gender, stage, grade, and morphology) were coded to remain in the model during the stepwise procedure without having to meet the entry and exit specifications of 0.05 as these were the primary variables under study. At the complet ion of the stepwise procedure, the final model was assessed by examining the p values for the main effect variables and the variable combinations for the interaction terms. Interaction terms were included in the final model so the effect of moderating var iables that could impact the relationship between the independent variables and survival were identified. Below in Table 61 a the results of the Type III testing for the final model are given. As shown Table 61 a there were a total of six main effects variables and ten interaction terms.

PAGE 246

229 Table 61 a : The Cox Proportional Hazards Model (CPHM 1 ) Type 3 Tests Final Model Effect DF Wald Chi Square Pr > ChiSq Gender 1 34.54 <.0001 Morphology 3 17.21 0.0006 Gender Morphology 3 8.61 0.035 0 Grade 3 3.81 0.2823 Grade Morphology 9 29.34 0.0006 Stage 3 92.72 <.0001 Stage Morphology 9 20.69 0.0141 Age Group 4 64.69 <.0001 Stage Age Group 12 22.99 0.0278 Race 2 4.14 0.1259 Treatment Type 7 20.54 0.0045 Gender Treatment Type 7 23.01 0.0017 Treatment Type Morphology 21 73.18 <.0001 Treatment Type Grade 21 61.24 <.0001 Treatment Type Stage 21 147.33 <.0001 Treatment Type Age Group 28 104.76 <.0001 Treatment Type Race 14 34.35 0.0018 Note: Age Group at defined on /at Date of Diagnosis indicates interaction term

PAGE 247

230 The extracted equation or expression from the Cox Proportional Hazards m odel would be: Hazard Ra te = exp ( 1 genderI + 2 morphologyI + 3 gradeI + 4 stage I + 5 age_groupI + 6 raceI + 7 treatment_typeI + 8 genderI morphologyI + 9 gradeI morphologyI + 10 stageI morphologyI + 11 stageI age_groupI + 12 genderI*treatment_typeI + 13 treatment_type I morphologyI + 14 treatment_typeI grade I + 15 treatment_typeI* stageI + 16 treatment_typeI* age_groupI + 17 treatment_typeI*raceI + others ) T he next three sections for Hypothesis III include 1) the interaction terms analysis, 2) the residuals analy sis (model assessment), and 3) gives the results for the overall interaction effect on the probability of survival. It was important to include the interaction terms analysis so a comparison of the results could be made to the overall interaction effect o n the outcome. Section 1: Cox Proportional Hazards Model Interaction Terms In the next six tables (Tables 62 67), the results of the final Cox Proportional Hazards Model (CPHM 1 ) are given for the interaction terms extracted from the full model that cont ained stati stically significant results. When there were association s between the independent variable and the outcome of interest, survival, the relationship

PAGE 248

231 varied at different levels dependent upon the effect modifier Interaction terms included in th e statistical model (when appropriate) allowed for the opportunity to examine a more complete overview of the relationship between an independent variable and the outcome and how that relationship changes based on the moderating variable (effect modifier). As stated previously during Hypothesis II testing, in the majority of the currently published literature of gender specific survival gender is evaluated as a main effect without interaction terms The survival estimates based on the information present ed in that literature are reported favorably for women relative to men 8 15 17 In testing Hypothesis II, the results were consistent with the published literature that found females have a survival advantage over men. Part of the research investigation of gender s pecific survival was to verify that the results obtained from the lung cancer data set were consistent with other research results published in the current literature. The investigation of gender specific survival was expanded during Hypothesis III testin g. After adjustment for covariates, the relationship between gender specific survival and the treatment received was analyzed. As shown in the following tables (Tables 62 through 67), there were increased hazard or increase risk of death based on treatmen t received The overall gender effect reported later in this section on survival will include the results for the terms containing gender; therefore the statistics for Table 62 will not be reviewed here.

PAGE 249

232 Table 62: Hazard Ratios and 95% Confidence Inter vals Interaction Term of Treatment Group and Gender 1 ) Interaction Terms ** Hazard Ratio 95% LCI 95% UCI Gender (Moderator) Treatment Group Female Radiation 1.18 1.09 1.29 Female Che motherapy 1.16 1.07 1.26 Female No Treatment 1.13 1.05 1.21 Female Radiation + Surgery 1.13 0.96 1.34 Female Radiation + Chemotherapy 1.18 1.09 1.27 Female Surgery + Chemotherapy 1.11 0.93 1.33 Female Radiation + Surgery + Chemotherapy 1.07 0.92 1.25 Male* Surgery* 1.00* LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, = designates reference ** A djusted for morphology, gender morphology, grade, grade morphology, stage, stage morphology, age group, stage age group, race morphology treatment type, grade treatment type, stage treatment type, age group treatment type, and race treatment type The results displayed in Table 6 3 are for the interaction term of treatment group and stage; stage is the moderator in this interaction term extracted from the full CPH model. For the first treatment group of radiation therapy, lung cancer cases with stage 3 were at a 20.0% decreased risk for death than those lung cancer cases receiving surgery after controlling for gend er, morphology, gender morphology, grade, grade morphology, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, age group treatment type, and race treatment type. Th ere are no trends or overall significant findings demonstrated for an increased or decreased risk in this treatment group (radiation) based on stage 1 or stage 2 for this disease. In the case of chemotherapy (treatment group 2),

PAGE 250

233 there is a decreasing risk of death as stage increases. For stage 1 lung cancer cases receiving chemotherapy the risk of death was 1.68 times greater than those lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade, grade morpho logy, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, age group treatment type, and race treatment type. Stage 2 lung cancer cases receiving chemotherapy were 1.44 times more likely to die than those lung cancer cases receiving surgery after adjustment. Later stage lung cancer cases (stage 3) receiving chemotherapy were 7.0% less likely to die than those lung cancer cases receiving surgery after adjustment, although this was not statistically significant (HR = 0.93, 95% CI 0.81 1.06). After controlling for gender, morphology, gender morphology, grade, grade morphology, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, age group treatment type, and race treatment type, stage 1 (HR = 1.68, 95% CI 1.41 1.99) and stage 2 (HR = 1.44, 95% CI 1.19 1.74) lung cancer cases were shown to be at a greater risk or hazard for death as opposed to lung cancer cases receiving surgery alone These findings are of particular interest as clinically early stage disease is associated with a decreased hazard for death and later stage disease has decreased survival. Evaluation of the effects o f the moderator (stage) in the relationship between treatment and survival after adjustment in the full model for this data set demonstrated that early stage lung cancer cases receiving chemotherapy had decreased

PAGE 251

234 survival. For the next treatment group of (stage 1) that received no treatment were 16.0% more likely to die (HR = 1.16, 95 % CI 1.02 1.32) than those lung cancer cases receiving surgery alone after controlling for gender, morphology, gender morph ology, grade, grade morphology, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, age group treatment type, and race treatment type. The risk of death increased by 4 6.0% for stage 2 lung cancer cases as compared to other lung cancer cases (HR = 1.46, 95 %CI 1.25 1.70) versus receiving surgery alone after adjustment. Although the hazard ratio decreased for stage 3 disease, the result shown in Table 6 3 for stage 3 lu ng cancer cases receiving no treatment was not statistically significant (HR = 0.93, 95% CI = 0.82 1.06). Once again it would be expected that later stage disease would have increased mortality; this was association was not demonstrated for those lung c ancer cases in the no treatment group. The only statistically significant hazard ratio in treatment group 4 (radiation in combination with surgery) in Table 63, was for stage 3 lung cancer cases. Lung cancer cases receiving radiation plus surgery were 26 .0% less likely to die than those lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade, grade morphology, stage morphology, age group, stage age group, race, gender treatment type, morphology tr eatment type, grade treatment type, age group treatment type, and race treatment type. This result indicated that for

PAGE 252

235 later stage disease the treatment combination of radiation and surgery improved survival. For early stage disease (Stage 1 and 2) lung cancer cases receiving chemotherapy plus radiation did not demonstrate a decrease in the risk of death (Table 63). On the contrary, stage 1 lung cancer cases receiving this treatment combination were 43.0% more likely to die with stage 2 lung cancer cases being 31.0% more likely to die than those lung cancer cases receiving surgery after adjustment. The trend of increasing stage having a decreased risk for death for the radiation plus chemotherapy treatment group was further demonstrated as stage 3 h ad an 11.0% decrease in the risk for death versus lung cancer cases receiving surgery after adjustment. These results suggest that stage moderated the relationship between the treatment group and survival: earlier stage lung cancer cases have decreased su rvival when the treatment for the disease consists of radiation in combination with chemotherapy. For the last two treatment groups (surgery + chemotherapy and radiation + chemotherapy + surgery) in Table 63, the only significant hazard ratio was for stag e 1 lung cancer cases receiving all three treatment modalities of radiation, chemotherapy and surgery (HR = 1.79, 95% CI 1.31 2.44) as compared to lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade, g rade morphology, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, age group treatment type, and race treatment type. This treatment group (radiation + surgery + che motherapy) with stage 1 disease had the highest risk for death (79.0%) as compared to all the other treatment groups as shown in Table 63.

PAGE 253

236 Table 6 3 : Hazard Ratios and 95% Confidence Intervals Interaction Term of Treatment Group and Stage Extracted from 1 ) Interaction Terms ** Hazard Ratio 95% LCI 95% UCI Treatment Group Stage (Moderator) Radiation I 1.03 0.89 1.19 Chemotherapy I 1.68 1.41 1.99 No Treatment I 1.16 1.02 1.32 Radiation + Surgery I 1.28 0.97 1.68 Radiation + Chemotherapy I 1.43 1.22 1.66 Surgery + Chemotherapy I 1.25 0.95 1.64 Radiation + Surgery + Chemotherapy I 1.79 1.31 2.44 Surgery* IV 1.00 Radiation II 1.10 0.92 1.32 Chemotherapy II 1.44 1.19 1.74 No Treatment II 1 .46 1.25 1.70 Radiation + Surgery II 1.07 0.82 1.41 Radiation + Chemotherapy II 1.31 1.11 1.54 Surgery + Chemotherapy II 1.01 0.72 1.41 Radiation + Surgery + Chemotherapy II 1.27 0.95 1.71 Surgery* IV 1.00 Radiation III 0.80 0.69 0.92 Chemother apy III 0.93 0.81 1.06 No Treatment III 0.93 0.82 1.06 Radiation + Surgery III 0.74 0.59 0.92 Radiation + Chemotherapy III 0.89 0.78 1.01 Surgery + Chemotherapy III 0.90 0.71 1.14 Radiation + Chemotherapy III 1.31 1.11 1.54 Radiation + Surgery + Chem otherapy III 0.95 0.77 1.17 Surgery* IV 1.00 LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, = designates reference ** A djusted for gender, morphology, gender morphology, grade, grade morphology, stage morphology, age group stage age group, race, gender treatment type, morphology treatment type, grade treatment type age group treatment type, and race treatment type Table 64 exhibit s only one statistically significant hazard ratio for the interaction term o f treatment group and grade. Grade I lung cancer cases receiving chemotherapy had an increased risk of death with a hazard ratio of 1.51 (95% CI 1.12 2.03) relative to

PAGE 254

237 lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade morphology, stage, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, age group treatment type, and race treatment type. For this data set and as shown in Table 64, there was only one statistically significant hazard ratio which may have been due to chance alone versus being truly significant for the relationship between treatment modality of chemotherapy and decreased survival moderated by the g rade of disease. The hazard ratios and the 95% confidence intervals for the interaction term of treatment group and morphology are displayed in Table 65. Morphology moderated the relationship between the independent variable treatment group and the outcom e, survival. The finding that morphology acted as a moderator was consistent with clinical practices of treating a disease based on cell type with a particular treatment regimen. The statistics demonstrated that the treatment selection based on cell type could decrease survival or may not increase survival. There were only three treatment groups (radiation, chemotherapy, and radiation in combination with chemotherapy) with statistically significant results meaning for a particular treatment group the haza rd ratio and the confidence interval for that hazard ratio did not include 1. For each of those three treatment groups, the hazard ratios demonstrated an increased risk for death or decreased survival for the lung cancer cases. Lung cancer cases with ade nocarcinoma receiving radiation were 45.0% more likely to die (HR = 1.45, 95% CI 1.03 2.04) as compared to

PAGE 255

238 the other treatment groups with adenocarcinoma after controlling for gender, gender morphology, grade, grade morphology, stage, stage morphol ogy, age group, stage age group, race, gender treatment type, stage treatment type, grade treatment type, age group treatment type, and race treatment type. The hazard ratio increased by 65% (HR = 1.65, 95% CI 1.17 2.33) for squamous cell lu ng cancer cases receiving radiation with a hazard ratio for large cell carcinoma increasing by 61% (HR = 1.61, 95% CI 1.14 2.28) relative to lung cancer cases receiving surgery after adjustment. For adenocarcinoma lung cancer cases receiving chemotherap y, the risk of death increased by 42.0% than those lung cancer cases receiving surgery after adjustment. The hazard ratio for squamous cell lung cancer receiving chemotherapy was 1.51 meaning there was a 51.0% increase in the risk of death as compared to lung cancer cases receiving surgery after controlling for gender, gender morphology, grade, grade morphology, stage, stage morphology, age group, stage age group, race, gender treatment type, stage treatment type, grade treatment type, age gr oup treatment type, and race treatment type. Radiation in combination with chemotherapy demonstrated the same trend of decreased survival for the lung cancer morphological types of adenocarcinoma, squamous cell, and large cell carcinoma. Adenocarcino ma lung cancer cases receiving radiation plus chemotherapy were 48.0% (HR = 1.48, 95% CI 1.08 2.02) more likely to die with large cell lung cancer cases having a 62.0% increase risk for death (HR = 1.62, 95% CI 1.18 2.24) than those lung cancer cases r eceiving surgery after adjustment. According to the statistics, chemotherapy, radiation, and chemotherapy in combination

PAGE 256

239 with radiation decreased survival for all three lung cancer types. In Table 65, the hazard ratios for radiation in combination with s urgery was the one treatment group that did squamous, and large cell) as compared lung cancer cases receiving surgery but these results were not statistically significant as the 95% confidence intervals did include 1. For all other treatment groups (no treatment, surgery plus chemotherapy, and radiation in combination with surgery plus chemotherapy) listed in Table 65, the results were not statistically significant theref ore an association between the treatment group and survival moderated by morphology was not demonstrated.

PAGE 257

240 Table 6 4: Hazard Ratios and 95% Confidence Intervals Interaction Term of Treatment Group and Grade 1 ) Interaction Terms ** Hazard Ratio 95% LCI 95% UCI Treatment Group Grade (Moderator) Radiation I 1.03 0.76 1.41 Chemotherapy I 1.51 1.12 2.03 No Treatment I 1.05 0.81 1.38 Radiation + Surgery I 0.66 0.34 1.27 Radiat ion + Chemotherapy I 1.26 0.94 1.67 Surgery + Chemotherapy I 1.22 0.69 2.17 Radiation + Surgery + Chemotherapy I 0.92 0.5 1.71 Surgery* IV 1 Radiation II 0.93 0.71 1.23 Chemotherapy II 1.09 0.84 1.41 No Treatment II 1.06 0.84 1.36 Radiation + S urgery II 0.89 0.52 1.53 Radiation + Chemotherapy II 1.12 0.88 1.44 Surgery + Chemotherapy II 0.97 0.6 1.59 Radiation + Surgery + Chemotherapy II 0.97 0.61 1.54 Surgery* IV 1 Radiation III 0.92 0.7 1.2 Chemotherapy III 0.89 0.7 1.14 No Treatmen t III 1 0.79 1.26 Radiation + Surgery III 0.91 0.54 1.55 Radiation + Chemotherapy III 0.98 0.77 1.24 Surgery + Chemotherapy III 0.98 0.61 1.56 Radiation + Surgery + Chemotherapy III 0.82 0.52 1.28 Surgery* IV 1 LCI = Lower Confidence Interval, U CI = Upper Confidence Interval, = designates reference ** A djusted for gender, morphology, gender morphology, grade morphology, stage, stage morphology, age group, stage age group, race, gender treatment type, morphology treatment type, stag e treatment type, age group treatment type, and race treatment type

PAGE 258

241 Table 6 5: Hazard Ratios and 95% Confidence Intervals Interaction Term of Treatment Group and Morphology 1 ) Interaction Terms ** Hazard Ratio 95% LCI 95% UCI Treatment Group Morphology (Moderator) Radiation Adenocarcinoma 1.45 1.03 2.04 Chemotherapy Adenocarcinoma 1.42 1.04 1.95 No Treatment Adenocarcinoma 1.07 0.79 1.46 Radiation + Surgery Adeno carcinoma 0.72 0.31 1.65 Radiation + Chemotherapy Adenocarcinoma 1.48 1.08 2.02 Surgery + Chemotherapy Adenocarcinoma 1.03 0.61 1.72 Radiation + Surgery + Chemotherapy Adenocarcinoma 1.53 0.92 2.55 Surgery* Small Cell* 1 Radiation Large Cell 1.61 1.14 2.28 Chemotherapy Large Cell 1.43 1.03 1.97 No Treatment Large Cell 1 0.72 1.37 Radiation + Surgery Large Cell 0.82 0.35 1.91 Radiation + Chemotherapy Large Cell 1.62 1.18 2.24 Surgery + Chemotherapy Large Cell 0.97 0.57 1.66 Radiation + Surgery + Chemotherapy Large Cell 1.43 0.84 2.42 Surgery* Small Cell* 1 Radiation Squamous 1.65 1.17 2.33 Chemotherapy Squamous 1.51 1.1 2.07 No Treatment Squamous 1.16 0.85 1.58 Radiation + Surgery Squamous 0.84 0.36 1.93 Radiation + Chemotherapy Squam ous 1.63 1.19 2.23 Surgery + Chemotherapy Squamous 1.04 0.62 1.77 Radiation + Surgery + Chemotherapy Squamous 1.58 0.94 2.64 Surgery* Small Cell* 1 LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, = designates reference ** A djust ed for gender, gender morphology, grade, grade morphology, stage, stage morphology, age group, stage age group, race, gender treatment type, grade treatment type, stage treatment type, age group treatment type, and race treatment type The hazard ratios for the interaction term of treatment group and age group in Table 66 displayed that age was a moderator in the association between the treatment received and survival. Also in Table 66, the hazard ratios did not exhibit an increasing or

PAGE 259

242 decreasing trend between the type of treatment received and survival. For all levels of age group (4, 5, 6, and 7) lung cancer cases receiving radiation therapy as a treatment modality were at increased risk for death as compared to lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade, grade morphology, stage, stage morphology, stage age group, race, gender treatment type, stage treatment type, grade treatment type, morphology treatment type, and race treatment type. Lung cancer cases in age group 4 that received radiation were 1.38 times more likely to die than those lung cancer cases receiving surgery after adjustment. All other age groups, 5 (HR = 1.51, 95% CI 1.27 1.81), 6 (HR = 1.27, 95% CI 1.10 1.46), and 7 (HR = 1.19, 95% CI 1.05 1.36) were at increased risk for death versus lung cancer cases receiving surgery after adjustment. The risk of death increased by 22.0% for lung cancer cases in age group 5 receiving chemother apy (HR = 1.22, 95% CI 1.03 1.45) versus lung cancer cases receiving surgery after adjustment (Table 66). The variation of that risk was as small as 3.0% to a maximum risk for death of 45.0%. In Table 66, the no treatment group demonstrated a trend of decreasing risk of death with the HRs ranging from 1.65 to 1.23; as the lung cancer case became older the hazard ratio decreased but this was not statistically significant. The only other treatment group with a statistically significant hazard ratio was t he treatment group of radiation in combination with chemotherapy (Table 66). Lung cancer cases that received radiation in combination with chemotherapy were 31.0% more likely to die (HR = 1.31, 95% CI 1.11 1.56) relative to lung cancer

