|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
Gene Expression Profiles and Clinical Parameters for Survival Prediction in Stage II and III Colorectal Cancer Patients by Mubeena Begum, M.D A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Public Health Department of Epidemiology and Biostatistics College of Public Health University of South Florida Major Professor: Thomas J. Mason, Ph.D. James F. Helm, M.D, Ph.D. Heather G. Stockwell, Ph.D. Tao Wang, Ph.D Date of Approval: February 1st, 2006 Keywords: Survival time, Prognostic fact ors, Molecular markers, SAM, mRNA. Copyright 2006, Mubeena Begum
Dedication I dedicate this research to my parents: Amartur Maqhbool Ahmed and Basheera Khatoon for being an inspiration to me all my life.
Acknowledgements I would like to thank my advisor, Dr. Th omas J. Mason, for his tremendous support and invaluable guidance during the time of this thesis work, not only in the research projects, but also in my life beyond scienc e. It has been a great pleas ure and honor to work and study under his mentorship I am grateful to Dr. James Helm for hi s guidance throughout the preparation of my thesis. My many thanks to my thesis comm ittee members Dr. Heather G. Stockwell and Dr. Tao Wang for their valuable suggestions and fo r taking time to serve on my thesis. I would like to extend my appreciation to our collaborators at Moffiitt Cancer Hospital and Research Institute: Dr. Thimothy Yeatman, Dr. Erin Siegel, Dr. Alan B. Cantor and Steve Eschrich. Helen Lewis in Moffitt Cancer Registry for providing me with patient information. Sincere thanks to Dr. Noreen Poor, Paul Tate, Hillary Straye r, Marilyn Williams, Silvia Calderon and Dr. Connie Mizak for thei r constant encouragement during my study at USF. A special thank you to my racquetball coach and friend Tom Va nn, for always being there for me, for his moral support and confidence in me. Finally, I would like to thank my parent s and family members for their love, encouragement and support. It is because of them that I moved forward.
i Table of Contents List of Tables iii List of Figures iv List of Abbreviations v Abstract vi Chapter 1: Introduction 1.1 Background 1 1.2 Structure and function of colon and rectum 2 1.3 Classification of disease 3 1.4 Statement of Problem 5 1.5 Study Purpose 6 1.6 Research Questions 7 1.7 Hypothesis 7 Chapter 2: Review of Literature 2.1 Descriptive characteristics 8 2.2 Prognostic Factors 8 2.2.1 Tumor related prognostic factors 9 2.2.2 Patient related prognostic factors 10 Chapter 3: Methods 3.1 IRB Approval 17 3.2 Study Design and Study Population 17 3.2.1 Inclusion Criteria 18 3.2.2 Exclusion Criteria 18 3.3 Gene Expression Profiles a nd Molecular Classification 19 3.4 Censoring 20 3.5 Statistical Analysis 20 Chapter 4: Data Analysis 4.1 Descriptive Analysis 21 4.2 Statistical Analysis of Research Questions 22
ii Chapter 5: Results 5.1 Descriptive Analysis 24 5.1.1 Demographic Data 24 5.1.2 Histopathologic Characteristics 28 5.2 Statistical Analysis 30 5.2.1 Univariate Analysis 30 5.2.2 Multivariate Analysis 38 5.2.3 Predictor Model 41 Chapter 6: Discussion 6.1 Important Findings 43 6.2 Strengths and Weakness 45 6.3 Consistency with Literature 47 6.4 Conclusion 48 6.5 Public Health Importance 48 6.6 Future Directions 49 List of References 50 Appendices Appendix 1: Colorectal Cancer Prevention 54 Appendix 2: ACS Screening Guidelines 57 Appendix 3: Clinical Chart Review Data Entry Form 58
iii List of Tables Table 1 AJCC/UICC Staging System for Colon and Rectal Cancer 4 Table 2 30 months and 60 months Re lative Survival rate by AJCC 6 Sixth Edition System Table 3 Literature Review of Clinical and Pathological factors assessed by 13 Multivariate Analysis Table 4 Survival Status at the end of 3-year follow-up 26 Table 5 Cancer Status at the Da te of Last Contact 26 Table 6 Clinical Characteristics of Patients, Sixth Edition System 27 Table 7 Histopathological Characte ristics of Patients 29 Table 8 Univariate Analysis 36 Table 9 3-year and 5-year Survival Estimates of the Study Population 37 Following curative Surgery, Based on the Prognostic Variables Table 10 Significant Prognostic Factors of Mortality (Overall Patient Survival) 39 Determined by Multivariate Analysis
iv List of Figures Figure 1 Anatomic Division of the Larg e Intestine and Rectum and Tumor 3 Penetration Figure 2 Dendrogram 20 Figure 3 Kaplan Meier Curve for Overa ll Survival of the Study Population 24 Figure 4 Age Distribution of the Study Population 25 Figure 5 Kaplan Meier Survival Cu rve for TNM Staging of Tumor 31 Figure 6 Kaplan Meier Survival Curve for Molecular Risk 32 Figure 7 Kaplan Meier Survival Curve for BMI of Patients 33 Figure 8 Kaplan Meier Survival Curve for Age of Patients 33 Figure 9 Kaplan Meier Survival Curv e for Staging and Molecular Risk 40 Figure 10 Survival Distribution fo r Prediction Model with Stage 42 Figure 11 Survival Distribution for Prediction Model containing Stage, 42 Molecular risk, age and body mass index.
v List of Abbreviations ACS American Cancer Society BMI Body Mass Index CEA Carcinoembriyonic Antigen CI Confidence Interval CIN Chromosomal Instability CRC Colorectal Cancer FAP Familial Adenomatous Polyposis FDR False Discovery Rate GEP Gene Expression Profiles HNPCC Hereditary Nonpolyposis Colorectal Cancer LN Lymph Node MSI Microsatellite Instability NSAIDS Non Steroidal Anti Inflammatory Drugs RMA Robust Microarray SAM Significance Analysis of Microarrays SE Standard Error VEGF Vascular En dot helial Growth Factor
vi Gene Expression Profiles and Clinical Para meters for Survival Prediction in Stage II and III Colorectal Cancer Patients Mubeena Begum ABSTRACT Prediction of outcome in colorectal cancer (CRC) is currently based on the TNM staging classification; however, histopathological clas sification alone is insufficient for accurately predicting survival in stage II and III patien ts. Studies indicate that microarray gene expression profiles can predict survival in CRC. We hypothesize that tumor gene expression in combination with clinical parameters, is a better predictor of outcome in stage II and III colorectal cancers than the TNM stage classification alone. Clinical records and follow-up data were retrospectively reviewed for 58 Stage II and Stage III patients with prim ary colorectal cancer, who di d not receive any neoadjuvant therapy preoperatively and whose samples had b een previously analyzed for gene expression profiles using the Affymetrix U 133a Gene chip. For molecular classification of patients as being at high or low risk for poor survival samples were divided into two clusters by hierarchical cluster analysis of genes selected by SAM. Univariate and multivariate analyses
vii using Cox proportional hazard models were don e to identify significant prognostic factors The 3-year and 5-year survival esti mates were 72.41% (SE=5.8%) and 55.17% (SE=6.7%), respectively, for all 58 patients. Univariate analysis showed that advanced stage, older age, high-risk molecula r classification, positive lym ph nodes were the statistically significant prognostic factors of poor survival (p<0.05), while gender, preoperative CEA level, and family history of CRC in first degree relatives were not statistically significant. In multivariate analysis molecu lar classification, age and body mass index were independent significant prognostic factors. In Cox proportional hazard model, the estimated hazard ratios for Stage III vs II was 2.45 (95%CI: 0.85-7.04), for high vs low molecular risk was 3.83 (95%CI: 1.22-12.06) and old vs young age wa s 3.72 (95%CI: 1.2-11.49). Model containing clinical stage in c onjunction with molecular risk, body mass index, and age was a stronger indicator of clinical outcome (p= 0.0056) than model with clinical stage alone. Gene expression profiles pred ict survival independent of clinical parameters, and the addition of gene expression profile s to stage is more predictive of survival than stage alone. Further analysis needs to be done to validate the molecular cl assification on an independent dataset.
1 CHAPTER ONE INTRODUCTION 1.1 Background Colorectal cancer (CRC), cancer of colon and rectum, the third most common cancer worldwide and the second leading cause of cancer-re lated mortality in the US [1,2]. It is the second most site-specific can cer affecting both men and wo men (lung cancer is first, affecting both men and women, breas t is the leader in women and prostate in men) . The lifetime probability of developi ng colon cancer in men is 1 in 17 and in women it is 1 in 18 . A study conducted by Parkin, DM et al, 199 9  discusses that approximately, 6% of the American population will eventually develop invasive CRC and over 6 million Americans who are alive today will die of the disease. 75% of patients with CRC have sporadic disease, with no evidence of having inherited the disorder and 25% have a family history of CRC suggesting a genetic contribution. Majority of the colorectal cancers aris e due to malignant transformation of an adenomatous polyp. The malignant tumor arises from colonic epithelial cells that line the mucosa. Transition from normal epithelium to adenoma and to carcinoma is due to acquired molecular events , that is, 85% of CRC ar e due to events which lead to chromosomal instability (CIN) and 15% are due to microsatel lite instability (MSI). These events alter chromosomes 5q (APC), 18q (DCC) and 17p (TP53) involved in DNA repair.
2 1.2 Colon and Rectum: Structure and function Colon refers to the upper six feet of the large intestine and rectum to the last five to six inches. Together colon and rectum make up the large intestine. The colon is made of four sections: ascending colon, transverse colon, desc ending colon and the sigmoid colon. Cancer can develop in any of the four sections of the colon and in the rectum. The distributions of colorectal cancer in the large intestine are: ascending colon and cecum-25%, transverse colon-15%, descending colon-5%, sigmoid co lon-25%, rectum-25% and rectosigmoid junction-10% . Tumors on the right side of the colon near the cecum usually gr ow large enough to be painful and cause bleeding. As a result they commonly present with anemia from chronic blood loss. Polyps commonly appear on the left side of the colon. Ca ncer on the left colon usually grows around the colon wall and encirc les it. Common symptoms of a tumor on the left side include constipation a nd change in bowel habits. Ca ncer can grow inward toward the hollow part of the colon or rectum, and /or outward through the wall of the colon or rectum. In untreated cases the cancerous cells break away from the pr imary site and spread to distant organs through bloodstream or ly mphatic system. This process is called metastases. 95% of CRC are carcinomas, and 95% of these are adenocarcinomas.
3 1.3 Classification of Disease : The currently used staging system for CRC is UICC-AJCC TNM staging system. The AJCC TNM staging system is considered to be more useful for clinical deci sion-making, due to its precise stratification. It cons ists of three independent progno stic variables: the depth of tumor invasion into the bowel wall (T), the pr esence or absence of lymph node involvement (N), and the presence or absen ce of distant metastases (M) [ 21, 53]. The pathologic staging is assigned after the resection of the primary tumor, removal and examination of regional lymph nodes and analysis of the surgical specimen. The AJCC/ UICC and Dukes classification system is shown in table 1. The survival rates for the different staging systems are summarized in table 2. Fig1. Anatomic Division and Tumor Penetration (AJCC 6th Edition System)
4 Table1. AJCC/UICC Staging Sytem for Colon and Rectal Cancer AJCC/UICC Description of AJCC Staging System Dukes MAC Stage 0 and Tis Carcinoma in situ: intraepithelial or invasion of the lamina propria. Stage I (T1 N0 M0) Tumor invades through the submucosa. No metastases to regional nodes or distant metastases. Dukes A A Stage I (T2 N0 M0) Tumor invades into the muscularis propria. No metastases to regional nodes or distant metastases. Dukes A B1 Stage IIA (T3 N0 M0) Tumor invades through the muscularis propria into the subserosa, or into non-peritonealized pericolic or perirectal tissues, no metastases to regional nodes or distant metastases. Dukes B B2 Stage IIB (T4 N0 M0) Tumor directly invades other organs or structures, and/or perforates visceral peritoneum, no metastases to regional nodes or distant metastases. Dukes B B3 Stage IIIA (T1-2 N1 M0) Tumor invades submucosa or muscularis propria. Metastases in 1 to 3 regional lymph nodes. No distant metastases. Dukes C C1 Stage IIIB (T3-4 N1 M0) Tumor invades through the muscularis propria into the subserosa, or into non-peritonealized pericolic or perirectal tissues or tumor directly invades other organs or structures, and/or perforates visceral peritoneum. Metastases in 1 to 3 regional lymph nodes. No distant metastases. Dukes C C2/C3 Stage IIIC (Any T, N2, M0) Any extent of tumor invasion. Metastasis in 4 or more regional. No distant metastases. Dukes C C1/C2/ C3 Stage IV Any extent of tumor invasion or number of metastases to regional nodes. Dist ant metastases present. Dukes D D
5 1.4 Statement of Problem Prediction of outcome is an important aspect in cancer research. In patients with colorectal cancer limited to the mucosa (TNM Stage I) and in patients with distant metastases (TNM Stage IV), the 5-year survival rates are 9095% and <10%, respectively [2, 32]. In these patients histopathologic criteria (TNM staging) are good predictors of survival. However, in patients with invasion of the colon wall or adja cent structures (stage II) and those who have regional nodal metastases (stage III), the probabi lity of survival is about 70-85% and 40-80% respectively based on the TNM st aging . Moreover, patient s with same tumor stages may show different prognosis indicating that conventional staging procedures may be unable to precisely predict cancer risk. Thus, it is in stage II and III patients that a better predictor of survival is needed. As an alternative to cl inical staging, recently developed microarray technology has permitted the develo pment of multiorgan cancer cl assifier, identification of tumor subclass, discovery of progression marker s, and prediction of disease outcome in many types of cancer. Unlike clinicopathological staging, molecula r staging is able to better predict the long-term outcome of an individual based on the ge ne expression profile of the tumor at diagnosis . Preliminary studies indicate that microarra y gene expression prof iles have been most accurate to date in predicting the overall surv ival in CRC . Howe ver, most research examining gene expression profile has not taken clinical parameters (s tage, age, sex, grade, preoperative CEA levels) into consideration.
