PLASMA ORGANOCHLORINES, INTERACTION BETWEEN THE ARYL HYDROCARBON RECEPTOR GENE AND ORGANOCHLORINES, AND RISK OF NON-HODGKIN LYMPHOMA by CARMEN HOI-MAN NG B.Sc, University of Toronto, 2003 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Health Care and Epidemiology) THE UNIVERSITY OF BRITISH COLUMBIA August 2007 © Carmen Hoi-Man Ng 2007 Abstract The association between plasma organochlorine exposure and increased risk for non-Hodgkin lymphoma (NHL) has been documented, although results are not consistent. No studies have yet explored gene-environment interactions with organochlorine exposure, which may provide more insight into etiology of disease. One candidate gene to study such gene-environment interactions is the aryl hydrocarbon receptor (AHR) gene, which is involved in detoxification of organochlorines in the body. This case-control study was conducted to measure the association between plasma organochlorines and NHL risk, and how variants in the AHR gene may modify this risk. All HIV-negative NHL cases aged 20-79 were diagnosed between March 2000 and February 2004, and resided in the Greater Vancouver or Greater Victoria regions. Controls frequency matched by age, sex and region were identified from the Client Registry of the BC Ministry of Health. Demographic information was collected by telephone interview. 791 cases and 797 controls provided a DNA sample for genotyping of AHR, and seven single nucleotide polymorphisms (SNPs) were genotyped. Pre-chemotherapy blood samples from 422 cases and 460 controls were utilized for organochlorine measurement. Levels of 14 polychlorinated biphenyl (PCB) congeners and 11 organochlorine pesticide analytes were determined. Logistic regression was used to evaluate the association between AHR SNPs and risk of NHL, lipid-adjusted organochlorine levels and risk of NHL, and interaction between AHR SNPs and organochlorines. We found significant associations with several PCB congeners and other organochlorines. Organochlorines significantly associated with NHL include PCB congeners 99, 118, 138, 153, 156, 170, 180, and 187 as well as pesticides p-HCCH, DDE, hexachlorobenzene, mirex, oxychlordane and trans-nonachlor. The odds ratios for the highest versus lowest exposure categories for these organochlorines ranged from 1.4 to 2.7. The AHRIVS1+4640G/A SNP (rsl 7722841) was found to increase the risk of NHL. The odds ratio for the G/A or A/A allele compared to the G/G allele was 1.3 (95% CI=1.1-1.6). Significant interactions were also found between this allele and the organochlorines PCB 118, oxychlordane, and trans-nonachlor. For oxychlordane the odds ratio for the highest versus lowest exposure categories was 3.2 (95% CI=1.8-5.5) in the G/G genotype, but was not significant in the G/A or A/A genotypes (OR 1.3, 95%CI=0.5-3.4). The strong association observed for various PCB congeners and other organochlorines confirms the increased risk of NHL from organochlorine exposure. Results suggest the role of organochlorines in NHL etiology may involve the AHR pathway. - i i i -Table of Contents Abstract ii List of Tables vii List of Figures x List of Abbreviations xi Glossary xiii Acknowledgements xiv Dedications xv Chapter 1: Introduction 1 1.1 Purpose 1 1.2 Research Objectives 2 1.3 The Following Sections 3 Chapter 2: Background 4 2.1 Overview 4 2.2 NON-HODGKIN LYMPHOMAS 4 Description 4 Epidemiology 7 2.3 Risk Factors of Non-Hodgkin Lymphomas 7 Age, Sex, and ethnicity 7 Immunity 8 2.4 ORGANOCHLORINES 10 Description 10 Pesticides 10 Aldrin and Dieldrin 10 Hexachlorocyclohexane (HCCH) and Lindane 11 Hexachlorobenzene (HCB) 12 DDT and DDE 12 Mirex 13 Chlordane/ Oxychlordane, nonachlor, heptachlor/ heptachlor epoxide 13 Non-pesticide Organochlorines 14 Polychlorinated biphenyls (PCBs) 14 Dioxins and Furans 15 Dioxin-like effects of PCBs and Toxic Equivalency Factors (TEFs) 15 Health Effects from Organochlorine Exposure 18 Sources of Organochlorine Exposure in the General Population 20 Exposure Assessment Methods 22 - iv -2.5 Organochlorines and Risk of Non-Hodgkin Lymphoma 24 Occupational exposure and risk of NHL 24 Lifetime background exposure and risk of NHL 25 2.6 Gene-Environment Interactions 26 Genetic association to disease 26 Gene-Environment Interactions 29 The aryl hydrocarbon receptor gene (AHR) 31 2.7 Summary 32 Chapter 3: Methods 41 3.1 My Role in the Study 41 3.2 Study Design 41 Cases 42 Controls 44 Sample size and power calculations 45 Interviews 45 Organochlorine measurement 46 DNA Extraction 47 PCR Amplification and Sequencing 48 Whole Genome Amplification 49 Genotyping 51 3.3 Organochlorine Main Effects 52 Summary Descriptive Statistics 52 Data Coding 53 Bivariate analyses 54 The Logistic Regression Model 55 Selection of Confounders for Control 56 The Final Model 57 Additional Analyses 58 3.4 Genetic Main Effects 60 Testing for Hardy-Weinberg Equilibrium 60 SNP Analysis 61 Additional Analyses 63 Haplotype Analysis 63 3.5 Gene-Environment Interactions 64 Gene-Environment Interactions for Individual SNPs and Organochlorines 64 Gene-Environment Interactions for Haplotypes and Organochlorines 65 Chapter 4: Results 70 4.1 Overview 70 4.2 Organochlorine Main Effects 71 Population Characteristics 71 Organochlorine Exposure 72 Organochlorines and association to risk of NHL 75 4.3 Genetic Main Effects 79 SNP characteristics . 79 Association between AHR variants and risk of NHL.... 80 AHR haplotypes and association with risk of NHL 82 4.4 Gene-Environment Interactions 83 Chapter 5: Discussion 123 Chapter 6: Conclusions 153 References 156 - vi -List of Tables Table 2.1: The WHO TEF scheme for dioxin-like PCBs 35 Table 2.2: Correlation of serum and adipose tissue PCB concentrations from previous studies 36 Table 2.3: Correlation of serum and adipose tissue pesticide concentrations from previous studies 37 Table 2.4: Studies of the association between biological measures of organochlorine exposure and risk of NHL 38 Table 3.1: Minimum detectable odds ratios for organochlorine exposure and gene-environment interactions on risk of NHL based on sample size of 450 cases and controls 68 Table 3.2: Minimum detectable odds ratios for genetic exposure and risk of NHL based on sample size of 800 cases and 800 controls (alpha=0.05, power=90%) 68 Table 4.1: Characteristics of cases and controls [frequency (percentage)]. P-values for chi-square test comparing cases and controls included in organochlorine study analysis 85 Table 4.2: Disposition of laboratory measurements of organochlorine compounds in blood plasma 87 Table 4.3: Descriptive summary of organochlorine concentrations (p.g/kg lipid) by case/control status 89 Table 4 . 4 : Spearman rank correlation for PCB congeners and total sum of PCBs... 90 Table 4 . 5 : Spearman rank correlation for pesticides/ pesticide metabolites and total sum of PCBs 91 Table 4 . 6 : Descriptive summaries of lipid concentrations for cases/ controls andp-value from the Wilcoxon rank-sum test 92 Table 4 . 7 : Organochlorine and lipid concentrations for cases included and excluded from the analyses, and Wilcoxon rank-sum test for differences in concentrations 93 Table 4 . 8 : Median organochlorine concentrations by categories of potential confounders and p-values for the Wilcoxon rank-sum tests 94 Table 4 . 9 : Confounder variables included in final logistic regression models 98 - vii -Table 4.10: Total PCB concentrations, dioxin-like PCB concentrations and association with non-Hodgkin lymphoma 99 Table 4.11: Non dioxin-like PCB concentrations and association with non-Hodgkin lymphoma 100 Table 4.12: Plasma organochlorine pesticide/ pesticide metabolite concentrations and association with non-Hodgkin lymphoma 101 Table 4.13: Statistically significant interactions between organochlorines and covariates 102 Table 4.14: Forward stepwise selection of organochlorine exposures and association with risk of NHL 103 Table 4.15: Odds Ratios (95% CI) for the largest versus smallest organochlorine exposure categories by histologic subtype of NHL 104 Table 4.16: Characteristics of AHR gene variants included in study 105 Table 4.17: Single nucleotide polymorphisms (SNPs) of the AHR gene, genotype frequencies, and p-value for the test of Hardy-Weinberg equilibrium in European controls ,. 107 Table 4.18: Percentage of genotype data determined from whole genome amplified samples and reliability of whole genome amplification 109 Table 4.19: Genotype distribution of SNPs in the AHR gene by categories of covariates, and chi-square tests 110 Table 4.20: Covariates included in logistic regression models measuring the association between AHR variants and NHL risk chosen by change in estimate criterion (5% threshold) 112 Table 4.21: Variants the AHR gene and association with Non-Hodgkin lymphoma 113 Table 4.22: Variants of the AHR gene and association with Non-Hodgkin lymphoma in Europeans using additive models of phenotypic expression 114 Table 4.23: Odds ratios (95% confidence intervals) from additive models of phenotypic expression within NHL subtypes 114 Table 4.24: Statistically significant interactions between AHR SNPs and confounder variables 115 - vii i -T a b l e 4 .25 : Estimated haplotypes and haplotype frequencies from seven AHR SNPs (0=maj or allele, 1 =minor allele) 116 T a b l e 4 .26 : Association between AHR haplotypes and risk of Non-Hodgkin lymphoma, in all ethnicities and in Europeans only 118 T a b l e 4 .27 : Log likelihood ratio test (LRT) p-values for models with gene-environment interactions between the IVS1+4640G/A SNP and organochlorine exposure H9 T a b l e 4 .28 : Odds ratio estimates for interactions between the AHR IVS1+4640G/A SNP and organochlorine exposure 120 T a b l e 4 .29 : Odds ratio estimates for interactions between the AHR IVS1+4640G/A SNP and organochlorine exposure in Europeans 121 T a b l e 4 .30 : Global log likelihood ratio test (LRT) p-values for interaction between haplotypes and organochlorine exposure 122 - I X -List of Figures Figure 2.1: Age-standardized incidence rates of NHL in Canada by sex, 1978-2007 33 Figure 2.2: Models of interaction between a genotype and an environmental risk factor on the risk of disease, from Ottman, R, 1996 34 Figure 3.1: Subject recruitment for parent NHL study and included in current organochlorine study 67 Figure 3.2: Location of genotyped SNPs on the AHR gene 69 Figure 4.1: Pairwise r2 values of linkage disequilibrium between the AHR SNPs, based on genotyping data 106 - x -List of Abbreviations 3-MC 3-Methylcholanthrene A T L L Adult T-cell leukemia/ lymphoma AHR Aryl hydrocarbon receptor gene AhR Aryl hydrocarbon receptor AIDS Acquired immunodeficiency syndrome BMI Body mass index CLL/SLL Chronic lymphocytic leukemia/ small lymphocytic lymph CRD Capital Regional District (Greater Victoria) CTL Cytotoxic T lymphocytes CYP Cytochrome P450 DDE l,l,-dichloro-2,2'-bis(p-chlorophenyl)-ethylene DDT 1 ,l,l-trichloro-2,2'-bis(p-chlorophenyl)ethane DLBC Diffuse large B-cell DNA Deoxyribonucleic acid EBV Epstein-Barr virus EGFR Epidermal growth factor receptor FDR False discovery rate GVRD Greater Vancouver Regional District HCB Hexachlorobenzene P-HCCH beta-Hexachlorocyclohexane y -HCCH gamma-Hexachlorcyclohexane HCV Hepatitis C virus HHV-8 Human herpes virus-8 - xi -HIV Human immunodeficiency virus H T L V - 1 Human T-cell lymphotropic virus-1 H W E Hardy-Weinberg equilibrium IUPAC International Union of Pure and Applied Chemistry IVS Intervening sequence L D Linkage disequilibrium M A L T Mucosa-associated lymphoid tissue M A P K Mitogen-activated protein kinase N H L Non-Hodgkin lymphoma PB Phenobarbital , PCB Polychlorinated biphenyls PCDDs Polychlorinated dibenzo-para-dioxins PCDFs Polychlorinated dibenzofurans P C R Polymerase chain reaction SNP Single nucleotide polymorphism SV40 Simian virus 40 T C D D 2,3,7,8- tetrachlorodibenzo-p-dioxin TEFs Toxic equivalency factors W H O World Health Organization X Exon X R E Xenobiotic response elements - xii -Glossary Allele - when there is variation of a DNA sequence at a given locus, each distinct version of the DNA sequence is an allele. When there are two alleles in the population, the less frequent one is the minor allele, and the more common one is the major allele. Heterozygotes - individuals who have different alleles on different chromosomes at a certain locus. Homozygotes - individuals who have the same allele on different chromosomes at a certain locus. Ligand - an extracellular molecule that binds to a cellular protein receptor. Locus - chromosomal location of a gene or DNA sequence. The plural of locus is loci. Genotype - refers to the genetic composition of an individual on a genome level, or to the specific alleles at a certain locus Phenotype - the physical and measurable characteristics that are a manifestation of a person's genotype and environmental influences. They include physical traits like eye and hair colour, disease states (presence and absence), and quantitative traits such as blood pressure. Recombination - genetic process by which a combination of alleles on a chromosome is rearranged during germ cell formation, so that the combination of alleles in the offspring can be different from that of a parent. - xi i i -Acknowledgements I would like to thank my supervisor John Spinelli for all the help and encouragement during my graduate training, and for patiently answering my numerous statistics questions. I would also like to thank Agnes Lai, who provided a tremendous amount of support for me throughout my time at the BC Cancer Agency. I also want to acknowledge Payal Sipahimalani for helping me with some of the bioinformatics tools online, and for answering my genetics questions. I would like to acknowledge the Michael Smith Foundation for Health Research for their financial support. - xiv -Dedications Firstly, I would like to dedicate this thesis to Barry Triggs, who supported me emotionally throughout the last two years. I thank him for patiently dealing with me at times of stress, and trying to help out whenever possible. I also want to dedicate this to my family, who provide endless support and always allowed me to make my own decisions. And lastly, I dedicate this thesis to Agnes Lai, who helped me in so many ways, and was always providing encouragement. She made the long hours spent working on this thesis tolerable, and I look forward to our graduation together. - X V -Chapter 1: Introduction 1.1 Purpose Non-Hodgkin lymphomas (NHL) are a collection of cancers of the lymphatic system, and are distinct from Hodgkin lymphomas because Hodgkin lymphomas have specific cells known as Reed-Sternberg cells (1). The incidence of NHL has doubled in Canada and the United States over the last few decades (1, 2). The same pattern can be observed in developed countries in all parts of the world. However, research over this same time period has given little insight into etiological factors and their role in disease pathogenesis. Some studies have implicated organochlorines, a group of chemicals used in industry and agriculture, as an environmental risk factor for NHL. Many of these studies looked at occupational exposure to organochlorines, while a smaller group of studies have looked at low background exposure in the general population. It is the exposure in the general population that is of interest because the rising incidence of NHL is seen globally, across ethnic groups, for all ages, and in males and females. Due to limitations in the studies, they have yielded weak or inconclusive results. But the studies still suggest a possible positive relationship between organochlorines and NHL, and whether or not the hypothesis is true remains to be confirmed. The role of environmental factors on the risk of NHL is complicated by genetics. Cancers are caused by both environmental and genetic factors, and a documented effect of family history on risk of NHL (3, 4) indicates a genetic cause to the disease. It is therefore important to look at both types of risk factors, and the interplay between them, in order to paint a complete picture of NHL pathogenesis. This thesis looks at the risk of NHL associated with plasma organochlorine concentrations, and how a gene involved in the metabolism of these compounds can affect the risk of NHL. Organochlorines have long been banned from production and use in North America, but they remain in the environment and continue to pollute the air, waterways, and food supply. If NHL risk is associated with exposure to organochlorines, there are broad implications for environmental protection and cancer prevention. Continued production and use of some organochlorines in developing countries may need to be reconsidered, and public awareness will be needed to reduce intake of foods with high concentrations of organochlorines. These steps are not possible unless adequate risk estimates are available. The research hypothesis of this study is that lifetime background exposure to organochlorine residues through diet is a significant risk factor for NHL. Plasma levels of these residues are assumed to reflect this lifetime exposure, and so increasing concentrations of plasma organochlorines are hypothesized to increase risk for NHL: If there is an increase in risk, the risk may be modified by variations in the aryl hydrocarbon receptor gene (AHR). The AHR gene is involved in the metabolism of a subset of organochlorines including dioxins, furans, and dioxin-like polychlorinated biphenyls, and is responsible for some of the biological responses to organochlorine exposure. 