PAGE 260

243 cases receiving s urgery after adjustment. For all age groups in the other three treatment groups of 1) radiation plus surgery, 2) surgery in combination with chemotherapy, and 3) radiation plus surgery plus chemotherapy, the risk of death or hazard ratio was not statistic ally significant as compared to lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade, grade morphology, stage, stage morphology, stage age group, race, gender treatment type, stage treatment typ e, grade treatment type, morphology treatment type, and race treatment type. Age group moderated the relationship between treatment group and survival with the no treatment group overall having the highest risk of death. Possible explanations of th e no treatment group having the highest risk would include 1) those particular lung cancer cases did not receive one of the seven treatments for lung cancer because they were too sick for treatment or 2) the lung cancer case may have refused treatment for their disease. Of the last four treatment groups (Radiation + Surgery, Radiation + Chemotherapy, Surgery + Chemotherapy, and Radiation + Chemotherapy + Surgery) shown in Table 66, the only treatment group moderated by age group that demonstrated a statis tically significant hazard ratio was for the radiation in combination with chemotherapy treatment group. Lung cancer cases in age group 5 were at a 31.0 % increased risk versus lung cancer cases receiving surgery after adjustment. The last table (Table 67) in the Hypothesis 3 section 1 lists the hazard ratios for the interaction term of treatment group and race extracted from the full model. After controlling for gender, morphology, gender morphology, grade, grade morphology,

PAGE 261

244 stage, stage morpho logy, age group, stage age group, gender treatment type, morphology treatment type, grade treatment type, stage treatment type, and age group treatment type, white and black lung cancer cases receiving radiation had an increased hazard for de ath or decreased survival with white lung cancer cases having a 38.0% (HR = 1.38, 95% CI 1.04 1.84) increased risk for death and black lung cancer cases having a 27.0% (HR = 1.51, 95% CI 1.27 1.81) increased risk for death than other lung cancer case s receiving surgery. For lung cancer cases receiving chemotherapy, black lung cancer cases demonstrated the only statistically significant relationship between the treatment type and survival. After adjustment, black lung cancer cases receiving chemother apy alone were 22.0% more likely to die as compared to other lung cancer cases receiving surgery alone (HR = 1.22, 95% CI 1.03 1.45). For white lung cancer cases receiving no treatment there was a 65.0% increase in the risk of death (HR = 1.65, 95%CI 1. 28 2.14) and for black lung cancer cases receiving no treatment the hazard ratio was 1.60 (95%CI 1.37 1.87) as compared to other lung cancer cases receiving surgery after controlling for gender, morphology, gender morphology, grade, grade morpholog y, stage, stage morphology, age group, stage age group, gender treatment type, morphology treatment type, grade treatment type, stage treatment type, and age group treatment type. Once again as demonstrated in Table 66 (interaction term o moderator between treatment group and survival.

PAGE 262

245 Black lung cancer cases receiving a combination of radiation and chemotherap y had a 1.31 times increase in the risk of death with a hazard ratio varying as low as 11.0% to a maximum of 56.0% (95% confidence interval of 1.11 to 1.56) versus other lung cancer cases receiving surgery after adjustment. For all other treatment groups, i.e. radiation + surgery, surgery + chemotherapy, and radiation + surgery + chemotherapy, the results wer e not statistically significant. A lthough when just evaluating the hazard ratios for radiation in combination with surgery and chemotherapy, there wa s a decreased risk of death for both white (HR = 0.73, 95%CI 0.39 1.35) and black lung (HR = 0.72, 95%CI 0.41 1.27) cancer cases.

PAGE 263

246 Table 66: Hazard Ratios and 95% Confidence Intervals Interaction Term of Treatment Group and Age Group Extrac 1 ) Interaction Terms ** Hazard Ratio 95% LCI 95% UCI Treatment Group Age Group (Moderator) Radiation 4 1.38 1.04 1.84 Chemotherapy 4 1.1 0.84 1.43 No Treatment 4 1.65 1.28 2.14 Radiation + Surgery 4 0.93 0.56 1.54 Radiation + Chemotherapy 4 1.24 0.96 1.61 Surgery + Chemotherapy 4 0.99 0.57 1.72 Radiation + Surgery + Chemotherapy 4 0.73 0.39 1.35 Surgery* 8* 1 Radiation 5 1.51 1.27 1.81 Chemotherapy 5 1.22 1.03 1.45 No Treatment 5 1.6 1.37 1.87 Radiation + Surgery 5 1.23 0.87 1.73 Radiation + Chemotherapy 5 1.31 1.11 1.56 Surgery + Chemotherapy 5 1.05 0.67 1.65 Radiation + Surgery + Chemotherapy 5 0.72 0.41 1.27 Surgery* 8* 1 Radiation 6 1.27 1.1 1.46 Chemotherapy 6 1.1 0.95 1.28 No Treatment 6 1.53 1.35 1.73 Radiation + Surgery 6 1.04 0.77 1.41 Radiation + Chemotherapy 6 1.13 0.98 1.32 Surgery + Chemotherapy 6 1.01 0.66 1.54 Radiation + Surgery + Chemotherapy 6 0.75 0.43 1.3 Surgery* 8* 1 Radiation 7 1.19 1.0 5 1.36 Chemotherapy 7 1 0.87 1.15 No Treatment 7 1.23 1.1 1.38 Radiation + Surgery 7 1.07 0.8 1.44 Radiation + Chemotherapy 7 1.08 0.93 1.24 Surgery + Chemotherapy 7 0.88 0.57 1.34 Radiation + Surgery + Chemotherapy 7 0.64 0.37 1.11 Surgery* 8* 1 LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, = designates reference Age Group s: 4 ( > 40 < 50 years old) 5 ( > 50 < 60 years old) 6 ( > 60 < 70 years old) 7 ( > 70 < 80 years old ), 8 ( > 80 < 9 0 years old) ** A djusted for gender, morphology, gender morphology, grade, grade morphology, stage, stage morphology, stage age group, race, gender treatment type, morphology treatment type, grade treatment type, stage treatment type, and race treatment type

PAGE 264

247 Table 67: Hazard Ratios and 95% Confidence Intervals Interaction Term of Treatment Group and Race 1 ) Interaction Terms ** Hazard Ratio 95% LCI 95% UCI Treatment Group Race (Moderator) Radiation White 1.38 1.04 1.84 Chemotherapy White 1.1 0.84 1.43 No Treatment White 1.65 1.28 2.14 Radiation + Surgery White 0.93 0.56 1.54 Radiation + Chemotherapy White 1.24 0.96 1.61 Surgery + Chemotherapy White 0.99 0.57 1.72 Radiation + Surgery + Chemotherapy White 0.73 0.39 1.35 Surgery* Other* 1 Radiation Black 1.51 1.27 1.81 Chemotherapy Black 1.22 1.03 1.45 No Treatment Black 1.6 1.37 1.87 Radiation + Surgery Black 1.23 0.87 1.73 Radiation + Chemotherapy Black 1.31 1.11 1.56 Surge ry + Chemotherapy Black 1.05 0.67 1.65 Radiation + Surgery + Chemotherapy Black 0.72 0.41 1.27 Surgery* Other* 1 LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, = designates reference ** A djusted for gender, morphology, gender morphology, grade, grade morphology, stage, stage morphology, age group, stage age group, gender treatment type, morphology treatment type, grade treatment type, stage treatment type, and age group treatment type Section 2: Residuals The next two figures are the residual plots for the independent variables versus the difference between the actual value and the estimated value for all units or ind ividuals generated to assess the overall fit of the Cox Proportional Hazard model to the lung cancer data. When the requirements or assumptions for the semi parametric Cox

PAGE 265

248 proportional Hazards model are met, the model would be appropriate or correctly estimate the behavior of the data. Validating or corroborating the final Cox Proportional Hazard model results via residual analysis established that the initial non prop ortional covariates (age group at diagnosis, race, and treatment group) were independent of survival time for that period under study. If the residuals exhibited a trend, e.g. increased over time for the covariates of interest, the hazard ratio or relativ e risk could be overestimated (overestimation could lead to inflated coefficient estimates) and those covariates would not be time invariant across the study period. As the residual plots for the covariates over survival time for the 5 year study time did not demonstrate any trends, the use of the Cox Proportional Hazards model was appropriate. Residual analysis was also performed to evaluate the proportionality and constant hazard assumptions for the remaining covariates and interaction terms no trends were seen with residuals and the values fell about zero ( Figures 17 and 18 ) In Figure 1 7 the Schoenfeld residuals for the independent variable (gender) versus the log of survival time in months are displayed; the residuals produced during the statisti cal testing were weighted and smoothed over time. There is no trend of increasing or decreasing residual patterns for gender over the log of survival time meaning the requirement of time invariance held true; the model accurately estimated the lung cancer data behavior for females and males. As n oted on the x axis for Figure 17 time was transformed into the log of survival time due to the nature of survival data (non normalcy) due to the effects of censoring. If the transformation of survival time was n ot

PAGE 266

249 done, the residual pattern could be inappropriately displayed and interpreted incorrectly. Figure 1 7 : Residual Testing of the Lung Cancer Distribution Gender versus the Log of Time in Months The next figure (Figure 18 ) includes the S choenfeld residuals plots for all eight independent research variables. Although the displays in Figure 1 8 of the residual plots are minimized as compared to the display of the single variable, gender as shown in Figure 1 7 the intent was to illustrate an y overall trend in the residual plots for gender, stage, grade, morphology, race, age group, treatment group, and marital status versus the log of survival time. The residual patterns did not increase or decrease over time (no slope) and were centered abou t zero as expected; the difference on average between the estimated and actual data point for the residual should fall or be located along the zero

PAGE 267

250 axis value. There is no trend of increasing or decreasing residual patterns for the individual independent v ariables over the log of survival time meaning the assumption of proportionality and constant hazard was not violated and that the model accurately estimated the lung cancer data behavior for females and males. Figure 1 8 : Residual Tes ting for the Lung Cancer Distribution Independent Variables versus the Log of Time in Months Section 3 : Overall Effect of Interaction on Survival The final assessment for the overall effect on survivorship is presented in this section. Utilizing the SAS contrast Hazard Ratios and 95% Confidence Intervals were calculated for the variables of gender,

PAGE 268

251 morphology, stage, grade, race, and treatment type and the statistically significant interaction terms In the fol lowing equation, the variables that were evaluated for the overall effect are given. The variables given in the equation were extracted from the full (Table 61 a) that contained statistically significant interaction terms : Hazard Rate = exp ( 1 genderI + 2 morphologyI + 3 gradeI + 4 stage I + 5 age_groupI + 6 raceI + 7 treatment_typeI + 8 genderI morphologyI + 9 gradeI morphologyI + 10 stageI morphologyI + 11 stageI age_groupI + 12 genderI*treatment_typeI + 13 treatment_type I morphologyI + 14 treatment_typeI grade I + 15 treatment_typeI* stageI + 16 treatment_typeI* age_groupI + 17 treatment_typeI*raceI + others) In Table 68, the statistically significant Hazard Ratio c o mbinations for the statistically significant inter action terms are listed by gender, morphology, stage, grade, race, and treatment type. Following Table 68, an example of the method used to calculate the Hazard Ratios for the overall effect of gender on the probability or risk for death for a given morph ological lung cancer type is reviewed

PAGE 269

252 Table 68: Overall Effect on Survival Hazard Rat io Combinations Overall Effect Variable Interaction Term Hazard Ratios Gender Gender 1 *Morphology 1 HR = exp ( 1 + 8 ) Morphology Gender 1 and Treatment Type 1 Morphology 1 and Gender 1 Morphology 1 and Grade 1 HR = exp ( 1 + 12 ) HR = exp ( 2 + 8 ) HR = exp ( 2 + 9 ) Grade Grade 1 and Morphology 1 Grade 1 and Treatment Type 1 HR = exp ( 3 + 9 ) HR = exp ( 3 + 17 ) Stage St age 1 and Morphology 1 HR = exp ( 4 + 10 ) Stage 1 and Treatment Type 1 Stage 1 and Age Group 1 HR = exp ( 4 + 15 ) HR = exp ( 4 + 11 ) Age Group 1 and Treatment Type 1 Age Group 1 and Stage 1 HR = exp ( 5 + 16 ) HR = exp ( 5 + 11 ) Race Race 1 and Treatment Type 1 HR = exp ( + 17 ) Treatment Type Treatment Type 1 and Gender 1 Treatment Type 1 and Morphology 1 Treatment Type 1 and Grade 1 Treatment Type 1 and Stage 1 Treatment Type 1 and Age Group 1 Treatment Type 1 and Race 1 HR = exp ( 7 + 12 ) HR = exp ( 7 + 13 ) HR = exp ( 7 + 14 ) HR = exp ( 7 + 15 ) HR = exp ( 7 + 16 ) HR = exp ( 7 + 17 ) N ote: Considering only Gender 1 (female) Stage I, Grade I, Race I (white) Age Group 1 (Age Group IV ( > 40 < 50 yrs old) ) Morphology 1, Treatment Type 1 ( radiation ) in this example.

PAGE 270

253 Evaluating the Overall Effect of Gender on Survival From the equation below, the gender and the statistically significant interaction terms containing gender are identified (bolded). Hazard Rate = exp ( 1 genderI + 2 morphologyI + 3 gradeI + 4 stage I + 5 age_groupI + 6 raceI + 7 treatment_typeI + 8 genderI morphologyI + 9 gradeI morphologyI + 10 stageI morphologyI + 11 stageI age_groupI + 12 genderI*treatment_typeI + 13 treatment_typeI morphologyI + 14 treatment_typeI grade I + 15 treatment_typeI* stageI + 16 treatment_typeI* age_groupI + 17 treatment_typeI*raceI + others ) From the equation above the following equation results when examining of the overall effect o f gender on survival : Hazard Rate = exp ( 1 genderI + 2 morphologyI + 7 treatment_typeI + 8 genderI morphologyI + 12 genderI*treatment_ typeI) Gender and Morphology Female : HR (gender =1, morphologyI) = exp ( 1 + 2 morphologyI + 7 treatment_typeI + 8 morphologyI + 12 *treatment_typeI ) M ale : HR (gender =0, morphologyI) = exp ( + 2 morphologyI + 7 treatment_typeI ) Subtracting male from f emale given morphologyI, the following equation is given as: Female : HR (gender =1, morphologyI) = exp ( 1 + 8 morphologyI + 12 *treatment_typeI)

PAGE 271

254 Then looking at morphologyI for morphology 1 = 1 and morphologyI for morphology 4 = 0, the following equation is given as: Female : HR (gender =1, morphologyI = 1) = exp ( 1 + 8 + 12 *treatment_typeI) F emale : HR (gender =1, morphologyI = 0) = exp ( 1 + 12 *treatment_typeI) Note that any treatment type does not affect the outcome under the conditions as stated above The Hazard Ratio is then calculated for females as compared to males adjusti ng for morphology and controlling for treatment group as: HR = exp ( 1 + 8 ) In Table 69 a the effect of gender on the probability of survival based on the morphological lung cancer type demonstrate an increase survival for all three non small ce ll lung cancer types fo r females as compared to males controlling for treatment type. After adjustment, f emales versus males with large cell lung cancer are 25 % more likely to survive (HR = 0.75, 95% CI 0.7 0 0.81) whereas females with squamous cell carc inoma are 18% more likely to survive as compared to males with squamous cell carcinoma (HR = 0.8 2, 95% CI 0.76 0.87 ) The effect of gender on the probability of survival (Table 69 b) shows an overall increase in survival given the treatment type s of c hemotherapy, no treatment, radiation + chemotherapy, and radiation + chemotherapy + surgery. The hazard rati os range from a minimum of 0.83 to a maximum of 0.92. Females versus males receiving radiation in combination with chemotherapy and surgery are 17 % more likely to survive (HR = 0.83, 95% CI 0.7 2 0.97) and 8 % more likely to die when females as compared to males

PAGE 272

255 receive radiation plus chemotherapy after adjustment The statistics demonstrate that for a particular treatment combination statistically significant survivorship is exhibited for females versus males. Table 69 a: Overall Effect of Gender on Survival Hazard Ratios and 95% Confidence Intervals Gender and Gender*Morphology Gender Morphology HR 95% LCI 95% UCI Female Adenocarcinoma 0.76 0.71 0.80 Squamous Cell 0.82 0.76 0.87 Large Cell 0.75 0.70 0.81 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Gender = Male Reference for Morphology = Small Cell Lung Cancer Table 69 b: O verall Effect of Gender on Survival Hazard Ratios and 95% Confidence Intervals Gender and Gender* Treatment Type Gender Treatment Type HR 95% LCI 95% UCI Female Chemotherapy 0.91 0.85 0.97 No Treatment 0.88 0.82 0.94 Radiation + Chemotherapy 0.92 0.8 6 0.98 Radiation + Chemotherapy + Surgery 0.83 0.72 0.97 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Gender = Male Reference for Treatment Type = Surgery

PAGE 273

256 Evaluating the Overall Effect of M orphology on Survival In Tables 69 c through 69 e the overall effect of morphology on the risk of survival are given. After adjustment, the risk of death is decreased for squamous cell carcinoma lung cancer cases by 4 2 5% and 40% for large cell lung can cer cases as compared to small cell lung cancer cases that are female (Table 69 c). In Table 69 d, large cell (HR = 0. 50, 95% CI 0.28 0.88 ) and squamous cell carcinoma (HR = 0.51, 95% CI 0.29 0.89 ) are at an increase risk for survival given that those patients are grade 1. For the overall effect of morphology on the probability of survival in Table 69 e, four of five the treatment type s are statistically significant for an increased survival with the HRs ranging from a minimum of 0.44 to a maximum H R of 0.65 The only HR in Table 69 e that demonstrates a decreased survival or increase risk of death is the morphologic lung cancer type of adenocarcinoma when those patients receive radiation in combination with chemotherapy (HR = 1 .24, 95% CI 1.04 1. 49 ) after adjustment. Table 69 c: Overall Effect of Morphology on Survival Hazard Ratios and 95% Confidence Intervals Morphology and Morphology *Gender Morphology Gender HR 95% LCI 95% UCI Squamous Cell Female 0.59 0.42 0.82 Large Cell 0.60 0.43 0 .83 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, Reference for Morphology = Small Cell Lung Cancer Reference for Gender = Male