6 Table 2: 30 months and 60 months rel ative survival rate by AJCC sixth Edition System . The 5-year survival rate refers to the percenta ge of people who live at least 5 years after their cancer is diagnosed. 1.5 Study Purpose The purpose of the study was to investigate whether clinicopatholog ical (TNM staging) based outcome (survival) prediction can be improved by combining microarray gene expression profiles together with other clinical predictors (age, sex, grade, preoperative CEA level). Stage ---------Survival Stage+ GEP+ age + gender + body mass index (BMI)+ family history + location of tumor+ grade+ preopera tive CEA level --------------Survival + total resected lymph nodes Stage 30 months (%) 60 months % (5-year relative survival rate) Stage I 96.1 93.2 Stage IIA 91.0 84.7 Stage IIB 80.2 72.2 Stage IIIA 91.4 83.4 Stage IIIB 77.3 64.1 Stage IIIC 59.1 44.3 Stage IV 17.3 8.1
7 1.6 Research Question: 1. What is the predictive value of traditional c linical predictorsage, sex, race, stage and grade in determining prognosis (survi val) in Stage II and III CRC patients? 2. What is the predictive value of gene e xpression profiles (GEP) by itself in this sample? 3. What does the addition of GEP together with clinical parameters (age, sex, family history, BMI, grade, location of tumor, pr eoperative CEA levels) contribute to the usual clinical predictiveness of outcome (s urvival) in stage II and III colorectal cancer? 1.7 Hypothesis Gene expression profiles (GEP ) and clinicopathological fact ors (age, sex, BMI, family history, grade, preoperative CEA level) add to the predictive va lue of staging in predicting the postoperative outcome (survival) in stage II and III CRC.
8 CHAPTER TWO REVIEW OF LITERATURE 2.1 Descriptive Characteristics Currently CRC constitutes 10% of new cancer cases in men and 11% new cancer cases in women. Estimated new cases and de aths from CRC in the US in 2005 are 104,950 new cases and 56,290 deaths . SEER data for 1998-2002 show the overall incidence of CRC is higher in men (62.1/100,000) than in women (24.8/100,000), and this holds true for mortality rates, in men (46.2/100,000) and women (17.4/100,000). The median age at diagnosis in the United States, during this period is 70 for men and 74 for women. Incidence is higher among African Ameri can men and women (62.4%) compared to White men and women (52.5%), and so are the mortality rates in Af rican Americans (20%) to Whites (27.9%). The risk of CRC increases af ter the age of 50-55 years and continues to rise exponentially w ith increasing age. Between 1998 and 2001, the incidence rate has declined by 2.9% and 5-year su rvival rate has increased by 7.3%, which could be due to advances in detection and scr eening and the increasing use of combination therapies . Although the exact cause of CRC cancer is unknown, several fact ors play a crucial role in the development and prognosis of CRC su rvival, which can be classified as prognostic factors (tumor related, host related and envir onmental related factors) and risk factors.
9 Definition of Prognostic Factor: In epidemiological lite rature prognostic factor refers to the probability of future event in pa tients who currently have a disease. It implies prediction of an event that will occur in the fu ture. It can be consider ed in the context of probability of cure or prolongation of survival Knowledge of prognostic factors helps us to understand the progress of the disease . Definition of a Risk Factor: A clearly defined occurrence or characteristic that has been associated with the increased rate of a subsequently occurring dise ase. It is limited to those who dont have a disease . 2.2 Prognostic Factors 2.2.1 Tumor related prognostic factors Pathological staging : Is the most important predicto r of outcome in patients with newly diagnosed colorectal cancer [39, 52], which depends on the degree of penetration of tumor through the bowel wa ll, presence of or absence of nodal involvement and presence or absence of di stant metastasis. Majority of the CRC are adenocarcinomas. Histological grade : Tumor prognosis correlates w ith histological grade: poor differentiation has a worse prognosis than a high degree of diffe rentiation [39, 45]. Large studies have shown that histologi cal grade correlates with survival and recurrence, with low-grade tumors having be tter survival . Venous, lymphatic and perineural invasion have also shown to decrease survival a nd increase the risk of local recurrence [21, 52].
10 Surgical margins and radial margins : The presence of positive surgical or radial margin is a poor prognostic factor, with local failure rate in all st ages increasing from 3%-85% . Molecular markers : The use of molecular markers that have prognostic significance aids in identifying high-risk patients who can benefit from adjuvant chemotherapy and avoiding those who have low risk from the toxicities of ad juvant chemotherapy. Presence of microsatellite instability ha s shown to improve prognosis in sporadic colorectal cancers [20, 22]. Three studies have independently shown unfavorable prognosis of patients with the loss of 18q in stage II and III CRC [24, 33, 41]. Carcinoembryonic antigen (CEA): CEA levels are used to monitor the course of colorectal cancer. Elevated pre-operative levels of CEA at diagnosis, has shown to be an independent prognostic factor for surv ival and recurrence in CRC patients . Lymph nodes examined: AJCC and NCI recommend that at least 12 lymph nodes should be examined in patients with colorect al cancer to confirm the absence of nodal involvement by tumor . It is a reflecti on of the aggressiveness of lymphovascular mesenteric dissection during surgery and pat hological identification of nodes in the specimen. Retrospective studies have demons trated that the numb er of LN examined in CR surgery may be associated with patie nts outcome . A study by Berger. A. C et al., 2005, have demonstrated that lymph node resection is a statistically important prognostic factor for determining overall survival and di sease free survival. 2.2.2 Patient-related prognostic factors: Genetic Syndromes : Patients with Hereditary Nonpolyposis colorectal cancer
11 (HNPCCis a familial syndrome in which individuals develop CRC before 50 years of age), chronic ulcerative co litis (UCinflammatory condi tion of the larg e intestine) and Familial adenomatous polyposis (FAPan autosomal dominant condition characterized by multiple polyps with high potential to progress to cancer) are at increased risk of colorectal cancer . Hereditary Nonpolyposis Co lorectal Cancer (HNPCC) : Also called Lynch Syndrome. It is an autosomal domin ant condition caused by mutation in the DNA mismatch repair genes (hMSH2, hMLH1) HNPCC accounts for about 3-5% of all CRC . They have increased risk of developing adenom as at an early age, the average age of CRC diagnosis in HNPCC s yndrome patients is around 44 years . They also have an increased risk of deve loping other cancers such as endometrial, ovarian, small intestine, pancreatic renal pelvis and brain tumors. Familial Adenomatous Polyposis (FAP) : is an autosomal dominant condition with a prevalence of 1/8,000 . FAP is due to mutation in the APC gene on chromosome 5q21. The disease is characterized by hundr eds of polyps in the colon and rectum, which develop after the first decade of life. By the age of 20 and 30 the probability of developing colonic adenoma increases by 75% and 90%, respectively . Familial Colon Cancer: A positive family history is an important risk factor developing CRC. A two to three-fold increase in risk is seen if an individual has a first degree relative with CRC and the risk increases if more relatives are affected . Family history of CRC is seen in 1015 % of persons with CRC. This increases the persons risk to develop CRC by 2-6 fol d. With a family history, the risk of CRC increases earlier in life (less than 45 years) than later.
12 Age : 3 % of the CRC arise before age of 30 years, and 11% have predisposing conditions such as FAP and UC. The risk of CRC increases with age and it is most common in men and women above 55 years. Studies have repor ted poor prognosis in CRC patients who are less than 40 years . Tumors with microsat ellite instability (MSI) have better prognosis irrespective of age . Racial difference: Racial differences in the overa ll survival were observed in few studies although some studies have shown that co-morbid conditions play a role in the survival outcome in different patient popul ation. Jews of European decent have a higher rate of CRC due to genetic mutation. Weight: Many studies have reported an increased risk of colorectal cancer with increasing body mass index . Having exce ss fat in the waist are (intra-abdominal fat) increases the risk more than having th e same amount of fat distributed in other areas (thighs, hips). Obesity placed men more than women at increased risk for colon cancer. This association in me n is due to greater waist circumference in men and the protective effect of estrogen in wo men, which decreases CRC risk. A study by Giovanucci et al., 2001  found a correla tion between colon cancer and type-2 diabetes. Obesity predisposes a premenopausal woman to the same risk as does for men in general . Geography: rates of colorectal cancer vary geographically. The disease is more common among industrialized nations U SA, Western Europe, Australia and uncommon in Asia, Africa and South America .
13 Table: 3 Literature Review of Clinical and Pathologi cal Factors Assessed by Cox Multivariate Analysis Author, year Study Features Factors Analyzed Findings and Strengths/Weakness Bertucci, F et al. 2004 Prospec tive study Gender, age, site of tumor, grade, Significance of prognostic classification made by AJCC  Single institution, France tumor penetration, LN invol vement, stage and the obtained gene set was compared. Sample size: 50 samples vascular invasion, stage, surger y, Classification based on AJ CC stage was significant but (26 patients) between less than that made by gene expression profiles. 1990-1998. prognostic impact of gene set persisted when applied to All tumor sections and pati ents without metastasis at diagnosis and patients Medical records were reviewed without metastasis and LN involvement. prior to analysis. Strengths: Accuracy of prediction of molecular Unsupervised hierchical metastatic signature was estimated by leave one clustering was used to investigate out proc edure. DNA microarray was able to identify relationship between samples clinically relevant tumor s ubgrous based on the gene and genes. cluster. Weakness: Multivariate analysis was not done to determine significant clinical and pathological factors affecting survival. Barrier, A. et al., 2005 Prospective study Gender, stage, grade and location. 70 gene predictor was built, 35 were overexpresse d in  Sample size: 18patients Prognostic prediction was build patients who developed a re currence and 35 in patients Stage II and II CRC patients based on microarray gene expression who were disease free for 5 years. Follow-up: measure for stage II and II CRC Strengths: Double cross validation was done by splitting Every 3 years-1st postoperative patients. For each dataset a total the data into testing and training set. The results of the thereafter. of 150 prognosis study suggest the ability to build a prognosis predictor year and every 6 months predictors were considered and for both stage II and III CRC, based on either T or NM Performance was assessed gene expressi on profiles. The accuracy of the 70-gene using six-fold cross-validation. NM based pr edictor was greater than that of 30-gene based predictor (83 vs 78%). Weakness: The study did not analyze the effect of Clinico pathological factor in association with gene Expression profiles in predicting the survival.