1.2 Research Objectives The specific objectives of the research study are to: • determine whether plasma levels of organochlorine compounds, including a number of PCB congeners and pesticides, are related to risk of NHL, and whether the risk is associated with increasing levels of combined or specific organochlorine compounds; • determine if genetic variations in the AHR gene, in the form of single nucleotide polymorphisms (SNPs) and haplotypes, confer an increased risk of NHL; -2-• and determine if NHL risk is associated with interaction between organochlorine exposure and genetic variants in the AHR. 1.3 The Following Sections Chapter 2 reviews the literature and summarizes the evidence for organochlorines' involvement in NHL carcinogenesis. The chapter begins with an introduction to NHL, its biology, classification, and epidemiology. Then the known risk factors for NHL are discussed. Next, organochlorines are reviewed, with information on their chemistry and uses, their metabolism in humans, their biological effects in humans, and sources of exposure. The evidence supporting the hypothesis that organochlorines are a risk factor for NHL is then discussed. Subsequently, there is a section to introduce gene-environment interactions and their importance in research, and finally, the candidate gene, AHR, is reviewed. Chapter 3 presents the methods of the study. The section contains details on study design, case and control selection, data collection, organochlorine measurement, DNA genotyping, and analysis methods. Chapter 4 presents the study results, starting with the main effects of organochlorine exposure on the risk of NHL, followed by the main effects of the AHR gene on the risk of NHL, and lastly, interaction effects between organochlorines and genetic variants. Chapter 5 discusses the study results and how they fit into the knowledge of existing literature. This chapter will also discuss possible mechanisms of disease and limitations of the study. Chapter 6 provides a summary, and provides implications for future research and research questions. Chapter 2: Background 2.1 Overview The rising incidence of non-Hodgkin lymphoma over the last decades remains largely unexplained. Aside from conditions causing immunodeficiency, risk factors remain unknown. One possible risk factor is exposure to organochlorines. Organochlorines persist in the environment and contaminate food supplies. Humans then accumulate the substances in the body over their lifetimes. Epidemiological studies looking at the association between background exposure to organochlorines and risk of NHL suggest a positive association. However, results are inconclusive as to which organochlorines are responsible for the increased risk. These studies lacked statistical power to detect weak associations between specific organochlorines and NHL. These studies also did not look at gene-environment interactions, which are important in obtaining better risk estimates and gaining insight into etiology. The AHR gene is involved in the processing of organochlorine residues in the body, and is a good candidate to study gene-environment interactions and risk of NHL. The following sections will first discuss general information on NHL and organochlorines, secondly the possible relationship between the two, and finally the role of genetics within the relationship. 2.2 NON-HODGKIN LYMPHOMAS Description The name lymphomas suggest they are cancers involving the lymphatic system. The main functions of the lymphatic system are fluid balance in the internal environment and immunity. It is composed of organs, lymph nodes, and lymph vessels. Lymph organs include the spleen, thymus, and bone marrow. The spleen is connected to the circulatory system and is important in filtering out microorganisms and infections in the blood. Lymphocytes, white blood cells involved in the immune system, are produced in the bone marrow. B-lymphocytes (B-cells) mature in the bone marrow while T-lymphocytes (T-cells) mature in the thymus gland. Other blood cells such as monocytes and leukocytes are also produced in the bone marrow. Lymph nodes are areas of connective tissue filled with lymphocytes. Lymphocytes also circulate throughout the body via lymph vessels and capillaries, which are closely associated with the circulatory system. Lymphomas are a collection of cancers that originate in lymphocytes. Lymphomas are classified either as Hodgkin lymphoma or non-Hodgkin lymphoma. Hodgkin lymphoma is distinguished by the presence of Reed-Sternberg cells (1). The more common non-Hodgkin lymphoma (NHL) is actually a variety of subtypes based on morphology, immunophenotypes, and somatic genetics. The current classification used internationally is the World Health Organization (WHO) classification (5). The WHO classification, published in 2001, is an update of the Revised European-American Classification of Lymphoid neoplasms (REAL) created in 1994, and includes all hematological and lymphoid tumors (6, 7). The WHO classification initially groups non-Hodgkin lymphomas based on cell lineage, into B-cell neoplasms, and T/natural-killer cell neoplasms. Within these two groups, they are divided into precursor (lymphoblastic) neoplasms and mature (peripheral) neoplasms. Even though the distribution of NHL subtypes differs by geographic location, the two most common subtypes by far in Western countries are diffuse large B-cell (DLBC) lymphoma, and follicular lymphoma. Both of these subtypes fall into the classification of -5 -mature B-cell neoplasms. In Vancouver, British Columbia, each of these two subtypes contributes to approximately 30% of all NHL cases. Other common subtypes of NHL include marginal zone B-cell lymphoma of mucosa-associated lymphoid tissue (MALT), mantle cell lymphoma, chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), and peripheral T-cell lymphoma. Although these subtypes are less common than diffuse large B-cell and follicular lymphomas, they also contribute to a large percentage of all NHL cases (8). MALT lymphoma, mantle cell lymphoma, and CLL/SLL are all under the classification of mature B-cell neoplasms. Peripheral T-cell lymphoma is classified as a mature T-cell neoplasm. In order to classify patients into their probability of survival and thus determine the treatment decision, the neoplasms are also classified into one of four stages according to the Ann Arbor Classification. The stages represent distribution and extent of disease, from stage I involving a single site to stage IV involving multiple organs or sites. All stages are further divided into absence (A) or presence (B) of the following systemic symptoms: fever, night sweats, or weight loss of more than 10% body weight. B-symptoms occur in approximately one third of patients (6, 7). Prior to the development of the WHO classification and still in popular use today, NHL is often categorized as indolent, localized indolent, aggressive, or highly aggressive. These categories are not part of the WHO classification but are nevertheless useful for conceptual grouping of NHL. Indolent lymphomas, which include follicular and marginal zone lymphomas, are slow-growing and patients can survive for many years without treatment. However, they are almost always incurable. Indolent lymphomas are often stage III/rV disease, which means that the tumour has spread to numerous sites on both sides the body. Localized indolent lymphomas are similar to indolent lymphomas except they are -6-limited to a single site or one side of the body ie. stage I/II disease. Aggressive and highly aggressive lymphomas indicate rapidly growing tumours and are often high stage disease (5). The presence of systemic B-symptoms are less common in indolent lymphomas, and are associated with more aggressive lymphomas such as diffuse large B-cell lymphoma (7). Treatment for NHL includes radiation therapy, single-agent or combination chemotherapy, immunotherapy using monoclonal antibodies directed at lymphocytes, and stem cell transplantation. Treatment usually involves a combination of these therapies. The type of treatment varies depending on the lymphoma subtype, the stage of disease and the patient's condition. Treatment outcome and prognosis are also variable and major determinants include the morphology and histology of the lymphoma (7). Epidemiology Non-Hodgkin lymphoma is the 5th most common cancer in Canada, with an estimated 6600 newly diagnosed cases and 3000 deaths in 2006 (9). The incidence rate of NHL has doubled over the last two decades (Figure 2.1), with an age-standardized incidence rate of 18.9 per 100 000 for males, and 13.6 per 100 000 for females in the year 2003 (10). Similar increasing trends have been seen in developed nations around the world (1, 10-12), making NHL an increasingly important contributor to disease burden. The rising incidence rates in developed countries have been documented in males and females, African-American and Caucasian populations, and across age groups. NHL incidence rates are lowest in Asian and African countries, and highest in North America and Australia (1, 11-13). 2.3 Risk Factors of Non-Hodgkin Lymphomas Age, Sex, and ethnicity The risk of NHL increases exponentially with age (1, 13), with a peak incidence at age 80-84 in Canada for the years of 1994-2003 (14). NHL is more common in men than in women, with an estimated incidence rate of 3600 per 100 000 for men compared to 3000 per 100 000 for women in Canada in 2006 (9). In the United States, incidence rates are higher in Caucasians than in African-Americans. From the period of 1978 to 1995, the incidence rate in Caucasians males was 17.1 per 100 000 compared to 12.6 per 100 000 in African-American males, while the incidence rate in Caucasian females was 11.5 per 100 000 compared to 7.4 per 100 000 in African-American females (13). Immunity NHL is a malignancy originating from lymphocytes, cells involved with the immune system. Thus, it is not surprising that many conditions affecting immunity are known risk factors of NHL. These conditions include congenital immunodeficiencies, autoimmune diseases, immunosuppression due to drugs, and infections. Children or young persons with congenital immune deficiency disorders, such as Wiskott-Aldrich syndrome, ataxia telangiectasia, and severe combined immunodeficiency are at an excess risk of developing NHL compared to the general population. Autoimmune diseases, in which the body's immune system fails to recognize and attacks itself, have also been associated with NHL. The data is best described for rheumatoid arthritis, where there is a two-fold increase in risk (11, 15). Other autoimmune diseases shown to be associated with NHL include Sjogren's syndrome, celiac disease, and systemic lupus erythematosis (11,12). Immune suppression can also occur by taking therapeutic drugs, immunosuppressive drugs are taken after organ transplantation to prevent rejection of the donor organ. A strong association between NHL risk and patients receiving immunosuppressive therapy after organ transplantation has been observed. The risk increases with the degree and duration of immunosuppression, with heart transplant patients having the highest risk because they -8-receive the strongest dosage of immunosuppressive drugs compared to other transplant patients (16). Lymphomas in transplant patients are often high grade and positive for Epstein-Barr virus (EBV) infection (11). EBV is a highly prevalent herpes virus that is found worldwide. The virus infects B-cells, and transforms them to proliferate clonally. The infection remains latent in healthy individuals because proliferating B-cells infected by EBV are targeted by the immune system. However, for immune suppressed individuals, the infected B-cells are allowed to proliferate uncontrollably and may become precursors to tumor cells (1, 11, 12). Another important infectious agent linked to NHL is the human immunodeficiency virus (HIV). HIV infects CD4+ T-cells and depletes these lymphocytes in the body. The subsequent loss in immune function leads to acquired immunodeficiency syndrome (AIDS). The relationship between HIV infection/ALDS and NHL has been well described. NHL is a one of the most common malignancies following development of AIDS. The increase in risk of NHL in ALDS patients has been reported to be within the range of 60 fold to over 200 fold. AIDS related NHL is consistently B-cell derived, high grade, and often extranodal (ie. not in the lymph nodes) in nature. However, a variety of subtypes have been observed (17). Other infections have been associated with specific subtypes of NHL. Human T-cell lymphotropic virus-1 (HTLV-1), endemic to Southwestern Japan, the Caribbean, and Southeastern United States, has been found to cause adult T-cell leukemia/lymphoma (ATLL). The bacterium Helicobacter pylori, which is found in the stomach, has been associated with gastric MALT (mucous-associated lymphoid tissue) lymphoma. Human herpes virus-8 (HHV-8), hepatitis C virus (HCV), and simian virus 40 (SV40) have also been implicated as risk factors for NHL(1, 11, 12). Immune impairment, due to the various reasons just described, is the only well known risk factor for NHL. However, the rise in the numbers of people with AIDS and organ transplantations cannot account for the entire rise in NHL incidence (13). Advancements in diagnosis and classification in the past may also have contributed to some of the increase. But the upward trends in NHL incidence continue to persist in some parts of the world, indicating that other unknown risk factors exist and are continuing to contribute to disease risk. 2.4 ORGANOCHLORINES Description Organochlorines (OC) are a diverse group of organic compounds that contain chlorine. They are very stable compounds and are also lipophilic, or fat-soluble. They are synthetic chemicals used in agriculture, industry or by-products of chemical processes in industry. The large-scale production of organochlorines occurred mainly after World War II (18) and were widely used until the 1970s when fear of environmental pollution and damage to public health lead to the ban of most of the substances. They can be broadly grouped as pesticides and non-pesticides. Pesticides Aldrin and Dieldrin Aldrin and dieldrin are both insecticides that were fist introduced commercially in 1950 and were used widely until the 1970s. Both were used for agricultural purposes. Dieldrin was also used in tropical countries to control the tsetse fly and other vectors of disease. Worldwide, their production and use were banned in the 1970s, with exceptions in some countries. For example, some countries still use the two insecticides for termite control, and in Japan, their manufacture is still allowed but controlled by the government -10-(19). In the United States, production of aldrin and dieldrin has been reported into the late 1990's (20), while in Canada, sales and production of these pesticides were reportedly discontinued by 1989 (21). Aldrin and dieldrin are metabolized in the liver of the human body and aldrin is converted to dieldrin (19). Aldrin is also converted to dieldrin in the environment (20). Thus, unless a large short-term exposure occurs, aldrin is not detected in blood or tissue samples. Dieldrin is degraded slowly and accumulates in the body. It is found in blood samples in approximately 10% of the US population (18). Hexachlorocyclohexane (HCCH) and Lindane Hexachlorocyclohexane (HCCH), also known as benzene hexachloride, is an insecticide that was introduced around the time of World War II. It was used widely in agriculture and also for vector control. The chemical compound HCCH actually has five configurations called the alpha (a), beta (P), gamma (y), delta (5), and epsilon (e) isomers. HCCH insecticide consists of a mixture of all five isomers. The most toxic isomer is y-HCCH, commonly known as lindane. Lindane, as a pure compound, was also used for household pest control, and for treatment of human lice (19). An important metabolite of lindane is hexachlorobenzene (HCB), which metabolizes slowly and is stored in the body. HCB will be discussed further below. In the late 1970's, use of HCCH as a mixture of isomers was stopped. However, pure lindane continues to be used in agriculture and in some medicinal and consumer products worldwide. Although direct exposure to lindane can still occur because of its continued use, the p-HCCH isomer is the predominant isomer detected in human blood or tissue samples. In the US, 17%) of the population has P-HCCH detected in the blood, compared to only 0.2% of the population with y-HCCH in the blood (18). This is due to the greater stability of J3-HCCH, allowing it to be stored for longer periods in the body (19). Hexachlorobenzene ( H C B ) As discussed, HCB is a metabolite of lindane insecticide (19) but is also a contaminant of other pesticides (22). It was also produced as an agricultural fungicide and was used commercially in Canada from 1948 to 1972 (22). It was used as an intermediate in dye manufacturing, a plasticizer for polyvinyl chloride, and formed as a byproduct in production of chlorinated solvents, pesticides, and other chlorinated compounds. Currently, emissions of HCB as a byproduct into the environment continue from the manufacture or use of certain chlorinated solvents and pesticides (22). HCB is detected in the blood of 4.9% of the US population (18). D D T and D D E DDT is known under the chemical names l,l,l-trichloro-2,2'-bis(/> chlorophenyl)ethane or dichlorodiphenylchloroethane. It was introduced around the time of World War II and was a heavily used insecticide in agriculture, residences, institutions, and public health programs. DDT was first registered in Canada in 1948 and its use was stopped in the 1970's. The pesticide was no longer registered after 1985, although collection and disposal programs carried on well into the 1990's (23). Most countries have prohibited the use of DDT since the 1970's. However, it is still used in some tropical countries for control of insects and pests (19). DDT insecticide is typically a mixture of three different DDT configurations: The para-para DDT (p,/?'