PAGE 274

257 Table 69 d: Overall Effect of Morphology on Survival Hazard Ratios and 95% C onfidence Intervals Morphology and Morphology*Grade Morphology Grade HR 95% LCI 95% UCI Squamous Cell I 0.50 0.28 0.88 Large Cell I 0.51 0.29 0.89 Large Cell III 0.54 0.32 0.92 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Conf idence Interval, Reference for Morphology = Small Cell Lung Cancer Reference for Grade = IV Table 69 e: Overall Interaction Effect of Morphology on Survival Hazard Ratios and 95% Confidence Intervals Morphology and Morphology*Treatment Type Morpholog y Treatment Type HR 95% LCI 95% UCI Adenocarcinoma Radiation + Chemotherapy 1.24 1.04 1.49 Squamous Cell No Treatment 0.65 0.53 0.80 Squamous Cell Radiation + Surgery 0.44 0.20 0.97 Squamous Cell Surgery + Chemotherapy 0.62 0.39 0.99 Large Cell Radiat ion + Surgery 0.45 0.20 0.99 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, Reference for Morphology = Small Cell Lung Cancer Reference for Treatment Type = Surgery

PAGE 275

258 Evaluating the Overall Effect of Gr ade on Survival In the next two tables (Table 69 f and Table 69 g) the effect of grade on the risk of death are given based on morphology and treatment type. Grade I (HR = 0.64, 95% CI 0.49 0.83 ) and grade III (HR = 0.68 95% CI 53 0.88 ) demonstrate a n increase risk of survival for patients that have adenocarcinoma as compared to small cell lung cancer after adjustment. The HR for Grade II adenocarcinoma lung cancer cases was not statistically significant. In Table 69 g, the effect of grade given the treatment type received shows a decrease survival for grade II and III lung cancer cases receiving chemotherapy and a decrease survival for grade III patients receiving radiation in combination with chemotherapy. The risk of death r anges from a minimum H R of 1.38 to a maximum HR of 1.66 Grade II versus grade IV lung cancer patients are 1.414 times more likely to die when they receive chemotherapy (HR = 1.41, 95% CI 1.03 1.9 5) versus receiving surgery for the treatment of their lung cancer. After adj ustment, fo r patients with Grade II lung cancer, the risk of death increases by 38 % when receiving radiation in combi nation with chemotherapy and 6 6% when receiving chemotherapy (Table 69 g). Table 69 f: Overall Effect of Grade on Survival Hazard Ratios and 95% Confidence Intervals Grade and Grade*Morphology Grade Morphology HR 95% LCI 95% UCI I Adenocarcinoma 0.64 0.49 0.84 III Adenocarcinoma 0.68 0.53 0.88 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, Ref erence for Morphology = Small Cell Lung Cancer Reference for Grade = IV

PAGE 276

259 Table 69 g: Overall Effect of Grade on Survival Hazard Ratios and 95% Confidence Intervals Grade and Grade*Treatment Type Grade Treatment Type HR 95% LCI 95% UCI II Chemotherapy 1 .41 1.03 1.95 III Chemotherapy 1.66 1.34 2.05 III Radiation + Chemotherapy 1.38 1.12 1.69 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, Reference for Treatment Type = Surgery Reference for Grade = IV Eva luating the Overall Effect of Stage on Survival In Tables 69 h, 69 i, and 69 j, the statistics for the effect of stage on the probability of survival given morphology, treatment type and age group at the time of diagnosis are given. In Table 69 h, for a ll stages of lung cancer (stages 1 through 3) there is an increase risk of survival for the morphologic types of adenocarcinoma, squamous cell, and large cell lung cancer as compared to small cell carcinoma after adjustment. The H Rs range from a minimum o f 0.41 for stage 1 adenocarcinom a patients to a maximum of 0.78 for stage 2 large cell lung cancer patients. Also for all stages of cases with squamous cell lung carcinoma the risk of an increased survival ranges from an HR of 0.52 to a maximum HR of 0.61 (Table 69 h). There is an increase risk of survival for stage 1 lung cancer cases receiving six of the seven possible treatment types based on surgery as the reference group and after adjustment (Table 69 i). Stage 1 lung cancer patients versus stage 4 lung cancer patients are 57 % more likely to survive when they

PAGE 277

260 receive radiation therapy alone and 31 % more likely to survive when they (stage I lung cancer patients) receive chemotherapy for the treatment of their lung cancer. Comparing the effect of stag e II lung cancer patients receiving radiatio n therapy alone, there is a 45 % incr eased risk of survival (HR 0.55, 95% CI 0.44 0.69 ) after adjustment. Also for stage III versus stage IV lung cancer case s, there is a 29 % increased risk of survival when rec eiving radiation therapy alone. Noted in T able 69 i the percent increased risk of survival decreases with increasing stage; this same trend is exhibited for the no treatment group. The HRs for the no tre atment group increase from 0.48 for stage 1, 0.62 for stage II and 0.79 for stage III lung cancer cases. This can be interpreted as the percent increase in survivorship values decreases with increasing stage. For stage I lung cancer case receiving no treatment there is a 52 % increased risk of survival, s tage II lung cancer cases have a 38 % increased risk of survival and for stage III l ung cancer cases there is a 21 % increased risk of survival (table 69 i). The statistically significant hazard ratios and 95% confidence intervals for the overall effect of stage on the risk of death given an age group is shown in Table 69 j. For all stage of lung cancer there is a decrease risk of death or increased survivorship for the four age groups 4 though 7 as compared to age group 8 and after adjustment. The HR s i n Table 69 j range from 0.33 to 0.67 with a trend of increasing HRs with increasing age group for each of the four age groups listed.

PAGE 278

261 T able 69 h: Overall Effect of Stage on Survival Hazard Ratios and 95% Confidence Intervals Stage and Stage*Morphology Stage Morphology HR 95% LCI 95% UCI I Adenocarcinoma 0.41 0.35 0.47 II 0.47 0.41 0.55 III 0.43 0.37 0.50 I Squamous Cell 0.52 0.41 0.67 II 0.61 0.47 0.78 III 0.55 0.43 0.71 I Large Cell 0.67 0.55 0.81 II 0.78 0.64 0.95 III 0.70 0.57 0. 86 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval, Reference for Morphology = Small Cell Lung Cancer Reference for Stage = IV

PAGE 279

262 Table 69 i: Overall Effect of Stage on Survival Hazard Ratios an d 95% Confidence Intervals Stage and Stage*Treatment Type Stage Treatment Group HR 95% LCI 95% UCI I Radiation 0.43 0.36 0.51 Chemotherapy 0.70 0.58 0.83 No Treatment 0.48 0.41 0.56 Radiation +Surgery 0.53 0.39 0.71 Radiation + Chemotherapy 0.59 0.50 0.70 Surgery + Chemotherapy 0.52 0.39 0.69 II Radiation 0.55 0.44 0.69 No Treatment 0.62 0.51 0.77 Radiation +Surgery 0.68 0.50 0.94 Radiation + Chemotherapy 0.76 0.61 0.96 Surgery + Chemotherapy 0.67 0.48 0.92 III Radiation 0.70 0.60 0.83 No Treatment 0.79 0.69 0.91 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Stage = Stage IV

PAGE 280

263 Table 69 j: Overall Effect of Stage on Sur vival Hazard Ratios and 95% Confidence Intervals Stage and Stage*Age Group Stage Age Group HR 95% LCI 95% UCI 1 4 0.35 0.26 0.47 5 0.33 0.28 0.40 6 0.36 0.30 0.43 7 0.41 0.35 0.49 2 4 0.45 0.33 0.63 5 0.43 0.34 0.55 6 0.46 0.37 0.58 7 0.53 0 .42 0.67 3 4 0.58 0.43 0.77 5 0.55 0.45 0.67 6 0.59 0.49 0.70 7 0.67 0.57 0.80 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Stage = Stage 4, Reference for Age Group = 8, Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs ) Evaluating the Overall Effect of Age Group at Time of Diagnosis on Survival In the next tables, 69 k and 69 l, the effect of age group on the risk of survival is given for treatment type and stage.

PAGE 281

264 Table 69 k: Overall Effect of Age Group on Survival Hazard Ratios and 95% Confidence Intervals Age Group and Age Group*Treatment Type Age Group Treatment Type HR 95% LCI 95% UCI 4 Radiation 0.83 0.69 0.99 Chemotherapy 0.66 0.57 0.75 Radiation +Surgery 0.56 0.35 0.89 Radiation + Chemotherapy 0.74 0.65 0.86 Radiation + Chemotherapy + Surgery 0.44 0.24 0.79 5 Chemotherapy 0.64 0.48 0.84 Radiation +Surgery 0.54 0.33 0.91 Radiation + Chemotherapy 0.72 0.55 0.95 Radiation + Chemotherapy + Surgery 0.43 0.23 0.80 6 Radiation + Chemotherapy + Surgery 0.51 0.28 0.95 7 No Treatment 1.35 1.05 1.74 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery, Reference for Age Group = 8, Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs )

PAGE 282

265 Table 69 l: Overall Effect of Age Group on Survival Haza rd Ratios and 95% Confidence Intervals Age Group and Age Group*Stage Age Group Stage HR 95% LCI 95% UCI 4 I 0.51 0.40 0.64 II 0.48 0.37 0.63 III 0.52 0.40 0.66 5 I 0.49 0.37 0.65 II 0.47 0.41 0.53 III 0.50 0.43 0.59 6 I 0.59 0.46 0.77 II 0.56 0.48 0.66 III 0.61 0.55 0.67 7 I 0.69 0.53 0.90 III 0.66 0.56 0.76 III 0.71 0.62 0.80 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Stage = Stage 4 Reference for Age Group = 8, Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs ) Evaluating the Overall Effect of Race on Survival In Table 69 m, the statistically significant HRs and 95%CIs for the overall effect of race on the probability of survival given treatment type are given. For white versus other lun g cancer cases, there is a 1.43 times increased risk of death when the patient r eceives chemotherapy and a 1.91 times increase in the risk of death when that cas e receives no treatment. For black lung cancer cases versus other lung cancer cases, the ri sk of death increases from 1.57 times when they receive chemotherapy alone to a 2.10

PAGE 283

266 times increase in the risk of death when the patient receives no treatment. Ta ble 69 m: Overall Effect of Race on Survival Hazard Ratios and 95% Confidence Intervals Race and Race*Treatment Type Race Treatment Type HR 95% LCI 95% UCI White Chemotherapy 1.43 1.07 1.90 No Treatment 1.91 1.46 2.50 Black Chemotherapy 1.57 1.15 2.1 4 No Treatment 2.10 1.57 2.82 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Race = 3 (or Other) Reference for Treatment Group = Surgery Evaluating the Overall Effect of Treatment Type on S urvival In the next five tables, the effect of treatment type received for the probability of survival given morphology, g r ade, stage, age group and race There were no statistically significant HRs and 95% CIs when the overall effect of treatment on the risk of survival was evaluated for gender. For the treatment type of chemotherapy (Table 69 n) the risk o f survival was increased by 48 % when the patient had large cell lung cancer versus small cell lung cancer (HR = 0.52, 95% CI 0. 30 0.91 ) In Table 69 o, for the treatment type effect of chemotherapy, there was a decrease in the risk of death for both grade 1 and grade 3 lung cancer. There was an increase in survivorship of 49.8% for grade 1 patients re ceiving chemotherapy (HR = 0.50, 95% CI 0.27 0 .94) and a 49 % increase in

PAGE 284

267 survival for grade 3 patients receiving chemotherapy. All other treatment types did not present statistically significant HRs with 95%CIs. The effect of treatment type on the risk of death given the stage (Table 69 p) of lung cancer demonstrated an decreased risk of survival for those stage 1 cases receiving chemotherapy alone (HR = 0. 51, 95% 0.2 8 0.90) but a 2. 10 times increased risk for death when stage 2 lung cancer cases re ceived radiation alone as their treatment for lun g cancer. For those lung cancer patient receiving chemotherapy in age groups 5 and 7, the risk of survival increased by as much as 79 % (Table 69 q). In age group 6, when the treatment was radiation therapy alone versus surgery, the risk of death i ncreas ed 2.07 times (HR = 2.07, 95% CI 1.04 4.13 ) after adjustment. In the last table (Table 69 r), the overall effect of treatment type on the risk of survival given raceI, shows that for black and white lung cancer cases, when the treatment was chemotherapy the risk of d eath decreases by as much as 77 %.

PAGE 285

268 Table 69 n: Overall Effect of Treatment Type on Survival Hazard Ratios and 95% Confidence Intervals Treatment Type and Treatment Type*Morphology Treatment Type Morphology HR 95% LCI 95% UCI Chemotherapy Large Cell 0.52 0.30 0.91 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Morphology = Small Cell Lung Cancer Table 69 o: Overall Effect of Treatment T ype on Survival Hazard Ratios and 95% Confidence Intervals Treatment Type and Treatment Type*Grade Treatment Type Grade HR 95% LCI 95% UCI Chemotherapy I 0.50 0.27 0.94 Chemotherapy III 0.51 0.28 0.94 Note: HR = Hazard Ratio, LCI = Lower Confidence Int erval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Grade = Grade IV Table 69 p: Overall Effect of Treatment Type on Survival Hazard Ratios and 95% Confidence Intervals Treatment Type and Treatment Type*Stag e Treatment Type Stage HR 95% LCI 95% UCI Chemotherapy I 0.51 0.28 0.90 Radiation II 2.10 1.08 4.06 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Treatment Group = Surgery Reference for Stage = Stage 4

PAGE 286

269 Table 69 q: Overall Effect of Treatment Type on Survival Hazard Ratios and 95% Confidence Intervals Treatment Type and Treatment Type*Age Group Treatment Type Age Group HR 95% LCI 95% UCI Chemotherapy 5 0.53 0.29 0.98 Radiation 6 2.07 1.04 4.13 Chemotherapy 7 0.45 0.21 0.97 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Stage = Stage 4 Reference for Age Group = 8, Age Groups: 4 = ( > 40 < 50 yrs ), 5 = ( > 50 < 60 yrs) 6 = ( > 60 < 70 yrs ), 7 = ( > 70 < 8 0 yrs ), 8 = ( > 80 < 90 yrs ) Table 69 r: Overall Effect of Treatment Type on Survival Hazard Ratios and 95% Confidence Intervals Treatment Type and Treatment Type*Race Treatment Type Race HR 95% LCI 95% UCI Chemotherapy Wh ite 0.41 0.23 0.75 Chemotherapy Black 0.53 0.38 0.75 Note: HR = Hazard Ratio, LCI = Lower Confidence Interval, UCI = Upper Confidence Interval Reference for Race = 3 Reference for Treatment Group = Surgery

PAGE 287

270 Hypothesis III Conclusion After evaluating the results generated during Hypothesis 3 testing, the decision was made to reject the null hypothesis because statistically significant differences existed in survival between women and men after controlling for covariates and interaction terms The final CP H model included stratification based on gender, stage, grade, morphology and treatment type as well as investigating main effects and the effect of moderating variables on the association between indepe ndent variables and survival. The additional inform ation obtained by utilizing a statistical model that included interaction terms served to reveal increased hazard ratios that demonstrated female and male lung cancer cases had survival patterns that were moderated by the treatment type received. When the overall effect was examined survival differences were exhibited Females as compared to males had an increased risk of survivorship specific to different treatment types and lung cancer type. Of all the treatment groups, the greatest increased survival was for women versus men being treated with radiation in combination with surgery and chemotherapy (HR = 0. 83 95% CI 0. 72 0.98 ). The hazard ratios based on the gender effect demonstrated an increase in survivorship for females versus males But when t he hazard ratios for the interaction term for gender and treatment type received were examined, females were at a decreased survival by as much as 18%. Without consideration of the overall effect of gender this survival advantage would not have been identi fied and the results could have been misinterpreted. In conclusion, f emale lung cancer cases have been reported in the literature to

PAGE 288

271 have increased survival as well as decreased survival as compared to males with lung cancer. Without the modeling approach presented in this research, specific treatment regimens and the importance of that treatment on survivorship would not have been ascertained for females as compared to males. The answer to the research question of impact gender presented for this data set. The research results found that gender does play a role in some of the lung cancer treatment selected and that selection impacts gender specific survival.

PAGE 289

272 C HAPTER V: DIS CUSSION Introduction Chapter 5 summarizes the results of this research with 1) an assessment of the major findings for the statistical analyses of the three hypotheses tested, 2) comparison of the key findings with current literature for consistency 3) an evaluation of the key findings for inconsistencies with current literature, 4) a review of the strengths and the weakness of the research study, 5) a presentation of the significant research findings as it relates to the importance in Public Health and last ly, 6) and future directions. The purpose of this research was to investigate if any differences in lung cancer treatment received were based on gender, and whether any associated treatment differences impac ted gender specific survival. To examine th e relationship between gender specific treatment and survival, the first research question to answer was if differences in lung cancer treatment (the outcome variable) existed based on gender (an independent variable). The question of gender specific lung cancer treatments was important to address as a first step in the investigation as there are no published quantitative results that show whether there is a statistically significant difference regarding the lung cancer treatment received by women as compa red to men 12, 40 45 The selec tion of a particular lung cancer treatment is a clinical decision based on standardized recommendations which considers several parameters including morphologic type, stage, and grade of lung cancer 4, 9, 13 Each morphologic type of lung cancer has its ow n medical intervention that can include any combination of surgery, radiation therapy, and/or chemotherapy. A

PAGE 290

273 particular lung cancer treatment may vary depending upon other differences such as co morbidities or regional differences, e.g. physician prefere nces, physician training, insurance. Assessment of the Major Findings The procedures and methods presented in this research concerning gender literature. Any dif ferences in lung cancer treatment for females relative to males were important to ascertain as those diffe rences could affect survival. The research questions were designed to investigate if gender specific treatment differences changed the relationship b etween female survival and male survival for the data analyzed in this dissertation. A novel approach was used to evaluate the overall interaction effect and that impact on the treatment received and survival. has not been published in the literature as it pertains to lung cancer treatment and survival. For example, the overall gender effect was calculated from the beta coefficients for the main effect of gender plus the beta coefficients of the statistically signi ficant interaction term containing gender. When the beta coefficients were added and exponentiated the Odds Ratios for treatment received and Hazard Ratios for survival were generated. From this information, differences in specific lung cancer treatments and gender specific survival could be ascertained. Upon examination of the Odds Ratios from the Multinomial Logistic Regression

PAGE 291

274 model, it was determined that the recommended standard of care for the treatment of primary lung cancer was not always adhere d to for females versus males For example, the ORs were statistically significant for stage I female lung cancer cases ; those patients had a s much as a 2.78 times increased probability of receiving radiation in combination with chemotherapy versus the st andard of care surgery Surgery is the primary modality for the treatment of early stage lung cancer (stage I) ; radiation plus chemotherapy is an adjunct or secondary treatment and the treatment modality could vary based on gender. When the hazard rati os were assessed some of the treatment decisions for the lung Depending upon the treatment received and morphologic lung cancer type males as compared to female s could have an increased hazard for death with the hazard ranging from 8 % to 29 %. These results demonstrated that males had a statistically significant decrease in survival when the overall gender effect was taken into account. Without the statistical approach utilized in this research, misinterpretation of gender specific treatment and survival could have been made. Evaluating just the main effects and/or interaction effects can give results that conflict with the overall effect a variable contributes to the outcome. Hypothesis I Hypothesis I stated that there was a treatment difference based on gender when adjusted for the research covariates. The outcome variable of interest was treatment