14Author, year Study Features Factors Analyzed Findings and Strengths/Weakness syndrome, cancer with IBD, Dukes stage, grade, micros atellite regulation and Dukes stage, LN status. VEGF rectal cancers were excluded status, resected nodes, postope rative overexpression correlated signi ficantly with Dukes stage. because their molecular feature, complications, recurrence. Strengths: Study compared levels of p53, p27, VEGF recurrence rate, and overall molecular and structural mark ersand MVC, which are involved in cycle regulation, differs from sporadic colon P53, P27, VEGF, microvessel count apoptosis and tumor neoangiogenesis, for normal cancers. and colon cancer cells with clinopathological variables. In conjuction with clinicopathological staging, molecular Expression markers p27, p53, VEGF, provide a stronger Indication of clinical outcome than with staging alone and help better select therapeutic option in colorectal cancer patients. Murphy T K et al. 2000 Large prospective study Age, sex, smoki ng and alcohol. The study was done to examine the association be tween BMI and  Population based, 50 States history, dietary history, exercise, colon cancer mortality in both men and women. The finding s in Sample size: 1,184,659 estrogen replacement therapy the stud y were that BMI was an independent risk factor for colon Follow up: 12 years and asprin use. cancer death in both sexes and the relationship is stronger and 1616 final sample size BMI was categorized for age more linear in men than in women. This could be due to central followinginclusion and and gender. obesity causing hyperinsulinemia and increased glycemic load exclusion criteria causing tumor growth. Alcohol intake significantly modi fied the association between BMI and colon cancer mortality. Strengths: Large prospective study, generalizeable to the Population. Weakness: lack of screening data, self reported measurements. Did not look into other clinicopa thological factor s which affect survival in CRC.
15Author, year Study Features Factors Analyzed Findings and Strengths/Weakness Ponz de Leon et al, 1992 Populatio n based study Age, site of tumor, family hist ory, Factors which were significant in univariat e analysis  Sample size: 132 interval between diagnosis and analysis were: age, pattern of growth and extent of Follow-up: 5 years surgery, interval between symptoms fibrosis. However, the only si gnificant factor related to Post operatively and diagnosis, st age, pattern of prognosis in mu ltivariate analysis was stage. Equal number of male growth, extent of fibrosis. Staging being the factor significant in multivariate and female patients analysis confirms the importance of stage in predicting survival in CRC. Prall, F et al. 2004  Retrospective study with review Clin ical and immunohistochemical In C ox multivariate analysis growth pattern of tumor, of clinical charts and medical tumor ma rker levels were estimated lymphohistiocy tic response, lymphatic permeation, and records. (p53, p27, p21 levels). extramural venous were f ound to be significant when Sample size: 184 tested against UICC stage. Mitotic index added to the All stages of CRC prognostic information to TNM stage in multivariate Follow-up history was obtained analysis. for 5 years postoperatively. Weakness of study: Effect of other prognostic markers in combination with tumor markers (p53, p27, p21) was not done on CRC patients in predicting survival. Rene A. C. et al. 2004 Retrospective st udy Age, Gender, Karnofsky Study determ ine better survival (5-year) rate  Sample size: 96 patients perfo rmance at admission, for patients undergoing curative surgery (58.3%) than From 1950 to 1990. site of tumor, type of surgery, palliative surgeries (0%). Follow-up: 3 years preoperative albumin level, Multivariate analysis showed Karnofsky performance All stages of CRC number of resect ed organs/ status was strongly rela ted to the risk of postoperative structure, hospital stay, grade, co mplications and postoperative deaths. stage, lymph node status, Factors which were related to better prognosis for CRC, lymphatic invasion, pe rineural were grade I and II, non metastatic LN, absence of invasion, tumor margins, vascular lymphatic, perineural invasion, clinical presentation. Poor prognostic factors were lympn node status and adjacent organ infiltration. Strengths: A significant decrease of postoperative deaths and complications from 1950 to 1990 could be due to improvement in staging methods pre and post operative and
16 Author, year Study Features Factors Analyzed Findings and Strengths/Weakness Ratto, Carlo et al, 1998 Prospective study Gender, Age (> 60 years) Factors significant and independently influencing  Sample size: 853 Location of primary out come were gender, lymph node involvement, Male and female patients history of bowel obstr uction history of metastasis and bowel obstruction. With CRC of all stages Tumor si ze, Stage, Grade, Strengths: Large sample size and good power, Follow-up: LN involvement, Metastasis, in cluding tumor markers similar results were Every 3months-1st year Preoperative CEA, tumor ploi dy observed in other studies too. Every 6 months in 2nd-3rd year, and vascular invasion. Weakness : Did not look into molecular markers once per year thereafter on survival prediction. Wang et al. 2004  Retrospective study w ith review Clinicopathological information was The study demonstrates the potential of DNA microarray based of clinical charts. collected. The sample was divided into gene expression pattern for the pr ediction of patients outcome Sample size: 74 Dukes B patients 2 groups testing and training set to select in col on cancer. This is likely to have an impa ct on the current Follow-up history for 3 years gene markers using the training se t and clinical practice for the eligibility of adjuvant chemo therapy on build a prognostic signature and va lidate treatment of Dukes B co lon cancer patients. The study it on the testing set. identified 23-gene si gnature that predicts recurrence in Dukes B patients. The signature was validat ed in 36 independent patients. The overall accuracy was 78%.
17 CHAPTER THREE METHODS 3.1 IRB Approval Prior to the initiation of the research, the study was approved by the Institutional Review Board (IRB) of the University of South Florida. 3.2 Study Design and Study Population Retrospective cohort study with the review of clinical charts of 116 CRC patients; who underwent curative surgery from 1/5/1993 to 5/1/2002, at Moffitt Cancer Center and had a follow-up history up to the date of last contact or death. Initial study started with the selection of 400 frozen tumor specimens of patients with any of the four colorectal cancer tumor st aging from Moffitt Cancer Center Tumor Bank (Tampa, Florida); such that all patients had postoperative follow-up for at least 36 months (because majority of patients who would die of CRC, will have done so by then) determined by Moffitt Cancer Registry. The sample si ze was reduced to 116 as gene expression profiling using mRNA technique was done to only 116 samples of the 400 CRC tumor samples. Retrospective review of inpatient charts, in cluding operative and pa thologic reports of the 116 stages II and III CRC wa s done to obtain clinicopath ological data. Data was
18 collected on patients demographi cs, clinical, pathological and su rvival data. A copy of the data entry form is provided in Appendix 1. Once the data was collected, in order to maintain the patient confidentia lity each patient was given a unique identification number generated by SAS Randomisation. These data were ente red on a standardized data entry form and entered into a database. Based on the exclusion criteria the final sample size was reduced to 58 patients consisting of confirmed stage II or stage III primary colorectal cancer only. 3.2.1 Inclusion Criteria: Confirmed Stage II and Stage III primary colorectal cancer patie nts, who did not receive any neo-adjuvant therapy preoperati vely and, who had a follow-up of at least 36 months and also had gene expression profiling done using the mRNA technique. 3.2.2 Exclusion criteria: As the study addressed only primary colorect al cancer 41 samples were excluded from the 116, who were Stage I or Stage IV colorectal cancer patients and also those for whom the cancer site were not colorectal such as (abdominal wall, periaortic lymph nodes, mesenteric lymph node, lung, liver, bladder, retroperitoneal lymph nodes, kidney, sm all intestine) after reviewing the histopthological re ports. Five patients had multiple primaries at the time of surgery and were assessed by the histological grade to determine the earliest primary and only the earliest primary was considered in th e sample size and the ot hers were excluded. Nine patients who received preoperative neoadjuv ant therapy such as radiated rectal cancers were dropped from the study, as preoperative neoadjuvant therapy would affect gene expression profiles, hence cannot be used as pr edictor. Histopathological information could not be reconfirmed for three patients and were excluded from the sample size. Clinical and histoathological characteristics of patients and their tumors are summarized in tables 6 and 7.
19 3.3 Gene expression profiles (GEP) and Molecular classification: The Gastrointestinal Tumor Program, Bioi nformatics and Biostatistics Core of Moffitt Cancer Center and Research Institute performed the GEP of the 116 colorectal cancer specimens. The microarray data was analyzed usi ng Significance Analysis of Microarrays program (SAM) with censored survival data [ 11]. SAM identified genes most correlated with survival time (3-year survival) and used permut ation analysis to estimate the False Discovery Rate (FDR). The first analysis was done fo r stage II and III CRC patients only. The data was preprocessed using Robust microarray (RMA) a nd the censored survival time was calculated for each sample. In this work, censoring occurred at time of last follow-up or for death in which there was no evidence of disease. SAM was used to calculate gene s that correlated to survival time. A threshold yielding a 10% me dian FDR was selected, which resulted in 53 overexpressed genes. The analysis was repeated fo r cases of all stages (stage I, II, III, IV). The data was again preprocessed using RMA and survival time was calculated. SAM was again used to select genes. A median FD R of 2.5% (the mini mum FDR possible) was selected, resulting in 30 genes. The gene expression for all 30 genes was ex tracted and the data was clustered. The Gene Cluster 3.0 program for clustering and Ja va Treeview for visualizing the heatmap was used. The genes were median centered and nor malized prior to clustering. Hierarchical clustering with the un-centered correlation simila rity metric and complete linkage was used. The resulting dendrogram can be seen below. Th is resulted in two main groupings (Clusters) of the sample (columns) in the data, which we re chosen as the prognostic groups for this work. These groups were listed as Cluster 1(Lo w risk) and Cluster 2 (High risk) based on 30-
20 gene cluster. Cluster Anal ysis of 30 SAM selected genes were performed. Red color represents over expressed genes rela tive to green, underexpressed genes. Fig 2. Dendrogram 3.4 Censoring Patients were considered for censoring when incomplete information was available about their survival time that is if they were lost to follow-up, a live with no evidence of CRC at the date of last contact or who died but not due to CRC. Patients who did not die of CRC were considered censored and patients who died of colorectal cancer at th e end of the study were considered as having experienced the outcome of interest (death) and were not censored. 3.5 Statistical Analysis Univariate analysis and Kaplan -Meier procedure was done to find predictive value of each prognostic variable. Multivariate analysis usi ng stepwise selection procedure was done to determine the effect of GEP in the presence of other clinical pred ictors. Final model selection was based on factors which met proportional hazard assumption and/or had biological or clinical significance. T T T T T T T T T T T T T T T T T T T T T T T T T T T 201525_at 203980_at 204938_s_at 202437_s_at 228335_at 205518_s_at 204793_at 233364_s_at 219789_at 204041_at 229228_at 202438_x_at 231964_at 227529_s_at 227530_at 236313_at 239903_at 206858_s_at 227318_at 220785_at 216056_at 217523_at 213364_s_at 1556221_a_at 225595_at 1556543_at 243835_at 235473_at 238593_at 242108_at Gene Cluster 1 ( n = 27 ) Gene Cluster 2 ( n = 31 )
21 CHAPTER FOUR DATA ANALYSIS 4.1 Descriptive Analysis: A total of 58 patients matched the inclusion criteria for the study and all patients had undergone curative surgery for cancer. Prelimin ary descriptive analyses were performed to characterize the sample. Frequency and pe rcentages for each variable were obtained accounting for the missing values. Table 6 and 7 summarizes the clinic al and pathological characteristics of patients and their tumors. Age at diagnosis was calculated as date of surgery minus the date of birth. Mean, median and age range at diagnosis were determ ined. Age at diagnosis was divided into 3 groups: less than 50 years, 50-70 years and above 70 years. To further determine the influence of advanced age on survival, age at di agnosis was divided into tertiles of upper 1/3 rd age group and lower 2/3 rd age group. Body mass index was calculated based on the he ight and weight at the time of surgery of index primary colorectal cancer and divide d into 4 groups: less th an 18.5 (underweight), 18.5-24.5 (normal), 25-29.9 (overweight) and above 30 ( obese). To further assess the effect increased BMI on overall survival, BMI was divided into 2 groups: 25 and >25. Location of primary cancer was merged in to three groups: proximal colon (cecum, ascending colon, hepatic flexure and transverse colon), distal colon (splenic flexure,
22 descending colon and sigmoid colon) and rectum and rectosigmoid junction. Tumor stage at the time of primary cancer was regrouped into tw o groups with Stage IIA and IIB grouped as stage II, and stage IIIA, IIIB and IIIC grouped as stage III, because there were very few patients in stage IIB and III A (table 6). Mean and range of Lymph nodes taken at the time of resection of primary tumor were determined and also grouped into 12 lymph nodes and greater than 12 lymph nodes. Survival differences for regional lymph node (LN) involvement were assessed for: No LN involvement, 1-3 group of LNs, 4 group of LN. Further analysis was done by collapsing the regional lymph node involveme nt into: lymph node positive group and lymph node negative group. Histological differentiation was categorized ba sed on the grading: well-differentiated, moderately differentiated and poor differentiatio n. Since, the number of patients in the welldifferentiated group were few, these patients we re merged with patients with moderate differentiation to compare survival with poor differentiated tumors. For preoperative CEA level was divided into < 5ng/ml and >5 ng/ml. Information on the vital status and cancer status at the date of last contact was obtained through the Moffitt cancer regist ry and summarized in table 5. 4.2 Statistical Analysis Once the descriptive analys is was performed on the st udy population, the research questions were analyzed using SAS 9.0 software. Survival distribution were estimated using the Kaplan-Meier procedure to assess the influence of individual predicto rs (clinical parameters and GEP) on the overall survival of patients in the study, to determine which para meters met the proportional hazard assumption
23 and if there was significant di fference between the strata for each parameter (predictor). Univariate analysis was done to determine sta tistically significant pr ognostic factors, the hazard ratios, confidence interval and p-values. Log-rank test was used to compare survival estimates for each stratified variable. Log-rank test was employed to evaluate the null hypothesis being tested that no overall survival difference exits between the strata for each variable. Hypotheses were tested using pvalues of 0.05 for stat istical significance. Multivariable analysis using stepwise Cox regression analysis was used to determine independent significant prognosti c variables. The likelihood ratio test based on maximum partial likelihood estimates was used to eliminate confounding va riables from the model. Variables were considered eligible for removal if the likelihood ratio test significance level was >0.05. The final model contained factors that me t the proportional hazard assumption and had a biological or clinical significance in predicting colorectal cancer survival. Survival rates were estimated by Cox proportional hazard model. Stepwise regression methods were used to build statistical model for the association of prognostic factor s with overall survival. Time dependent hazard ratios were estimated. By observing HR, 95% CI for each factor, and the change in the log likelihood stat istic, it was ascertained which variables should remain in the final model.