-DDT), ortho-para DDT (o,/?'-DDT), and ortho-ortho DDT (o,o'-DDT) isomers. The p,p'-DDT isomer makes up the largest percentage of the total mixture and is typically the most detected isomer (18). In the environment, DDT is degraded to DDE (1,1,-- 12-dichloro-2,2'-bis(p-chlorophenyl)-ethylene) by microorganisms. DDE is much more resistant to degradation than DDT and therefore has greater accumulation potential in humans. About 99% of the US population has detectable levels of DDE in the blood, compared to only 35% with detectable levels of DDT in the blood (18). Mirex Mirex, whose trade name is dechlorane, was used worldwide mainly for the control of fire ants, but was also used against termites and other insects. It has also been used as a flame retardant in plastics, rubber, paint, paper, and electrical goods. It was first used in the US around 1958, mainly for pesticide control, and was banned by the 1970's (19). In Canada, mirex was never registered for use as a pesticide but for non-agricultural purposes, and its use was banned by 1978 (24). Although mirex is a persistent environmental pollutant, it has been rarely detected in blood samples from the population (18, 19). However, higher levels have been detected in subpopulations near Lake Ontario and the St. Lawrence River who regularly consume polluted fish and fish-eating birds (24). Chlordane/ Oxychlordane, nonachlor, heptachlor/ heptachlor epoxide Chlordane has the chemical name 1,2,4,5,6,7,8,8-octachloro-2,3,3a,4,7,7a-hexahydro-4,7-methano-lH-indene. Technical chlordane was used as an insecticide and consists of a number of chemicals, including cis- and trans-nonachlor. Chlordane insecticide was used mainly for non-agricultural purposes, such as protecting buildings, lawns, ornamental trees, and drainage ditches from insects. By the 1980's, most use of chlordane has been limited. Since 1985, Canada has restricted the use of chlordane except for termite control (19). As of 1990, all uses of chlordane has been discontinued in Canada (25). Trans-nonachlor is a component of chlordane, but it is also one of the metabolites of chlordane. Another important metabolite of chlordane is oxychlordane. Chlordane, trans-- 13-nonachlor, and oxychlordane are all in the group of chemicals known as chlorinated cyclodienes. These chemicals are environmental contaminants and can persist for more than 30 years after their application (19). Heptachlor is also a chlorinated cyclodiene that is known chemically as 1,4,5,6,7,8,8-heptachloro-3a,5,7,7a-tetrahydro-4,7-methano-lH-indene. It was first discovered in 1946 and its use in Canada was discontinued in 1985 (19). Heptachlor was used widely for insect control and domestically for control of soil insects and termites. Heptachlor epoxide is a more persistent metabolite of heptachlor that is detected in human tissues (19). Non-pesticide Organochlorines Polychlorinated biphenyls (PCBs) Polychlorinated biphenyls (PCBs) are a group of chemicals with similar structure and properties. They consist of two benzene rings linked together with varying degree and positions of chlorination on the benzene rings. There are 209 possible variations of chlorination and they are called congeners. Each congener is given a number from 1 to 209 based on the chlorination structure. PCBs were used in industry beginning in 1929 and its uses were discontinued in the US by 1976 and shortly after in Canada in 1977 (19, 26). Aroclor is the trade name for some of the mixtures of different congeners used in industry. PCBs have important properties that make them useful. They are very stable and do not conduct heat well. Because of their insulating and dielectric properties, they were manufactured as coolants and lubricants in electrical equipment (27). Although PCBs are no longer manufactured, much of the equipment containing PCBs is still in service today. Thus, there is a continuing source of exposure to these organochlorines. PCBs can be detected in blood samples in almost everyone in the US population (18, 19). -14-Dioxins and Furans Polychlorinated dibenzo-para-dioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs) are commonly known as dioxins and furans. They are two families of chemicals with similar properties to PCBs. There are a number of congeners of PCDDs and PCDFs, based on the varying degree and positions of chlorination. PCDDs and PCDFs are not manufactured for use. They are contaminations from the production of other chemicals, such as PCBs, and from production of iron and steel. They are also formed from a number of other industrial processes, including the burning of waste (28). PCDDs and PCDFs are well known for their toxicity, and the most toxic congener is 2,3,7,8- tetrachlorodibenzo-p-dioxin (TCDD). Based on animal studies, PCDDs and PCDFs have been found to be carcinogenic, with the most common tumours formed in the liver. Animal studies have found dermal, immunological, endocrinologic, and major developmental defects in animals (28). In humans, the health effects of PCDDs and PCDFs are not as clear because most documented cases of high exposure are a result of a mixture of chemicals. PCDDs and PCDFs can be detected in 100% of blood samples from the US population (18). Dioxin-like effects of PCBs and Toxic Equivalency Factors (TEFs) PCDDs and PCDFs exert their responses by binding to the aryl hydrocarbon receptor (AhR) in cells. Induction of the AhR causes a signaling cascade in the cell, and as a result, the expression of a number of genes are increased or decreased (29). The functions of these genes exert the biological effects resulting from exposure, although the exact mechanisms and pathways are not clear. A number of PCB congeners have been shown to exhibit similar responses as PCDDs and PCDFs in animals. These PCBs also have similar structures as PCDDs and PCDFs, and - 15 -can bind to the AhR. The similarities suggest that PCDDs, PCDFs and PCBs all act via the same biological mechanisms (30). These PCBs are sometimes grouped as the dioxin-like PCBs. Dioxin-like PCBs include those with chlorine substitution in the non-ortho position and some with the mono-ortho position. PCBs with the non-ortho positioning, such as PCBs 77, 126, and 169, are also known as coplanar PCBs, because the atoms of these congeners line up in a single plane. The World Health Organization-European Centre for Environment and Health and the International Programme on Chemical Safety created the concept of Toxic Equivalency Factors (TEFs) to compare the toxicity of PCDDs, PCDFs, and dioxin-like PCBs (30-32). TEFs for each organochlorine were measured relative to the most toxic organochlorine, TCDD, and were based on available molecular, cell, and animal studies. Thirteen PCB congeners were originally considered to exhibit dioxin-like toxicity, and their TEFs range from 0.00001 to 0.1. The most current TEF scheme includes twelve PCB congeners with TEFs ranging from 0.00003 to 0.1. These TEFs are presented in table 2.1. Organochlorine Metabolism in Humans Organochlorines are known to be metabolized by cytochrome P450 (CYP) enzymes, proteins found in both prokaryotes and eukaryotes. Due to their ubiquitous nature, much of the information on their role in organochlorine metabolism comes from study in animals. The cytochrome P450 superfamily of monooxygenases is responsible for the conversion and metabolism of a variety of substances through various chemical reactions such as hydroxylation, epoxidation, and oxidation. Their substrates include compounds found in the body, such as fatty acids, sterols, and vitamins, as well as foreign chemicals such as pesticides and carcinogens (33). Metabolism by cytochrome P450 enzymes is normally divided into phase I and phase II reactions. Phase I reactions involve oxidation of the - 1 6 -substrates to form polar intermediates. Phase II reactions are known as conjugation reactions and form water-soluble products for excretion. Metabolism of organochlorines is typically very slow, allowing for their accumulation in tissues (34). The most well studied CYP enzymes of organochlorine metabolism are CYP1 Al and CYP1A2, which are induced by PCDDs, PCDFs, and dioxin-like PCBs via the AhR pathway. The CYP1A1 and CYP1A2 phase I enzymes are involved in hydroxylation and oxidation of PCDDs, and hydroxylation and sulfonation of PCDFs. The major metabolites of PCBs include hydroxylated products, methyl sulfonyl products, and methyl ether derivatives (34). The molecular pathway of CYP activation will be described later in this chapter when discussing AHR as a candidate gene for NHL susceptibility. CYP1A1 activation can also occur without the AhR pathway. Chen and Bunce (35) found that PCB congener 153 was able to induce CYP1A1 gene expression, even though PCB 153 cannot activate the AhR signal transduction pathway (36). The CYP1A1 gene has been postulated to be induced by other receptor-mediated signal transduction pathways (37). This signifies that many organochlorines, even those that do are not involved in the AhR pathway, may induce CYP1A1 expression. 3-Methylcholanthrene (3-MC) is a typical inducer of CYP1 Al (38) so organochlorines that are metabolized by CYP1A1 are sometimes classified as 3-MC type inducers in the literature. Organochlorines also activate other cytochrome P450 enzymes including the CYP2B enzymes (39), although this occurs independently of the AhR pathway. PCBs that are able to induce CYP2B genes are sometimes referred to as phenobarbital (PB)-like inducers, because PB is a typical and well studied inducer of these genes. The mechanism of CYP2B induction by PB and PB-like inducers remain a mystery, although the constitutive androstane receptor (CAR) pathway is thought to be involved (38, 40). Dioxin-like PCBs are typically metabolized by CYP1A1 - 17-enzymes (i.e. 3-MC type inducers), while non dioxin-like PCBs are metabolized by CYP2B enzymes (i.e. PB-like inducers) (38). There is less information available on the metabolism of individual pesticides, although studies in rats and mice have shown that DDT, DDE, dieldrin, chlordane, and mirex primarily induce CYP2B and 3A enzymes (41-43). Health Effects from Organochlorine Exposure PCBs and DDT make up the largest proportion of organochlorine residues measured in the body (44, 45) while PCDDs are the most toxic, and thus, the health effects associated with these compounds are the most intensively studied. The most recognized effect of these organochlorines is neurotoxicity from acute poisoning. Organochlorine pesticides are used to kill pests by attacking the nervous system and they affect humans in the same mechanism (46). There is little consensus on other health effects of the pesticides. The main reason is that exposure to pesticides often involves a mixture of chemicals, so it is hard to decipher which effects are attributable to each single pesticide residue. The same problem arises when studying the effects of PCBs and PCDDs, which typically consists of a mixture of different congeners. As well, PCDFs and PCDDs are contaminants of PCB production and are often highly correlated with PCB exposure (18, 47). The most documented human health effects of organochlorines come from accidental poisonings. These include two accidents in Asia involving PCB exposure, and a third accident in Italy involving dioxin exposure. The Yusho incident in Japan in 1968 affected 1700 people, and the second Yucheng incident in Taiwan in 1979 affected two-thousand people (18). The cause of these two poisonings was due to leakage of PCBs from equipment used for processing rice oil. PCDFs released from the PCBs also contaminated the rice oil, so it is unsure which of the effects are due to each contaminant. The most commonly - 18-observed- clinical outcomes were skin conditions such as chloracne and hyperpigmentation. Other manifestations included damages to the nervous system, immune system, and the liver. Follow up of the two populations documented an increase in mortality due to liver cancer and liver disease (18). The children born to mothers who were poisoned in these incidents were closely monitored for years after the acute exposure. The children also had hyperpigmentation of the skin, gums, and nails, and experienced many of the same symptoms as the adults (18, 47). Women who were pregnant at the time of exposure had children who weighed less at birth, and the children developed neurological and developmental problems (47, 48). In Seveso, Italy, an accident at a chemical plant exposed the population in the surrounding area with high levels of dioxin, particulary the most toxic congener TCDD. Immediately after the accident, the most well documented health effect was chloracne (49). Similar to the Yusho and Yucheng incidents, there were also documented alterations in immune, nervous system and liver function in Seveso victims. Long-term studies revealed increased mortality from cardiovascular disease, respiratory disease, diabetes, and specific cancers including non-Hodgkin lymphoma, although all-cause mortality and all-cause cancer did not increase (50). Groups with high exposure to organochlorines due to occupation or high consumption of heavily contaminated fish have been studied as well. Results have shown reproductive defects such as low semen quality or lower birth weigh of children (48), as well as immune changes (51, 52). Occupational exposure to PCBs has resulted in development of chloracne and abnormal liver function tests. DDT exposed workers also have experienced abnormal liver function tests (18). Aside from these observations, other health effects have not been confirmed. -19-Low organochlorine exposure in the general population has also been studied in relation to cancer and various reproductive, immunological, developmental, neurological and thyroid effects. Because organochlorines are persistent pollutants in the environment and in the food chain, the general population will continue to receive low dose exposure over the lifetime, predominantly through the diet. Whether this exposure has detrimental effects to humans will have impact on public health and public health guidelines. Aside from a subtle effect of background PCB and dioxin exposure on child neurological development (53, 54), most of this research has come up with weak or inconclusive results (18, 48, 54, 55). Sources of Organochlorine Exposure in the General Population Organochlorines have garnered much attention because of their persistent characteristics. They are stable and resist degradation both in the environment and within the body. This combined with their lipophilic properties allows them to accumulate in the fatty tissue of animals and humans. It has been noted that levels of organochlorines measured in the environment and human samples, with the exception of oxychlordane and HCB, have been decreasing since their discontinued use (19, 56). Nevertheless, exposure for any individual continues to increase over lifetime as organochlorines accumulate in fatty tissue. As well, organochlorines such as lindane and DDT are still being used in different parts of the world; organochlorines including HCB and PCDDs are released into the environment as byproducts; and PCBs continue to be released from older equipment. The long-range transport of organochlorine vapors in the atmosphere (57, 58) means that their distribution is worldwide and exposure is not localized at the source. So even with the ban of organochlorine use in North America since the 1970's, they continue to pose a threat to the environment and health. -20-Although organochlorines are detected in air, soil, and water, background exposure from these sources are negligible. The main source of exposure for the general population is from food, and a proportion of 80% (47) to 90% (59) of organochlorine exposure attributed to food intake have been