PAGE 292

275 group which included eight different lung cancer treatment o ptions. The eight lung cancer treatment groups were classified as 1) radiation therapy, 2) chemotherapy, 3) no treatment received, 4) radiation plus surgery, 5) radiation in combination with chemotherapy, 6) surgery plus chemotherapy, 7) radiation plus ch emotherapy plus surgery and 8) surgery. The statistics for the data analyzed in this dissertation demonstrated that some females as compared to males depending upon stage and marital status did not receive the same lung cancer treatment modality. The st andard of care as outlined in the article by Collins, et. al. (2007) is to treat a later stage lung cancer with chemotherapy or a combination of chemotherapy and radiation. Treatment decisions are based on the standards of care established by the medical community are overseen by several organizations such as the American Medical Association, the American College of Surgeons, the National Cancer Institute, and the American College of Radiology 33 The statistical analyses found that some of the treatment selection for stage I and III was gender dependent. The null hypothesis of no differences in treatment outcomes between men and women was rejected as there were statistically signi ficant differences between gender and the lung cancer treatment received. Hypothesis I I The major finding for Hypothesis II (there was a statistically significant

PAGE 293

276 difference in survival for female lung cancer cases as compared to the survival for male l ung cancer cases) was the unadjusted gender specific survival patterns were comparable to the published literature 14 17 Female lung cancer cases had an increased probability of survival (increased survivorship) as compared to males. T he mean survival time for females was 19.8 months whereas the mean survival time for males was 16.4 months; males; these results were statistically significa nt. Women were at a lower risk to experience the event (death) as compared to men for the selected five year time range Cancer in Women as Compared to Men: Stage, Tre 8 also found gender differences in survival. Women were found to live, on average, 12 months longer than men. The authors concluded there was a significant survival difference between men and women with lung cancer with women having a survival advantage over men. For the data analyzed in this dissertation, the null hypothesis of no difference was rejected as it was determined that there was a sta tistically significant increase in survival in women with lung cancer as compared to the survival of men with lung cancer. Hypothesis I II Univariate and multivariate survival analysis were included in the statistical methods to test Hypothesis Three. Hy pothesis III stated that w o men with the same histological type, stage, grade of lung cancer, and the same treatment modality differ

PAGE 294

277 significantly in survival as compared to men with the same histological type, stage, and grade of lung cancer, and the same treatment modality. The gender effect demonstrated a decreased risk of death for females versus males dependent upon the treatment received and the morpholo gic lung cancer type After adjustment females versus males with large cell lung cancer could ex hibit a 30 % increase in the probability of survival (HR = 0.75 95% CI 0. 70 0.81 ) and a 29% increase in survival for females versus males with adenocarcinoma Based on the statistical analyses with the overall gender effect, females exhibited a distinct survival advantage when the type of treatment received and morphological lung cancer type was examined. The majority of the literature 1 7, 27, 40 that was reviewed males are at increased risk of death as compared to females with lung cancer. The cited ar ticles do not take into account any effect of gender on survival. Comparisons were made to the statistics generated by models cited in the literature that did not mention any adjustment for interaction terms demonstrated females had a survival advantage a s compared to males 1 7, 27, 4 1 For example, in the article by Ringer, et. al. (2005), the statistics used for the primary outcome of survival were given as the Chi Square and Student t test. Although survival rates (%) were given for lung cancer patients by stage of disease, histologic type, and by gender, there was no mention of interaction. Fu, et. al. (2007) 12 reported on a model that included interaction terms; the only interaction term that was statistically significant was gender and age. The au thors found that women and men that were age 50 or greater demonstrated increased survival as compared to women less than 50 years of age.

PAGE 295

278 For Hypothesis III, the null hypothesis of no difference was rejected as the results of the statistical testing sup ported statistically significant differences in gender specific survival. Comparison and Consistency of Key Findings with the Current Literature When comparing the characteristics of the lung cancer cases, Radzikowska et. al., (2002), investigate d dem ographic factors (gender, age, and smoking) and factors connected with the disease (histology, performance status, stage, treatment and survival) for lung cancer patients Women were found to be more likely to have adenocarcinoma and SCLC as compared to m en. Squamous cell cancer was the predominant type of lung cancer among men, and less than ten percent of men had adenocarcinoma. This was consistent with research findings of this dissertation, Kowski, et al., (2010) ; adenocarcinoma was the most prevalen t histological type for women (17.8%) whereas sq uamous cell lung cancer included the greatest number of males (19.4%). Radzikowska et. al., (2002) found that 21.6% of all females had adenocarcinoma of the lung. There was a 2.2% difference for the number of females with adenocarcinoma when both lung cancer data sets were compared. Fu, et. al. (2007) 12 evaluated the survival rates for men and women who received one of five treatment groups (surgery alone, radiotherapy alone, surgery + radiotherapy, no s urgery or radiotherapy, unknown) utilizing the life table method. With this method, for patients receiving surgery as part of their treatment, females had an increased survival

PAGE 296

279 as compared to men 12 For females and males that underwent surgery in combin ation with radiation therapy, females had an increase in survival by as much as 66%. This increase in survivorship for women receiving radiotherapy in combination with surgery was also demonstrated by Kowski, et. al. (2010) with the Cox Proportional Hazar ds model. Women as compared to men receiving radiotherapy in combination with surgery were 1.08 times more likely to survive (HR = 0.92, 95% CI 0.86 0.98). Although each of the authors utilized different statistical methods, both found similar results of a decrease survivorship for men versus females when treatment type was evaluated. There were other areas of agreement (consistency) in this research with cited literature 8 15 17 27 167 that reported women with lung cancer survive longer than men with lung cancer. Analyses of data among females demonstrated statistically significant increased sur vivorship in the unadjusted survival rates as compared to males utilizing the Life Table method (non parametric). Lung cancer mortality rates are higher in men as compared to women 3, 10 This was consistent with the research findings during Hypothesis II testing. Comparison of the K ey F indings with the C urrent Literature for Inc onsistenc y 8 found gender differences in survival were not significant. This was inconsistent with the statistical analyses addressing Hypothesis II in this research which found females having a survival advantage. When stratified analysis

PAGE 297

280 based on stage of di sease was assessed by Ouellette, et. al. (1998); women were found to live, on average, 12 months longer than men. The authors concluded there was a statistically significant survival difference after adjusting by stage between men and women with lung canc er with women having a survival advantage over men. In comparison to the increased survivorship of women after adjustment for stage, published by Ouellette, et. al. (1998), this dissertation, Kowski, et. al. (2010) did not find a statistically significant relationship between gender, stage and survival when the overall gender effect was considered. These differences in survival were demonstrated utilizing a semi parametric statistical model the Cox Proportional Hazards model. When the statistics were a ssessed for the overall gender effect on survival, there were no statistically significant hazard ratios that included gender and stage. Ringer et al. (2005) in the article "Influence of sex on lung cancer histology, stage, and survival in a Midwestern U nited States Tumor Registry." identified differences between men and women with regard to lung cancer type, stage at diagnosis, and survival Women were found to have a decreased survival with late stage lung cancer as compared to men 27 but there was no e xpansion of the results based on any analysis that included the type of treatment received for women and men. Kowski, et. al. (2010) research results for the lung cancer cases demonstrated an inconsistency based on the treatment received. Females versus males with lung cancer were at a statistically significant increased risk of survival when they were treated with radiation in combination with surgery (HR = 0.92, 95% CI 0.86 0.98), chemotherapy alone (HR =

PAGE 298

281 0.91, 95% CI 0.85 0.97), radiation in combin ation with chemotherapy and surgery (HR = 0.83, 95% CI 0.72 0.97) or if no treatment was received (HR = 0.88, 95% CI 0.82 0.94) as compared to receiving surgery alone. Survival rates were shown to be independent of lung cancer morphology as cited i n the article by Visbal, et. al. (2004) 17 The survival rates presented in this dissertation demonstrated a statistically significant difference for the 4 major lung cancer morphologic types. When the gender effect was considered, females versus males with squamous cell lung cancer receiving surgery alone demonstrated an increase risk of survival (HR = 0. 82 95% CI 0. 76 0. 87 ). Also, females versus males with large cell lung cancer receiving radiation therapy alone were 1.1 1 times more likely to survi ve (HR = 0. 89 95% CI 0.82 0.96 ). T his research expanded the investigation to include the possible effect of stage, grade, treatment type, age group, marital status, and race for each morphologic lung cancer type. pidemiology, tumor biology, and by Belani, et.al. ( 2007 ) noted gender specific differences in cancer prognosis 41 Belani compared several studies examining histological types of lung cancer and gender differences. For example, the authors noted that the major hi stologic type of lung cancer was adenocarcinoma with ratio between males to females being 1.0 to 1.3 41 Belani, et. al. ( 2007 ) further reported that males as compared to females have a greater proportion of squamous cell carcinoma approximately 1.7 to 1.0. As reported in the previous section adenocar cinoma was the

PAGE 299

282 most prevalent histological type for females (17.8%) whereas squamous cell lung cancer include d the greatest number of males (19.4%). Th e difference in the distribution for the morphologic types could have resulted from differences in the d ata analyzed in this dissertation for Kowski, et. al. ( 2010 ) versus the data sets Belani, et. al. ( 2007 ) examined as those lung cancer cases were from different studies. Study Limitations Different types of bias or systematic error can be initiated in t he design phase, the data collection phase, the analysis phase or during the publication phases for the research study. Several possible limitations in this research were experienced in the initial phase of data collection. The data that were collected w as secondary data. Secondary data can be subject to measurement error. This bias could have been introduced by errors made during data collection by the cancer registries. As the cancer registries collected the data in a standardized format, this partic ular limitation was considered to be minimal. Initially, all cancer registries that were members of NAACCR from the four geographic regions of the US were possible candidates for inclusion into the study. From the NAACCR cancer registries, cancer regist ries were selected that met and maintained quality standards for the years of study (2000 2004). Once it was established that a cancer registry followed the standardized procedures for NAACCR, two state cancer registries were randomly selected from each geographic region. Also, there were differences for each state cancer registry IRB protocol for the release of data. This may

PAGE 300

283 have introduced a selection bias in that some of the states that were selected randomly from each geographic region would not r elease the data; another state had to be selected from the region. Selection bias is minimal when the samples are selected randomly and although this was the intent for this research, a completely random selection of the cancer registries that supplied lu ng cancer cases for the data set was not achievable; this may have limited the external validity of the study results due to a systematic or random error bias. Another limitation was the limited access over which variables could be obtained from the canc er registries. Patient anonymity was a major concern limiting the number of variables that could be obtained for the research study. Also variations due to changes in the characteristics of the lung cancer population may have been introduced by geographi c differences, e.g. different patterns of care specific to a region, environmental differences, e.g. an increase in lung cancer cases due to radon and these random variations may have limited the interpretation of the study results. When evaluating the st atistical analyses, one of the limitations could be identified as some of the lower bounds of the confidence intervals were minimally statistically significant, i.e. some of the lower bounds of the confidence intervals did approach one. The statistics wer e reported and standardized on the level of significance to the hundreds; therefore these results were still reported as statistically significant. Not all subgroup analyses resulted in statistically significant findings, which could be interpreted as a p ossible limitation if statistically significant results for all subgroups were anticipated.

PAGE 301

284 Without the subgroup analyses, the information on treatment and gender specific survival differences between could not be determined or examined. One of the major limitations for comparative purposes was that the current literature 8 15 17 27 167 does not ad dressed treatment differences based on gender. Contrary to the current literature, gender specific survival differences, were demonstrated when females and males were stratified by lung cancer type (morphology), stage, grade and treatment type and when t he effect of moderating variables were accounted for in the statistical models T he authors Belani, et. al. ( 2007) expressed an urgent need to increase research and funding to improve lung cancer care, in particular for women 41 their recommendation was based on limited information as difference in tr eatment modality by gender only included studies focused on surgery alone or radiation therapy alone. For this research presented in this dissertation, the major treatment types for lung cancer were critical for a valid assessment. As demonstrated in the statistical analyses, there are statistically significant differences in the treatments women receive as compared to men based on stage for the data analyzed in this dissertation Study Strengths The data from the eight cancer registries for this rese arch was acquired over a year and a half time period. Strength in the study design included the large population size (power) and quality of data. Each cancer registry that was included in the data set met national standards as outlined by NAACCR decreas ing any discrepancies with data

PAGE 302

285 collection and data quality. In the design phase, strength of this study was in the protocol for the selection of the state cancer registries which minimized selection bias by the random selection of the cancer registries. Error due to random variations in the characteristics of lung cancer cases was accounted for by assessing random effects during the statistical testing phase. Possible random variations in the lung cancer cases due to geographic or environmental differen ces that may have invalidated the results were addressed comparing a random effects model to a fixed effects model. No effect on the association between for the outcome and independent variables were seen for the data analyzed in this dissertation when th e two models were compared. A particular strength of the study statistical testing included a more complete assessment of gender specific survival adjusted for treatment type, stage, morphology, grade and interaction terms. Studies have not been publishe statistical modeling approach which included these variables with interaction terms. Temporal differences due to changes in treatment regimens for the treatment of lung cancer were minimized as t he time range of this study was 5 years (01 01 2000 though 12 31 2004), as the standards of care did not change over this time period. Also there were no coding changes introduced by NACCR for lung cancer during the study time range, so any misclassificat ion error would be thought as minimal. Expanding upon the strength of the methods utilized, initially, during the model criteria development, different classifications (strata) for lung cancer treatments were identified. Other independent variables were s elected for inclusion into the statistical

PAGE 303

286 model (the Multinomial Logistic Regression model) to answer question one. Stratification based on the independent variables of gender, morphology, stage, grade, age, race and marital status was utilized in the st atistical model but this approach is unlike statistical models in other currently published studies 12, 40 45 This research model identified possible moderating variables that could have affected a lung cancer treatment based on the overall gender effect. The fixed effects were accoun ted for in the first model for the multinomial logistic regression model (MLR 1 ) Another important aspect in answering Research Question I was to investigate possible random effects. A random effects component was included in a second multinomial logisti c regression (MLR 2 ) model. The decision to test for possible changes in the associations between the outcome and independent variables due to random effects was based on previous risk factors cited in the literature (Chapter Two) which included the enviro nment 10, 172 and geographic variations 10 Possible random effects due to these and other risk factors may have introduced differences in the lung cancer cases from the state cancer registries located in the four geographic regions of the United States. Any differences in the ass ociation between the outcome and covariates due to the fixed effects versus random effects would have to be identified as the resultant statistics could be biased and could have included invalid interpretations. After the assessment of gender specific tre atment differences as outlined in Question One during Hypothesis I testing, Research Questions Two and Three then expanded the study of survival based on gender differences and other covariates. Other

PAGE 304

287 covariates included the treatment received, age, morph ologic lung cancer type, grade and stage. Research Question Two examined survival rates between males and females without any statistical adjustment for additional covariates in the model. The investigation of the unadjusted lung cancer survival for the data analyzed in this dissertation over the five year time interval served a two fold purpose. First, an initial assessment of the unadjusted gender specific survival associated with the se data had to be made without the effects of the covariates on the o utcome. Secondly, an evaluation of the survivorship for these data was necessary so a comparison of the gender specific survival patterns reported in the literature could be made. In the literature, females with lung cancer have been reported to have a s urvival advantage relative to males with lung cancer 3, 10 ; consistency with the published literature would add to the external validity of the findings for this research For example, a lthough interpretations of the unadjusted results were limited in scope, the individuals comprising th e research lung cancer data set could be representative of the lung cancer cases in the US if the lung cancer data set survival patterns were consistent with gender specific lung cancer survival results published in the literature 3, 10 wer to the final research question during Hypothesis III statistical testing utilized univariate and multivariate survival analysis. Univariate survival analysis was comprised of evaluating the statistics and graphs generated during the non parametric tec hnique for the Life Table Method. Each independent variable was tested separately to evaluate the proportionality

PAGE 305

288 assumption between the strata of the independent variable versus survival time. The proportionality assumption infers that the hazard or ris k of failure (death) is constant over survival time. For the assumption of proportionality to hold true, the graphs between the survival curves and the strata of the independent variable will be parallel; any overlapping, diverging or converging lines of the graphs can be cause for concern as this could violate the basic statistical assumptions for a semi parametric model. The semi to assess the multivariable relationship between the outcome (survival time) and the independent variables with the inclusion of interaction terms. In order to obtain a model with the inclusion and exclusion of variables and variable combination for second order interaction terms, the stepwise procedure was used. Included in the evaluation of the model fit, any non proportionality concerns for a non constant hazard over survival time were addressed via residual analysis. Residual analysis was used to test for trends; any resultant trends in the residual pl ots for the individual variables would be displayed as increasing or decreasing slopes over the log of survival time. If a trend was displayed for a variable over the log of survival time, the model would be inappropriate for the variable selected or the variable was not constant over the survival time. The reported relationship in the literature between gender and survival is inconsistent. Contrary to some of the articles publishe d in the literature 8 15 17 27 167 with women having increased survivorship as compared to men, in some circumstances

PAGE 306

289 this research found female had a survival disadvantage as compared to males. Treatment differences based on gender were demonstrated and that those treatment differences changed the association for gender specific survival when adju stments for the covariates and interaction terms were taken into account in the model. When the overall gender effect was considered for the treatment received morphologic lung cancer type and survival, this research design and resultant findings suppor ts the literature 27 in which females have an increased survivorship as compared to men. Public Health Importance Finding the most effective treatment for increasing lung cancer survival has immense public health consequences. Finding the most effective treatment includes many factors that must be accounted for but c an be difficult to ascertain. Prior to investigating effective treatments that increase survival, the examination of treatment differences based on key factors for lung cancer would have to be made. This would include any treatment for lung cancer that differed on the basis of gender. The clinical pathways for the care of lung cancer patients is standard ized but when quantifiable techniques were utilized, differences in the standard of care for lung cancer patients were demonstrated to be gender dependent. For example, the standard of care for early stage lung cancer is surgery For this data set, surge ry was not consistently shown to be the first treatment choice for early stage lung cancer. For example, separated f emales with stage I lung cancer versus separated males with stage I were 2.82 times more likely to

PAGE 307

290 receive chemotherap y alone (OR = 2.82 9 5% CI 1.17 6.80 ) as compared to receiving surgery alone For later stage disease, divorced females as compared to divorced males with stage III lung cancer were 1.57 times more likely to receive radiation in combination with surgery and chemotherapy (OR = 1.57 95% CI 1. 07 2.30 ) versus receiving surgery alone For later stage disease, radiation therapy in combination with chemotherapy or chemotherapy alone is the standard treatment recommendation Building on this information of gender differences in lung cancer treatments, when the overall gender effect was assessed for survival lung cancer type and the treatment received males versus females had a statistically significant decrease in survival. G ender specific survivorship was demonstrated to be s tatistically significant when adjusted for grade, grade*morphology, stage, stage*morphology, age group, stage*age group, race, treatment type*morphology, treatment type*grade, treatment type*stage, treatment type*age group, and treatment type*race When t he gender effect for survival was assessed, females compared to males had a statistically significant survival advantage for six of the seven treatment groups. For the other treatment group of radiation therapy in combination with chemotherapy the result for the gender effect on survival w as not statistically significant. Generally, lung cancer cases receive a specific treatment for lung cancer regardless of ge nder; this was not the case for the data analyzed for this dissertation. For males versus fe ma le lung cancer cases, differences in the type of treatment received could increase the risk of death or decrease survival time. The a ssociated gender differences with treatment selection were tested with multiple

PAGE 308

291 modeling techniques resulting in the same conclusion, there is a statistically significant difference in the way female and male lung cancer cases are treated The methods and statistical analyses outlined in this research identify the impact of treatment decisions on female and male survival in particular for early stage lung cancer. The costs associated with lung cancer care are enormous according to the National Heart Lung & Blood Institute (NHLBI). Lung cancer costs shows medical expenditures as approximately 10 billion annually, according to the Centers for Medicare and Medicaid Services (CMS) 115 Over 13% of the total cancer care costs for 2006 were attributed to lung c ancer. The non medical total or personal care exceeded 250 billion for the same time period. If it is possible to assess the most effective treatment, there could be an increase in survival and a decreas e in healthcare costs, thereby improving Public Hea lth. Future Directions A first step in the future direction of this research would be a comparative analysis of an active versus passive cancer registry such as SEER. These data are collected and compiled independently by SEER registries. Further, the data are publicly available and issues of patient confidentiality will be minimized. An independent comparison and verification of the study results would be a necessary next step to verify that treatment differences based on gender exist. Lastly, a pos sible future direction, after validation of the research results presented in this dissertation would be the development

PAGE 309

292 of a task group to investigate treatment differences based on gender and the subsequent impact on gender specific survivorship. Sever al scientific and medical associations such as the American Medical Association, the American College of Surgeons, the National Cancer Institute, or the American College of Radiology might possibly accept this role.