24 CHAPTER FIVE RESULTS 5.1 Descriptive Analysis 5.1.1 Demographic data Table 6 displays the demogra phic (clinical) characteristic s of the 58 CRC subjects in the study. The median survival time for the study sample was 75.36 years and the 3-year and 5year overall survival estimates we re 72.41% (SE=5.8%) and 55.17% (SE=6.7%), respectively (Fig 3). Fig 3 Kaplan Meier Curve for Overall Survival of the Study Population 0.00 0.25 0.50 0.75 1.00 Survival in Months 020406080100120140 Censored Observations
25 Female patients had higher incidence rate of CRC (53%) than males (47%). In the study sample, CRC was more commonly seen am ong Caucasians (81%), followed by African Americans (4%) and Hispanics (2%). BMI of majority of the patients (58%) was above normal (>25) at surgery. History of weight loss at the time of surgery was reported by 45% of CRC patients. Positive family history for CRC in first degree relative was present in only 8 of the 58 patients. Based on age, 10% were in the <50 year s age group, 38% were 50-70 years age group and 52 % were 70 years, at diagnosis. The median surv ival time for <50 years, 50-70 years and >70 years at surgery was 4.8 years, 5.12 years and 3 years, respectivel y. Majority of the individuals diagnosed with CRC were of the between 70-79 years of age (35%). Fig.4 Age distribution of the study population At the end of 3 year follow-up 42 patient s (72%) were alive and 16 patients were dead (28%). At the date of last contact 29 pa tients (50%) were aliv e with no evidence of cancer, 18 patients (31%) were dead with evid ence of cancer, 4 patients (7%) died but with no evidence of cancer and cancer status could not be ascertained for 7 patients (12%). Age at Diagnosis0 5 10 15 20 2525-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94YearsPercent (%)
26 Table 4 Survival Status at the end of 3-year follow-up Vital Status n % Alive 42 72 Dead 16 28 Table 5 Cancer Status at the Date of Last Contact Cancer Status at Last Contact Vital Status at Last Contact No evidence of Cancer Evidence of Cancer Unknown Alive (n=31) 29 0 2 Dead (n=27) 4 18 5
27 Table 6 Clinical Charac teristics of Patients Characteristics No. of Patients (%) Total Sample Size (n) 58 Age at Diagnosis, years Mean 67.17 years Median 70.47 years Age Range 26.97 92.41 50 years 6 (10) 50-70 years 22 (38) 70 years 30 (52) Upper 1/3rd Age group ( 74.75 years) 19 (33) Lower 2/3rd Age group (< 74.75 years) 39 (67) Gender Female 31 (53) Male 27 (47) Race Caucasian 47 (81) African American 4 (7) Hispanic 2 (3) Asian 1 (2) Other/Unknown 4 (7) Body Mass Index <18.5 (underweight) 1 (2) 18.5-24.9 (normal) 18 (31) 25.9-29.9 (overweight) 21 (36) 30 (obese) 13 (22) Unknown 5 (9) History of Smoking Ever 25 (43) Never 28 (48) Unknown 5 (9) Ever Female (n=31) 15 (48) Ever Male (n=27) 10 (37) Family History of Cancer Present 34 (58) Absent 12 (21) Unknown 12 (21) Family History of Colorectal Cancer Present 8 (14) Absent 38 (65) Unknown 12 (21)
28 5.1.2 Histopathological Characteristics: Table 7 displays the histopa thologic characteristics of the sample population as described below. The three most common site descripti ons of CRC were sigmoid colon (27%), ascending colon (25%) and rectosigmoid junctio n (17%). Based on the AJCC classification6th edition, in the study sample, 26% were stage IIA, 3% were stage IIB, 2% were stage IIIB, 14% were stage IIIB and 13% were stage IIIC colorectal cancers. Stage IIA and IIB when combined together constituted 46% had St age II CRC and stage IIIA, IIIB, IIIC when combined together constituted 56% of color ectal cancers in the study. Of the patients observed, 47% had no regional lymph node i nvolvement, 29% had 1-3 group of regional node positivity and 21% had more than 4 group of lymph node involvement. Moderate differentiation of the tumor ( 76%) was the most common histological grade in the sample population followed by poor di fferentiation (16%) and well-differentiated tumor (7%). Pretreatment CEA levels were unknown for 59% of the patients. 17% had preoperative CEA level more than 5 ng/ml and 24% had levels < 5ng/ml. Based on the gene cluster analysis 47% were cl assified as low-risk group a nd 53% were of the high-risk group.Other histological featur es observed were vascular in vasion (13 patients), lymphatic invasion (7 patients) and perine ural invasion (3 patients). In addition to adenocarcinomatous histology, mucinous histology of tumor was obser ved in 5 patients. Perineural, lymphatic and vascular invasion was seen in 3 patients, 7 patients and 13 pa tients, respectively. Resected margins were not clear of cancer afte r surgery with curative intent in 4 patients only.
29 Table 7 Histopathological Characteristics of Patients Characteristics No. of Patients (%) Location of Primary Proximal colon 21 (36) Cecum & Ileocecal Valve 2 Acending colon 15 Hepatic flexure 0 Transverse colon 4 Distal Colon 20 (34) Splenic flexure 0 Descending colon 4 Sigmoid Colon 16 Rectum and rectosigmoid junction 17 (29) Rectosigmoid junction 10 Rectum 7 Stage IIA (T3 N0 M0) 26 (45) IIB (T4 N0 M0) 3 (5) IIIA (T1-2 N1 M0) 2 (3) IIIB (T3-4 N1 M0) 14 (24) IIIC (Any T N2 M0) 13 (23) Regional Lymph Node Metastasis No Lymph node involvement (N0) 27 (47) 1-3 Lymph node involvement (N1) 17 (29) 4 Lymph node involvement (N2) 12 (21) Could not be assessed (Nx) 2 (3) Total Lymph Nodes Resected Mean Lymph nodes resected 13.63 Range of Lymph nodes examined 2 35 12 28 (48) >12 30 (52) Grade/Differentiation Well 4 (7) Moderately 44 (76) Poor 9 (16) Unknown 1 (2) Preoperative CEA* level CEA 5.0 ng/ml 10 (17) CEA < 5.0 ng/ml 14 (24) Unknown 34 (59) Molecular Classification Low risk (Cluster 1) 27 (47) High risk (Cluster 2) 31 (53) Other Histological Features Mucinous Histology 5 Signet Ring Histology 0 Perineural Invasion 3 Lymphatic Invasion 7 Vascular Invasion 13 Resected Margins Not Clear of Cancer 4 CEA: Carcinoembriyonic Antigen
30 5.2 Statistical Analysis The research questions are restated in this se ction to facilitate c oherence and readability. The first two research questions are stated as follows: 1. What is the predictive value of traditional c linical predictorsage, sex, race, stage and grade in determining prognosis (survi val) in Stage II and III CRC patients? 2. What is the predictive value of gene e xpression profiles (GEP) by itself in this sample? 5.2.1 Univariate Analysis: Table 8 lists the prognostic vari ables, their hazard ratios, 95% CI and p-values for the comparisons of interest. 3-year and 5-year survival rates of the study sa mple are listed in table 9. General Hypothesis: H0: No difference between the survival distribution for Group1 and Group 2, (S1 (t) =S2 (t) for all t >0) HA: There is difference in survival distribution between Group1and Group 2, (S1 (t) S2 (t) for some t >0) Where, Group1: Survival function S1 (t) Group 2: Survival function S2 (t) If the p-value < al pha (0.05) then H0 is rejected and conclude d that group1 and group2 have different survival distributions. If p-value alpha, H0 is retained and conclude that ther e no sufficient evidence in the data suggesting the opposite is true.
31 A) TNM staging: 0.00 0.25 0.50 0.75 1.00 Survival in Months 0 20 40 60 80 100 120 140 Log-Rank p = 0.0513 S tage 2 ( n= 29) Stage 3 (n=29) Fig 5 Kaplan Meier Survival Curves for TNM Staging of Tumor H0: No difference between the survival di stribution for stage III and stage II CRC, (S1 (t) =S2 (t) for all t >0). HA: There is difference in survival distri bution between stage III and stage II CRC, (S1 (t) S2 (t) for some t >0). Patients with stage II CRC had better 3-year an d 5-year survival rates compared to patients with stage III CRC. The p-value < alpha (0.05) hence, H0 is rejected and concluded that stage III and stage II patients have different survival distributions. The hazard ratio for patient s with stage III CRC was twi ce as high than that for patients with stage II cancers. 3-year rate (%) 5-year rate(%) HR p-value Stage III 68.97 43.97 0.77 2.17 0.0513 Stage II 75.86 67.65
32 B) Molecular Risk: 000 025 050 0.75 100 Survival in Months 0 20 40 60 80 100 120 140 Log-Rank p = 0.0141 Low Risk (n=31) High Risk (n=27) Fig 6 Kaplan Meier Survival Curves for Molecular Risk H0: No difference between the survival distribut ion for high molecular risk patients and low molecular risk patients, (S1 (t) =S2 (t) for all t >0). HA: There is difference in survival distribution between high and low Molecular risk patients, (S1 (t) S2 (t) for some t >0). Patients of low molecular risk had better 3-ye ar and 5-year survival rates compared to patients of the high molecular risk cluste r. The p-value < alpha (0.05) hence, H0 is rejected and concluded that low molecular risk and hi gh molecular risk patients have different survival distributions. According to this univariate an alysis, the hazard for death for patients in the high-risk gene expression group was double than that for patients in the low-risk gene expression group and it is statis tically significant. 3-year rate (%) 5-year rate(%) HR p-value High Risk 58.06 41.29 0.987 2.68 0.0141 Low Risk 88.89 71.46
33 C) BMI: 000 025 050 0.75 100 SurvivalinMonths 0 20 40 60 80 100 120 140 Log-Rank p= 0.0041 BMI>25 ( n=33 ) BMI ) <25 ( n=20 ) Fig. 7 Kaplan Meier Survival Curves for BMI of Patients H0: No difference between the survival dist ribution between patients with body mass index >25 and <25, (S1 (t) =S2 (t) for all t >0). HA: There is difference in survival distribu tion between between patients with body mass index >25 and <25, (S1 (t) S2 (t) for some t >0). Patients with BMI > 25 were found to have better survival and lower hazard for death than patients with BMI <25 and an invers e relationship were s een between BMI and mortality. The hazard for death for patients in the BMI >25 group was 32% lower than that for patients in the BMI<25 group 3-year rate (%) 5-year rate(%) HR p-value BMI >25 84.85 71.46 -1.15 0.317 0.00041 BMI<25 55 33.39
34 D) Age 0.00 0.25 0.50 0.75 1.00 Survival in Months 0 20 40 60 80 100 120 140 Log-Rank p = 0.0555 Lower 2/3r d age (<74.25 years, n=39) Upper 1/3r d Age (>74.25 years, n=19) Fig. 8 Kaplan Meier Survival Curves for Age of Patients Patients in the lower 2/3rd age group had better 3-year a nd 5-year survival rates and lower hazard for death than patients who were in the upper 1/3rd age group. The p-value < alpha (0.05) hence, H0 is rejected and concluded that ol der age group patients (upper 1/3 rd tertile) have different surviv al distributions compared to younger patients (lower 2/3rd tertile). The hazard for death for patients in the upper 1/3rd age group was double than that for patients in the lower 2/3rd age group. 3-year rate (%) 5-year rate(%) HR p-value Upper 1/3rd 57.89 34.74 0.74 2.09 0.0555 Lower 2/3rd 79.49 65.59
35 Gender (HR: 1.58, p=0.2407), race (HR: 0.83, p=0.6932), family history of CRC (HR: 0.79, p=0.6568), grade of tumor (HR: 1.42, p=0.4816), preoperative CEA levels (HR: 2.3, p=0.1461), location of tumor (HR: 0.73, p= 0.2092), and total lymph nodes examined (HR: 0.84, p=0.6607), were not statistically significan tly associated with survival (Table 8). Table 9 summarizes the 3-year and 5-year survival estimates of the study population following surgery with curative intent based on the prognostic variable. Decreased survival rates were observed for patients in the younge r age group (<50 years) and the older age group (>70 years) as compared to those in the mi ddle age (50-70 years), with approximately 40-50 %, 5-year survival rate for patients in the young and old age group following diagnosis of CRC. Patients who were overweight and obese (BMI> 25) at the time of surgery had better survival rates compared to patients who were normal or underweight. Higher survival rates at year 3 and year 5 were observed for patients in the low molecular risk group as compared to those in the high risk group, whic h were statistically significant. Specifically Stage 2 with low molecu lar risk and stage 3 with low molecular risk. Patients with stage II CRC had better surviv al rates than Stage III patients and as cancer progressed, the survival rates showed a trend with decreased survival with time. Similar results were observed for lymph node involvement, with better survival rates for patients with no nodal involvement as compared to <3 group and > 3 group of nodes positive patients and survival rate decreased with time.