estimated. Rivers, oceans and lake water are sinks for organochlorines from urban and agricultural runoffs, sewage, and industrial discharges. These residues then accumulate through the aquatic food chain. Predatory fish such as salmon and trout, which are consumed by humans, are among the highest sources of dietary organochlorine exposure. In particular, freshwater fish have a much higher levels of organochlorine contaminants than ocean fish (60, 61). Hites et al (62) also found that farmed salmon is a higher source of organochlorines than wild salmon. Other foods high in fat content will also be high sources of organochlorine intake because the residues are dissolved in fat. Meat, butter, dairy, and processed foods all contain organochlorines (60, 63, 64). On the other hand, fruits and vegetables contain minimal organochlorine contaminants and contribute only a small percentage of the total daily organochlorine intake (60, 63, 65). Fruits and vegetables are generally low sources of exposure due to the low lipid content. Also, there is no evidence of root uptake of organochlorines in plants (65). Therefore, organochlorine contaminants on fruits and vegetables are mainly accumulated on the surface. However, in one study that measured organochlorine contaminants in food items in the US, fruits and vegetables were found to be a significant source of DDE, comparable to the levels detected in meat, fish and dairy (60). An additional source of dietary organochlorine intake for infants is breast milk. The contaminant load in breast milk will be dependent on the body burden of the mother. Patandin et al (59) found that about half of the daily intake of PCBs and dioxins in pre-- 2 1 -school children are from dairy products. Dietary exposure to organochlorines in children is of particular concern in public health because the weight-adjusted dosage that they receive is actually much higher than that in adults due to their small size. For example, in Vancouver, Canada, infants 0 to 1 month old receive the highest dietary intake of PCBs, at 9.65 ng/kg body weight per day. In contrast, adults over 40 years of age on average receive less than 2 ng/kg body weight per day of organochlorine intake (66). Exposure Assessment Methods Many studies use questionnaire data on occupation, dietary habits, and other proxy measures to measure past organochlorine exposure. Questionnaire data can be quite inaccurate due to difficulty in remembering information from years or decades ago. As well, many people in occupations exposed to organochlorines may not be aware of specific chemicals that they worked with. Measurement of organochlorine levels in adipose or fatty tissue is the most accurate biomarker of exposure. Because organochlorines accumulate in fat and degrade slowly, their levels in adipose tissue reflect lifetime exposure to the compounds. The difficulty and costs of obtaining adipose tissue samples from a large cohort have deterred most population-based studies from using this exposure biomarker. Instead, many studies measure organochlorine levels in blood plasma as a proxy of levels in adipose tissue. When the body is in equilibrium, organochlorine residues dissolved in fat should theoretically be distributed equally in all tissues (67). Thus, the organochlorine levels in the lipid portion of the blood should equal the levels in adipose tissue. It is important that measured levels in plasma are adjusted for the lipid levels in the plasma in order to obtain an accurate proxy for adipose tissue levels (67). - 2 2 -A number of studies have measured the correlation between organochlorine levels in adipose tissue and blood plasma. Three studies (45, 68, 69) only found some of the organochlorine residues to be correlated between adipose and serum samples. However, these studies had very small sample numbers, ranging from 10 to 35 subjects. Overall, significant correlations were found between the two types of samples (44, 68, 70-72), but the correlation coefficients ranged from as low as 0.3 to as high as 0.99. Table 2.2 contains the correlation coefficients for PCBs from results of these studies and Table 2.3 contains correlation coefficients for the pesticides/ pesticide metabolites from these studies. It appears that serum organochlorine concentrations serve as a better proxy measure of lifetime exposure for some residues but not for others, which may be due to transport of some organochlorines on other fractions in the blood, rather than in the lipid portion (73). For example, 5% of PCBs is carried on albumin protein versus half of DDT and dieldrin is carried on albumin in rat serum (67, 69). The partitioning of organochlorines to the lipid, protein, or water fractions in the blood depends on the hydrophobicity of the compound (73), which may vary between compounds. The issue of correlation is therefore complex and an overall relationship between adipose and serum concentrations for all organochlorines cannot be derived. Since organochlorines should theoretically distribute equally in adipose tissue and the lipid portion of serum, one would expect the measured concentrations from the two samples to have a ratio of one. In actuality, the measured concentrations are not on a 1:1 ratio (70, 72) and measured lipid-adjusted serum concentrations are lower than measured adipose tissue concentrations. Despite these issues, measuring organochlorine concentrations in blood is still the most practical biomarker of exposure for epidemiological studies. -23 -2.5 Organochlorines and Risk of Non-Hodgkin Lymphoma Occupational exposure and risk of NHL The hypothesis that exposure to organochlorines is a risk factor for NHL began with studies of agricultural workers. Farmers are generally healthier and experience less mortality than the general population. But early in the 1980's, farmers were observed to be at an excess risk of specific cancers including NHL (74-76). It was suggested that this increased risk could be due to pesticide use. A number of large population based studies followed, to determine the association between pesticides and NHL (77-81). Overall, these studies found a positive association between general pesticide use and risk of NHL. But none of the studies were able to confirm specific agents of risk because subjects typically had exposure to multiple pesticides. Studies looking at specific organochlorine have had inconclusive results. For example, the pesticides lindane (82) and DDT (83) were not found to be associated with NHL. However, in a subsequent study by McDuffie et al (84), both pesticides were found to significantly increase risk of NHL. Similar studies of occupational settings, such as workers in the rubber or paint industry, have consistently found increased mortality from NHL (85-88). The increased risk was attributed to the heavy use of various compounds such as phenoxy acids, chlorophenols, and organic solvents (86, 89, 90). Organic solvents can include some organochlorines such as PCBs or contain organochlorine contaminants such as PCDDs and PCDFs. These occupational studies provided additional support for organochlorines as a possible risk factor for NHL. A more detailed discussion of the studies involving occupational PCB exposure and risk of NHL can be found in a recent review by Engel et al (91). -24-Lifetime background exposure and risk of NHL In the mid 1990's, a Swedish group conducted a study measuring concentrations of organochlorines in adipose tissue samples of 28 NHL patients, compared to non-malignant controls (92, 93). They were using adipose tissue as a biomarker of lifetime exposure to organochlorines. Higher concentrations of PCBs and chlordane-related compounds cis-nonachlor, trans-nonachlor, and oxychlordane were found in cases compared to controls. No differences were seen for PCDDs, PCDFs, DDE, and HCB. Shortly after, a nested case-control study from the US studied serum concentrations of organochlorines and risk of NHL (94). A dose-response relationship between serum PCB concentrations and risk of NHL was found, and the relationship still existed after adjustment for confounders and other organochlorines. However, no association between chlordane and NHL was observed, in contrast to results of the Swedish group. As well, no associations between NHL and other organochlorines including DDT, lindane, P-HCCH, dieldrin and chlordane-related compounds were observed (94, 95). The authors regarded the results as hypothesis-generating and should be replicated, but nevertheless the results suggested that lifetime background exposure to organochlorines increases risk of NHL. Larger studies have since attempted to confirm these initial results. These include an analysis of a recent case-control study by De Roos et al (96), which had similar results to the original US cohort study. De Roos et al also measured serum concentrations of organochlorines, and found significant odds ratios and trends for specific PCB congeners, but no association was seen with any of the organochlorine pesticides. There was also an update of the Swedish study with additional cases and controls (total 82 cases and 83 controls) (97). They confirmed their results for chlordane and PCBs and furthermore found an increased risk of NHL for hexachlorobenzene (HCB) exposure. In contrast, a study by - 2 5 -Quintana et al (98) did not find PCB or HCB exposure to be associated with risk of NHL, but chlordane-related compounds were associated. In addition, this study found DDE and |3-HCCH also increased risk, which was not observed in the other studies. A summary of the results from these studies are presented in Table 2.4. Some of the inconsistency in results may be explained by the incomparability between studies. The US studies by Rothman et al, Cantor et al, and De Roos et al, which all found significant associations between serum PCB levels and risk of NHL, obtained pre-treatment serum samples. The original Swedish study used pre-treatment adipose tissue samples, while the study by Quintana et al obtained adipose tissue samples from post-treatment cadavers of NHL patients. Chemotherapy treatment has been shown to affect organochlorine levels in the body (99) and therefore will have affected results in the Quitana et al study. As well, obtaining samples from cadavers of NHL patients implies selection for cases with poor prognosis, so results from that study cannot be generalized for the general population (98). Although results from the various studies do not agree completely, there is still a strong suggestion of an association between organochlorine exposure and NHL risk. Another important note is that these studies did not have large numbers of samples. The consequence is an increase of the Type II error rate in hypothesis testing and a decrease in power. The studies did not have the power to detect weaker associations, which may still be clinically significant. • 2.6 Gene-Environment Interactions Genetic association to disease Since the start of the human genome project, a wealth of information about genetic markers has become available. Genetic markers are variations at specific locations in the genome, and they are used for gene mapping and detecting association with disease. The -26 -most common type of genetic variation is SNPs (single-nucleotide polymorphisms). SNPs account for more than 90% of all genetic variation and occur every 100 to 300 bases within the genome (100). They are a change in a single nucleotide, so that the vast majority of SNPs have two forms. A change in a single nucleotide is considered a SNP if it has a frequency of over 1%) in the population. The more commonly found nucleotide is the major allele, and the less commonly found nucleotide is the minor allele. SNPs can occur in protein-coding regions of a gene or in non-coding regions. SNPs in protein-coding regions can alter the gene product and therefore are more likely to have a functional effect that may be disease causing. SNPs in non-coding regions may also be disease causing if they lie in regulatory regions that control the activity or splicing of the disease-causing gene. In addition, SNPs may have no effect but can still be associated with the disease because they are in close proximity to the causal genetic variant. In this situation, the SNP is said to be in linkage disequilibrium (LD) with the causal variant. Linkage disequilibrium is when an allele at one locus is found together with a particular allele at another locus more often than would be expected from random segregation (101). LD allows researchers to find genetic association to a disease without knowing the actual genetic cause. Genotyping SNPs across a region of interest in the genome can then help narrow down the location of causal variant. One of the earliest measures of linkage disequilibrium is D, which is defined as D= PAB - PA X P B , Where PAB is the observed frequency of the haplotype consisting of alleles A and B, and PA and P B are the frequencies of the alleles (102). Two common measures of LD used currently in the literature are D' and r2, both of which are based on D. D' is the absolute value of D divided by its maximum value, while r2 is calculated by dividing D2 by the product of the - 2 7 -frequencies of the four haplotype combinations at the two loci. For both measures, a value equal to 1 indicates that the two loci are in perfect LD. The r value of LD has been considered the measure of choice for various reasons (102). One reason is that r values less than 1 are interpretable, whereas for D', values less than 1 are not interpretable and do not correspond to the strength of LD. This is because r2 values takes into account differing allele frequencies at the two loci. Another reason is that D' values are strongly affected by sample size, so that high values can be obtained even when two loci are in linkage equilibrium. Because of LD in the genome, association with genetic variation can be studied in terms of single SNPs or in terms of multiple SNPs in a single haplotype. A haplotype is a combination of specific alleles along a single chromosome and two or more SNPs can make up a haplotype (103). Haplotypes may more powerful for detecting genetic associations to disease because looking at more than one marker spanning a region of the chromosome can provide more information than looking at a single SNP. LD is important in understanding haplotypes. Regions of high LD or 'blocks' of LD give rise to a limited number of distinct haplotypes in the population, called a haplotype block. Regions of low LD with high recombination rates separate the haplotype blocks (103). Within haplotypes and haplotype blocks, genetic markers are highly correlated with each other and provide redundant information. Genotyping all SNPs in a haplotype will not provide any more information than genotyping a minimal number of representative SNPs. These representative SNPs are called tag-SNPs (104). Genotyping of tag-SNPs greatly reduces the number of SNPs required in a study, decreasing cost and increasing efficiency. -28 -Gene-Environment Interactions Cancer is a complex disease that arises from a combination of multiple genetic and environmental factors. These multiple factors can exert their effects independently or may work together to increase disease risk. A gene-environment interaction involves two risk factors, one of which is genetic and the other is environmental. In statistical terms, a gene-environment interaction occurs when the effect due to the environmental exposure differs in persons with different genotypes, or when the effect of a single genotype differs in individuals with different environmental exposures (105). Biologically, gene-environment interactions can be explained by the models put forth by Ottman (105) and illustrated in Figure 2.2. A hypothetical example using a genotype of interest, an environmental risk factor such as smoking, and their risk on lung cancer will illustrate these models. In the first model, a genotype can produce the environmental risk factor that causes disease. However, the environmental risk factor can occur even without the specific genotype. For example, smoking is a known risk factor for lung cancer. A person with a genotype that increases addictive behaviour will cause this person to smoke more frequently, which increases the risk for lung cancer. However, a person without this genotype may still smoke. In model B, the genotype strengthens or weakens the effect of the environmental exposure, while the genotype has no effect for people who are unexposed to the environmental factor. Using the lung cancer example, there might be a genotype in a gene of interest that degrades toxins from cigarette smoke more slowly. Thus for smokers, this genotype will increase risk because the body is being detoxified more slowly. But for non-smokers, the genotype will have no effect since there are no toxins from smoking in the body to degrade. - 2 9 -In model C, the environmental exposure strengthens or weakens the effect of the genotype, while the environmental exposure has no effect when the genotype is not present. Although such a scenario has not been described for the association