PAGE 310

293 REFERENCES 1. F CDS. Florida Cancer Data System. 2003 Florida Annual Cancer Report: Incidence and Mortality for 2000 [ 2. ACS, Society AC. Cancer Facts & Figures. 3. Alberg AJ, Samet JM. Epidemiology of lung cancer. Chest. Jan 2003;123(1 Suppl):21S 49S. 4. Collins LG, Hai nes C, Perkel R, Enck RE. Lung cancer: diagnosis and management. Am Fam Physician. Jan 1 2007;75(1):56 63. 5. IARC. International Agency for Research on Cancer. http://www.iarc.fr/index.html 6. Martini B. [L ung cancer -epidemiology, prognosis and therapy]. Med Monatsschr Pharm. Jun 2006;29(6):217 221. 7. Lillington GA. Lung cancer. Curr Opin Pulm Med. Jul 2004;10(4):239 241. 8. Ouellette D, Desbiens G, Emond C, Beauchamp G. Lung cancer in women compared with men: stage, treatment, and survival. Ann Thorac Surg. Oct 1998;66(4):1140 1143; discussion 1143 1144. 9. Rivera MP, Detterbeck F, Mehta AC. Diagnosis of Lung Cancer. CHEST CHEST J1 CHEST. 2003/01//Jan2003 Supplement 2003;123(1):129S. 10. Schottenfeld D, Fraumeni J. Cancer Epidemiology and Prevention : Oxford University Press; 2006. 11. SEER. Staging Summary. http://training.seer.cancer.gov 12. Fu JB, Kau TY, Severson RK, Kalemkerian GP. Lung cancer in women: analysis of the national Surveillance, Epidem iology, and End Results database. Chest. Mar 2005;127(3):768 777. 13. SEER. Lung Cancer Statistics. http://seer.cancer.gov/Publications/SummaryStage 14. Mayne S, Buenconsejo J, Janerich D. Previous lung disease and risk of lung cancer among men and women nonsmokers. Am. J. Epidemiol. January 1, 1999 1999;149(1):13 20. 15. Radzikowska E, Glaz P, Roszkowski K. Lung cancer in women: age, smoking, histology, performance status, stage, initial treatment and survival. Population based study of 20 561 cases. Ann Oncol. Jul 2002;13(7):1087 1093. 16. Thompson E. Latest advances and research in lung cancer. Drug News Perspect. Jul Aug 2005;18(6):405 411. 17. Visbal AL, Williams BA, Nichols FC, 3rd, et al. Gender differences in non small cell lung cancer survival: an analysis of 4,618 patients diagnosed between 1997 and 2002. Ann Thorac Surg. Jul 2004;78(1):209 215; discussion 215. 18. Zahm SH, Pottern LM, Lewis DR, Ward MH, White DW. Inclusion of women and minorities in occupational cancer epidemiologic research. J O ccup Med. Aug 1994;36(8):842 847.

PAGE 311

294 19. Abidoye O, Ferguson MK, Salgia R. Lung carcinoma in African Americans. Nat Clin Pract Oncol. Feb 2007;4(2):118 129. 20. Bradley CJ, Given CW, Roberts C. Disparities in cancer diagnosis and survival. Cancer. Jan 1 2001; 91(1):178 188. 21. Gadgeel SM, Kalemkerian GP. Racial differences in lung cancer. Cancer Metastasis Rev. Mar 2003;22(1):39 46. 22. Charloux A, Quoix E, Wolkove N, Small D, Pauli G, Kreisman H. The increasing incidence of lung adenocarcinoma: reality or art efact? A review of the epidemiology of lung adenocarcinoma. Int. J. Epidemiol. February 1, 1997 1997;26(1):14 23. 23. USDHEW. U.S Department of Health, Education, and Welfare: Smoking and Health. Report of the Advisory Committee to the Surgeon General 1964 DHEW Publication No. [PHS] 1103. 24. CDCP. Centers for Disease Control and Prevention Office of Smoking and Health: Women and Smoking A Report of the Surgeon General 2001 : Centers for Disease Control and Prevention Office of Smoking and Health; 2001. 2 5. Blumer G. Cigarette smoking and cancer of the lung. Ill Med J. Aug 1951;100(2):98 99. 26. Rauscher GH, Mayne ST, Janerich DT. Relation between Body Mass Index and Lung Cancer Risk in Men and Women Never and Former Smokers. Am. J. Epidemiol. September 15 2000 2000;152(6):506 513. 27. Ringer G, Smith JM, Engel AM, Hendy MP, Lang J. Influence of sex on lung cancer histology, stage, and survival in a midwestern United States tumor registry. Clin Lung Cancer. Nov 2005;7(3):180 182. 28. ACS ACS. Cancer Facts & Figures. 29. IARC. International Agency for Research on Cancer. http://www.iarc.fr/ 30. Devesa S, Bray F, Vizcaino AP, Parkin DM. International lung cancer trends by histologic type: male:female differences diminish ing and adenocarcinoma rates rising. Int J Cancer. Nov 1 2005;117(2):294 299. 31. Le Faou AL, Scemama O. [Epidemiology of tobacco smoking.]. Rev Mal Respir. Jun 9 2005. 32. Sobue T, Yamamoto S, Hara M, Sasazuki S, Sasaki S, Tsugane S. Cigarette smoking and subsequent risk of lung cancer by histologic type in middle aged Japanese men and women: the JPHC study. Int J Cancer. May 10 2002;99(2):245 251. 33. ALA. American Lung Association, Epidemiology and Statistic Unit, Research and Scientific Affairs, Trends in Lung Cancer Morbidity and Mortality; June 2004. 2004. 34. Gray. http://education.yahoo.com/reference/gray/illustrations/figure?id=970 http://education.yahoo.com/reference/gray/illustrations/figure?id=970 35. Price SA, Wilson LM. Pathophysiology: Clinical Concepts of Disease Processes.

PAGE 312

295 Vol Fourth. St. Louis: Mosby; 1992. 36. Basmaj ian JV. Grant's Method of Anatomy : Williams and Williams; 1982. 37. Price SA, McCarthy Wilson L. Pathophysiology: Clinical Concepts of Disease Processes Fourth ed. St. Louis: Mosby; 1992. 38. Anatomy L. http://www.talktransplant.com/Lung/Anatomy.aspx http://www.talktransplant.com/Lung/Anatomy.aspx 39. Gray. http://education.yahoo.com/reference/gray/illustrations/figure?id=970 40. Alberg AJ, Brock MV, Samet JM. Epidemiology of lung cancer: looking to the future. J Clin Oncol. May 10 2005;23(14):3175 3185. 41. Belani CP, Marts S, Schiller J Socinski MA. Women and lung cancer: Epidemiology, tumor biology, and emerging trends in clinical research. Lung Cancer. Jan 2007;55(1):15 23. 42. Chatkin JM, Abreu CM, Fritscher CC, Wagner MB, Pinto JA. Is there a gender difference in non small cell lung cancer survival? Gend Med. Aug 2004;1(1):41 47. 43. Hofmann HS, Bartling B. How will lung cancer be treated in the future? Future Oncol. Aug 2005;1(4):551 559. 44. Patel JD, Bach PB, Kris MG. Lung cancer in US women: a contemporary epidemic. Jama. Apr 14 2004;291(14):1763 1768. 45. Rivera MP, Stover DE. Gender and lung cancer. Clin Chest Med. Jun 2004;25(2):391 400. 46. Stabile LP, Siegfried JM. Sex and gender differences in lung cancer. J Gend Specif Med. 2003;6(1):37 48. 47. Wisnivesky JP, Bonomi M, Hens chke C, Iannuzzi M, McGinn T. Radiation therapy for the treatment of unresected stage I II non small cell lung cancer. Chest. Sep 2005;128(3):1461 1467. 48. Thomas L, Doyle LA, Edelman MJ. Lung cancer in women: emerging differences in epidemiology, biology and therapy. Chest. Jul 2005;128(1):370 381. 49. Baldini EH, Strauss GM. Women and lung cancer: waiting to exhale. Chest. Oct 1997;112(4 Suppl):229S 234S. 50. Brogdon CF. Women and cancer. J Intraven Nurs. Nov Dec 1998;21(6):344 355. 51. Carbone DP. The biology of lung cancer. Semin Oncol. Aug 1997;24(4):388 401. 52. Patel JD. Lung cancer in women. J Clin Oncol. May 10 2005;23(14):3212 3218. 53. Wisnivesky JP, Halm EA. Sex differences in lung cancer survival: do tumors behave differently in elderly women? J Clin Oncol. May 1 2007;25(13):1705 1712. 54. Anraku M, Waddell TK. Surgery for small cell lung cancer. Semin Thorac Cardiovasc Surg. Fall 2006;18(3):211 216. 55. Brundage MD, Davies D, Mackillop WJ. Prognostic Factors in Non small Cell Lung Cancer* : A Decade of Progress 10.1378/chest.122.3.1037. Chest. September 1, 2002 2002;122(3):1037 1057. 56. Haraguchi S, Hioki M, Koizumi K, et al. Characteristics of Multiple Primary

PAGE 313

296 Malignancies Associated with Lung Cancer by Gender. Respiration. May 11 2006. 57. Barbone F, Bovenzi M, Cavallieri F, Stanta G. Air Pollution and Lung Cancer in Trieste, Italy. Am. J. Epidemiol. June 15, 1995 1995;141(12):1161 1169. 58. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2006. CA Cancer J Clin. Mar Apr 2006;56(2):106 1 30. 59. MacCallum C, Gillenwater HH. Second line treatment of small cell lung cancer. Curr Oncol Rep. Jul 2006;8(4):258 264. 60. SEER. National Center for Health Statistics. Report of Final Mortality Statistics, 1979 2001 4. National Cancer Institute SEER Program: Cancer Statistics Review, 1973 2001. 2003. 61. Cetingoz R, Cetinayak HO, Sen RC, et al. Prognostic factors in limited stage small cell lung cancer of patients treated with combined modality approach. J Buon. Jan Mar 2006;11(1):31 37. 62. Donington JS, Le QT, Wakelee HA. Lung cancer in women: exploring sex differences in susceptibility, biology, and therapeutic response. Clin Lung Cancer. Jul 2006;8(1):22 29. 63. Padilla J, Calvo V, Penalver JC, et al. Survival and risk model for stage IB non small cell lung cancer. Lung Cancer. Apr 2002;36(1):43 48. 64. Sachs S, Fiore JJ. An overview of lung cancer. Respir Care Clin N Am. Mar 2003;9(1):1 25. 65. Wakai K, Inoue M, Mizoue T, et al. Tobacco smoking and lung cancer risk: an evaluation based on a systema tic review of epidemiological evidence among the Japanese population. Jpn J Clin Oncol. May 2006;36(5):309 324. 66. Caldarella A, Crocetti E, Comin CE, Janni A, Pegna AL, Paci E. Gender differences in non small cell lung cancer: A population based study. E ur J Surg Oncol. Feb 14 2007. 67. Bepler G. Lung cancer epidemiology and genetics. J Thorac Imaging. Oct 1999;14(4):228 234. 68. Jin Y, Xu Y, Xu M, Xue S. Increased risk of cancer among relatives of patients with lung cancer in China. BMC Cancer. 2005;5:14 6. 69. Ponz de Leon M. Genetic factors in lung cancer. Recent Results Cancer Res. 1994;136:146 161. 70. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of chemotherapeutics. Nat Med. Nov 2006;12(11):1294 1300. 71. Rom WN, Tchou Won g KM. Molecular and genetic aspects of lung cancer. Methods Mol Med. 2003;75:3 26. 72. Perez C, Brady L. Priciples and Practice of Radiation Oncology Second ed. Philadelphia: J.B. Lippincott Company; 1992. 73. Hennekens CH, Buring JE. Epidemiology in Medi cine ; 1987. 74. Rothman KJ, Greenland S. Modern Epidemiology : Lippincott Williams and Wilkins; 1998.

PAGE 314

297 75. Koop CE, Luoto J. "The Health Consequences of Smoking: Cancer," overview of a report of the Surgeon General. Public Health Rep. Jul Aug 1982;97(4):318 324. 76. MMWR. Women and smoking: a report of the Surgeon General. Executive summary. MMWR Recomm Rep. Aug 30 2002;51(RR 12):i iv; 1 13. 77. NCCDPHP. National Center For Chronic Disease Prevention and Health Promotion: Women and Smoking: A Report of the Su rgeon General 2001. (Stock Number 017 023 00207 4). 78. Lam WK, White NW, Chan Yeung MM. Lung cancer epidemiology and risk factors in Asia and Africa. Int J Tuberc Lung Dis. Sep 2004;8(9):1045 1057. 79. Bobak M, Gilmore A, McKee M, Rose R, Marmot M. Change s in smoking prevalence in Russia, 1996 2004. Tob Control. Apr 2006;15(2):131 135. 80. Caracta CF. Gender differences in pulmonary disease. Mt Sinai J Med. Sep 2003;70(4):215 224. 81. Zang EA, Wynder EL. Differences in lung cancer risk between men and wome n: examination of the evidence. J Natl Cancer Inst. Feb 21 1996;88(3 4):183 192. 82. Parkin DM. International variation. Oncogene. Aug 23 2004;23(38):6329 6340. 83. Alberg AJ, Samet JM. Epidemiology of Lung Cancer : Academic Press; 1998. 84. Yong LC, Brown CC, Schatzkin A, et al. Intake of vitamins E, C, and A and risk of lung cancer. The NHANES I epidemiologic followup study. First National Health and Nutrition Examination Survey. Am J Epidemiol. Aug 1 1997;146(3):231 243. 85. Yong L C, Brown CC, Schatzkin A, et al. Intake of Vitamins E, C, and A and Risk of Lung Cancer The NHANES I Epidemiologic Followup Study. Am. J. Epidemiol. August 1, 1997 1997;146(3):231 243. 86. USSG. United States Surgeon General Report. Reducing the health consequeneces: twenty five years of progress. Washington, D.C.: U.S. Government Printing Office; 1998. 87. Chandra S, Shiffman S, Scharf DM, Dang Q, Shadel WG. Daily smoking patterns, their determinants, and implications for quitting. Exp Clin Psychopharmacol. Feb 2007;15(1):67 80. 88. Ezzati M, Henley SJ, Lopez AD, Thun MJ. Role of smoking in global and regional cancer epidemiology: current patterns and data needs. Int J Cancer. Oct 10 2005;116(6):963 971. 89. Hyland A, Laux FL, Higbee C, et al. Cigarette purchase patterns in four countries and the relationship with cessation: findings from the International Tobacco Control (ITC) Four Country Survey. Tob Control. Jun 2006;15 Suppl 3:iii59 64. 90. Zang EA, Wynder EL. Differences in Lung Cancer Risk Between Men and Women: Examination of the Evidence 10.1093/jnci/88.3 4.183. J. Natl. Cancer Inst. February 21, 1996 1996;88(3 4):183 192. 91. Aronchick JM. Lung cancer: epidemiology and risk factors. Semin Roentgenol. Jan 1990;25(1):5 11.

PAGE 315

298 92. Bach PB, Kattan MW, Thornquist MD, et al. Variat ions in lung cancer risk among smokers. J Natl Cancer Inst. Mar 19 2003;95(6):470 478. 93. Boffetta P. Human cancer from environmental pollutants: the epidemiological evidence. Mutat Res. Sep 28 2006;608(2):157 162. 94. Broome CM, Borum M. Cancer and women Med Clin North Am. Mar 1998;82(2):321 333. 95. Doll R, Peto R. The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J Natl Cancer Inst. Jun 1981;66(6):1191 1308. 96. Stockwell HG, Armstrong AW, Leaverton P E. Histopathology of lung cancers among smokers and nonsmokers in Florida. Int J Epidemiol. 1990;19 Suppl 1:S48 52. 97. Payne S. 'Smoke like a man, die like a man'?: a review of the relationship between gender, sex and lung cancer. Soc Sci Med. Oct 2001;53 (8):1067 1080. 98. Itaya T, Yamaoto N, Ando M, et al. Influence of histological type, smoking history and chemotherapy on survival after first line therapy in patients with advanced non small cell lung cancer. Cancer Sci. Feb 2007;98(2):226 230. 99. Govind an R, Page N, Morgensztern D, et al. Changing epidemiology of small cell lung cancer in the United States over the last 30 years: analysis of the surveillance, epidemiologic, and end results database. J Clin Oncol. Oct 1 2006;24(28):4539 4544. 100. Bepler G. Lung cancer: provoking new concepts, generating novel ideas, and rekindling enthusiasm. Cancer Control. Jul Aug 2003;10(4):275 276. 101. Blumenthal SJ. Smoking v women's health: the challenge ahead. J Am Med Womens Assoc. Jan Apr 1996;51(1 2):8. 102. US DHHS. The Health Effects of Active Smoking: A Report of the Surgeon General 2004. 103. Agudo A, Esteve MG, Pallares C, et al. Vegetable and fruit intake and the risk of lung cancer in women in Barcelona, Spain. Eur J Cancer. Jul 1997;33(8):1256 1261. 104. Alavanja MCR, Dosemeci M, Samanic C, et al. Pesticides and Lung Cancer Risk in the Agricultural Health Study Cohort 10.1093/aje/kwh290. Am. J. Epidemiol. November 1, 2004 2004;160(9):876 885. 105. Blot WJ, McLaughlin JK. Are women more susceptible to lung cancer? J Natl Cancer Inst. Jun 2 2004;96(11):812 813. 106. Folsom AR, Kushi LH, Anderson KE, et al. Associations of general and abdominal obesity with multiple health outcomes in older women: the Iowa Women's Health Study. Arch Intern Med. Jul 24 2000;16 0(14):2117 2128. 107. Gasperino J, Rom WN. Gender and lung cancer. Clin Lung Cancer. May 2004;5(6):353 359. 108. Olson JE, Yang P, Schmitz K, Vierkant RA, Cerhan JR, Sellers TA. Differential Association of Body Mass Index and Fat Distribution with Three Ma jor Histologic Types of Lung Cancer: Evidence from a Cohort of Older Women. Am.