36 Table 8 Univariate Analysis Variable Parameter Estimate ChiSquare Hazard Ratio (95% CI) p-value Stage (Stage 3 vs Stage 2) 0.776 3.61 2.17 (0.98-4.8) 0.06 Molecular Risk GEP (High risk vs Low risk) 0.987 5.46 2.68 (1.2-6.15) 0.02 Age (Upper 1/3rd vs Lower 2/3rd age group) 0.740 3.51 2.09 (0.9-4.5) 0.06 Gender (Female vs Male) 0.456 1.35 1.58 (0.7-3.4) 0.24 Race (Caucasian vs Other) -0.183 0.16 0.83 (0.3-2.1) 0.64 BMI (Overweight and obese vs Normal) -1.150 7.45 0.317 (0.1-0.7) 0.006 Family History of Colorectal Cancer -0.276 0.20 0.79 (0.2-2.6) 0.66 Grade/Differentiation (Poor vs moderate and well) 0.352 0.49 1.42 (0.5-3.8) 0.48 Regional Lymph Node Involvement (N1-N2 vs N0) 0.784 3.40 2.09 (0.9-5.0) 0.06 Location of Primary Tumor -0.309 1.57 0.73 (0.5-1.2) 0.21 Preoperative CEA Level (>5ng/ml vs < 5 ng/ml) 0.834 1.99 2.3 (0.7-7.3) 0.16 CEA: Carcinoembriyonic antigen, GEP: gene expression profiles.
37 Table 9: 3-year and 5-year Surviv al Estimates of the Study Population Following Curative Surgery Based on the Prognostic Variable Variable No. of Patients (%) 3-year survival rate (%) 5-year survival rate (%) Age 50 years 6 10 66.67 50 50-70 years 22 38 86.36 75.4 70 years 30 52 63.33 42.04 BMI <18.5 (underweight) 1 2 50 18.5-24.9 (normal) 18 34 55.56 31.33 25.9-29.9 (overweight) 21 40 90 72.73 30 (obese) 13 25 76.92 68.83 >25 34 64 84.85 71.08 <25 19 36 55 33.39 Family History of CRC Yes 8 17 75 No 38 82 76.32 56 Molecular Risk Low 27 47 88.89 71.46 High 31 53 58.06 41.29 Stage Stage II 29 50 75.86 67.65 Stage III 29 50 68.97 43.97 IIA (T3 N0 M0) 26 45 76.92 67.57 IIB (T4 N0 M0) 3 5 66.67 66.67 IIIA (T1-2 N1 M0) 2 3 50 50 IIIB (T3-4 N1 M0) 14 24 78.57 48.21 IIIC (Any T N2 M0) 13 23 61.54 38.46 Lymph Node Involvement N0 27 48 77.78 68.87 N1 17 30 76.47 51.73 N2 12 21 58.33 33.33 Grade Well 4 7 75 NA* Moderately 44 77 77.27 55.24 Poor 9 16 55.56 44.44 Total LN Resected >12 28 48 73.33 57 <12 30 52 71.43 50 NA*: Censored, LN: Lymph Node, CRC: colorectal cancer
38 5.2.2 Multivariate analysis (S tepwise Model Selection): Research Question 3. What does the addition of molecular classi fication (GEP) together with clinical parameters (age, sex, family history, BMI, grade, location of tumor, preoperative CEA levels) contribute to the usual clinical predictiveness of outcome (survival) in stage II and III colorectal cancer? General Hypothesis: H0: i =0 (no difference of covariates) HA: i 0 (covariates influence survival) where, i is parameter estimate for multiple variables. If p-value >0.05, then H0 is not rejected. Is based on the Cox proportional hazard model: h (t,x) = h(t0) exp ( x). This model allowed the estimation of the eff ect of each covariate in the presence of the others. The hazard ratio for each variable is ad justed for the effects of all of the other variables in the multivariate model. The aim of the study was to establish whet her GEP with clinipat hological variables provides better prognostic information for patie nts with stage II and I II CRC in addition to that afforded by staging alone. In order to address this multivariate an alysis using stepwise Cox Regression method was used, with threshold of 0.05. Factors that were found to be significant in univariate analysis were tested for independent statisti cal significance in multivariate analysis. Three
39 prognostic variables: molecular risk, BMI and ag e were found to be independent predictors of overall survival (Table 10). Staging of tumor was found to be statistically not significant on multivariate model; however, it was retained in the final model in order to determine the association of other factors in the presence of clinical staging of tumor. None of the other variables were found to be significant. The lik elihood ratio test base d on maximum partial likelihood estimates was used for eliminati ng confounding. The results of the multivariate analysis for overall patient survival are shown in table 10. Table 10: Significant Prognostic Risk Factors for Mortality (Overall Patient Survival) Determined by Multivariate Analysis Variable Parameter Estimate ChiSquare Hazard Ratio (95% CI) p-value Stage (Stage 3 vs Stage 2) 0.897 2.78 2.45 (0.85-7.04) 0.09 Molecular Risk GEP (High risk vs Low risk) 1.343 5.26 3.83 (1.22-12.06) 0.02 Age (Upper 1/3rd vs Lower 2/3rd age group) 1.313 5.21 3.72(1.2-11.49) 0.02 BMI (overweight and obese vs normal) -1.22 5.78 0.29 (0.11-0.79) 0.016
40 E. Multivariate analysis of Molecular Risk and Clinical Stage: Fig. 9 Kaplan Meier Survival Curves for C linical Stage and Molecular risk combined. According to fig 9, that there is difference in overall survival when molecular risk and clinical staging were combined in multivariate analysis, and patients of the low molecular risk group had better survival outcome than patients with high molecular risk group within stage II and stage III clinical staging. Combining clinical stage with molecular risk classification we were able to differentiate patients into different strata. 0.00 0.25 0.50 0.75 1.00 Survival in Months 0 20 40 60 80 100 120 140 Lo g -Rank p = 0.0301 Stage 2 Low risk (n=16) Stage2, High Risk (n=13) Stage 3, Low Risk (n=11) Stage 3, High Risk (n=18) 3-year rate (%) 5-year rate (%) Stage II, Low Mol. Risk (n=16) 86.67 77.04 Stage II, High Mol. Risk (n=13) 58.33 50 Stage III, Low Mol. Risk (n=11) 89.34 65.63 Stage III, High Mol. Risk (n=18) 57.89 36.84
41 5.2.3 Predictor Model Factors, which were prognos tically significant (biological or clinical) and met the proportional hazard model by the assessment of log-minus log survival plot were included in the final model. Stepwise procedure was used and based on the assessment of log likelihood statistic final predictor model was selected. Hypothesis: Ho: Model with more variable s (predicted model) is similar to the model with fewer variables (basic model). HA: The two models are different. Basic Model: h (t,x) = h(t0) exp [( 1 (stage) Predicted model: h (t,x) = h(t0) exp [( 1 (stage)+ 2 (molecular risk)+ 3 (age)+ 4(body mass index)]. The Cox proportional hazard model: h (t,x) = h(t0) exp ( x), where, h(t0) is base line hazard function, h (t,x) is hazard at time t, is parameter estimate for the variable and x is value for each variable denoted by and for presence and absence of the prognostic factor. Based on the Cox Proportional Hazard Model the hazard ratios observed for stage, molecular risk, age and body mass are: h (t,x) = h(t0) exp [( 1 (stage)+ 2 (molecular risk)+ 3 (age)+ 4(body mass index)]. h (t, stage III) = exp [(0.897* (1))+ (1.343* (0))+ (1.313*(0))+ (-1.22* (0))]= 2.4522; h (t, stage III & high molecular risk) = exp[(0.897* (1))+ (1.343* (1))+ (1.313*(0))+ (-1.22* (0))]= 9.39; h (t, stageIII & high molrisk & old age& BMI >25) = exp [(0.897* (1))+ (1.343* (1))+ (1.313*(1))+ (-1.22* (1))]= 10.27; The hazard ratio for stage alone= 2.452. And the addition of molecular risk to stage increased the hazard to 9.39 and for the final m odel containing stage, molecular risk, age and
42 BMI was 10.27. So, there is evidence that th e advanced tumor stage (stage III), high molecular risk, old age and body mass index >25 result in higher the hazard ratio and decreased survival. Predictor model with clinical stage, mol ecular risk, age and BMI was compared with the model containing clinical stag e alone. The test statistic is equal to the difference between the 2LogL value in the model fit statistic for both models The test statistic = 125.256112.659= 12.597, which is greater than Chi-squa re, 3 degree of freedom and at 95% significance = 7.815. By observing HR, 95% CI for each factor and the change in the log likelihood statistic, it was ascertained th at clinical stage, mo lecular risk, age and BMI should remain in the final model. Model containing clinical st age in conjunction with molecular risk, BMI, and age was a stronger indicator of clinical outcome (p= 0.0056) than model with clinical stage alone. Fig.10 Survival Distribution for Fig.11 Survival Distribution for Prediction Model with Stage Prediction Model Containing Stage, Molecular Risk, Age and BMI
43 CHAPTER SIX DISCUSSION 6.1 Important Findings The mean age at diagnosis was 67.17 y ears (SE= 14.32%). The median age was observed to be 70.47 years, and the age range at diagnosis was 26.97 to 92.41 years. 33% of patients were in the upper 1/3rd age tertile and 67% we re in the lower 2/3rd age tertile. The median survival after surgery for Stage IIA and IIB was 4.6 and 5 years, respectively. And for stage III A, IIIB and IIIC the median su rvival was 3.7, 4.5 and 4.2 years, respectively. The median survival in years following surg ery for the low molecular risk group was 4.9 years and 4.1 years for the high risk group. The risk of mortality increased with a dvance in tumor stage and with age. A progressive decrease in 5-year survival rates was evident as cancer progressed from TNM stage II A to stage IIIC (table 9). Low mol ecular risk patients had be tter survival outcome and 5-year survival rates compared to high molecular risk patient. The inverse relationship seen with between BMI >25 and mortality could be due to the fact that most patients were overweight before cancer diagnosis and had significant weight loss from the time of diagnosis to surg ery at which time BMI was recorded and were less healthy due to cancer. Hence, BMI <25 indicates significant weight loss in these patients and poor survival rates due to adva nce in cancer, co-morbid condition and poor
44 immune status. Patients with BMI >25 at su rgery had 32% lower hazard for death than patients with BMI<25. Although, family history of CRC did not show to be significantly associated with poor outcome, information on history of CRC in the family and the size of the patient cohort may be insufficient to identify positive family history of CRC as a prognostic factor in this study. A larger patient cohort will be required to definitely determine whether family history of CRC at diagnosis improves the ac curacy of outcome prediction. Significant difference in survival outco me was not observed based on the race, because majority of the patients in the study were Caucasians and could not be generalized to the population. Grade, an important factor in survival prediction was not found to be a statistically significant prognostic f actor, as has been reported in previous studies. In some studies significant difference in survival was seen for preoperative CEA level, lymph node resection, regional lymph node i nvolvement, however, in this study no significant difference in survival was observed. Univariate analysis of the study showed 4 out of 10 variables were statistically significantly associated with overall survival (table 8). Among the clinical variables analyzed, gene expression profiling analysis distinguished two gr oups with significantly different survival outcomes (high risk vs low risk, HR: 2.68, p=0.0151, fig6). Borderline statistical significance was obs erved for TNM staging of tu mor (stages III vs II, HR: 2.17, p=0.0513, fig5). The strong measure of association indicates that the sign ificance is not due to chance alone. Being older (>74.27 years) doubled the hazar d for death compared to those younger (<74.25 years). Patients with BMI > 25 were f ound to have better survival and lower hazard