between smoking and lung cancer, an example of this model would include the autosomal dominant disease called porphyria variegata. The disease is caused by mutations in genes encoding- enzymes for heme production, and manifests as skin problems. Exposure to barbiturates or alcohol can cause much more severe effects in individuals with the disease, including paralysis or death. But exposure to barbiturates or alcohol in the general population will not cause these health effects (105). In model D, the presence of both factors is needed to increase risk of disease, while the presence of only one of the factors does not affect risk. It is difficult to imagine a scenario where smoking by itself does not increase lung cancer risk, so again an example for this model with come from another disease. For the most part, people with glucoses-phosphate dehydrogenase (G6PD) deficiency are asymptomatic. But ingestion of fava beans can cause acute hemolytic anemia in individuals with disease, while such dietary exposure would not cause such a response in individuals without G6PD deficiency (105). In model E, each factor contributes to disease risk independently, while the presence of both factors will further increase or decrease the effect. For example, smoking is known to increase risk for lung cancer. Specific genotypes in DNA repair genes such as XRCC1 have also been documented to increase risk. Hu et al (105, 106) found that the odds ratio for the XRCC1 -77T>C polymorphism alone is 1.55 (95%CI=1.21, 1.98), whereas the odds ratio for those who smoke/ever smoked compared to non-smokers was 4.07 (95%CI=2.85, 5.81). For people who smoke/ever smoked and have the -77T>C polymorphism, the odds ratio increases to 9.82 (95%CI=5.66, 17.02). - 3 0 -It is increasingly clear that common diseases such as cancer are caused not only by specific factors alone but also by a complexity of inter-dependent factors. The study of gene-environment interactions will be integral to future epidemiological research of complex diseases. It will help to obtain better estimates of disease risk based on the joint effects of the genetic and environmental factors, rather than looking at each factor separately. It will also provide more information than the study of single factors in understanding the web of causation for any disease. There has already been a focus on studying gene-environment interactions in colorectal cancer (107) and breast cancer (108). In contrast, there is little research in this area for NHL, and no information exists for genetic interaction with organochlorine exposure and risk of NHL. One candidate gene that may have joint effects with organochlorine exposure on NHL risk is AHR. The aryl hydrocarbon receptor gene (AHR) The AHR gene encodes the aryl hydrocarbon receptor (AhR). The AhR binds to a number of natural and synthetic ligands, of which the most potent ligand known is 2,3,7,8-PCDD (TCDD). Furans (PCDFs) and dioxin-like PCBs are also known to bind to the receptor, albeit with less affinity (109). The receptor normally resides in a dormant state in the cytoplasm of the cell. Upon binding to a ligand, the AhR translocates to the nucleus where it forms a complex with the aryl hydrocarbon receptor nuclear translocator (ARNT) protein. This AhR/ARNT complex then binds to specific DNA sequences called xenobiotic response elements (XREs), which are found only in genes activated via this pathway (29). Genes containing XREs that are induced by AhR include those encoding for the enzymes CYP1A1, 1A2, 1B1, and NADH dehydrogenase (110). As mentioned previously, the CYP1A1 and 1A2 enzymes are involved in the metabolism of PCDDs, PCDFs, and dioxin-like PCBs. The subsequent pathways that elicit the clinical responses observed, after -31 -organochlorine exposure is unknown. Despite this lack of knowledge, it is still accepted that the AhR pathway is the main mechanism of toxicity for dioxins, furans, and dioxin-like PCBs. This understanding is based on genetic studies with AHR-deficient mice (29). The unknown genetic pathways activated by AhR may also trigger other biological responses that are not observed clinically. These responses are equally important to understand as the ones causing clinical toxicity. If organochlorines are a risk factor for NHL, understanding all the downstream responses for AhR will help in determining the mechanism of disease. The AhR pathway has also been implicated in the regulation of the cell cycle and cell proliferation (29, 111). Alterations in the cell cycle or cell proliferation can be a stage of development of normal cells into cancerous cells. For these reasons, the AHR gene is a good candidate for looking at gene-environment interactions with organochlorine exposure and association with risk of NHL. A recent study by De Roos et al (112) studied the relationship between the R554K (X10+1661G/A ) (rs2066853) variant of the AHR gene and the risk of NHL, but found no association. However, this study only looked at one variant within the gene, and it is possible that other variants may be associated. 2.7 Summary Epidemiological studies looking at the relationship between organochlorines and NHL to date have not been large enough to detect weak associations, which may still be clinically important. As well, no study has yet looked at gene-environment interactions with regard to NHL. For a complex disease like NHL, it is necessary to look at joint effects of genetic and environmental risk factors in order to obtain risk estimates and understand the biological mechanisms of disease. One logical genetic involvement would be genes involved in metabolizing organochlorines in the body, such as the AHR gene. - 3 2 -Figure 2.1: Age-standardized incidence rates of NHL in Canada by sex, 1978-2007 (10). o - I — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — < -Year estimated. Figure 2.2: Models of interaction between a genotype and an environmental risk factor the risk of disease, from Ottman, R, 1996 (105). GENOTYPE MODEL A MODEL B ENVIRONMENTAL RISK FACTOR" 4. DISEASE S GENOTYPE ENVIRONMENTAL RISK FACTOR OiSEASE MODEL C MODEL D ENVIRONMENTAL RISK FACTOR GENOTYPE , ENVIRONMENTAL RISK FACTOR* DISEASE GENOTYPE MODEL E ENVIRONMENTAL RISK FACTOR •+ DISEASE -34-Table 2.1: The WHO TEF scheme for dioxin-like PCBs (30-32). Chlorine Substitution Chlorine Positions IUPAC No. TEF 1 (1994) TEF 2 (1997) TEF 3 (2005) Non-ortho substituted: 3,3',4,4'-TetraCB 77 0.0005 0.0001 0.0001 3,4,4',5-TetraCB 81 0.0001 0.0003 3,3',4,4',5-PentaCB 126 0.1 0.1 0.1 3,3',4,4',5,5'-HexaCB 169 0.01 0.01 0.03 Mono-ortho substituted: 2,3,3',4,4'-PentaCB 105 0.0001 0.0001 0.00003 2,3,4,4',5-PentaCB 114 0.0005 0.0005 0.00003 2,3',4,4',5-PentaCB 118 0.0001 0.0001 0.00003 2',3,4,4',5-PentaCB 123 0.0001 0.0001 0.00003 2,3,3\4,4\5-HexaCB 156 0.0005 0.0005 0.00003 2,3,3'AA',5'-HexaCB 157 0.0005 0.0005 0.00003 2,3,,4,4',5,5'-HexaCB 167 0.00001 0.00001 0.00003 2,3,3',4,4,,5,5'-HeptaCB 189 0.0001 0.0001 0.00003 Di-ortho substituted: 2,2',3,3',4,4',5-HeptaCB 170 0.0001 2,2',3,4,4,,5,5'-HeptaCB 180 0.00001 -35 -Table 2.2: Correlation of serum and adipose tissue PCB concentrations from previous studies. PCB Congeners Rusiecki, JA et al (24)a Archibeque-Engle, SL et al (25)* Mussa to-Rauhamaa, H (43)ab Stellman, SD et al (44) n Spearman r n Pearson r n Pearson r n Pearson r 74 83 0.305** 99 76 0.524*** 118 90 0.688*** 138 100 0.508*** 146 69 0.318** 153 26 0.69*** 15 0.377* 99 0.271** 156 60 0.203 167 30 0.052 170 77 0.234* 172 6 -0.454 178 24 0.65*** 35 -0.337* 180 80 0.325** 183 43 0.358* 187 83 0.392*** Total PCBs 10 0.39 Concentrations included in calculation of the correlation coefficient were all above the detection limit. * significance atp<0.05, ** significance atp<0.01, ***siginifcance atpO.OOl a Lipid-adjusted serum concentrations were used in calculating the correlation coefficient b Log-transformed concentrations were used in calculating the correlation coefficient Table 2.3: Correlation of serum and adipose tissue pesticide concentrations from previous studies. Pesticide/ Pesticide Metabolite Rusiecki, JA etal (70)ac Archibeque-Engle, SL et al (45)abc Mussalo-Rauhamaa, H (69)abc Stellman, SD etal (44)c Waliszewski, SM et al (71)a Needham, LL etal(12fhcA Mes, J (68)ad n Spearman r n Pearson r n Pearson r n Pearson r n Pearson r n Pearson r n Pearson r p,p'-DDE 25 092*** 35 0.81* 8 0.74* 101 0.73*** 64 0.92* 293 0.86 25 0.88 o,p'-DDD 67 0 79*** p,p'-DDD 64 0.39* o,p'-DDT 64 0.19 p,p'-DDT 25 094*** 31 0.70*** 64 0.72* 25 0.43 Lindane 10 -0.94** HCB 11 0.33 99 0.84*** 64 0.97* 284 0.53 25 0.788 a-HCCH 64 0.51* P-HCCH 25 090*** 78 0.58*** 64 0.56* 134 0.73 25 0.989 Y-HCCH 64 0.33* a-chlordane 13 0.23 y-chlordane 16 0.032 Oxychlordane 97 0.33** 213 0.70 16 0.656 trans-Nonachlor 99 0.66*** 270 0.68 Dieldrin 68 0.53 Heptachlor 1 Epoxide 233 0.66 * significance at/?<0.05, ** significance atp<0.0\, ***siginifcance at/?<0.001 a Lipid-adjusted serum concentrations were used in calculating the correlation coefficient b Log-transformed concentrations were used in calculating the correlation coefficient c Only concentrations above the detection limit were included. d No p-values were reported. Table 2.4: Studies of the association between biological measures of organochlorine exposure and risk of NHL. Authors Rothman et al. (94), Cantor et al. (95) Hardell et al. (92,93) Hardell et al. (97) Quintana et al. (98) De Roos et al. (96) Year of Publication 1997,2003 1996 2001 2004 2005 Sample Collection Period 1974-1989 (CLUE I/CLUE II studies) 1994-1995 1994-1999 1969-1983 (NHATS study) 1998-2000 Cases/Controls 74/147 28/17 82/83 175/481 100/100 Matching race, sex, date of birth, participation in CLUE study, date of blood-sample donation, participation in census, location of stored blood specimen frequency matched by age, sex frequency matched by age, sex age, sex, region of hospital, race frequency matched by age, date of blood draw, sex, study site Sample type Pre-treatment blood Pre-treatment adipose tissue Adipose tissue or blood post diagnosis/ post-mortem adipose tissue Pre-chemotherapy blood Organochlorines included in analysis {statistically significant Wilcoxon test} [statistically significant OR for highest exposure category, adjusted where available] Summed PCBs sum of 28 congeners {p=0.001} [OR 4.1 (1.4-11.9)] sum of 10 dioxin-like congeners sum of 36 congeners; sum of 9 'immunotoxic' congeners [OR 3.2 (1.4-7.4)] Aroclor 1254 and 1260 sum of 12 noncoplanar; low degree of chlorination; medium degree of chlorination; high degree of chlorination Table 2.4 (continued): Studies of the association between biological measures of organochlorine exposure and risk of NHL. Authors Rothman et al., Cantor et al. Hardell et al. Hardell et al. Quintana et al. De Roos et al. Individual PCBs 34 congeners (PCBs 74,156, 157, 182, 171, 172, 170, 189, 201, 202, 194, 208, 0.015% and/or the SNP constitutes an amino acid residue change predicted to affect the resulting protein structure and function. The SIFT (121, 122) and PolyPhen (123) programs were used to predict changes in protein structure. Additionally, three tag-SNPs were chosen from the public database HapMap (124). A total of 8 SNPs were chosen for genotyping of AHR. Whole Genome Amplification Due to the low yield of DNA collected by the Ficoll technique, some samples in the parent NHL study were subjected to whole genome amplification before the AHR gene was genotyped. Whole genome amplification is a method used to amplify a limited amount of DNA into a larger quantity. Of the 797 cases and 791 controls genotyped for AHR variants in the parent NHL study, genotyping was done using original samples in 468 (58.7%) cases and 535 (67.6%) controls. Genotyping was done only with whole genome amplified samples for 127 (15.9%) cases and 128 (16.2%) controls. Both original and whole genome amplified -49-samples were genotyped in 202 cases (25.3%) and 128 controls (16.2%), and these include the quality control sample pairs. For subjects where original and whole genome amplified samples did not result in the same genotype, the genotype from the original sample was used, although this situation was uncommon. For the organochlorines study, genotyping of the AHR gene was done using original samples in 265 (62.8%) cases and 333 (72.5%) controls; using whole genome amplified samples in 56 (13.3%) cases and 66 (14.4%) controls; and using both original and whole genome amplified samples in 101 (23.9%) cases and 60 (13.1%) controls. The WGA method was done with the REPLI-g Midi Kit (QIAGEN) in batches of 100 reactions, of which 4 were negative controls. Over lOng of genomic DNA template re-suspended in TE buffer was used for each reaction. 5pi of Buffer DI was added to 5pi of DNA and incubated at room temperature for 3 minutes to denature the DNA. lOul of Buffer NI was then added to the mixture to stop denaturation. lul of REPLI-g Midi DNA polymerase in 29pl of REPLI-g Midi reaction buffer was then added to the denatured DNA for a final reaction volume of 50ul. After incubation at 30°C overnight for 16 hours, the REPLI-g Midi DNA polymerase was heat-inactivated at 65°C for 3-10 minutes. The amplified genomic DNA was stored at 4°C for quantitation, and subsequently stored at -20°C. Two sets of quality control (QC) samples were used to assess the integrity of whole genome amplification. 97 randomly chosen samples with a high quantity of DNA were whole genome amplified. The QC pairs consist of the genotypes from the original sample and the whole genome amplified sample. The error rate is defined as the percentage of discordant genotypes for QC pairs where both genotypes were valid (i.e. not undetermined). The error rate of this QC set would reflect the error rate of the whole genome amplification - 5 0 -method, and would be the minimum error rate if all DNA samples were of high quantity. The second set of QC pairs consisted of 99 subjects who provided two DNA samples, where the first sample had a low DNA yield due to DNA extraction from lymphocytes, and a second sample with a high DNA yield from extraction with whole blood. The error rate will be calculated from the number of discordant genotypes between the whole genome amplified low quantity DNA samples and the high DNA quantity samples within the same subject. Again the error rate only includes QC pairs where both genotypes were determined. The error rate of the second QC set will reflect the error rate of whole genome amplification of low quantity DNA, and is expected to be higher than the error rate of the first QC set. Genotyping The naming convention for SNPs used in this thesis will consist of the gene region, followed by the SNP position within the gene, and the nucleotide change from the major allele to the minor allele (major allele/minor allele). The gene region may be an exon (X) or an intervening sequences (IVS). The SNPs chosen for genotyping in the AHR gene, in genomic order, include: X1-459G/A (rs7796976), IVS1+1204C/G (rs713150), IVS1+4640G/A (rsl7722841), and IVS1-3946G/A (rs2282885), X2+132T/C(rsl7779352), IVS7+33G/T (rs2074113), X10+1661G/A (rs2066853), and X10+2356A/G. The IVS1+1240C/G, IVS1+4640G/A, and IVS1-3946G/A SNPs are tag-SNPs. For the X10+2356A/G SNP, only a single heterozygote was observed in the entire dataset, and so it was excluded from analysis due to its rarity. A diagram of the AHR gene and the location of the seven SNPs that were genotyped are given in Figure 3.2. The diagram was created by Rozmin Janoo-Gilani from the BC Cancer Agency Genome Sciences Centre. Genotyping was performed with the TaqMan allelic discrimination assay designed using the Assays-by-DesignSM service (Applied Biosystems, Foster City, CA, USA). The -51 -reaction system was run in MJ Tetrad 384-well PCR machines, containing 10 ng of DNA; 0.125 ul of 2x TaqMan Universal PCR master mix; 2.5 ul of 40x TaqMan probes/primers mix and 2.375 pi of nuclease free water in a total reaction volume of 5 pi. Genotyping was followed by reading of the fluorescent products in the ABI 7900HT instrument. 