PAGE 316

299 J. Epidemiol. October 1, 2002 2002;156(7):606 615. 109. Risch HA, Miller AB. Re: Are women more susceptible to lung cancer? J Natl Cancer Inst. Oct 20 2004;96(20):1560; author reply 1560 1561. 110. Zang EA, Wynder EL. Cumulative tar exposure. A new index for estimating lung cancer risk among cigarette smokers. Cancer. Jul 1 1992;70(1):69 76. 111. Arca JA, Ramos MA, de la Infanta RG, Lopez CP, Perez LG, Lopez JL. [Lung cancer dia gnosis: hospitalization costs]. Arch Bronconeumol. Nov 2006;42(11):569 574. 112. Au DH, Udris EM, Fihn SD, McDonell MB, Curtis JR. Differences in health care utilization at the end of life among patients with chronic obstructive pulmonary disease and patie nts with lung cancer. Arch Intern Med. Feb 13 2006;166(3):326 331. 113. Molinier L, Combescure C, Chouaid C, et al. Cost of lung cancer: a methodological review. Pharmacoeconomics. 2006;24(7):651 659. 114. Ramsey SD, Clarke L, Kamath TV, Lubeck D. Evaluati on of erlotinib in advanced non small cell lung cancer: impact on the budget of a U.S. health insurance plan. J Manag Care Pharm. Jul Aug 2006;12(6):472 478. 115. NIH. NIH Cost of Illness Report to the U.S. Congress National Health Care Expenditures Projec tions: 2003 2013. http://www.cms.hhs.gov/statistics/nhe/projections 2003 116. Alberts WM. The future and lung cancer: room for optimism? Cancer Control. Jan Feb 2000;7(1):13 14. 117. Bach PB, Niewoehner DE, Black WC. Screening for lung cancer: the guidelines. Chest. Jan 2003;123(1 Suppl):83S 88S. 118. Bach PB, Kelley MJ, Tate RC, McCrory DC. Screening for lung cancer: a review of the current literature. Chest. Jan 2003;123(1 Suppl):72 S 82S. 119. Reich JM. Lung cancer screening: contumacy vs mendacity. Chest. Mar 2003;123(3):963 964. 120. Brown ML, Lipscomb J, Snyder C. The burden of illness of cancer: economic cost and quality of life. Annual Review of Public Health. 2001;22:91 113. 12 1. Shimosato Y. [Pathology of early lung cancer]. Gan No Rinsho. Aug 1988;34(10):1373 1377. 122. Carroll R. The histology of lung cancer. Ir J Med Sci. Aug 1960;416:374 382. 123. Black C, de Verteuil R, Walker S, et al. Population screening for lung cancer using computed tomography, is there evidence of clinical effectiveness? A systematic review of the literature 10.1136/thx.2006.064659. Thorax. February 1, 2007 2007;62(2):131 138. 124. Vauclair R, Kobusch T, Morissette N, Simard A. [Screening for pulmonar y cancer: general considerations and recent data from the literature]. Union Med Can. May 1982;111(5):457 462, 467 458. 125. Buccheri G, Ferrigno D. Lung cancer: clinical presentation and specialist referral time 10.1183/09031936.04.00113603. Eur Respir J. December 1, 2004

PAGE 317

300 2004;24(6):898 904. 126. Hamilton W, Sharp D. Diagnosis of lung cancer in primary care: a structured review 10.1093/fampra/cmh605. Fam. Pract. December 1, 2004 2004;21(6):605 611. 127. Hamilton W, Peters TJ, Round A, Sharp D. What are the clinical features of lung cancer before the diagnosis is made? A population based case control study 10.1136/thx.2005.045880. Thorax. December 1, 2005 2005;60(12):1059 1065. 128. Salomaa E R, Sallinen S, Hiekkanen H, Liippo K. Delays in the Diagnosis and Treatment of Lung Cancer 10.1378/chest.128.4.2282. Chest. October 1, 2005 2005;128(4):2282 2288. 129. Munden RF, Swisher SS, Stevens CW, Stewart DJ. Imaging of the patient with non small cell lung cancer. Radiology. Dec 2005;237(3):803 818. 130. Janssen He ijnen MLG. Epidemiology of lung cancer. Tumors of the Chest: Biology, Diagnosis and Management ; 2006:3 12. 131. De Wever W, Stroobants S, Verschakelen JA. Integrated PET/CT in lung cancer imaging: history and technical aspects. Jbr Btr. Mar Apr 2007;90(2): 112 119. 132. Isobe K, Takagi K, Hata Y, et al. [Usefulness of FDG PET for the diagnosis of postoperative recurrence of lung cancer]. Nihon Kokyuki Gakkai Zasshi. May 2007;45(5):377 381. 133. Gauger J, Patz EF, Jr., Coleman RE, Herndon JE, 2nd. Clinical st age I non small cell lung cancer including FDG PET Imaging: sites and time to recurrence. J Thorac Oncol. Jun 2007;2(6):499 505. 134. Spiro SG, Porter JC. Lung Cancer -Where Are We Today?: Current Advances in Staging and Nonsurgical Treatment 10.1164/rccm. 200202 070SO. Am. J. Respir. Crit. Care Med. November 1, 2002 2002;166(9):1166 1196. 135. Spira A, Ettinger DS. Multidisciplinary management of lung cancer. N Engl J Med. Jan 22 2004;350(4):379 392. 136. Corner J, Hopkinson J, Fitzsimmons D, Barclay S, Mue rs M. Is late diagnosis of lung cancer inevitable? Interview study of patients' recollections of symptoms before diagnosis 10.1136/thx.2004.029264. Thorax. April 1, 2005 2005;60(4):314 319. 137. Tockman M, Mulshine J. Early Lung Cancer Detection: Status a nd New Strategies. Primary Care and Cancer. 1998;Supplement 1, Vol 18. 138. Tockman M, Mulshine J, Piantadosi S, et al. LCEDWG Investigators, YTC Investigators. Prospective detection of preclinical lung cancer: results from two studies of hnRNP overexpre ssion. Clin Cancer Res. 1997;3:2237 2246. 139. Duarte RL, Paschoal ME. Molecular markers in lung cancer: prognostic role and relationship to smoking. J Bras Pneumol. Jan Feb 2006;32(1):56 65. 140. Gohagan J, Marcus P, Fagerstrom R, Pinsky P, Kramer B, Pror ok P. Baseline Findings of a Randomized Feasibility Trial of Lung Cancer Screening With Spiral CT Scan vs Chest Radiograph: The Lung Screening Study of the National Cancer

PAGE 318

301 Institute 10.1378/chest.126.1.114. Chest. July 1, 2004 2004;126(1):114 121. 141. Oke n MM, Marcus PM, Hu P, et al. Baseline chest radiograph for lung cancer detection in the randomized Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial. J Natl Cancer Inst. Dec 21 2005;97(24):1832 1839. 142. Oken MM, Marcus PM, Hu P, et al. Basel ine Chest Radiograph for Lung Cancer Detection in the Randomized Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial 10.1093/jnci/dji430. J. Natl. Cancer Inst. December 21, 2005 2005;97(24):1832 1839. 143. Sone S, Li F, Yang Z, et al. Characteris tics of small lung cancers invisible on conventional chest radiography and detected by population based screening using spiral CT. Br J Radiol. February 1, 2000 2000;73(866):137 145. 144. Thomas L, Doyle LA, Edelman MJ. Lung Cancer in Women: Emerging Diffe rences in Epidemiology, Biology, and Therapy 10.1378/chest.128.1.370. Chest. July 1, 2005 2005;128(1):370 381. 145. Ruckdeschel JC. The problem of cancer: lung cancer as a paradigm -keynote address summary. Ethn Dis. Winter 2005;15(1 Suppl 1):S1 2. 146. Sc hmitz R, Haverdink B. Lung cancer in women. Adv Nurse Pract. Apr 2006;14(4):17. 147. Petriat, Cornet, Castaing, Tessier, Chevais, Martin. [Diagnosis and therapy of lung cancer.]. Bord Chir. Apr 1953;2:65 67. 148. Peto R, Darby S, Deo H, Silcocks P, Whitley E, Doll R. Smoking, smoking cessation, and lung cancer in the UK since 1950: combination of national statistics with two case control studies. Bmj. Aug 5 2000;321(7257):323 329. 149. Doll R. Smoking and lung cancer. Br Med J. Feb 28 1953;1(4808):505 506. 150. Doll R, Gray R, Hafner B, Peto R. Mortality in relation to smoking: 22 years' observations on female British doctors. Br Med J. Apr 5 1980;280(6219):967 971. 151. Doll R, Peto R. Cigarette smoking and bronchial carcinoma: dose and time relationships a mong regular smokers and lifelong non smokers. J Epidemiol Community Health. Dec 1978;32(4):303 313. 152. Lengauer C. Cancer biology. Curr Opin Oncol. Jan 2001;13(1):57. 153. Blanchon F, Grivaux M, Asselain B, et al. 4 year mortality in patients with non s mall cell lung cancer: development and validation of a prognostic index. Lancet Oncol. Oct 2006;7(10):829 836. 154. Brueckl WM, Herbst L, Lechler A, et al. Predictive and prognostic factors in small cell lung carcinoma (SCLC) -analysis from routine clinica l practice. Anticancer Res. Nov Dec 2006;26(6C):4825 4832. 155. Chen K, Wang PP, Sun B, et al. Twenty year secular changes in sex specific lung cancer incidence rates in an urban Chinese population. Lung Cancer. Jan 2006;51(1):13 19. 156. Pelletier MP, Edw ardes MD, Michel RP, Halwani F, Morin JE. Prognostic markers in resectable non small cell lung cancer: a multivariate analysis. Can J

PAGE 319

302 Surg. Jun 2001;44(3):180 188. 157. Song LH, Song XR, Liu MQ, et al. [Prognostic factors in patients with stage III and IV non small cell lung cancer]. Zhonghua Zhong Liu Za Zhi. Jun 2004;26(6):345 348. 158. Oken MM, Creech RH, Tormey DC, et al. Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol. Dec 1982;5(6):649 655. 159. Gwede CK, John son DJ, Daniels SS, Trotti A. Assessment of toxicity in cooperative oncology clinical trials: the long and short of it. J Oncol Manag. Mar Apr 2002;11(2):15 21. 160. Nakaya N, Goto K, Saito Nakaya K, et al. The Lung Cancer Database Project at the National Cancer Center, Japan: Study Design, Corresponding Rate and Profiles of Cohort 10.1093/jjco/hyl015. Jpn. J. Clin. Oncol. May 1, 2006 2006;36(5):280 284. 161. Wigren T. Confirmation of a prognostic index for patients with inoperable non small cell lung cance r. Radiother Oncol. Jul 1997;44(1):9 15. 162. Doll R, Peto R, Wheatley K, Gray R, Sutherland I. Mortality in relation to smoking: 40 years' observations on male British doctors. Bmj. Oct 8 1994;309(6959):901 911. 163. Godtfredsen NS, Prescott E, Osler M. E ffect of smoking reduction on lung cancer risk. Jama. Sep 28 2005;294(12):1505 1510. 164. Ederer F, Mersheimer WL. Sex differences in the survival of lung cancer patients. Cancer. Mar Apr 1962;15:425 432. 165. Satcher D, Thompson TG, Koplan JP. Women and s moking: a report of the Surgeon General. Nicotine Tob Res. Feb 2002;4(1):7 20. 166. Kelly A, Blair N, Pechacek TF. Women and smoking: issues and opportunities. J Womens Health Gend Based Med. Jul Aug 2001;10(6):515 518. 167. Aitakov ZN, Savchenko AA. [Bron chiolo alveolar cancer and its surgical treatment (a practical view)]. Khirurgiia (Mosk). 1998(8):31 33. 168. Ben Zaken Cohen S, Pare PD, Man SF, Sin DD. The growing burden of chronic obstructive pulmonary disease and lung cancer in women: examining sex di fferences in cigarette smoke metabolism. Am J Respir Crit Care Med. Jul 15 2007;176(2):113 120. 169. Etzel CJ, Amos CI, Spitz MR. Risk for Smoking Related Cancer among Relatives of Lung Cancer Patients. Cancer Res. December 1, 2003 2003;63(23):8531 8535. 1 70. Yach D, Wipfli H. A century of smoke. Annals of Tropical Medicine and Parasitology. Jul Sep 2006;100(5 6):465 479. 171. Zang EA, Wynder EL. Smoking trends in the United States between 1969 and 1995 based on patients hospitalized with non smoking relate d diseases. Prev Med. Nov Dec 1998;27(6):854 861. 172. Samet JM. Epidemiology of Lung Cancer. Vol 74. New York, New York: Marcel DekKer; 1994. 173. Stockwell HG, Goldman AL, Lyman GH, et al. Environmental tobacco smoke

PAGE 320

303 and lung cancer risk in nonsmoking wo men. J Natl Cancer Inst. Sep 16 1992;84(18):1417 1422. 174. Henschke CI, Yip R, Miettinen OS. Women's susceptibility to tobacco carcinogens and survival after diagnosis of lung cancer. Jama. Jul 12 2006;296(2):180 184. 175. Willsie SK, Foreman MG. Disparit ies in lung cancer: focus on Asian Americans and Pacific Islanders, American Indians and Alaska Natives, and Hispanics and Latinos. Clin Chest Med. Sep 2006;27(3):441 452, vi. 176. Kolonel LN, Henderson BE, Hankin JH, et al. A Multiethnic Cohort in Hawaii and Los Angeles: Baseline Characteristics. Am. J. Epidemiol. February 15, 2000 2000;151(4):346 357. 177. Ahrendt SA, Hu Y, Buta M, et al. p53 mutations and survival in stage I non small cell lung cancer: results of a prospective study. J Natl Cancer Inst. Jul 2 2003;95(13):961 970. 178. Gazdar AF. DNA repair and survival in lung cancer -the two faces of Janus. N Engl J Med. Feb 22 2007;356(8):771 773. 179. Page GP, Green JL, Lackland D. Epidemiology of lung cancer with special reference to genetics, bioassa ys, women, and developing countries. Semin Respir Crit Care Med. 2000;21(5):365 373. 180. Pisani P, Srivatanakul P, Randerson Moor J, et al. GSTM1 and CYP1A1 polymorphisms, tobacco, air pollution, and lung cancer: a study in rural Thailand. Cancer Epidemio l Biomarkers Prev. Apr 2006;15(4):667 674. 181. Schwartz AG, Yang P, Swanson GM. Familial Risk of Lung Cancer among Nonsmokers and Their Relatives. Am. J. Epidemiol. September 15, 1996 1996;144(6):554 562. 182. Schwartz AG, Rothrock M, Yang P, Swanson GM. Increased cancer risk among relatives of nonsmoking lung cancer cases. Genet Epidemiol. 1999;17(1):1 15. 183. Lichtenstein P, Holm NV, Verkasalo PK, et al. Environmental and Heritable Factors in the Causation of Cancer -Analyses of Cohorts of Twins from Sweden, Denmark, and Finland 10.1056/NEJM200007133430201. N Engl J Med. July 13, 2000 2000;343(2):78 85. 184. Braun M, Caporaso N, Page W, Hoover R. A cohort study of twins and cancer. Cancer Epidemiol Biomarkers Prev. July 1, 1995 1995;4(5):469 473. 185. Wakai K, Nagata C, Mizoue T, et al. Alcohol Drinking and Lung Cancer Risk: An Evaluation Based on a Systematic Review of Epidemiologic Evidence among the Japanese Population 10.1093/jjco/hyl146. Jpn. J. Clin. Oncol. March 1, 2007 2007:hyl146. 186. Prescott E, Grnbk M, Becker U, Srensen TIA. Alcohol Intake and the Risk of Lung Cancer: Influence of Type of Alcoholic Beverage. Am. J. Epidemiol. March 1, 1999 1999;149(5):463 470. 187. Riboli E, Norat T. Epidemiologic evidence of the protective effect of frui t and vegetables on cancer risk. Am J Clin Nutr. Sep 2003;78(3 Suppl):559S 569S.

PAGE 321

304 188. Smith Warner SA, Spiegelman D, Yaun SS, et al. Fruits, Vegetables and Lung Cancer: A Pooled Analysis of Cohort Studies. Int J Cancer. Dec 20 2003;107(6):1001 1011. 189. A ICR. American Institute for Cancer Research Expert Panel: Food, Nutrition, and the Prevention of Lung Cancer A Global Perspective 1997. 190. Brennan P, Hsu CC, Moullan N, et al. Effect of cruciferous vegetables on lung cancer in patients stratified by ge netic status: a mendelian randomisation approach. Lancet. Oct 29 Nov 4 2005;366(9496):1558 1560. 191. Skuladottir H, Tjoenneland A, Overvad K, et al. Does insufficient adjustment for smoking explain the preventive effects of fruit and vegetables on lung ca ncer? Lung Cancer. Jul 2004;45(1):1 10. 192. Feskanich D, Ziegler RG, Michaud DS, et al. Prospective study of fruit and vegetable consumption and risk of lung cancer among men and women. J Natl Cancer Inst. Nov 15 2000;92(22):1812 1823. 193. Swanson CA, Ma o BL, Li JY, et al. Dietary determinants of lung cancer risk: results from a case control study in Yunnan Province, China. Int J Cancer. Apr 1 1992;50(6):876 880. 194. Knekt P, Jarvinen R, Seppanen R, et al. Dietary antioxidants and the risk of lung cancer Am J Epidemiol. Sep 1 1991;134(5):471 479. 195. De Waart F, Schouten E, Stalenhoef A, Kok F. Serum carotenoids, {{alpha}} tocopherol and mortality risk in a prospective study among Dutch elderly. Int. J. Epidemiol. February 1, 2001 2001;30(1):136 143. 19 6. Goodman GE, Thornquist MD, Balmes J, et al. The Beta Carotene and Retinol Efficacy Trial: incidence of lung cancer and cardiovascular disease mortality during 6 year follow up after stopping beta carotene and retinol supplements. J Natl Cancer Inst. Dec 1 2004;96(23):1743 1750. 197. Duffield Lillico AJ, Begg CB. Reflections on the landmark studies of beta carotene supplementation. J Natl Cancer Inst. Dec 1 2004;96(23):1729 1731. 198. ATBC. The effect of vitamin E and beta carotene on the incidence of lun g cancer and other cancers in male smokers. The Alpha Tocopherol, Beta Carotene Cancer Prevention Study Group. N Engl J Med. Apr 14 1994;330(15):1029 1035. 199. Stram DO, Huberman M, Wu AH. Is residual confounding a reasonable explanation for the apparent protective effects of beta carotene found in epidemiologic studies of lung cancer in smokers? Am J Epidemiol. Apr 1 2002;155(7):622 628. 200. Omenn GS, Goodman GE, Thornquist MD, et al. Risk factors for lung cancer and for intervention effects in CARET, th e Beta Carotene and Retinol Efficacy Trial. J Natl Cancer Inst. Nov 6 1996;88(21):1550 1559. 201. Hennekens CH, Buring JE, Manson JE, et al. Lack of effect of long term supplementation with beta carotene on the incidence of malignant neoplasms and cardiova scular disease. N Engl J Med. May 2 1996;334(18):1145 1149. 202. Albanes D, Heinonen OP, Taylor PR, et al. Alpha Tocopherol and beta carotene