45 for death than patients with BMI <25 and an inverse relationship were seen between BMI and mortality. The hazard for death for patients in the BMI >25 group was 32% lower than that for patients in the BMI<25 group (Table 8) The hazard ratio for patients with lymph node involvement was two times higher than patients without nodal involvement. The purpose of the study was to investigat e the hypothesis that combined assessment of clinicopathological factors a nd gene expression profiles will allow increased accuracy of survival prediction for patients with CRC over the use of staging al one. A significant difference in survival was observed between the starta when staging and molecular risk were combined (p=0.0301, fig 9). In this study it was found that GEP (mol ecular risk) are power ful predictors of survival independent of clinicopathological pr edictors. Stage, molecular risk, age and body mass index, have been identified as significant prognosticators of surviv al in this study. Combining molecular classificati on, BMI and age of patients to cl inical staging, was a better indicator of clinical outcome than clinical staging alone. Four prognostic variables: TNM staging, molecular risk, BMI and age were found to be independent predictors of overall survival in multivariate analysis. 6.2 Strengths and Weakness of the study: Strength : Study had highly specific inclusion and exclusion crit eria and limited to only primary colorectal cancer of stage II and III pa tients. All surgical and pathology reports for the study population were individua lly verified. To minimize e rror due to case abstraction, majority of the patients charts were reviewed by same individual and for the few patients whose history was abstracted by other abstract ors, the information collected by them was
46 cross validated. Certain risk fact ors that could affect the survival were controlled for in this study in multivariate analysis by stratifying the variables based on presence or absence of the risk factor (postmenopausal hormone use, sm oking history, alcohol hi story). In the study only 10% of the patients were lost to follow-up. This type of study is useful to obtain information on colorectal cancer disease patterns over time and associate such patterns to the distribution of time to death among patients diag nosed with colorectal cancer. A study of this nature can be applied to public health prac tice by identifying factors that are important in cancer control and intervention, which can mi nimize mortality and morbidity by screening procedures and early diagnosis and treatment among the general population. The study was able to determine the contri bution of gene expression profile as a predictor along with other parame ters in survival prediction in colorectal cancer patients. Weakness: The study is a retrospective cohort st udy from which information on patients demographics and histopathologi cal characteristic were gather ed, which could contribute to misclassification. Due to the small sample si ze the study lacked power and hence significant differences in survival outcomes for certain prognostic variables such as preoperative CEA levels, grade and smoking and alcohol history, coul d not be determined. Cancer status at the end of follow up was ascertained from Moffitt Can cer Registry. Since 7 patients were lost to follow up and this is related to outcome (survival) and this could have cau sed attrition bias in the study. It is possible that differential miscla ssification of tumor stage and cause of death could have occurred either by re cording deaths from other causes as deaths from colorectal cancer or vice versa. Since most patients were Caucasians the resu lts of this study cannot be generalized to the general population, as the study sample does not represent the population in general. The
47 influence of confounding factors such as sm oking and alcohol consum ption could not be determined in this study, as detailed information could not be obtained from clinical charts. Other factors, which could modify the risk of developing CRC such as diet, postmenopausal hormone use and physical activity, were not take n into consideration because of inadequate information. Certain factors that could influe nce survival outcome such as body mass index at the baseline (before cancer diagnosis) and the effect of treatment following surgery were not taken into considerati on during data analysis. 6.3 Consistency with literature: As in this study; similar results were seen in other studies that identified stage, molecular risk and age as significant prognosti cators of survival. Data from literature and present study suggest clinical parameters, particularly st age [9, 39, 52], mo lecular risk [11,20, 22], age [1,2, 31] and body mass index [12,16,27] are related to patient survival rate and are most reliable prognostic factors. Prognostic factors that were found to be si gnificant in other studies but were not significant in this study were fa mily history [2, 40], histologi cal grade of tumor [21, 39, 52], and preoperative CEA levels [21, 30]. Differen ces between our study and others are mostly due to differences in the numb er of patients, lack of power length of follow-up, and grouping of continuous variables also because the influe nce of comorbid conditions on survival which could not be analyzed in this study, family history, personal history of cancers, dietary history, multivitamin use, physical activit y, screening history were not taken into consideration because of inadequate information.
48 6.4 Conclusion Because of the frequency of the disease, th e ability to identify high risk groups, better survival of patients with early-stage lesion, and the relative simplicity and accuracy of the screening tests, screening for col on cancer should be part of rou tine care for all adults at the age of 50 years, especially those with family hi story of colorectal cance r. Periodic evaluation following treatment of CRC helps to identi fy and manage recurrent disease. 6.5 Public Health Importance: Colorectal cancer presents a major health probl em with an annual estimated incidence rate of 106,680 new cases of colon cancer and 41,930 rectal cancers in 2006 and together they will cause 55,170 deaths in the US. Due to the long natural history of cancer there is time for early diagnosis and treatment before it reach es an advanced and incurable stage. Since CRC is highly treatable and detected early, effective preventive approaches help reduce the morbidity and mortality asso ciated with due the disease. Removal of premalignant lesion (adenoma) at the time of screening (colonoscopy) may be an effective form of primary prevention. Certain behavioral factors which modify th e risk of developing CRC such as dietary habits, physical activ ity, alcohol and cigarette smoki ng represent poten tial means of prevention. According to the ACS a guideline listed in Appendix 2, screening helps in early detection and reduces the risk of dying from CRC. People who have no identifiable risk factors (other than age) shoul d begin regular screening at th e age of 50 years. Those who have a family history or othe r risk factors for CRC polyps or cancer should start screening at
49 a younger age and more frequently. Following diagnosis and treatment for pe rsons with CRC, periodic evaluation by monitoring preoperative CEA leve ls, aids in earlier identif ication and management of recurrent disease and cancer progression. 6.6 Future directions For future studies it will be important to validate the gene expr ession cluster obtained in this study and the predictor model on an i ndependent dataset. To conduct a prospective study of a larger sample size and include pe ople of differences in ethnicity, study the prognostic ability of GEP and other factors su ch as family history, comorbid conditions, treatment following surgery, histopathologi cal features (p27, p53, VEGF), vascular, perineural and lymphatic invasion and the in fluence of comorbid conditions on outcome prediction. Analyze the inter action between clinicopathologi cal factors and GEP and other outcome measures such as time to relapse and treatment response. To determine if differences exists in CRC survival on the ba sis of screening between those persons who underwent regular screening and those w ho did not undergo regular screening as recommended by the American Cancer Society.
50 List of References 1. National Cancer Institute (NCI), 2005 2. American Cancer Society: Cancer Facts and Figures 2005. Atlanta, GA., 2005 3. Barrier, A. and S. Dutoit (2005). "Col on cancer prognosis prediction by gene expression profiling." Oncogene 24 : 6155-6164. 4. Berger, A. C. and e. al (2005). "Colon Can cer Survival is associated with decreasing ratio of metastatic to examin ed lymph nodes." J Clin Oncol 23 (34): 8706-12. 5. Bertucci, F., S. Salas, et al. (2004). "Gene expression profiling of colon cancer by DNA microarrays and correlation with hi stoclinical parameters." Oncogene 23 : 13771391. 6. Bostcik, R., J. Potter, et al. (1993). "Reduced risk of col on cancer with high intake of vitamin E: the Iowa womens' Health St udy." Cancer Epidemiol Biomarkers Prev 5 (11): 897-900. 7. Charlson, M. E. and R. C. MacKenzie (1987). "A new meth od of classifying prognostic comorbidity in longitudinal studie s: development and validation." J Chron Dis 40 (5): 373-383. 8. Concato, J. and A. R. Feinstein (1995). "Importance of events per independent variable in proportional hazards analysis Background, goals, and general strategy." J Clin Epidemiology 48 (12): 1495-1501. 9. Compton CC, Greene FL (2004): The stagi ng of colorectal cancer: 2004 and beyond. CA Cancer J Clin 54 (6): 295-308. 10. Doria-Rosa, V. P. et. al (2006). "Body mass index and the risk of death following the diagnosis of colorectal cancer in postm enopausal women (United States)." Cancer Causes and Control 17 : 63-70. 11. Eschrich, S. and T. J. Yeatman (2005). "Mol ecular staging for survival prediction of colorectal cancer patients." Journal of Clinical Oncology 23 (15): 3526-3535. 12. Frezza, E. E., M. S. Wachtel, et al. (2 006). "Influence of obesity on the risk of developing colon cancer." Gut 55 : 285-291.
51 13. Friedenreich, C. (2001). "Physical activity a nd cancer prevention: from observational to intervention research." Cancer Epidemiol Biomarkers Prev 10 (4): 287-301. 14. Fuchs, C., E. Giovannucci, et al. (1999). "D ietary fiber and the risk of colorectal cancer and adenoma in women." N Engl J Med 340 (3): 169-76. 15. Galizia, G., E. Lieto, et al (2004). "Determination of mo lecular marker expression can predict clinical outcome in colon carcinomas." Clinical Cancer Research 10 : 3490-3499. 16. Giovannucci, E. (2001). "Insulin, Insulin like growth factors and colon cancer: a review of the evidence." J. Nutr 131 : 3109-20s. 17. Giovannucci, E., D. Nikolova, et al. (1998). "Multivitamin use, folate and colon cancer in women in the Nurses' Health Study." Ann Intern Med 129 (7): 517-524. 18. Giovannucci, E., E. Rimm, et al. (1994). "A spirin use and the risk for colorectal cacner and adenoma in male health professionals." Ann Intern Med 121 (4): 241-246. 19. Gospodarowicz, M. K., D. E. Henson, et al. (2001). Prognostic Factors in Cancer New York, A John Wiley & Sins, inc. 20. Gryfe, R. and H. Kim (2000). "Tumor micros atellite instability and clinical outcome in young patients with colore ctal cancer." N Engl J Med 342 : 69-70. 21. Guerra, A. and F. J. Jimenez (1998). "Multi variate analysis of prognostic factors in resected colorectal cancer: A new prognostic index." Eur J Gast Hepatol 10 : 51-8. 22. Halling, K. and A. French (1999). "microsate llite instability and 8p allelic imbalance in stage B2 and C colorectal caners." J. NAtl Cancer Inst 91 : 1295-1303. 23. Iwatsuki, S. and T. E. Star zi (1999). "Hepatic resection for metastatic colorectal adenocarcinoma: a proposal of a prognostic scoring system." Journal of American College of Surgeons 189 (3): 291-299. 24. Jen, J. and H. Kim (1994). "Allelic loss of chromosome 18q and prognosis in colorectal cancer." N Engl J Med 331 : 213-21. 25. Johnston, P. G. (2004). "Of what value genom ics in colorectal cancer? Opportunities and challenges." Journa l of Clinical Oncology 22 (9): 1538-1539. 26. Kampman, E. and E. Giovannucci (1994). "Cal cium, Vitamin D, dairy foods and the occurrence of colorectal adenoma am ong men and women in two prospective studies." Am J Epidemiology 139 (1): 16-29. 27. Kune, G., S. Kune, et al. ( 1990). "Body weight and physical activity as predictors of colorectal cancer risk." Nutr Cancer 13 (1-2): 9-17.