3.3 Organochlorine Main Effects Summary Descriptive Statistics All statistical analyses were performed using SPSS 13.0 (125) or SAS 9.1 (126). Characteristics of cases and controls were tabulated. Frequencies of cases and controls by categories of the characteristics age, sex, region, education level, ethnicity, body mass index one year before study participation, presence or absence of family history of NHL, income group, per capita income group, farming, and having had a previous transplant are presented. The farming variable is whether or not the subject has ever worked/lived on a farm. Exposure to organochlorines, particularly pesticides, may be higher for those who have worked/lived on a farm. The frequencies for the variables are reported for cases and controls in the parent NHL study, cases and controls in the organochlorine study, and cases and controls that were included in analysis for the organochlorine study. Frequencies of cases by histology were also calculated. DLBC and follicular lymphomas, which are both B-cell derived, were included in separate categories because they constitute a large proportion of NHL cases. The number of other NHL subtypes was low, and they were grouped either into the other B-cell category or the T-cell category. Differences in the distribution of cases and controls within the analyzed organochlorine population by categories of the demographic variables were tested with the chi-square test. Differences in the distribution of the demographic variables in cases that were included in analyses and cases excluded from - 5 2 -analyses were also tested using the chi-square test. Missing values were excluded from chi-square tests. For each PCB congener, pesticide, and pesticide metabolite, the number of samples with detectable levels and the number of samples with non-detectable levels were calculated. For PCB congeners 52, 101, and 128, aldrin, a-chlordane, and y-chlordane, less than 5% of samples had detectable levels. These organochlorines were not included in further analyses. Organochlorine concentrations measured did not follow normal distributions and are skewed to the left. Concentration distributions were described using the median, 25th and 75th percentiles. The median, 25th, and 75th percentiles for each organochlorine compound were calculated by case/control status. Data Coding Lipid-adjusted concentrations for organochlorine residues were categorized according to their distribution in control samples. For organochlorines where there were less than 25% of samples below the detection level, the concentrations were categorized into quaitile variables. Non-detectable samples are thus included in the lowest quartile. Organochlorines with over 25% of samples below the detection level were categorized into two or three exposure groups. PCB 183 had 38.4% of samples non-detectable, and the concentrations were categorized into three categories: The lowest category for non-detects, a second category with detectable values below the median level in controls, and the third category with detectable values above the median level in controls. PCB 28, PCB 105, cis-nonachlor, p,p '-DDT, and mirex had more than half of the samples below the detection limit. Concentrations for these residues were categorized either as above or below the detection limit. The lowest quartile or lowest exposure group was used as the reference category. - 5 3 -Variables for total sum of all PCB congeners, summed dioxin-like PCB congeners, and summed non dioxin-like PCB congeners were created for summary measures of PCB exposure. Dioxin-like and non dioxin-like PCBs possibly act via different mechanisms; so separate variables were created for the two groups. Lipid adjusted concentrations were computed by summing the whole-weight measurement and then dividing by the total lipid value. Non-detectable values for whole-weight measurements were set to the detection limit divided by V2 (127). Total PCBs, total dioxin-like PCBs, and total non dioxin-like PCBs were then categorized into quartiles based on the distribution in controls. The environmental exposure variables used in further analyses were the categorized variables for individual organochlorines, and the summary variables total PCBs, summed dioxin-like PCBs, and summed non dioxin-like PCBs. The outcome variable of disease state is a binary numeric variable with a value of one for cases and zero for controls. The covariates age, sex, region, ethnicity, income, per capita income, education level, family history of NHL, and farming were all categorical variables. Bivariate analyses Spearman rank correlation matrices were calculated first for the PCB congeners and total sum of PCBs, and second for the pesticides/ pesticide metabolites. In the correlation matrix for the pesticides, total sum of PCBs was also included to measure the correlation between the PCBs and the pesticides. Lipid-adjusted concentrations were used for calculating the Spearman rank correlation coefficients. Non-detectable values were set at the detection limit divided by V2 then lipid-adjusted. Wilcoxon rank-sum tests were performed to test case-control differences in lipid concentrations, which were used to calculate lipid-adjusted organochlorine levels. Differences in organochlorine exposure and lipid concentrations between cases included in - 5 4 -the analyses and cases that were excluded due to significant weight loss were also tested with the Wilcoxon rank-sum test. Wilcoxon tests comparing organochlorine exposures between the two groups of cases included the non-detectable organochlorine concentrations, which were set at the detection limit divided by V2 and lipid-adjusted. The relationships between organochlorine exposure and the various covariates age, region, sex, education, ethnicity, BMI, farming, and family history of NHL were also explored. For simplicity, the covariates age (<60 vs 60+), ethnicity (European vs. others), education (less than high school vs. more than high school) and BMI (<25.3 vs 25.3) were categorized into two categories. The median organochlorine concentrations across categories of covariates and the p-values of the resulting Wilcoxon rank-sum test are reported. The Logistic Regression Model Unconditional logistic regression was used to model the association between organochlorine exposure and the likelihood of NHL, with the odds ratio as a measure of the association. The exposure is individual categorical organochlorine variables, and the outcome is case/control status. Due to the high correlation between different organochlorine residues, each organochlorine was analyzed in a separate logistic regression model. Categorical exposure variables were used, so that the regression estimates showed a change in log odds across categories, with the lowest exposure group as the reference. The likelihood ratio test was used to assess the significance of the association fitted by the model. Significance of individual regression estimates was tested by Wald statistics. To perform tests for trend across quartiles, a continuous variable equal to the median value for controls within each exposure category was created for organochlorines with quartile variables and for the 3-category PCB 183 variable. The significance of the trend test was based on the likelihood ratio test for the logistic regression model with the trend -55-variable. For organochlorines with two exposure categories, the significance of trend was based on the Wald statistic of the categorical exposure variable in the logistic regression model. Selection of Confounders for Control Greenland and Rothman (128) suggest the change-in-estimate criterion to select confounders for control in the final logistic regression model. For this study, a 5% change in the odds ratio estimate was set as the cut-off for an important change. The possible confounders included age, sex, region, ethnicity, income, per capita income, education level, family history of NHL, BMI one year before study participation and farming. Two age group variables were tested. One had 12 categories of five-year ranges and the second had four categories of age ranges 20-49, 50-59, 60-69, and 70+, which are approximate quartiles of age. Both age variables gave similar regression estimates when entered in the logistic regression model. The age variable with fewer categories was used for further regression modeling to improve efficiency. Due to the large number of missing data for income level and per capita income, they were not tested by the change in estimate criterion. There were 53 cases and 64 controls with missing income data, and 55 cases and 66 controls with missing per capita income data. Analyses of variables with missing data by creating a special missing category or by exclusion of subjects (complete-subject-analysis) are both biased approaches (129). Multiple imputation and weighted estimates methods are more valuable but are complex and difficult to accomplish for categorical variables. On the other hand, the education variable had few missing data and it also serves as a measure of socioeconomic status (130). Thus, education level was used as a proxy variable for income. Those with missing education information, 9 cases and 5 controls, were excluded in the analysis. - 5 6 -To test family history of NHL, farming, and BMI as confounders, those with missing information were excluded. There were 19 cases and 1 control with missing family history data, 19 cases and 2 controls with missing farming data, and 14 cases and 11 controls with missing BMI data. When testing ethnicity, subjects with missing information were placed into the 'other/mixed' category. The list of possible confounders tested by the change-in-estimate criterion included age, sex, region, education, BMI, ethnicity, farming, and family history of NHL. The change-in-estimate criterion was applied as follows. Firstly, a full logistic regression model with the exposure variable and all 8 covariates was run to obtain an initial regression estimate for comparison. Each confounder variable was then removed one-at-a-time in separate models. The model with the removed covariate that caused the smallest odds ratio estimate change for the exposure variable in comparison to the full model was chosen to replace the full model. The remaining covariates were then removed one-at-a-time in separate models and tested by the same criterion. This process repeated until all the remaining covariates, when removed, changed the odds ratio estimates for the exposure variable by more than 5%. These remaining covariates were included in the final model for the specified exposure variable. A final model for every organochlorine exposure variable was chosen using this criterion. The Final Model The final logistic regression models to study the association between organochlorine exposure and NHL are reported. For models adjusting for BMI, education, farming, or family history of NHL, subjects with missing information were excluded. The false discovery rate (131) was used to address the problem of multiple comparisons in the organochlorine analysis. The FDR controls the expected proportion of false positives rather - 5 7 -than controlling the chance of any false positives, as the Bonferroni correction does. The FDR threshold is determined from the p-value distribution in the data. The FDR was applied to all organochlorine compounds included in analysis, including the summary PCB variables. The p-values distribution was based on the p-values of the trend test for each organochlorine. Briefly, the FDR method was applied as follows. Firstly, the p-values were ranked in increasing order. The FDR threshold was calculated as the p-value multiplied by its rank, and divided by a=0.05. P-values that are below the FDR threshold imply statistical significance under FDR control. The FDR is known to be very conservative when there is a high correlation between exposure variables. Interaction between the confounder variables and organochlorine exposure variables were assessed using the trend variables. Interaction was tested for the potential confounders age (<60 vs 60+), sex, region, ethnicity (European vs. others), education and BMI (<25.3 vs 25.3). Interaction terms between the confounder variable and the organochlorine trend variable were entered into the logistic regression model and significance was based on the likelihood ratio test for the interaction term. Interaction between each confounder and organochlorine was tested in separate models. For significant interactions, logistic regression models were fitted with the categorical organochlorine variables within categories of the confounder in order to obtain odds ratio estimates. Additional Analyses The forward stepwise selection method in SPSS 13.0 was used to determine if combinations of organochlorines in the logistic regression model give better risk estimates than individual organochlorines. The selection method uses the significance of the score statistic for entry of variables, and removal of variables based on the likelihood-ratio statistic. The stepwise selection method was performed for all organochlorines analyzed in - 5 8 -this study, then separately for the PCB congeners, and pesticides. Trend variables of organochlorine exposure were entered into the models for stepwise selection. Estimates across categories of organochlorine exposure were subsequently obtained once the final models were chosen. Any covariates included in individual regression models were included in the model for stepwise selection. For stepwise selection with all organochlorines and with PCBs, the covariates age, sex, region, education, BMI, ethnicity, farming, and family history of NHL were included in the model. For stepwise selection with pesticides, the covariates age, sex, BMI, ethnicity, farming, and family history of NHL were included. Association between organochlorines and NHL using logistic regression were also performed within histological subtypes. Follicular lymphoma and diffuse large B-cell lymphoma comprise the majority of NHL cases in the study and were analyzed as individual groups. The numbers of cases for other NHL subtypes were very low and not analyzed. Instead, the other subtypes were grouped as either T-cell or B-cell and these two groups were analyzed. Two groups of B-cell lymphomas were analyzed: The first comprising of all B-cell cases including the follicular and DLBC subtypes ("all B-cell"), and second a group comprising of B-cell cases excluding follicular and DLBC subtypes ("other B-cell"). For simplicity of comparison, only odds ratios comparing the highest versus lowest exposure categories will be reported in the results. Conditional logistic regression was performed in SAS 9.1 (126) to determine whether there would be differences in the regression estimates as compared to the estimates from unconditional logistic regression. Estimates were found not to change, and only results for unconditional logistic regression will be presented. Twenty-four cases were excluded from analyses due to B-symptom weight loss and another eight cases were excluded due to unavailable B-symptom weight loss information. A - 5 9 -sensitivity analysis was performed with the entire data set including these thirty-two cases. Generally, odds ratios increased by roughly 10% when these cases were included. Inclusion of the cases with high concentrations of organochlorines due to B-symptom weight loss results in bias away from the null hypothesis. The more conservative results from analyses excluding these cases will be presented. 3.4 Genetic Main Effects Linkage disequilibrium between the AHR SNPs genotyped in this study was measured using r2 values, based on genotyping data. The r2 values were calculated and plotted by Payal Sipahimalani from the BC Cancer Agency Genome Sciences Centre using the Haploview software (132). Testing for Hardy-Weinberg Equilibrium The Hardy-Weinberg equilibrium (HWE) principle states that under certain assumptions, allele frequencies in a randomly mating population at a given gene locus will remain fixed. These assumptions include random mating, an infinite population size, and no selection, mutation, or migration. For example, at a locus with a dominant allele A and a recessive allele a, and allele frequencies of p and q, respectively, the population frequencies are p + q=l. The genotypic frequencies of' AA' homozygotes would be p2, ' Aa' heterozygotes would be 2pq, and 'aa' homozygotes would be q . Under HWE, p2 + 2pq + q2 = 1. Deviations of alleles of each SNP from HWE were tested using the chi-square test. Testing for HWE was only done on controls, which approximate a sample from the general population. The assumptions underlying HWE testing may not necessarily hold for cases. -60-Controls genotyped in the parent NHL study were used to test HWE. Tests of HWE were only performed for Europeans, as the number of subjects within other ethnicities was small. SNP alleles that are not in HWE may indicate a population substructure, biased sampling of individuals or genotyping error. Genotype frequencies of all SNPs and tests of HWE are reported. Genotype frequencies were calculated for both the population included in the parent NHL study and the population included in the organochlorine study. The percentage of genotypes in each variant that were obtained from whole genome amplified samples was reported. Error rates for WGA using the two sets of quality control sample pairs were also calculated for each SNP. SNP Analysis Genotype frequencies in the parent NHL population by age, sex, region, education and ethnicity, and the chi-square test were used to assess differences between categories. For simplicity, two category variables for age (<60 vs. >60) and ethnicity (European vs. other) were used. These covariates were the possible confounders tested in the SNP analyses. Education was included as a possible confounder because there was a significant difference between cases and controls for education level in the parent NHL study (p=0.011). To test the association between individual SNPs and the risk of NHL, genotype data from the parent NHL population was analyzed. Unconditional logistic regression was used to model the association between individual SNPs and the likelihood of