PAGE 322

305 supplements and lung cancer incidence in the alpha tocopherol, beta carotene cancer prevention study: effects of b ase line characteristics and study compliance. J Natl Cancer Inst. Nov 6 1996;88(21):1560 1570. 203. Wynder EL, Hebert JR, Kabat GC. Association of dietary fat and lung cancer. J Natl Cancer Inst. Oct 1987;79(4):631 637. 204. De Stefani E, Fontham ET, Chen V, et al. Fatty foods and the risk of lung cancer: a case control study from Uruguay. Int J Cancer. May 29 1997;71(5):760 766. 205. Xie JX, Lesaffre E, Kesteloot H. The relationship between animal fat intake, cigarette smoking, and lung cancer. Cancer Cau ses Control. Mar 1991;2(2):79 83. 206. Alavanja MC, Field RW, Sinha R, et al. Lung cancer risk and red meat consumption among Iowa women. Lung Cancer. Oct 2001;34(1):37 46. 207. Hursting SD, Thornquist M, Henderson MM. Types of dietary fat and the incidenc e of cancer at five sites. Prev Med. May 1990;19(3):242 253. 208. Knekt P, Seppanen R, Jarvinen R, et al. Dietary cholesterol, fatty acids, and the risk of lung cancer among men. Nutr Cancer. 1991;16(3 4):267 275. 209. Mohr DL, Blot WJ, Tousey PM, Van Dore n ML, Wolfe KW. Southern cooking and lung cancer. Nutr Cancer. 1999;35(1):34 43. 210. Swanson CA, Brown CC, Sinha R, Kulldorff M, Brownson RC, Alavanja MC. Dietary fats and lung cancer risk among women: the Missouri Women's Health Study (United States). Ca ncer Causes Control. Nov 1997;8(6):883 893. 211. Veierod MB, Laake P, Thelle DS. Dietary fat intake and risk of lung cancer: a prospective study of 51,452 Norwegian men and women. Eur J Cancer Prev. Dec 1997;6(6):540 549. 212. Goodman MT, Hankin JH, Wilken s LR, Kolonel LN. High fat foods and the risk of lung cancer. Epidemiology. Jul 1992;3(4):288 299. 213. Henley SJ, Flanders WD, Manatunga A, Thun MJ. Leanness and lung cancer risk: fact or artifact? Epidemiology. May 2002;13(3):268 276. 214. Xiang Y, Gao Y Zhong L, et al. [A case control study on relationship between body mass index and lung cancer in non smoking women]. Zhonghua Yu Fang Yi Xue Za Zhi. Jan 1999;33(1):9 12. 215. Wannamethee G, Shaper AG. Body weight and mortality in middle aged British men: impact of smoking. Bmj. Dec 16 1989;299(6714):1497 1502. 216. Knekt P, Heliovaara M, Rissanen A, et al. Leanness and lung cancer risk. Int J Cancer. Sep 9 1991;49(2):208 213. 217. Kark JD, Yaari S, Rasooly I, Goldbourt U. Are lean smokers at increased ris k of lung cancer? The Israel Civil Servant Cancer Study. Arch Intern Med. Dec 11 25 1995;155(22):2409 2416. 218. Kanashiki M, Sairenchi T, Saito Y, Ishikawa H, Satoh H, Sekizawa K. Body mass index and lung cancer: a case control study of subjects participa ting in a mass screening program. Chest. Sep 2005;128(3):1490 1496. 219. Kabat GC, Wynder EL. Body mass index and lung cancer risk. Am J Epidemiol.

PAGE 323

306 Apr 1 1992;135(7):769 774. 220. Eichholzer M, Bernasconi F, Jordan P, Stahelin HB. Body mass index and the r isk of male cancer mortality of various sites: 17 year follow up of the Basel cohort study. Swiss Med Wkly. Jan 8 2005;135(1 2):27 33. 221. Ogden C, et. al. Mean Body Weight, Height, and Body Mass Index, United States 1960 2002. Advanced Data from Vital an d Health Statistics http://www.cdc.gov/nchs/data/ad/ad347 222. Drinkard CR, Sellers TA, Potter JD, et al. Association of body mass index and body fat distribution with risk of lung cancer in older wom en. Am J Epidemiol. Sep 15 1995;142(6):600 607. 223. Goodman MT, Wilkens LR. Relation of body size and the risk of lung cancer. Nutr Cancer. 1993;20(2):179 186. 224. Gordis L. Epidemiology : W. B. Saunders Company; 2000. 225. NIOSH. NIOSH CARCINOGEN LIST. http://www.cdc.gov/niosh/npotocca.html 226. Checkoway H, Pearce N, Crawford Brown D. Research Mehtods in Occupational Epidemiology Second ed: Oxford Univesity Press; 1994. 227. Radford EP, Renard KG. Lung cancer in Swedish iron miners exposed to low doses of radon daughters. N Engl J Med. Jun 7 1984;310(23):1485 1494. 228. Hinds MW, Kolonel LN, Lee J. Application of a job exposure matrix to a case control study of lung cancer. J Natl Cancer Inst. Aug 1985;75(2):193 197. 229. Schwartz AG, Prysak GM, Murphy V, et al. Nuclear estrogen receptor beta in lung cancer: expression and survival differences by sex. Clin Cancer Res. Oct 15 2005;11(20):7280 7287. 230. Han W, Pentecost BT, Pietropaolo RL, Fasco MJ, Spivack SD. Estrogen receptor alpha increases basal and cigarette smoke extract induced expression of CYP1A1 and CYP1B1, but not GSTP1, in normal human bronchial epithelial cells. Mol Carcinog. Nov 2005;44(3):202 211. 231. Mao Y, Hu J, Ugnat A M, Semenciw R, Fincham S. Socioeconomic status and lung cancer risk in Canada. Int. J. Epidemiol. August 1, 2001 2001;30(4):809 817. 232. Singh GK, Miller BA, Hankey BF. Changing Area Socioeconomic Patterns in U.S. Cancer Mortality, 1950 1998: Part II -Lung and Colore ctal Cancers 10.1093/jnci/94.12.916. J. Natl. Cancer Inst. June 19, 2002 2002;94(12):916 925. 233. Vineis P, Hoek G, Krzyzanowski M, et al. Lung cancers attributable to environmental tobacco smoke and air pollution in non smokers in different European coun tries: a prospective study. Environmental Health. 2007;6(1):7. 234. Ramanakumar AV, Parent M E, Siemiatycki J. Risk of Lung Cancer from Residential Heating and Cooking Fuels in Montreal, Canada 10.1093/aje/kwk117. Am. J. Epidemiol. March 15, 2007 2007;165( 6):634 642. 235. Schabath MB, Delclos GL, Martynowicz MM, et al. Opposing Effects of Emphysema, Hay Fever, and Select Genetic Variants on Lung Cancer Risk 10.1093/aje/kwi063. Am. J. Epidemiol. March 1, 2005 2005;161(5):412 422.

PAGE 324

307 236. Angeletti C, Mussi A, J anni A, et al. Second primary lung cancer and relapse: treatment and follow up. Eur J Cardiothorac Surg. July 1, 1995 1995;9(11):607 611. 237. Carney DN. Lung cancer -time to move on from chemotherapy. N Engl J Med. Jan 10 2002;346(2):126 128. 238. Venuta F, Anile M, Diso D, et al. Operative complications and early mortality after induction therapy for lung cancer. European Journal of Cardio Thoracic Surgery. 2006;In Press, Corrected Proof. 239. Cho BC, Craig T. More optimal dose distributions for moving lu ng tumours: a planning study. Radiother Oncol. Apr 2006;79(1):122 130. 240. Giraud P, Simon L, Saliou M, et al. [Respiratory gated radiotherapy: the 4D radiotherapy]. Bull Cancer. Jan 2005;92(1):83 89. 241. Giraud P, Reboul F, Clippe S, et al. [Respiration gated radiotherapy: current techniques and potential benefits]. Cancer Radiother. Nov 2003;7 Suppl 1:15s 25s. 242. Underberg RWM, Lagerwaard FJ, Slotman BJ, Cuijpers JP, Senan S. Benefit of respiration gated stereotactic radiotherapy for stage I lung canc er: An analysis of 4DCT datasets. International Journal of Radiation Oncology*Biology*Physics. 2005/6/1 2005;62(2):554 560. 243. Sheski FD, Mathur PN. Endoscopic Treatment of Early Stage Lung Cancer. Cancer Control. Jan./Feb. 2000 2000;7(1). 244. Chute JP, Chen T, Feigal E, Simon R, Johnson BE. Twenty years of phase III trials for patients with extensive stage small cell lung cancer: perceptible progress. J Clin Oncol. Jun 1999;17(6):1794 1801. 245. Pirker R, Ulsperger E, Messner J, et al. Achieving full do se, on schedule administration of ACE chemotherapy every 14 days for the treatment of patients with extensive small cell lung cancer. Lung. Sep Oct 2006;184(5):279 285. 246. O'Brien ME, Ciuleanu TE, Tsekov H, et al. Phase III trial comparing supportive car e alone with supportive care with oral topotecan in patients with relapsed small cell lung cancer. J Clin Oncol. Dec 1 2006;24(34):5441 5447. 247. Hong YS, Lee HR, Park S, et al. Three week schedule of irinotecan plus cisplatin in patients with previously untreated extensive stage small cell lung cancer. Br J Cancer. Dec 18 2006;95(12):1648 1652. 248. Saito H, Takada Y, Ichinose Y, et al. Phase II study of etoposide and cisplatin with concurrent twice daily thoracic radiotherapy followed by irinotecan and c isplatin in patients with limited disease small cell lung cancer: West Japan Thoracic Oncology Group 9902. J Clin Oncol. Nov 20 2006;24(33):5247 5252. 249. Board RE, Thatcher N, Lorigan P. Novel therapies for the treatment of small cell lung cancer: a time for cautious optimism? Drugs. 2006;66(15):1919 1931. 250. Bepler G. Using translational research to tailor the use of chemotherapy in the treatment of NSCLC. Lung Cancer. Oct 2005;50 Suppl 1:S13 14. 251. Bepler G, Kusmartseva I, Sharma S, et al. RRM1 modu lated in vitro and in vivo

PAGE 325

308 efficacy of gemcitabine and platinum in non small cell lung cancer. J Clin Oncol. Oct 10 2006;24(29):4731 4737. 252. Olaussen KA, Dunant A, Fouret P, et al. DNA Repair by ERCC1 in Non Small Cell Lung Cancer and Cisplatin Based Ad juvant Chemotherapy 10.1056/NEJMoa060570. N Engl J Med. September 7, 2006 2006;355(10):983 991. 253. Smythe WR. Treatment of Stage I Non small Cell Lung Carcinoma 10.1378/chest.123.1_suppl.181S. Chest. January 1, 2003 2003;123(90010):181S 187. 254. Ohashi R, Takahashi K, Miura K, Ishiwata T, Sakuraba S, Fukuchi Y. Prognostic factors in patients with inoperable non small cell lung cancer -an analysis of long term survival patients. Gan To Kagaku Ryoho. Nov 2006;33(11):1595 1602. 255. Lam S, Hung JY, Kennedy SM, et al. Detection of dysplasia and carcinoma in situ by ratio fluorometry. Am Rev Respir Dis. Dec 1992;146(6):1458 1461. 256. Lam S, Kennedy T, Unger M, et al. Localization of bronchial intraepithelial neoplastic lesions by fluorescence bronchoscopy. Ch est. Mar 1998;113(3):696 702. 257. Gerasin VA, Levashov YN, Shafirovscy BB, Berezin YD, Jurba VM, Palamarchuk GF. Bronchoscopic laser photocoagulation of superficial cancer of the bronchi. Chest. Jul 1990;98(1):235 236. 258. NAACCR. North American Associat ion of Central Cancer Registries. http://www.naaccr.org/ 259. NPCR. National Program of Cancer Registries (NPCR). http://www.cdc.gov/cancer/npcr/ 260. Porojnicu AC Robsahm TE, Dahlback A, et al. Seasonal and geographical variations in lung cancer prognosis in Norway Does Vitamin D from the sun play a role? Lung Cancer. Mar 2007;55(3):263 270. 261. Stewart SL, King JB, Thompson TD, Friedman C, Wingo PA. Cancer morta lity surveillance -United States, 1990 2000. MMWR Surveill Summ. Jun 4 2004;53(3):1 108. 262. Grann V, Troxel AB, Zojwalla N, Hershman D, Glied SA, Jacobson JS. Regional and racial disparities in breast cancer specific mortality. Soc Sci Med. Jan 2006;62(2 ):337 347. 263. Edwards BK, Brown ML, Wingo PA, et al. Annual report to the nation on the status of cancer, 1975 2002, featuring population based trends in cancer treatment. J Natl Cancer Inst. Oct 5 2005;97(19):1407 1427. 264. Folsom AR, French SA, Zheng W, Baxter JE, Jeffery RW. Weight variability and mortality: the Iowa Women's Health Study. Int J Obes Relat Metab Disord. Aug 1996;20(8):704 709. 265. Osborne C, Ostir GV, Du X, Peek MK, Goodwin JS. The influence of marital status on the stage at diagnosis treatment, and survival of older women with breast cancer. Breast Cancer Res Treat. Sep 2005;93(1):41 47.

PAGE 326

309 266. US_Census_Bureau. 2007. 267. BUFFLER PA, COOPER SP, STINNETT S, et al. AIR POLLUTION AND LUNG CANCER MORTALITY IN HARRIS COUNTY, TEXAS, 1979 19 81. Am. J. Epidemiol. October 1, 1988 1988;128(4):683 699.

PAGE 327

3 10 APPENDICES

PAGE 328

311 A ppendix I : State Demographics T able 70 : G eographic Area: Florida Profile of Sex and Age Characteristics: 2000 Subject Number Percent Total population 15,982,378 100.0 SEX AND AGE Male 7,797,715 48.8 Female 8,184,663 51.2 Under 5 years 945,823 5.9 5 to 9 years 1,031,718 6.5 10 to 14 years 1,057,024 6.6 15 to 19 years 1,014,067 6.3 20 to 24 years 928,310 5.8 25 to 34 years 2,084,100 13.0 35 to 44 years 2,485,247 15.5 45 to 54 years 2,069,479 12.9 55 to 59 years 821,517 5.1 60 to 64 years 737,496 4.6 65 to 74 years 1,452,176 9.1 75 to 84 years 1,024,134 6.4 85 years and over 331,287 2.1 Median age (years) 38.7 (X) 18 years and over 12,336,038 77.2 Male 5,926,729 37.1 Female 6,409,309 40.1 21 years and over 11,736,378 73.4 62 years and over 3,245,806 20.3 65 years and over 2,807,597 17.6 Male 1,216,647 7.6 Female 1,590,950 10.0 Legend: (X) Not Applicable Source: U.S. Census Bureau, Census 2000.

PAGE 329

312 Table 71 : Geographic Area: Idaho Profile of Sex and Age Characteristics: 2000 Subject Number Percent Total population 1,293,953 100.0 SEX AND AGE Male 648,660 50.1 Female 645,293 49.9 Under 5 years 97,643 7.5 5 to 9 years 100,756 7.8 10 to 14 years 104,608 8.1 15 to 19 years 110,858 8.6 20 to 24 years 93,994 7.3 25 to 34 years 169,433 13.1 35 to 44 years 192,968 14.9 45 to 54 years 170,248 13.2 55 to 59 years 60,024 4.6 60 to 64 years 47,505 3.7 65 t o 74 years 75,970 5.9 75 to 84 years 51,889 4.0 85 years and over 18,057 1.4 Median age (years) 33.2 (X) 18 years and over 924,923 71.5 Male 458,934 35.5 Female 465,989 36.0 21 years and over 860,220 66.5 62 years and over 173,097 13. 4 65 years and over 145,916 11.3 Male 64,161 5.0 Female 81,755 6.3 Legend: (X) Not Applicable Source: U.S. Census Bureau, Census 2000.

PAGE 330

313 Table 72 : Geographic Area: Indiana Profile of Sex and Age Characteristics: 2000 S ubject Number Percent Total population 6,080,485 100.0 SEX AND AGE Male 2,982,474 49.0 Female 3,098,011 51.0 Under 5 years 423,215 7.0 5 to 9 years 443,273 7.3 10 to 14 years 443,416 7.3 15 to 19 years 453,482 7.5 20 to 24 years 425,731 7.0 25 to 34 years 831,125 13.7 35 to 44 years 960,703 15.8 45 to 54 years 816,865 13.4 55 to 59 years 294,169 4.8 60 to 64 years 235,675 3.9 65 to 74 years 395,393 6.5 75 to 84 years 265,880 4.4 85 years and over 91,558 1.5 Median age (years) 35.2 (X) 18 years and over 4,506,089 74.1 Male 2,174,756 35.8 Female 2,331,333 38.3 21 years and over 4,221,426 69.4 62 years and over 888,688 14.6 65 years and over 752,831 12.4 Male 303,797 5.0 Female 449,034 7.4 Legend: (X) Not A pplicable Source: U.S. Census Bureau, Census 2000.

PAGE 331

314 Table 73 : Geographic Area: Massachusetts Profile of Sex and Age Characteristics: 2000 Subject Number Percent Total population 6,349,097 100.0 SEX AND AGE Male 3,058,816 48.2 Fe male 3,290,281 51.8 Under 5 years 397,268 6.3 5 to 9 years 430,861 6.8 10 to 14 years 431,247 6.8 15 to 19 years 415,737 6.5 20 to 24 years 404,279 6.4 25 to 34 years 926,788 14.6 35 to 44 years 1,062,995 16.7 45 to 54 years 873,353 13.8 55 to 5 9 years 310,002 4.9 60 to 64 years 236,405 3.7 65 to 74 years 427,830 6.7 75 to 84 years 315,640 5.0 85 years and over 116,692 1.8 Median age (years) 36.5 (X) 18 years and over 4,849,033 76.4 Male 2,289,671 36.1 Female 2,559,362 40.3 21 years and over 4,587,935 72.3 62 years and over 997,277 15.7 65 years and over 860,162 13.5 Male 341,539 5.4 Female 518,623 8.2 Legend: (X) Not Applicable Source: U.S. Census Bureau, Census 2000.

PAGE 332

315 Table 74 : Geogr aphic Area: Nebraska Profile of Sex and Age Characteristics 2000 Subject Number Percent Total population 1,711263 100.0 SEX AND AGE Male 843,351 49.3 Female 867,912 50.7 Under 5 years 117,048 6.8 5 to 9 years 123,445 7.2 10 to 14 years 128,934 7.5 15 to 19 years 134,909 7.9 20 to 24 years 120,331 7.0 25 to 34 years 223,273 13.0 35 to 44 years 263,834 15.4 45 to 54 years 225,754 13.2 55 to 59 years 77,584 4.5 60 to 64 years 63,956 3.7 65 to 74 years 115,699 6.8 75 to 84 years 82,543 4.8 85 years and over 33,953 2.0 Median age (years) 35.3 (X) 18 years and over 1,261,021 73.7 Male 612,965 35.8 Female 648,056 37.9 21 years and over 1,180,859 69.0 62 years and over 269,893 15.8 65 years and over 232,195 13.6 Male 95,630 5.6 Female 136,565 8.0 Legend: (X) Not Applicable Source: U.S. Census Bureau, Census 2000.