52 28. Longley, D., D. Harkin, et al. (2003). "5 -Fluorouracil: Mech anism of action and clinical strategies." Nat Rev Cancer 3 : 330-338. 29. Mackillop, W. J. and C. F. Quirt (19 97). "Measuring the acc uracy of prognostic judgement in oncology." J Clin Epidemiology 50 (1): 21-29. 30. Moertel, C. and V. Go (1986). "The pre operative carcinoembriyonic antigen test in the diagnosis, staging and prognosis of colorectal cancer." cancer 58 (603). 31. Murphy, T. K. and e. al (2000). "Body mass index and colon cancer mortality in large prospective study." Am J Epidemiology 152 (9): 847-854. 32. O'Connell, J. B., M. A. Maggard, et al. ( 2004). "Colon cancer surviv al rates with the new American Joint Committee on Cancer si xth edition staging." Journal of the National Cancer Institute 96 (19): 1420-1425. 33. Ogunbiyi, O. and P. Goodfellow (1998). "Conf irmation that 18q allelic loss in colon cancer is a prognostic indicator." J Clin Oncol 16 : 427-33. 34. Parkin, D., P. Pisani, et al. (1999). "Global cancer stat istics." CA Cancer J Clin 49 (33). 35. Ponz de Leon et al. (1992) "Clinical and pathological prognostic Indicators in Colorectal Cancer". Cancer 69 : 626-635. 36. Prall, F. and e. al (2004). "Expression Profiling of Colorectal Carcinomas Using Tissue Microarray: Cell Cycle Regul atory Proteins p21, p27, and p53 as Immunohistochemical Prognostic Markers in Un ivariate and Multivariate Analysis." Applied Immunohistochemistry and Molecular Morphology 12 (2): 111-121. 37. Quirke, P. and P. Durdy (1986). "Local recu rrence of rectal ade nocarcinoma due to inadequate surgical resection. Histopathol ogical study of lateral tumor spread and surgical excision." Lancet 1 : 996-999. 38. Ratto, C. and F. Crucitti (1998). "Prognostic fa ctors in colorectal cancer. Literature review for clinical appl ication." Dis Colon Rectum 41 (8): 1033-49. 39. Rene A.C. et al (2004). "Prognostic factors in locally advanced colon cancer treated by extended resection". Rev. Hosp. Clin. Med. S. Paulo 59 (6): 361-368. 40. Sandler, R. S. (1996). "Epidemiology and risk factors for colorectal cancer." Gastroenterology Clinics of North America 25 (4): 717-734. 41. Shibata, D. and M. Reale (1996). "The D CC protein and prognosis in colorectal cancer." N Engl J Med 335 : 1727-32. 42. Slattery, M., M. Schumacher, et al. (1988). "P hysical activity diet and risk of colon cancer in Utah." Am J Epidemiology 128 (5): 989-99.
53 43. Sleisenger, M., M. Feldman, et al. (2004) Gastrointestinal and Liver diseases, Pathophysiology/ diagnosis and management Saunders. 44. Society, A. C. (2005). "Can cer Facts and Fiures." 45. Steinberg, S. and K. Barwick (1986). "Importance of tumor pathology and morphology in patients with surgically resected colorectal cancer." cancer 58 : 1340. 46. Strumer, T., R. Glynn, et al. (1998). "Aspirin use and colorectal cancer: post-trial follow-up data from the Physicians' Health Study." Ann Intern Med 128 (9): 713-20. 47. Tepper JE, O'Connell MJ, Niedzwiecki D, et al.(2001): Impact of number of nodes retrieved on outcome in patients with rect al cancer. J Clin Oncol 19 (1): 157-63. 48. Trichopoulos, D. and e. al (1997). Epideimiology of Cancer Philadelphia, LippincottRaven. 49. Vieira, R. A. C. and C. A. Melo (2004). "P rognostic factors in lo cally advanced colon cancer treated by extended resec tion." Rev. Hosp.Clin. Fac. Med 59 (6): 361-368. 50. Wang, Y., T. Jatkoe, et al. (2004). "Gene expression profiles a nd molecular markers to predict recurrence of Dukes B colon cancer." Jour nal of Clinical Oncology 22 (9): 1564-1571. 51. Watson, P. and K. Lin (1998). "Colorectal carcinoma survival among hereditary nonpolyposis colorectal carcinoma family members." Cancer 83 : 259. 52. Wolmark, N. and B. Fisher (1986). "the pr ognostic value of the modification of the Dukes' C class of colorectal cancer." Ann Surgery 203 : 115. 53. Wolters, U. and H. W. Keller (1996). "Col orectal cancerA multivariate analysis of prognostic factors." Eur J Surg Oncol. 22 : 592-7
54 APPENDIX 1 Colorectal Cancer Prevention Colorectal cancer occurs as a result of co mplex interaction between a persons inherited susceptibility (genetic) and environmental factors. Epid emiological and clinical investigations suggest that diet high in fat, calories, protei n, alcohol and meat and low in calcium and folate are associated with in creased with increased incidence of CRC. Modifiable factors: Diet: Diet high in fat, meat (both red and wh ite), alcohol and low in calcium and folate are associated with increased incidence of CRC. Evidence on whether diet high in fiber exerts a protective role in reducing the inci dence of CRC is mixed. A high-fiber diet is thought to be protective, because it accelerat es the rate at which fats pass through the bowel and reducing the exposure/ contact of the large intes tine to carcinogens [26, 40]. However, conflicting results are seen in some studies. NSAIDS: Some studies reported a reduction in colon cancer in cidence with the use of aspirin, with a 30% overall reduction in colo rectal cancer, includi ng a 50% reduction in advanced cases . However, in a followup study there was no association between the use of aspirin and the incidence of CRC . The use of NSAIDS as a primary prevention measure is being considered a nd will depend on the dose and duration of intake.
55 Cigarette smoking: Cigarette smoking is associated with an increased tendency to form adenomas and develop into colorectal cancer [2 ]. Most case control studies of cigarette exposure and adenomas have found an elevat ed risk for smokers. In the Cancer Prevention study II, a large nationwide cohor t study, multivariate-adjusted colorectal cancer mortality rates were highest among cu rrent smokers, intermediate among former smokers, and lowest in never smokers, with increased risk after 20 or more years of smoking history among both men and women. Ba sed on this study data, it was estimated that 12% of colorectal cancer deaths in th e US population in 1997 were attributable to smoking (Chao, A et al, 2000). A positive relati onship between alcohol intake and large bowel cancers was seen in some studies . Vitamins: An inverse association was found betw een the risk of CRC and intake of vitamins E ; the RR for the highest compared to the lowest quartile was 0.3 (95% CI, 0.19-0.54) . A similar association wa s seen for Vitamin D and folic acid intake and risk of CRC . Calcium: Several studies have observed an inve rse relationship between calcium intake and cancer risk. Orally ingested calcium binds with bile acids and fatty acids released into the intestine following a high fat diet to form insoluble compounds which are not harmful to the colonic mucosa and thereby re duces the exposure to the toxic effects of bile acids . Post menopausal female hormones: Epidemiologic Studies have suggested a decreased risk of colon cancer among users of postme nopausal female hormone supplements 
56 Physical activity: An inverse relationship is seen between level of physical activity and colon cancer incidence. A sedentary lifestyle ha s been associated with an increased risk of colorectal cancer in some st udies [13, 44] but not all . Colonscopy: Colonscopy with the removal of ad enomatous polyps helps in reducing the risk of CRC.
57 APPENDIX 2 ACS Screening Guidelines Beginning at age 50, men and women should ha ve 1 of the 5 screening option below: 1. A fecal occult blood test (FOBT) or fecal immunochemical test (FIT) every year 2. Flexible sigmoidoscopy every 5 years 3. FOBT/FIT every year plus sigmoidoscopy every 5 years 4. Double contrast barium enema every 5 years or 5. Colonoscopy every 10 years.
58 APPENDIX 3 CLINICAL CHART REVIEW DATA ENTRY FORM Note: Dates are entered in MM/DD/YYYY format. If month or day is unknown, specify value as 01. If entire date is unknown, specify value as 01/01/1111 2. Moffitt medical record number 3-4. Tissue for study: Note: This tissue should be the earliest colorectal primary; if colorectal primary tissue not available, then earliest available other site 1 First colorectal primary 2 Second colorectal primary 3 Local recurrence 4 Appendiceal adenocarcinoma 5 Hepatic metastasis 6 Pulmonary metastasis 7 Other metastasis: Site _________________________ (4) 5. Date of collection / / 6. Surgical Accession Number S _____ _________________ 7. Gender 0 Male 1 Female 8. Race/ethnicity 1 Caucasian 2 African American 3 Hispanic 4 Asian 5 Other/unknown 9. Date of birth (MM/DD/YYYY) / /
5910-18. Clinical presentation with index primary colorectal cancer: None recorded Asymptomatic Melena Hematochezia Change in bowel habits Abdominal pain Weight loss Clinically Obstructed Hct at presentation 0 No 1 Yes (10) 0 No 1 Yes (11) 0 No 1 Yes (12) 0 No 1 Yes (13) 0 No 1 Yes (14) 0 No 1 Yes (15) 0 No 1 Yes (16) 0 No 1 Yes (17) ____________ (18) 19. Weight loss in lbs prior to presentation with index primary colorectal cancer (use 999 if unknown) 20. Weight in lbs at presentation with index primary colorectal cancer (use 999 if unknown) 21. Height in inches (use 999 if unknown) 22-40. Charlson Index comorbidities : Myocardial infarct Congestive heart disease Peripheral vascular disease Cerebrovascular disease Dementia Chronic pulmonary disease Connective tissue disease Peptic ulcer disease Mild liver disease Diabetes Hemiplegia Moderate/severe renal disease Diabetes w/ end organ disease Non-metastatic cancer, other than colon cancer Metastatic cancer, other than colon cancer 0 No/unknown 1 Yes (22) 0 No/unknown 1 Yes (23) 0 No/unknown 1 Yes (24) 0 No/unknown 1 Yes (25) 0 No/unknown 1 Yes (26) 0 No/unknown 1 Yes (27) 0 No/unknown 1 Yes (28) 0 No/unknown 1 Yes (29) 0 No/unknown 1 Yes (30) 0 No/unknown 1 Yes (31) 0 No/unknown 1 Yes (32) 0 No/unknown 1 Yes (33) 0 No/unknown 1 Yes (34) 0 No/unknown 1 Yes (35) 0 No/unknown 1 Yes (36) 0 No/unknown 1 Yes (37)
60 Leukemia Lymphoma Moderate/severe liver disease AIDS 0 No/unknown 1 Yes (38) 0 No/unknown 1 Yes (39) 0 No/unknown 1 Yes (40) 41-44. Smoking history Packs per day Years Pack-years If any quantity unknown, specify 999. Give only numbers from chart. 0 No 1 Yes 2 Unknown (41) __________ (42) __________ (43) __________ (44) 45. History of alcohol use 0 No 1 Yes 2 Unknown 46. History of hormone replacement therapy 0 No 1 Yes 2 Unknown 47 55. Personal history of cancer other than the index primary colorectal cancer Cancer #1 Cancer #2 Cancer #3 ________ (47) ______ (48) ______ (49) CODING KEY additional primary cancers 0 Primary of 5 Urothelium unknown origin (renal pelvis, ureter, bladder) 1 2nd Colorectal 6 Endometrium 2 Stomach 7 Ovary 3 Small Intestine 8 Brain 4 Ampulla 9 Other 99 None Cancer #1 ___________________________
61 If cancer is 9 Other, specify primary: Did the additional primary cancer coexist with the index primary colorectal cancer? Cancer #1 Cancer #2 Cancer #3 (50) Cancer #2 ___________________________ (51) Cancer #3 ___________________________ (52) 0 No 1 Yes 2 Unknown (53) 0 No 1 Yes 2 Unknown (54) 0 No 1 Yes 2 Unknown (55) 56-92. Family history recorded in chart? Family history of cancer: Family member # 1 # 2 # 3 # 4 # 5 # 6 # 7 # 8 # 9 0 No 1 Yes (56) Relation 2nd Primary Primary to Patient Degree? Cancer #1 Cancer #2 _______ (57) _____ (58) _____ (59) _____ (60) _______ (61) _____ (62) _____ (63) _____ (64) _______ (65) _____ (66) _____ (67) _____ (68) _______ (69) _____ (70) _____ (71) _____ (72) _______ (73) _____ (74) _____ (75) _____ (76) _______ (77) _____ (78) _____ (79) _____ (80) _______ (81) _____ (82) _____ (83) _____ (84) _______ (85) _____ (86) _____ (87) _____ (88) _______ (89) _____ (90) _____ (91) _____ (92) CODING KEY Relation to Patient 0 Unknown 7 Grandmother 1 Mother 8 Grandfather 2 Father 9 Aunt 3 Sister 10 Uncle 4 Brother 99 No cancer