NHL, with the odds ratio as a measure of the association. Genetic exposure variables were categorical numeric variables representing each SNP. Levels of the variables take the values of 0, 1, or 2, representing homozygotes for the major allele, heterozygotes, and homozygotes for the minor allele, respectively. The outcome was case/control status. By fitting the models, regression coefficients for levels of each SNP variable were obtained. The regression -61 -estimates show a change in log odds across categories, with the major allele homozygote group as the reference. Logistic regression using co-dominant, additive, dominant and recessive models of phenotypic expression were compared. In the co-dominant model, having one copy of the minor allele confers a different risk than having two copies of the minor allele. In an additive model, the effect of the minor allele on the risk of disease is changed by a factor equal to the number of copies of the minor allele. In other words, the mean level of the heterozygotes is an average of the means of the homozygotes. In a dominant model, any of one or two copies of the minor allele confer risk of NHL, but having two copies does not confer any greater risk, so minor allele homozygotes and heterozygotes are grouped into one exposure category. In a recessive model, no difference in risk is observed with one copy of the minor allele, but a difference in risk is only observed for two copies, so major allele homozygotes and heterozygotes are grouped into one exposure category. Additive, dominant, and recessive models were compared to the co-dominant model using the likelihood ratio test statistic. The change-in-estimate criterion was used to select confounders for control in the logistic regression model of each individual SNP. The same process as described earlier for organochlorine main effects was employed. The confounders tested by the criterion were age, sex, region, education and ethnicity. The co-dominant model of phenotypic expression was used in the models, except for the AHR +67T/C SNP, where the dominant model of expression was used because there were very few subjects with the homozygote minor allele genotype. Confounders included in final logistic regression models and model estimates are reported separately. The FDR was used to address the problem of multiple comparisons in the SNP analyses. The FDR was applied to the multiple SNPs within the AHR gene included in this -62-study. The p-value distribution was based on the trend p-values of the additive model of phenotypic expression for each SNP. Additional Analyses Interaction between the confounder variables and AHR SNPs were assessed using the SNP trend variables. Interaction was tested for the potential confounders age (<60 vs 60+), sex, region, education and ethnicity (European vs. others). Interaction terms were entered into the logistic regression model and significance was based on the likelihood ratio test for the interaction term. For significant interactions, logistic regression models were fitted with the SNP variables within categories of the confounder in order to obtain odds ratio estimates. The additive models of phenotypic expression were used, as the additive models were found not to differ significantly from the co-dominant models in the main analysis. Logistic regression looking at the association between individual AHR SNPs and risk of NHL was done within European, Asian, and Indo-Pakistani ethnic groups to determine if estimates differ with ethnicity. Regression models were also fitted for NHL subtypes, but for all ethnicities. The NHL subtypes are the same as those analyzed previously for the organochlorine main effects: All B-cell lymphomas, diffuse large B-cell lymphoma, follicular lymphoma, other B-cell lymphomas, and T-cell lymphomas. For fitting these regression models, the additive models of phenotypic expression were used. Haplotype Analysis The Hapassoc package (133, 134) for the R programming environment (135) was used to estimate haplotype frequencies and perform haplotype analysis. Data from the parent NHL population was included in this analysis. Hapassoc uses a likelihood approach to infer haplotype effects, non-genetic effects and their interactions in generalized linear models of disease risk, where some genotypes can be unknown. Parameter estimates are obtained by -63 -use of the expectation-maximization (EM) algorithm and standard errors are calculated using Louis' formula. Assumptions include independence of haplotype and non-genetic covariates and HWE in the population. The input variables to estimate haplotype frequencies were diallelic SNP variables for the seven AHR SNPs and haplotypes with frequencies less than 0.05 were grouped into a pooled estimate for further analyses. A generalized linear model was then used to measure the association between the inferred haplotypes and risk of NHL. In the model, case/control status was the outcome and individual and pooled haplotypes were the exposure variables. The most common haplotype was the reference. Haplotype variables were categorical numeric variables with values of 0, 1, or 2 depending on the number of copies of the haplotype each subject carries. Any confounder variables that were included in regression models of individual SNPs for each gene were included in the generalized linear model. The likelihood ratio test was used to assess the significance of the association fitted by the model with haplotypes with frequencies over 0.05 and the pooled haplotype. A p-value was calculated for a chi-square distribution with six degrees of freedom. Significance of individual regression estimates was tested by Wald statistics. Haplotype analysis was conducted with all ethnic groups included and for Europeans only. Subgroup analyses for Asians and Indo-Pakistanis were not done because of the low number of subjects. 3.5 Gene-Environment Interactions Gene-Environment Interactions for Individual SNPs and Organochlorines Testing for gene-environment interactions were limited to organochlorines and SNPs with significant main effects before FDR control, in order to decrease the number of tests performed. Summary PCB variables, individual PCBs, and pesticides/pesticide metabolites -64-with significant main effects were tested in combination with SNPs with significant main effects separately. Tests of interaction were performed using the trend variables for the organochlorines and SNPs. A logistic regression model was fitted with main effects of both exposures. Covariates included in this model were any confounders included in the regression model for the organochlorine or the SNP. Interaction terms reflecting the product of the SNP and organochlorine trend variables were entered into the logistic regression model, and significance of interaction was assessed by the likelihood-ratio test of the models with and without the interaction term. The FDR was applied to control for multiple tests of interaction. The p-value distribution used for the FDR method was based on the p-values of each likelihood ratio test used to assess interaction between an organochlorine and a SNP. For organochlorines and SNPs with significant interactions, odds ratio estimates for categories of organochlorine exposure were obtained within each SNP genotype. Analyses of gene-environment interaction were also done in the European ethnic group to determine if ethnicity affects estimates. Gene-Environment Interactions for Haplotypes and Organochlorines Gene-environment interactions were tested for organochlorines with significant main effects and AHR haplotypes. Gene-environment interactions were tested separately for each organochlorine. First, a generalized linear model was fitted with the main effects of AHR haplotypes and the organochlorine. Individual and pooled haplotypes were entered in the model, with the most common haplotype as the baseline. The trend variable for organochlorine exposure was used. Any covariates included in previous models were also included. A second generalized linear model was fitted with additional interaction terms between all AHR haplotypes and the organochlorine trend variable. The log likelihoods of - 6 5 -the two models were compared as a global test of interaction, and p-values were calculated using a chi-square distribution with six degrees of freedom. The FDR method was applied to the global tests of interaction to deal with multiple comparisons. Statistical significance of individual interaction terms was assessed using Wald statistics. -66-Figure 3.1: Subject recruitment for parent NHL study and included in current organochlorine study. Total no. eligible: 3636 Eligible cases= 1263 Contacted cases = 1068 (84.6%) Deceased before contact = 133 (10.5%) No contact = 62 (4.9%) Consenting cases = 840 (81.7%) Refusals = 147 (13.7%) Poor Health = 73 (6.8%) Language = 8 (0.7%) I Cases in parent study = 828 (98.6%) Cases not classified/ Cases with prior transplantation = 12 (1.4%) Cases for genetic analysis AHR= 797 (100%) Whole genome amplification n= 329 (41.2%) Blood samples = 769 (92.9%) Mouthwash sample only = 28 (3.4%) No sample = 31 (3.7%) Pre-chemotherapy blood samples = 455 (59.2%) Cases for organochlorine analysis = 422 (92.7%) B-symptom weight loss = 24 (5.3%) No information on B-symptom weight loss = 8 (1.8%) No lipid analysis = 1 (0.2%) Cases for gene-environment interaction analysis = 422 (100%) Eligible controls = 2373 Contacted controls = 1856 (78.2%) Deceased = 13 (0.5%) No contact = 504 (21.2%) Consenting controls = 848 (46.9%) Refusals = 856 (46.1%) Poor Health = 103 (5.5%) Language = 49 (2.6%) Controls in parent study = 848 Blood samples = 679 (80.2%) Mouthwash sample only = 113 (13.3%) No sample = 55 (6.5%) Blood samples = 463 (frequency matched by age, sex, region) Controls for genetic analysis AHR= 791 (99.7%) Insufficient DNA material = 1 (0.1%) Administrative error = 1 (0.1%) Whole genome amplification= 256 (32.4%) Controls for organochlorine analysis = 460 (99.4%) No lipid analysis = 3 (0.6%) Controls for gene-environment interaction analysis= 459 (99.8%) Insufficient DNA material 1 (0.2%) - 6 7 -Table 3.1: Minimum detectable odds ratios for organochlorine exposure and gene-environment interactions on risk of NHL based on sample size of 450 cases and controls. Sample Size Main Effect of organochlorine* Gene-environ ment Interaction** Above vs. Below Median Exposure Upper vs. Lower Quartile of Exposure Above vs. Below Median Exposure Upper vs. Lower Quartile of Exposure 450:450 1.50 1.86 2.64 4.50 alpha=0.05, power=90% alpha=0.05, power=90%, SNP frequency=0.2, OR(high exposure in absence of SNP variant)=1.5, OR(SNP in absence of exposure)=1.0. Table 3.2: Minimum detectable odds ratios for genetic exposure and risk of NHL based on sample size of 800 cases and 800 controls (alpha=0.05, power=90%). Allele Frequency in Controls Minimum detectable odds ratio (heterozygotes and minor allele homozygotes vs. major allele homozygotes) 0.005 3.33 0.010 2.53 0.020 2.03 0.050 1.65 0.100 1.48 0.200 1.40 0.300 1.39 -68--459G>A ||1 IVS1+1204OG I V S 1 + 4 6 4 0 G > A , IVS1-3946G>A • Ex2+132T>C UZ IVS7+33G>T ExlO+1661G>A Chapter 4: Results 4.1 Overview The results show a statistically significant increased risk of NHL from increasing exposure to organochlorines. Total sum of PCBs, total dioxin-like PCBs, total non dioxin-like PCBs, PCBs 118, 153, 156, 170, 180, and 187 were all associated with NHL risk, even after controlling for multiple comparisons. Pesticides associated with NHL risk included p,p'-DDE, HCB, mirex, oxychlordane and trans-nonachlor, after controlling for multiple comparisons. The change in risk ranged from an increased risk of 39% for those in the highest exposure group for mirex up to a 2.7 fold risk for those in the highest exposure group for oxychlordane. Age was a confounder of the association between organochlorine exposure and NHL risk in many cases. Associations were fairly consistent across NHL subtypes. Within the other B-cell subgroup, several non dioxin-like PCBs showed a stronger association, with over 10% increase in odds ratios. Several OCs also showed an increase in odds ratios within the follicular subtype, including HCB (OR for the highest exposure quartile 2.4, 95% CI 1.2-4.6) and oxychlordane (OR 3.3, 95% CI 1.7-6.4). Of the seven SNPs of the AHR gene included in this study, the tag-SNP IVS1+4640G/A was found to be significantly associated with risk of NHL, although the association is not significant after application of the false discovery rate for multiple comparisons. Any increase in the number of copies of the A allele confers an odds ratio of 1.24 (95% CI 1.02-1.50, p=0.029) on NHL risk. Estimates were similar within Europeans, and within subtypes of NHL. None of the 7 SNP haplotypes provided better estimates than looking at individual SNPs. There were significant interactions between the 1YS1+4640G/A SNP and PCB congener 118, oxychlordane, and trans-nonachlor. There was an increased risk of NHL for increasing -70-exposure to these organochlorines in the major allele homozygote group, but there was no increased risk for those with at least one copy of the minor allele. 4.2 Organochlorine Main Effects Population Characteristics Table 4.1 compares frequencies of cases and controls by selected demographic variables and other characteristics considered potential confounders in the current analysis, the complete organochlorine study population and the parent NHL study population. There is a smaller percentage of DLBC lymphoma cases within the organochlorine study in comparison to the parent NHL study. DLBC lymphoma is an aggressive subtype that is often treated with chemotherapy relatively quickly after diagnosis compared to indolent lymphomas such as follicular lymphoma, which are commonly low grade. This difference in treatment resulted in an under-representation of DLBC lymphoma cases in the organochlorine study, due to difficulties in obtaining pre-treatment blood samples from these cases. The study had slightly more males (55% for cases and 56.1% for controls) than females. The large majority of subjects were from the GRVD, which is a reflection of the larger population size compared to Greater Victoria. The large majority of subjects were also categorized as European (80.3% of cases and 80.2% of controls), while the remainder were composed of Asians, Indo-Pakistanis, and unidentified ethnicities. Cases and controls did not differ demographically. There were slightly more cases with family history of NHL (21 cases vs. 14 controls, p=0.11), and more cases that worked or lived on a farm (17 cases vs. 11 controls, p=0.13), but differences were not statistically significant. Family history of NHL and farming confounded the association between organochlorine exposure and risk of NHL for some of the organochlorine compounds, based on confounder selection methods -71 -discussed in Chapter 3. These variables were included in the logistic regression models for those compounds. Based on the chi-square test, there were also no significant differences between frequencies of cases included in and cases excluded from analyses with respect to the various demographic variables age, region, sex, BMI, education, income, per capita income, ethnicity, family history of NHL, and having lived/worked on a farm (p-values 0.19730 73 (17.3) 65 (14.1) 80 (17.6) 65 (14.0) 145 (17.5) 123 (14.5) Unknown 14(3.3) 11 (2.4) 15 (3.3) 11 (2.4) 23 (2.8) 20 (2.4) Family history of NHL (p=0.11) Yes 21 (5.0) 14 (3.0) 21 (4.6) 14 (3.0) 34 (4.1) 21 (2.5) No 382 (90.5) 445 (96.7) 414(91.0) 448 (97.0) 763 (92.1) 814(96.0) Missing 19 (4.5) 1 (0.2) 20 (4.4) 1 (0.2) 31 (3.7) 13(1.5) Income (p=0.68) <$25 000 88 (20.9) 86(18.7) 95 (20.9) 86(18.6) 174 (21.0) 164(19.3) $25 000 - $50 000 103 (24.4) 130 (28.3) 117(25.7) 130(28.1) 218(26.3) 226 (26.7) $50 000 - $75 000 87 (20.6) 86(18.7) 91 (20.0) 87(18.8) 157(19.0) 174 (20.5) $75 000-$100 000 46 (10.9) 49 (10.7) 48 (10.5) 50(10.8) 90 (10.9) 93 (11.0) >$100 000 45 (10.7) 45 (9.8) 47(10.3) 45 (9.7) 87(10.5) 79 (9.3) Unknown 53 (12.6) 64(13.9) 57(12.5) 65 (14.0) 102(12.3) 112(13.2) Per capita income (p=0.90) <$15 000 107 (25.4) 106 (23.0) 115(25.3) 106 (22.9) 224 (27.1) 221 (26.1) $15 000-$30 000 110(26.1) 126 (27.4) 120 (26.4) 127 (27.4) 204 (24.6) 228 (26.9) $30 000 - $45 000 110(26.1) 118(25.7) 118(25.9) 118(25.5) 215 (25.9) 208 (24.5) >$45 000 40 (9.5) 44 (9.6) 42 (9.2) 45 (9.7) 76 (9.2) 76 (9.0) Unknown 55 (13.0) 66 (14.3) 60(13.2) 67(14.5) 109(13.2) 115 (13.6) Ever worked or lived on a farm (p=0.13) Yes 17 (4.0) 11 (2.4) 18(4.0) 11 (2.4) 27 (3.3) 16(1.9) No 386 (91.5) 447 (97.2) 416(91.4) 450 (97.2) 772 (93.2) 828 (97.6) Missing 19 (4.5) 2 (0.4) 21 (4.6) 2 (0.4) 29 (3.5) 4 (0.5) Previous transplant Yes 0 2 (0.4) 0 2 (0.4) 0 3 (0.4) No 422 458 (99.6) 455 461 (99.6) 829 845 (99.6) - 8 6 -Table 4.2: Disposition of laboratory measurements of organochlorine compounds in blood plasma. Study sample (n=882) Quality control samples (n=69 pairs) Measured Number of Median above Below QC pairs Average Intraclass ICC ICC detection limit detection detection with values intrabatch correlation lower upper PCBs (ng/g-lipid) limit (n[%]) limit (n[%]) >DL variation (CV) (ICC) bound bound 28 2.22 158(17.9) 724 (82.1) 17 0.053 0.977 0.939 0.992 52 3.70 11(1.2) 871 (98.8) 1 0.149 99 2.22 721 (81.7) 161 (18.3) 53 0.083 0.977 0.961 0.987 101 2.22 30 (3.4) 852 (96.6) 3 0.040 0.932 -0.155 0.998 105 2.22 275 (31.2) 607 (68.8) 22 0.053 0.841 0.656 0.931 118 2.22 825 (93.5) 57 (6.5) 64 0.034 0.997 0.996 0.998 128 2.22 2 (0.2) 880 (99.8) 0 138 2.22 874 (99.1) 8 (0.9) 69 0.025 0.999 0.998 0.999 153 2.22 880 (99.8) 2 (0.2) 69 0.017 0.999 0.999 1.000 156 2.22 787 (89.2) 95 (10.8) 64 0.035 0.992 0.987 0.995 170 2.22 846 (95.9) 36(4.1) 68 0.024 1.000 0.999 1.000 180 2.22 878 (99.5) 4(0.5) 68 0.019 1.000 1.000 1.000 183 2.22 543 (61.6) 339 (38.4) 41 0.023 0.999 0.997 0.999 187 2.22 841 (95.4) 41 (4.6) 67 0.021 1.000 0.999 1.000 Table 4.2 (continued): Disposition of laboratory measurements of organochlorine compounds in blood plasma. Study sample (n=882) Quality control samples (n=69 pairs) Measured Number of Pesticides/ Median above Below QC pairs Average Intraclass ICC ICC Pesticide detection limit detection detection with values intrabatch correlation lower upper Metabolites (ug/kg lipid) limit (n[%]) limit (n[%]) >DL variation (CV) (ICC) bound bound Aldrin 2.22 0 882 0 p-HCCH 3.70 779 (88.3) 103 (11.7) 58 0.048 0.998 0.997 0.999 a-Chlordane 2.22 0 882 0 y-Chlordane 2.22 1 (0.1) 881 (99.9) 0 cis-Nonachlor 2.22 250 (28.3) 632 (71.7) 16 0.044 0.995 0.986 0.998 p, p'-DDE 3.70 882 0 69 0.031 0.999 0.999 1.000 p,p'-DDT 3.70 288 (32.7) 594 (67.3) 22 0.075 1.000 0.999 1.000 HCB 2.22 881 (99.9) 1(0.1) . 