PAGE 333

316 Table 7 5 : Geographic Area: Oregon Profile of Sex and Age Characteristics: 2000 Subject Number Percent Total population 3,42 1,399 100.0 SEX AND AGE Male 1,696,550 49.6 Female 1,724,849 50.4 Under 5 years 223,005 6.5 5 to 9 years 234,474 6.9 10 to 14 years 242,098 7.1 15 to 19 years 244,427 7.1 20 to 24 years 230,406 6.7 25 to 34 years 470,695 13.8 35 to 44 years 52 6,574 15.4 45 to 54 years 507,155 14.8 55 to 59 years 173,008 5.1 60 to 64 years 131,380 3.8 65 to 74 years 219,342 6.4 75 to 84 years 161,404 4.7 85 years and over 57,431 1.7 Median age (years) 36.3 (X) 18 years and over 2,574,873 75.3 M ale 1,262,405 36.9 Female 1,312,468 38.4 21 years and over 2,429,348 71.0 62 years and over 513,663 15.0 65 years and over 438,177 12.8 Male 186,477 5.5 Female 251,700 7.4 Legend: (X) Not Applicable Source: U.S. Census Bureau, Censu s 2000.

PAGE 334

317 Table 76 : Geographic Area: Rhode Island Profile of Sex and Age Characteristics: 2000 Subject Number Percent Total population 1,048,319 100.0 SEX AND AGE Male 503,635 48.0 Female 544,684 52.0 Under 5 years 63,896 6.1 5 to 9 years 71,905 6.9 10 to 14 years 71,370 6.8 15 to 19 years 75,445 7.2 20 to 24 years 71,813 6.9 25 to 34 years 140,326 13.4 35 to 44 years 170,310 16.2 45 to 54 years 141,863 13.5 55 to 59 years 49,982 4.8 60 to 64 years 39,007 3.7 65 to 7 4 years 73,684 7.0 75 to 84 years 57,821 5.5 85 years and over 20,897 2.0 Median age (years) 36.7 (X) 18 years and over 800,497 76.4 Male 376,436 35.9 Female 424,061 40.5 21 years and over 748,445 71.4 62 years and over 175,111 16.7 65 years and over 152,402 14.5 Male 60,002 5.7 Female 92,400 8.8 Legend: (X) Not Applicable Source: U.S. Census Bureau, Census 2000.

PAGE 335

318 Table 7 7 : Geographic Area: South Carolina Profile of Sex and Age Characteristics: 2 000 Subject Number Percent Total population 4,012,012 100.0 SEX AND AGE Male 1,948,929 48.6 Female 2,063,083 51.4 Under 5 years 264,679 6.6 5 to 9 years 285,243 7.1 10 to 14 years 290,479 7.2 15 to 19 years 295,377 7.4 20 to 24 years 281,714 7. 0 25 to 34 years 560,831 14.0 35 to 44 years 625,124 15.6 45 to 54 years 550,321 13.7 55 to 59 years 206,762 5.2 60 to 64 years 166,149 4.1 65 to 74 years 270,048 6.7 75 to 84 years 165,016 4.1 85 years and over 50,269 1.3 Median age (years) 35. 4 (X) 18 years and over 3,002,371 74.8 Male 1,432,413 35.7 Female 1,569,958 39.1 21 years and over 2,814,131 70.1 62 years and over 581,573 14.5 65 years and over 485,333 12.1 Male 196,734 4.9 Female 288,599 7.2 Legend: (X) Not Applicable Source: U.S. Census Bureau, Census 2000.

PAGE 336

319 Appendix II: Lung Cancer Distribution Tables Table 41: Lung Cancer Distribution Treatment Group versus Race Treatment Group Race Frequency Percent Radiation White 3921 90.1 I Black 401 9.2 Other 29 0.7 Chemotherapy White 6026 93.1 II Black 383 5.9 Other 63 1.0 Surgery White 11967 94.0 III Black 659 5.2 Other 102 0.8 Radiation + Surgery White 990 93.1 IV Black 64 6.0 Other 9 0.9 Radiation + Chemotherapy White 7262 91.3 V Black 627 7.9 Other 66 0.8 Surgery + Chemotherapy White 1166 93.4 VI Black 75 6.0 Other 8 0.6 Radiation + Surgery + White 1254 93.0 Chemotherapy Black 81 6.0 VII Other 13 1.0 No Radiation, Surgery, Whi te 8872 91.5 and/or Chemotherapy Black 752 7.8 VIII Other 73 0.8 There were no obvious differences in treatment groups versus and distribution of race see Table 41 in Appendix B. The majority of lung cancer cases are White ranging from 91.3% of al l lung cancer cases in Group V (Radiation and Chemotherapy) to 94% of

PAGE 337

320 contained the least amount of lung cancer cases for each treatment group with each Treatment Group ha ving a minimum of approximately one percent within each treatment classification (I VIII).

PAGE 338

321 Table 42: Lung Cancer Distribution Treatment Group vs. Marital Status at Diagnosis Treatment Group Marital Status Frequency Percent Single 482 11.1 Radiation Married 2309 53.1 I Separated 46 1.1 Divorced 486 11.2 Widowed 1028 23.6 Single 652 10.1 Chemotherapy Married 3879 59.9 II Separated 63 1.0 Divorced 734 11.3 Widowed 1144 17.7 Single 1014 8.0 Surgery Married 8014 63.0 III Separated 76 0.6 Divorced 1301 10.2 Widowed 2323 18.3 Single 93 8.8 Radiation + Surgery Married 693 65.2 IV Separated 6 0.6 Divorced 108 10.2 Widowed 163 15.3 Radiation + Single 784 9.9 Chemotherapy Married 4991 62.7 V Separated 86 1.1 Divorced 982 12.3 Widowed 1112 14.0 Surgery + Single 103 8.3 Chemotherapy Married 873 69.9 VI Separated 9 0.7 Divorced 122 9.8 Radiation + Surgery + Single 98 7.3 Chemotherapy Married 934 69.3 VII Separated 11 0.8 Divorced 148 11.0 Widowed 157 11.7

PAGE 339

322 No Radiation, Surgery, Single 1201 12.4 and/or Chemotherapy Married 5066 52.2 VIII Separated 70 0.7 Divorced 1039 10.7 Widowed 2321 23.9 For all treatment groups in Table 42 (Treatme nt Group vs. Marital Status at Diagnosis) the greatest percentage of the lung cancer cases were married at the time of diagnosis ranging from 52.2 percent for Treatment Group VIII to maximum percentage of 69.9 percent for surgical and chemotherapy, Group VI.

PAGE 340

323 Table 43: Lung Cancer Distribution Treatment Group vs. Age Group at Diagnosis Treatment Group Age Group at Diagnosis Frequency Percent > 40 < 50 189 4.3 Radiation > 50 < 60 573 13.2 I > 60 < 70 1093 25. 1 > 70 < 80 1675 38.5 > 80 < 90 821 18.9 > 40 < 50 404 6.2 Chemotherapy > 50 < 60 1192 18.4 II > 60 < 70 2061 31.8 > 70 < 80 2225 34.4 > 80 < 90 590 9.1 > 40 < 50 442 3.5 Surgery > 50 < 60 1662 13.1 III > 60 < 70 3893 30.6 > 70 < 80 5366 42.2 > 80 < 90 1365 10.7 > 40 < 50 59 5.6 Radiation + Surgery > 50 < 60 178 16.8 IV > 60 < 70 358 33.7 > 70 < 80 395 37.2 > 80 < 90 73 6.9 Radiation + > 40 < 50 666 8.4 Chemo therapy > 50 < 60 1743 21.9 V > 60 < 70 2705 34.0 > 70 < 80 2361 29.7 > 80 < 90 480 6.0 Surgery + > 40 < 50 92 7.4 Chemotherapy > 50 < 60 287 23.0 VI > 60 < 70 479 38.4 > 70 < 80 350 28.0

PAGE 341

324 > 80 < 90 41 3.3 Radiation + Surgery + > 40 < 50 133 9.9 Chemotherapy > 50 < 60 345 25.6 VII > 60 < 70 485 36.0 > 70 < 80 366 27.2 > 80 < 90 19 1.4 No Radiation, Surgery, > 40 < 50 367 3.8 and/or Chemotherapy > 50 < 60 1199 12.4 VIII > 60 < 70 2462 25 .4 > 70 < 80 3666 37.8 > 80 < 90 2003 20.7 The greatest percentage for age group IV ( > 70 < 80) were found in Treatment Groups I (38.5), II (34.4), III (42.2), IV (37.2), V (29.7), and VIII (37.8). The remaining two treatment groups had the h ighest percentage in the third age group, > 60 < 70, Treatment Group VI (38.4) and Treatment Group VII (36.0).

PAGE 342

325 A ppendix I II : Chemotherapy Agents Table 7 8 : Chemotherapy Agents for Lung Can c er Source: Alexander Spira, M.D., Ph.D., and David S. Ettinger, M.D.; N Engl J Med 2004; 350:379 92.

PAGE 343

326 A ppendix IV : Calculation of the Overall Interaction Effect Calculation of the overall effect for the treatment type received (variables extracted for the Multinomial Logistic Regression Model ). Stage The i nteraction terms containing stage with the main effects are included in the following equation that was originally extracted from the full model. Logit (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 g ender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI + 10 morphologyI + 1 1 raceI The terms that contain stage are identified and include the main effect: Logit (Y= Treatment | X) = + 1 genderI + 2 stageI + 3 marita l_statusI + 4 gradeI + 5 age_groupI + 6 genderI*stageI + 7 genderI* marital_statusI + 8 stageI*gradeI + 9 stageI* age_groupI + 10 morphologyI + 1 1 raceI The following equation results: Logit (Y= Treatment | X) = + 1 genderI + 2 stageI + 4 gra deI + 5 age_groupI + 6 genderI*stageI + 8 stageI*gradeI + 9 stageI* age_groupI These terms must be evaluated separately to assess the effect of stage on the outcome. In other words, as there are three interaction terms with stage, three separate equat ions containing stage are calculated for gender, grade and age group. In Part I below, the example treatment is radiation, stageI (stageI = stage 1 coded as 1 and stage IV coded as 0 (reference). Part II will examine stage and grade and Part III will ass ess stage and age group. Part I: Evaluating stage and gender Stage I: Logit (Y= Radiation | stageI = 1, genderI ) = + 1 gender I + 2 + 4 gradeI + 5 age_groupI + 6 genderI + 8 *gradeI + 9 age_groupI

PAGE 344

327 Stage IV: Logit (Y= Radiation | stageI = 0, genderI ) = + 1 gender I + 4 gradeI + 5 age_groupI Subtracting stage I from stage IV, the following is given: Logi t (Y= Radiation | stageI = 1, genderI ) = 2 + 6 genderI + 8 *gradeI + 9 age_groupI Looking at females as compared to males with gender = 1 for females and gender = 0 for males. Logit (Y= Radiation | stageI = 1, genderI = 1 ) = 2 + 6 + 8 *gradeI + 9 age_groupI Logit (Y= Radiation | stageI = 1, genderI = 0 ) = 2 + 8 *gradeI + 9 age_groupI The Odds Ratio for females with stage 1 lung cancer (grade and age group are fixed or controlled for) is given as: OR = exp ( 2 + 6 ) Part I I : Evaluating stage and grade Stage I: Logit (Y= Radiation | stageI = 1, genderI ) = + 1 gender I + 2 + 4 gradeI + 5 age_groupI + 6 genderI + 8 *gradeI + 9 age_groupI Stage IV: Logit (Y= Radiation | stageI = 0, genderI ) = + 1 gender I + 4 grad eI + 5 age_groupI Subtracting stage I from stage IV, the following is given: Logit (Y= Radiation | stageI = 1, genderI ) = 2 + 6 genderI + 8 *gradeI + 9 age_groupI Looking at grade I as compared to grade IV with gradeI = 1 for grade I and gradeI = 0 for grade IV. Logit (Y= Radiation | stageI = 1, gradeI = 1 ) = 2 + 6 genderI + 8 + 9 age_groupI Logit ( Y= Radiation | stageI = 1, gradeI = 0 ) = 2 + 6 genderI + 9 age_groupI The Odds Ratio for stage 1 grade 1 lung cancer (gender and age group are fixed or

PAGE 345

328 controlled for) is given as: OR = exp ( 2 + ) Part I I I : Evaluating stage and age group Stage I: Logit (Y= Radiation | stageI = 1, age_groupI ) = + 1 gender I + 2 + 4 gradeI + 5 age_groupI + 6 genderI + 8 *gradeI + 9 age_groupI Stage IV: Logit (Y= Radiation | stageI = 0, age_groupI = + 1 gender I + 4 gradeI + 5 age_groupI Subtracting stage I from stage IV, the following is given: Logit (Y= Radiation | stageI = 1, age_groupI ) = 2 + 6 genderI + 8 *gradeI + 9 age_groupI Looking at grade I as compared to grade IV with gradeI = 1 for grade I and gradeI = 0 for grade IV. Logit (Y= Radiation | stageI = 1, age_groupI = 1 ) = 2 + 6 genderI + 8 *gradeI + 9 Logit ( Y= Radiation | stageI = 1, age_groupI = 0 ) = 2 + 6 genderI + 8 *gradeI + The Odds Ratio for stage 1 in age group I (gender and grade are fixed or controlled for) is given as: OR = exp ( 2 + ) Grade Given the equation extracted from the full model: Logit (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 gender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI + 10 morphologyI + 1 1 raceI The terms that contain grade are identified and include the main effec t: Logit (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 gender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI The following equation results:

PAGE 346

329 Logit (Y= Treatment | X) = + 2 stageI + 4 gradeI + 8 stageI*gradeI Next, the effect of grade on the probability of receiving radiation therapy as a treatment, given that the patient is at stageI, is determined as: Grade I : Logit (Y=Radiation| grade =1, s tage I) = + 2 s tageI + 4 + 8 *stageI Grade IV : Logit (Y=Radiation| grade =0, stage I) = + 2 stageI By subtracting the Logit for grade IV from Logit for grade I, the following equation is given as : Logit (Y=Radiation | grade =1, stageI) = 4 + 8 *stageI A t the variab le stage I which is coded as 1 for stageI and 0 for stageI (V = reference), the results are given as : Logit (Y=Radiation |gr ade =1, stage I =1) = 4 + 8 Logit (Y=Radiation |g r ade =1, stage I =0) = 4 Estimating the overall effect of grade ( grade I as compared to grade IV) on the probability of receiving radiation treatment, after adjusting for stageI (stage=1) results in the following equation for the Odds Ratio is given as: OR = exp ( 4 + 8 ). Marital Status The interaction terms conta ining marital status with the main effects are included in the following equation that was originally extracted from the full model. Logit (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 gender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI + 10 morphologyI + 1 1 raceI The terms that contain stage are identified and include the main effect: Logit (Y= Treatment | X) = + 1 genderI + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 genderI*stageI + 7 genderI* marital_statusI + 8 stageI*gradeI + 9 stageI* age_groupI

PAGE 347

330 The following equation results: Logit (Y= Treatment | X) = + 1 genderI + 3 marital_statusI + 7 genderI* marital_statusI Evaluatin g marital_statusI for marital status = I (single) and marital_statusI for marital status = V (widowed), the following is given: Marital Status I: Logit (Y= Radiation | marital_statusI = 1, genderI ) = + 1 genderI + 3 + 7 genderI* Marital Status V: L ogit (Y= Radiation | marital_statusI = 0, genderI) = + 1 genderI Subtracting marital status I from marital status IV, the following equation results: Logit (Y= Radiation | marital_statusI = 1, genderI ) = 3 + 7 genderI* Looking at females as comp ared to males with genderI = 1 for females and gradeI = 0 for males. Logit (Y= Radiation | marital_statusI = 1, genderI = 1 ) = 3 + 7 Logit ( Y= Radiation | marital_statusI = 1, genderI = 0 ) = 3 The Odds Ratio for the overall interaction effect giv en marital status for females as compared to males is given as: OR = exp ( 3 + 7 ) Age Group The interaction terms containing age group with the main effects are included in the following equation that was originally extracted from the full model. Lo git (Y= Treatment | X) = + 1 gender I + 2 stageI + 3 marital_statusI + 4 gradeI + 5 age_groupI + 6 gender I *stageI + 7 gender I marital_statusI + 8 stageI*gradeI + 9 stage I age_groupI + 10 morphologyI + 1 1 raceI The terms that contain age group are identified and include the main effect: Logit (Y= Treatment | X) = + 1 genderI + 2 stageI + 3 marital_statusI + 4 gradeI +

PAGE 348

331 5 age_groupI + 6 genderI*stageI + 7 genderI* marital_statusI + 8 stageI*gradeI + 9 stageI* age_groupI The following equation results: Logit (Y= Treatment | X) = + 2 stageI + 5 age_groupI + 9 stageI* age_groupI For the next example, the treatment group still remains as Radiation alone. Evaluating age_groupI for age group I = 1 ( > 40 < 50 year ) and age_groupI for age group V = 0 ( > 80 < 90 years ), the following is given: Marital Status I: Logit (Y= Radiation | age_groupI = 1, stageI ) = + 2 stageI + 5 + 9 stageI Marital Status V: Logit (Y= Radiation | age_groupI = 0, stageI) = + 2 stageI Subtractin g age group I from age group V, the following equation results: Logit (Y= Radiation | age_groupI = 1, stageI ) = 5 + 9 stageI Looking at stage I and stage IV with stageI = 1 for stage I and stageI = 0 for stage IV. Logit (Y= Radiation | marital_status I = 1, stageI = 1 ) = 5 + 9 Logit ( Y= Radiation | marital_statusI = 1, stageI = 0 ) = 5 The Odds Ratio for the overall interaction effect given age group I controlling for stage is given as: OR = exp ( 5 + 9 ) Logit (Y= Radiation | stageI = 1, gr ade =1 ) = + 1 gender + 2 stageI + 4 gradeI + 5 age_groupI + 6 gender*stageI + 8 stageI*gradeI + 9 stage* age_groupI Logit (Y= Radiation | stageI = 1, age_group =1 ) = + 1 gender + 2 stageI + 4 gradeI + 5 age_groupI + 6 gender*stageI + 8 stag eI*gradeI + 9 stage* age_groupI

PAGE 349

332 ABOUT THE AUTHOR My story starts out when I was thirteen (13) years old. My father d ied of c ancer. The unintentional consequences of my career path has lead me on a course through c ancer diagnosis and therapy to where I am today, redesigning my career to include Cancer Care research My education, experiences and professional journey began, after high school, with technical training in Radiology and Nuclear Medicine which are modalities used in the d iagnosis of Cancer. I continued on g rowing professionally with a Masters Degree in Medical Physics the knowledge I gained allowed me to enter into cancer treatment through Radiation Therapy Industry Along the way, in 1972 I married my husband Stan and in 1985 we had a son J Stephen I am very grateful for their help and support during my life, especially their encouragement as I reshape my career once again. Without their confidence, love and caring I could not have made this rewarding personal progress, I thank them. My P hD from the Department of Epidemiology and Biostatistics will make me uniquely qualified to research issues that includ e my education and experience in Radiological Physics. I find Radiation Epidemiology is receiving great respect and becoming a specialty area that I can assimilate easily. Finally, I do not know how to adequately thank, Dr. Mason my Mentor, Advisor, Co Major Professor, and Friend. I would also like to thank my Co Major Professor, Dr. Stockwell, and other committee members, Dr. Dagne and Dr. Zhukov for all their hard work, direction, guidance and sustained confidence throughout my research to complete my PhD.