62 5 Daughter 6 Son CODING KEY2nd Degree? 0 1st degree or unknown 1 maternal 2nd degree relative 2 paternal 2nd degree relative Note: 2nd degree relatives are grandparents, aunts and uncles CODING KEY Primary Cancer 0 Primary of 5 Urothelium unknown origin (renal pelvis, ureter, bladder) 1 Colorectum 6 Endometrium 2 Stomach 7 Ovary 3 Small Intestine 8 Brain 4 Ampulla 9 Other 93. Location of index primary colorectal adenocarcinoma in patient 1 Cecum & Ileocecal Valve 2 Appendix 3 Ascending colon (Right colon) 4 Hepatic flexure of colon 5 Transverse colon 6 Splenic flexure of colon 7 Descending colon (Left colon) 8 Sigmoid colon 9 Colon, NOS 10 Rectosigmoid junction (Rectosigmoid colon) 11 Rectum 94-97. TNM tumor stage at time of surgery for index primary colorectal cancer Primary Tumor (T ) (94) 0 T0 No evidence of primary tumor 1 T1 Tumor invades submucosa 2 T2 Tumor invades muscularis propria 3 T3 Tumor invades through the muscularis propria into the subserosa, or into non-peritonealized pericolic or perirectal tissues
63 TNM staging (continued) 4 T4 Tumor directly invades other organs or structures, and/or perforates visceral peritoneum Note 1 : Direct invasion in T4 includes invasion of other segments of the colorectum by way of the serosa; for example, invasion of the sigmoid colon by a carcinoma of the cecum Note 2 : Tumor that is adherent to other organs or structures, macroscopically, is classified as T4 5 TX Primary tumor cannot be assessed 6 Tis Carcinoma in situ; intraepithelial or invasion of lamina propria Note : Tis includes cancer cells confined within the glandular basement membrane (intraepithelial) or lamina propria (intramucosal) with no extension through the muscularis mucosa into the submucosa Regional Lymph Nodes (N) (95) 0 N0 No regional lymph node metastasis 1 N1 Metastasis in 1 to 3 regional lymph nodes 2 N2 Metastasis in 4 or more regional lymph nodes 3 NX Regional lymph nodes cannot be assessed Distant Metastasis (M) (96) 0 M0 No distant metastasis 1 M1 Distant metastasis 2 MX Distant metastasis cannot be assessed
64Radial Margins and Residual Tumor (R) (97) 0 R0 Complete resection, margins histologically negative, no residual tumor left after resection 1 R1 Incomplete resection, margins histologically involved, microscopic tumor remains after resection of gross disease 2 R2 Incomplete resection, margins involved or gross disease remains after resection 3 Resection margins cannot be assessed 98. Tumor stage at time of surgery for index primary colorectal cancer (use TNM classification to determine stage) 0 Stage 0 Tis N0 M0 1 Stage I T1 N0 M0 T2 N0 M0 2 Stage IIA T3 N0 M0 3 Stage IIB T4 N0 M0 4 Stage IIIA T1-T2 N1 M0 5 Stage IIIB T3-T4 N1 M0 6 Stage IIIC Any T N2 M0 7 Stage IV Any T Any N M1
658 Tumor stage cannot be assessed 99. Histologic grading of index primary adenocarcinoma 1 G1 Well differentiated 2 G2 Moderately well differentiated 3 G3 Poorly differentiated 4 G4 Undifferentiated 9 GX Grade not determined or unspecified 100. Histology Mucinous component? 0 No mucinous component noted 1 Yes 101. Histology Signet ring cells? 0 No signet ring cells noted 1 Yes 102. Histology Perineural invasion? 0 No perineural invasion noted 1 Yes 103. Histology Lymphatic invasion? 0 No lymphatic invasion noted 1 L Microscopic lymphatic invasion 104. Vascular invasion? 0 No vascular invasion noted 1 V1 Microscopic vascular invasion 2 V2 Macroscopic vascular invasion 105108. Metastases present at time of surgery for index primary colorectal cancer: Liver Lung Metastasis other than liver or lung 0 No 1 Yes (105) 0 No 1 Yes (106) 0 No 1 Yes (107) Site(s): __________________________ (108) 109. Date of surgery for index primary colorectal cancer (MM/DD/YYYY) If month or day is unknown, specify value as 01. If entire date is unknown, specify value / /
66as 01/01/1111. 110. Type of surgery for index primary colorectal cancer 0 Not resected 1 Subtotal colectomy 2 Right hemicolectomy 3 Transverse colectomy 4 Left hemicolectomy 5 Sigmoid resection 6 Low anterior resection (rectal cancer) 7 Abdominoperineal resection (rectal cancer) 111. Were resection margins clear on index primary colorectal cancer? 0 No 1 Yes 2 Primary not resected 3 Unknown 112113. Size of index primary colorectal cancer from pathology report Length along axis of colon _________ cm (112) Maximum diameter __________ cm (113) 114115. Number of nodes taken at resection of index primary colorectal cancer Number of these nodes that were positive for cancer ____________ (114) ____________ (115) 116123. Treatment for hepatic metastasis present at time of surgery for index primary colorectal cancer: Surgical treatment Wedge resection Segmentectomy Lobectomy Radio frequency ablation Infusion pump Timing of this treatment for hepatic metastasis present at surgery for index primary colorectal cancer: 0 No 1 Yes (116) 0 No 1 Yes (117) 0 No 1 Yes (118) 0 No 1 Yes (119) 0 No 1 Yes (120) 0 No 1 Yes (121) (122) 0 None of the above treatments done for hepatic metastasis at time of surgery or later 1 Treatment done at time of surgery
672 Treatment delayed until a later time Date ______ / ______ / ______ (123) 124126. Treatment for pulmonary metastases at time of surgery for index primary colorectal cancer: Surgical treatment Wedge resection Lobectomy 0 No 1 Yes (124) 0 No 1 Yes (125) 0 No 1 Yes (126) 127. Preoperative treatment for index primary colorectal cancer 0 None 1 Chemotherapy 2 Radiation therapy 3 Both 128. Intra-operative radiation therapy at time of treating index primary? 0 No 1 Yes 129. Postoperative treatment for index primary colorectal cancer 0 None 1 Chemotherapy 2 Radiation therapy 3 Both 130. Recurrence of index primary colorectal cancer (radiographic) 0 No 1 Yes 2 Unknown 131135. Sites of recurrent colorectal cancer: Local Liver Lung Sites other than liver or lung 0 No 1 Yes (131) 0 No 1 Yes (132) 0 No 1 Yes (133) 0 No 1 Yes (134) Site(s): ____________________________ (135) 136. Date recurrence discovered / /
68(MM/DD/YYYY) If month or day is unknown, specify value as 01. If entire date is unknown, specify value as 01/01/1111. 137139. Treatment of recurrent disease: Surgical resection Radio frequency ablation (RFA) Chemotherapy 0 No 1 Yes 2 Unknown (137) 0 No 1 Yes 2 Unknown (138) 0 No 1 Yes 2 Unknown (139) 140. Date of treatment of recurrence by resection or RFA. If month or day is unknown, specify value as 01. If entire date is unknown, specify value as 01/01/1111. / / 141. Date of last recorded contact with patient in chart If month or day is unknown, specify value as 01. If entire date is unknown, specify value as 01/01/1111. / / 142209. CEA: Date drawn Level ______ /______ / ______ (142) ________ (143) ______ /______ / ______ (144) ________ (145) ______ /______ / ______ (146) ________ (147) ______ /______ / ______ (148) ________ (149)
69 CEA (continued) ______ /______ / ______ (150) ________ (151) ______ /______ / ______ (152) ________ (153) ______ /______ / ______ (154) ________ (155) ______ /______ / ______ (156) ________ (157) ______ /______ / ______ (158) ________ (159) ______ /______ / ______ (160) ________ (161) ______ /______ / ______ (162) ________ (163) ______ /______ / ______ (164) ________ (165) ______ /______ / ______ (166) ________ (167) ______ /______ / ______ (168) ________ (169) Date drawn Level ______ /______ / ______ (170) ________ (171) ______ /______ / ______ (172) ________ (173) ______ /______ / ______ (174) ________ (175) ______ /______ / ______ (176) ________ (177) ______ /______ / ______ (178) ________ (179) ______ /______ / ______ (180) ________ (181) ______ /______ / ______ (182) ________ (183) ______ /______ / ______ (184) ________ (185) ______ /______ / ______ (186) ________ (187) ______ /______ / ______ (188) ________ (189) ______ /______ / ______ (190) ________ (191) ______ /______ / ______ (192) ________ (193)
70______ /______ / ______ (194) ________ (195) ______ /______ / ______ (196) ________ (197) ______ /______ / ______ (198) ________ (199) ______ /______ / ______ (200) ________ (201) ______ /______ / ______ (202) ________ (203) ______ /______ / ______ (204) ________ (205) ______ /______ / ______ (206) ________ (207) ______ /______ / ______ (208) ________ (209) 210. Second tissue for study: Note: This tissue cannot be a first colorectal primary; if second tissue available, then it should be the next earliest available site 1 none 2 Second colorectal primary 3 Local recurrence 4 Appendiceal adenocarcinoma 5 Hepatic metastasis 6 Pulmonary metastasis 7 Other metastasis: Site _________________________ (211) 212. Date of collection, second tissue / / 213. Surgical Accession Number, second tissue S ______________________ 214. Third tissue for study: Note: This tissue cannot be a first colorectal primary; if second tissue available, then it should be the next earliest available site 1 none 2 Second colorectal primary 3 Local recurrence 4 Appendiceal adenocarcinoma 5 Hepatic metastasis 6 Pulmonary metastasis 7 Other metastasis: Site _________________________ (215) 216. Date of collection, third tissue / /
71217. Surgical Accession Number, third tissue S ______________________
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 Ka
controlfield tag 001 001795293
007 cr mnu|||uuuuu
008 070702s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001554
Gene expression profiles and clinical parameters for survival prediction in stage II and III colorectal cancer
h [electronic resource] /
by Mubeena Begum.
[Tampa, Fla] :
b University of South Florida,
ABSTRACT: Prediction of outcome in colorectal cancer (CRC) is currently based on the TNM staging classification; however, histopathological classification alone is insufficient for accurately predicting survival in stage II and III patients. Studies indicate that microarray gene expression profiles can predict survival in CRC. We hypothesize that tumor gene expression in combination with clinical parameters, is a better predictor of outcome in stage II and III colorectal cancers than the TNM stage classification alone. Clinical records and follow-up data were retrospectively reviewed for 58 Stage II and Stage III patients with primary colorectal cancer, who did not receive any neoadjuvant therapy preoperatively and whose samples had been previously analyzed for gene expression profiles using the Affymetrix U 133a Gene chip. For molecular classification of patients as being at high or low risk for poor survival, samples were divided into two clusters by hierarchical cluster analysis of genes selected by SAM. Univariate and multivariate analyses using Cox proportional hazard models were done to identify significant prognostic factors. The 3-year and 5-year survival estimates were 72.41% (SE=5.8%) and 55.17% (SE=6.7%), respectively, for all 58 patients. Univariate analysis showed that advanced stage, older age, high-risk molecular classification, positive lymph nodes were the statistically significant prognostic factors of poor survival (p < 0.05), while gender, preoperative CEA level, and family history of CRC in first degree relatives were not statistically significant. In multivariate analysis molecular classification, age and body mass index were independent significant prognostic factors. In Cox proportional hazard model, the estimated hazard ratios for Stage III vs II was 2.45 (95%CI: 0.85-7.04), for high vs low molecular risk was 3.83 (95%CI: 1.22-12.06) and old vs young age was 3.72 (95%CI: 1.2-11.49). Model containing clinical stage in conjunctionwith molecular risk, body mass index, and age was a stronger indicator of clinical outcome (p= 0.0056) than model with clinical stage alone. Gene expression profiles predict survival independent of clinical parameters, and the addition of gene expression profiles to stage is more predictive of survival than stage alone. Further analysis needs to be done to validate the molecular classification on an independent dataset.
Thesis (M.A.)--University of South Florida, 2006.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 71 pages.
Adviser: Thomas J. Mason, Ph.D.
x Public Health
t USF Electronic Theses and Dissertations.