69 0.042 0.999 0.998 0.999 Mirex 2.22 328 (37.2) 554 (62.8) 19 0.082 0.976 0.938 0.991 Oxychlordane 2.22 848 (96.1) 34 (3.9) 68 0.051 0.993 0.989 0.996 trans-Nonachlor 2.22 863 (97.8) 19 (2.2) 68 0.026 0.998 0.997 0.999 Lipids total C g/L 882 0 69 0 1 free C g/L 882 0 69 0 1 trigl g/L 882 0 69 0 1 phospho g/L 882 0 69 0 1 total lipids g/L 882 0 69 0 1 Table 4.3: Descriptive summary of organochlorine concentrations (pg/kg lipid) by case/control status. Cases (n=422) Controls (n=460) N 25th Per-centile Med ian 75th Per-centile N 25th Per-centile Med ian 75th Per-centile Organochlorines PCBs: 28 74 2.67 3.50 5.36 84 2.32 3.12 4.20 99 350 4.09 6.42 9.90 371 4.10 5.73 8.80 105 133 2.71 3.66 5.21 142 2.68 3.34 4.90 118 398 5.46 9.26 15.90 427 5.06 8.16 13.39 138 418 12.29 21.55 33.66 456 11.70 19.45 29.79 153 421 27.76 45.39 71.22 459 25.29 38.77 59.07 156 381 4.55 6.66 9.83 406 4.23 6.15 8.66 170 409 8.08 12.53 19.61 437 7.76 11.71 17.67 180 422 24.67 39.57 62.28 456 22.25 35.76 54.84 183 260 3.23 4.42 6.72 283 2.96 3.95 5.68 187 404 7.30 11.25 18.84 437 6.56 10.27 16.07 Total PCBs 422 110.62 172.20 268.42 460 100.88 155.59 220.01 Dioxin-like PCBs 422 11.71 18.10 27.70 460 10.12 15.35 23.72 Non dioxin-like PCBs 422 97.78 150.53 237.97 460 88.57 136.19 196.38 Pesticides: P-HCCH 375 12.17 18.32 32.66 404 10.76 16.09 28.64 cis-Nonachlor 126 2.71 3.30 4.68 124 2.33 3.07 4.19 p, p'-DDE 422 133.64 294.11 607.63 460 136.03 263.91 512.02 p,p'-DDT 133 7.02 10.10 18.52 155 7.00 10.17 23.46 HCB 422 12.57 18.21 27.99 459 11.47 16.12 22.79 Mirex 174 3.02 4.03 5.73 154 2.56 3.30 5.16 Oxychlordane 410 7.44 11.40 16.48 438 6.71 10.01 13.99 trans-Nonachlor 415 9.88 15.78 24.46 448 9.31 14.34 21.08 -89-Table 4.4: Spearman rank correlation for PCB congeners and total sum of PCBs. Total P C B s P C B 28 P C B 99 P C B 105 P C B 118 P C B 138 P C B 153 P C B 156 P C B 170 P C B 180 P C B 183 P C B 187 Total P C B s 1 .126** .700** .403** .697** .884** .964** .899** .939** .924** .854** .940** P C B 28 1 .204** .532** .242** .125** .087** 0.058 0.028 0.036 .164** .079* P C B 99 1 .548** .767** .845** .731** .582** .490** .460** .723** .591** P C B 105 1 .635** .463** .377** .323** .226** .211** .428** .302** P C B 118 1 .762** .704** .604** .509** .483** .611** .578** P C B 138 1 .924** .805** .738** .692** .866** .785** P C B 153 1 .867** .879** .857** .871** .891** P C B 156 1 .888** .848** .703** .815** P C B 170 1 .982** .751** .928** P C B 180 1 .727** .929** P C B 183 1 .825** P C B 187 1 * Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed). Table 4.5: Spearman rank correlation for pesticides/ pesticide metabolites and total sum of PCBs. Organochlorine Total PCBs p-HCCH p,p'-DDE p,p'-DDT cis-Nonachlor HCB Mirex Oxy-chlordane trans-Nonachlor Total PCBs 1 .328** .438** .102** .362** .594** .263** .649** .689** p-HCCH 1 .598** .422** .348** .555** 0.038 .392** .430** p, p'-DDE 1 471 * * .278** .482** 0.038 .440** .453** p,p'-DDT 1 .462** .171** .173** .109** .192** cis-Nonachlor 1 .318** .402** .409** .585** HCB 1 0.065 .664** .584** Mirex 1 \ 7 4 * * .255** Oxychlordane 1 .870** trans-Nonachlor 1 ** Correlation is significant at the 0.01 level (2-tailed). Table 4.6: Descriptive summaries of lipid concentrations for cases/ controls and p-value from the Wilcoxon rank-sum test. Cases (n=422) Controls (n=460) 25th Per-centile Median 75th Per-centile 25th Per-centile Median 75th Per-centile p-value Lipids (g/L): Total cholesterol 1.66 1.91 2.21 1.79 2.04 2.30 <0.01 Free cholesterol 0.38 0.46 0.54 0.43 0.49 0.57 <0.01 Triglycerides 0.92 1.36 1.84 1.01 1.37 2.05 0.06 Phospholipids 1.88 2.12 2.40 2.04 2.25 2.44 <0.01 Total lipids 5.69 6.55 7.57 6.14 6.89 7.88 <0.01 -92-Table 4.7: Organochlorine and lipid concentrations for cases included and excluded from the analyses, and Wilcoxon rank-sum test for differences in concentrations. Cases analyzed (n=422) Cases excluded (n=32) N detect-able 25th Per-centile Median 75th Per-centile N detect-able 25th Per-centile Median 75th Per-centile p-value Organochlorines PCBs (ng/kg lipid): 28 74 2.67 3.50 5.36 5 2.84 4.17 7.79 0.467 99 350 4.09 6.42 9.90 30 6.24 9.71 14.32 0.001 105 133 2.71 3.66 5.21 16 3.19 4.43 7.48 0.049 118 398 5.46 9.26 15.90 32 7.29 13.05 25.32 0.002 138 418 12.29 21.55 33.66 32 20.72 32.75 51.42 <0.001 153 421 27.76 45.39 71.22 32 45.35 61.21 107.65 O.001 156 381 4.55 6.66 9.83 31 5.68 8.88 14.99 0.002 170 409 8.08 12.53 19.61 31 11.70 19.66 32.33 0.001 180 422 24.67 39.57 62.28 32 34.72 69.16 113.63 0.002 183 260 3.23 4.42 6.72 30 3.66 5.43 9.88 O.001 187 404 7.30 11.25 18.84 31 10.05 22.04 42.30 0.001 Total PCBs 422 110.62 172.28 268.42 32 160.35 246.90 425.37 0.001 Total dioxin-like PCBs 422 11.71 18.10 27.70 32 14.95 23.51 47.40 0.002 Total non dioxin-like PCBs 422 97.70 150.50 207.97 32 143.84 225.01 379.53 0.001 Pesticides (ug/kg lipid): p-HCCH 375 12.17 18.32 32.66 32 16.55 25.70 45.61 0.001 cis-Nonachlor 126 2.71 3.30 4.68 21 2.58 3.40 4.88 O.001 p, p'-DDE 422 133.64 294.11 607.63 32 271.80 518.57 1. 08 x 103 0.002 p,p'-DDT 133 7.02 10.10 18.52 14 8.50 11.15 24.73 0.098 HCB 422 12.57 18.21 27.99 32 19.06 24.96 35.12 <0.001 Mirex 174 3.02 4.03 5.73 20 3.20 4.85 8.63 0.003 Oxychlordane 410 7.44 11.40 16.48 31 13.35 19.24 26.29 O.001 trans-Nonachlor 415 9.88 15.78 24.46 32 20.04 27.67 36.56 O.001 Lipids (g/L): Total cholesterol 422 1.66 1.91 2.21 32 1.47 1.82 2.06 0.08 Free cholesterol 422 0.38 0.46 0.54 32 0.35 0.47 0.56 0.97 Triglycerides 422 0.92 1.36 1.84 32 0.91 1.30 2.17 0.91 Phospholipids 422 1.88 2.12 2.40 32 1.64 1.94 2.23 0.03 Total lipids 422 5.69 6.55 7.57 32 4.88 6.18 7.91 0.24 -93-Table 4.8: Median organochlorine concentrations by categories of potential confounders and p-values for the Wilcoxon rank-sum tests. Confounder Age Sex Category <60 60+ Male Female N total 393 489 490 392 N detect-able Median (ug/kg) N detect-able Median (rlg/kg) p-value N detect-able Median N detect-able Median p-value Total PCBs 393 110.6 489 196.2 O.001 490 155.7 392 167.8 0.261 Dioxin-like PCBs 393 11.7 489 20.9 <0.001 490 14.7 392 19.3 <0.001 Non dioxin-like PCBs 393 99.6 489 170.8 <0.001 490 140.8 392 143.8 0.609 PCB 28 47 3.2 111 3.3 0.228 64 3.2 94 3.3 0.327 PCB 99 291 4.8 430 7.2 O.001 387 5.8 334 6.4 0.011 PCB 105 73 3.3 202 3.6 O.001 117 3.6 158 3.4 0.005 PCB 118 343 6.5 482 10.9 <0.001 443 7.7 382 10.4 O.001 PCB 138 390 14.2 484 25.4 <0.001 486 19.5 388 21.6 0.041 PCB 153 392 29.1 488 51.8 <0.001 489 40.7 391 42.8 0.498 PCB 156 310 4.8 477 7.1 O.001 431 6.4 356 6.5 0.571 PCB 170 362 8.8 484 14.6 <0.001 466 12.4 380 11.8 0.477 PCB 180 389 25.9 489 47.2 <0.001 489 37.8 389 36.3 0.395 PCB 183 177 3.4 366 4.5 <0.001 292 4.2 251 4.1 0.543 PCB 187 358 7.5 483 13.2 <0.001 466 10.7 375 10.8 0.547 P-HCCH 321 15.9 458 17.5 <0.001 417 15.8 362 18.5 O.001 p, p'-DDE 393 201.9 489 363.3 O.001 490 242.9 392 336.5 O.001 p,p'-DDT 121 11.1 167 10.0 0.839 158 10.0 130 10.4 0.082 cis-Nonachlor 71 3.3 179 3.2 <0.001 147 3.2 103 3.1 0.005 HCB 392 13.5 489 20.1 <0.001 489 15.0 392 20.8 O.001 Mirex 127 3.5 201 3.7 0.15 216 3.8 112 3.4 O.001 Oxychlordane 364 7.6 484 13.1 <0.001 470 10.0 378 11.3 0.005 trans-Nonachlor 377 10.4 486 17.6 O.001 480 15.6 383 14.4 0.276 Table 4.8 (continued): Median organochlorine concentrations by categories of potential confounders and p-values for the Wilcoxon rank-sum tests. Confounder Region BMI Category GVRD CRD <25 25+ N total 759 123 375 482 N detect-able Median (ug/kg) N detect-able Median (^ g/kg) p-value N detect-able Median (ug/kg) N detect-able Median (ug/kg) p-value Total PCBs 759 166.3 123 144.7 0.269 375 166.3 482 159.0 0.961 Dioxin-like PCBs 759 16.8 123 15.2 0.685 375 16.6 482 17.5 0.176 Non dioxin-like PCBs 759 145.3 123 127.1 0.224 375 147.3 482 140.3 0.796 PCB 28 139 3.2 19 4.0 0.8 56 3.2 101 3.3 0.715 PCB 99 621 6.1 100 5.7 0.442 292 5.8 413 6.3 0.018 PCB 105 239 3.4 36 3.6 0.694 109 3.3 162 3.6 0.792 PCB 118 709 8.8 116 8.2 0.831 343 8.2 460 9.2 0.003 PCB 138 753 20.9 121 18.7 0.482 372 19.9 479 21.1 0.123 PCB 153 758 42.4 122 38.9 0.558 375 43.1 481 41.1 0.967 PCB 156 671 6.6 116 5.8 0.624 333 6.8 436 6.1 0.039 PCB 170 725 12.3 121 11.3 0.323 357 12.9 467 11.7 0.172 PCB 180 756 37.9 122 34.4 0.22 375 39.7 480 35.9 0.162 PCB 183 471 4.3 72 3.7 0.152 213 4.1 318 4.1 0.275 PCB 187 721 11.0 120 9.5 0.15 356 11.4 462 10.4 0.72 P-HCCH 669 17.4 110 16.2 0.674 316 16.5 440 17.6 0.014 p, p'-DDE 759 287.9 123 213.4 0.014 375 255.3 482 292.5 0.006 p,p'-DDT 257 10.3 31 9.7 0.066 113 10.4 164 10.0 0.436 cis-Nonachlor 223 3.3 27 2.7 0.116 98 3.3 147 3.1 0.253 HCB 759 17.1 122 17.3 0.932 375 15.6 482 18.1 0.001 Mirex 291 3.7 37 3.5 0.05 162 3.8 160 3.4 O.001 Oxychlordane 728 10.8 120 10.0 0.446 354 10.0 473 11.0 0.002 trans-Nonachlor 742 15.0 121 15.0 0.784 366 14.3 475 15.3 0.041 Table 4 .8 (continued): Median organochlorine concentrations by categories of potential confounders and p-values for the Wilcoxon rank-sum tests. Confounder Ethnicity Education r" Category European Other high school high school+ N total 708 146 444 424 N detect-able Median (ug/kg) N detect-able Median (ug/kg) p-value N detect-able Median (ug/kg) N detect-able Median (ug/kg) p-value Total PCBs 708 166.5 146 120.9 O.001 444 166.1 424 156.2 0.191 Dioxin-like PCBs 708 17.5 146 14.2 0.003 444 16.7 424 16.5 0.524 Non dioxin-like PCBs 708 146.0 146 109.5 O.001 444 146.4 424 137.7 0.161 PCB 28 138 3.3 16 2.8 0.966 71 3.2 85 3.4 0.339 PCB 99 587 6.1 114 5.6 0.245 362 6.4 347 5.7 0.187 PCB 105 225 3.4 44 4.4 0.223 139 3.4 132 3.6 0.693 PCB 118 665 8.8 134 8.1 0.17 417 8.8 394 8.6 0.634 PCB 138 702 21.2 144 16.4 0.026 439 21.1 421 19.5 0.36 PCB 153 706 42.8 146 35.6 0.002 443 43.1 423 40.3 0.344 PCB 156 656 6.5 106 6.0 O.001 402 6.5 373 6.3 0.402 PCB 170 692 12.4 127 10.7 <0.001 426 12.5 407 11.6 0.082 PCB 180 707 39.1 143 26.3 O.001 443 38.0 421 35.4 0.087 PCB 183 456 4.1 70 4.5 0.047 278 4.3 255 4.0 0.3 PCB 187 688 10.7 126 10.6 0.002 421 11.4 406 10.0 0.096 P-HCCH 621 15.8 136 125.0 <0.001 398 17.2 367 17.2 0.128 p, p'-DDE 708 248.8 146 700.0 <0.001 444 299.5 424 249.2 0.007 p,p'-DDT 182 8.3 99 24.8 <0.001 148 11.7 136 9.5 0.55 cis-Nonachlor 183 2.9 61 4.9 O.001 135 3.2 110 3.2 0.552 HCB 707 17.3 146 15.7 0.05 444 17.8 423 16.1 0.002 Mirex 269 3.6 52 3.9 0.434 165 4.0 155 3.4 0.267 Oxychlordane 686 10.9 134 8.9 <0.001 430 11.8 404 9.6 O.001 trans-Nonachlor 699 14.9 136 16.3 0.902 433 15.8 416 14.0 0.002 Table 4.8 (continued): Median organochlorine concentrations by categories of potential confounders and p-values for the Wilcoxon rank-sum tests. Confounder Ever farmed Family history of NHL Category no yes no yes N total 833 28 827 35 N detect- Median N detect-able Median p-value N detect-able Median N detect-able Median p-value able (ug/kg) (ug/kg) (ug/kg) (ug/kg) Total PCBs 833 159.1 28 174.3 0.152 827 158.9 35 164.1 0.662 Dioxin-like PCBs 833 16.6 28 19.1 0.159 827 16.6 35 16.3 0.908 Non dioxin-like PCBs 833 141.4 28 154.1 0.166 827 141.4 35 140.4 0.626 PCB 28 152 3.3 2 3.5 0.103 143 3.2 10 4.0 0.049 PCB 99 683 6.0 22 7.1 0.259 676 6.0 29 5.9 0.942 PCB 105 257 3.5 13 3.2 0.218 260 3.5 10 3.5 0.51 PCB 118 780 8.5 25 10.4 0.25 772 8.6 34 9.4 0.476 PCB 138 826 19.9 27 24.7 0.065 819 20.1 35 20.3 0.789 PCB 153 831 41.2 28 44.2 0.119 825 41.6 35 41.1 0.856 PCB 156 740 6.5 28 5.8 0.294 736 6.5 33 5.7 0.757 PCB 170 798 12.1 28 11.7 0.41 792 12.1 35 10.7 0.449 PCB 180 829 37.2 28 38.0 0.302 823 37.0 35 34.3 0.456 PCB 183 511 4.2 20 3.8 0.269 509 4.2 22 3.8 0.71 PCB 187 792 10.7 28 9.9 0.557 786 10.7 35 10.3 0.51 p-HCCH 733 17.4 25 15.3 0.358 726 17.0 34 17.8 0.117 p, p'-DDE 833 274.5 28 387.3 0.415 827 275.2 35 286.7 0.298 p,p'-DDT 267 10.1 11 8.3 0.678 268 10.1 13 7.2 0.165 cis-Nonachlor 232 3.2 11 3.3 0.289 230 3.3 11 2.6 0.402 HCB 832 17.1 28 17.1 0.22 826 17.1 35 16.8 0.478 Mirex 312 3.7 7 3.3 0.072 307 3.7 13 3.6 0.778 Oxychlordane 799 10.5 28 13.9 0.024 793 10.5 35 11.2 0.39 trans-Nonachlor 814 14.6 28 19.0 0.02 808 14.7 35 17.3 0.162 Table 4.9: Confounder variables included in final logistic regression models. Confounder variables included in logistic Organochlorine regression model PCBs Total summed PCBs Age, family history of NHL Summed Dioxin-like PCBs Age, farming PCB 105 Education PCB 118 Age, BMI, farming PCB 156 Age, farming Summed Non dioxin-like PCBs Age, region, family history of NHL PCB 28 None PCB 99 Age, farming PCB 138 Age PCB 153 Age PCB 170 Age PCB 180 Age, sex, BMI, ethnicity, family history of NHL PCB 183 Age PCB 187 Age, education, family history of NHL Pesticide P-HCCH Age, ethnicity, family history of NHL p, p'-DDE Age, BMI, ethnicity p,p'-DDT None cis-Nonachlor None HCB Age, sex, BMI Mirex Age, sex Oxychlordane Age, BMI trans-Nonachlor Age - 9 8 -Table 4.10: Total PCB concentrations, dioxin-like PCB concentrations and association with non-Hodgkin lymphoma. Congener Quartiles fag/kg) Cases Controls Odds ratio (95% confidence interval) Pfor trend Total summed <100.9 81 115 1.00 0.001* >100.9-155.6 103 114 1.41 (0.93-2.14) >155.6-220.0 77 115 1.11 (0.71-1.74) >220.0-6571 142 115 2.14 (1.38-3.30) Dioxin-like <10.12 82 115 1.00 <0.001* Summed >10.12-15.35 96 114 1.41 (0.91-2.16) >15.35-23.72 82 115 1.57 (1.00-2.46) >23.72-280.8 143 115 2.40 (1.53-3.77) PCB 105 Not Detected 281 316 1.00 0.675 >1.32-37.65 132 139 1.06 (0.80-1.42) PCB 118 <4.57 82 109 1.00 0.004* >4.57-7.78 88 113 1.12 (0.74-1.69) >7.78-12.85 95 114 1.23 (0.81-1.88) >12.85-202.1 129 113 1.77 (1.15-2.72) PCB 156 <3.65 85 114 1.00 0.004* >3.65-5.51 85 114 .1.10 (0.72-1.68) >5.51-8.32 105 115 1.43 (0.93-2.21) >8.32-113.3 128 115 1.77 (1.14-2.74) * statistically significant after controlling for false discovery rate. -99-Table 4.11: Non dioxin-like PCB concentrations and association with non-Hodgkin lymphoma. Congener Exposure Categories (Ug/kg lipid) Cases Controls Odds ratio (95% confidence interval) P for Trend Non <88.57 85 115 1.00 O.001* dioxin- >88.57-136.2 96 115 1.30 (0.85-1.97) like >136.2-196.4 93 115 1.19 (0.76-1.86) Summed >196.4-6445 148 115 2.18 (1.41-3.38) PCB 28 Not detected 348 376 1.00 0.779 >1.38-54.47 74 84 0.95 (0.67-1.34) <3.06 106 113 1.00 0.045 >3.06-4.83 82 115 0.78 (0.52-1.15) PCB 99 >4.83-7.78 85 115 0.81 (0.54-1.21) >7.78-61.34 130 115 1.27 (0.86-1.87) <11.61 100 115 1.00 0.02* >11.61-19.28 90 115 0.93 (0.62-1.38) PCB 138 >19.28-29.72 94 115 0.99 (0.66-1.50) >29.72-289.4 138 115 1.46 (0.98-2.18) <25.29 90 115 1.00 0.002* >25.29-38.68 86 115 1.04 (0.68-1.57) PCB 153 >38.68-59 106 115 1.34 (0.87-2.04) >59-735.9 140 115 1.79 (1.17-2.72) <7.16 88 115 1.00 0.005* >7.16-11.17 93 115 1.17 (0.77-1.79) PCB 170 >11.17-17.23 107 115 1.41 (0.91-2.18) >17.23-901.5 134 115 1.80 (1.16-2.79) <21.93 85 111 1.00 0.005* >21.93-35.63 94 110 1.28 (0.82-2.00) PCB 180 >35.63-54.72 89 115 1.25 (0.78-2.00) >54.72-3787 126 113 1.91 (1.19-3.07) Not detected 162 177 1.00 0.113 PCB 183 >1.87-3.95 107 142 0.83 (0.59-1.18) >3.95-84.86 153 141 1.22 (0.87-1.71) <5.93 88 114 1.00 0.003* >5.93-9.82 98 114 1.27 (0.83-1.95) PCB 187 >9.82-15.46 79 114 1.04 (0.66-1.63) >15.46-833.2 136 112 1.92 (1.23-2.98) statistically significant after controlling for false discovery rate. -100-Table 4.12: Plasma organochlorine pesticide/ pesticide metabolite concentrations and association with non-Hodgkin lymphoma. Analyte Exposure Categories (jig/kg lipid) Cases Controls Odds ratio (95% confidence interval) P for Trend <9.14 87 111 1.00 0.043 p-HCCH >9.14to 14.68 80 113 0.95 (0.63-1.44) >14.68 to 26.06 129 115 1.54 (1.03-2.30) >26.02 to 7034.84 112 110 1.59 (1.01-2.49) cis-Nonachlor Not detected 296 336 1.00 0.339 >0.79 to 19.28 126 124 1.15 (0.86-1.55) <134.41 103 112 1.00 0.027* p, p'-DDE >134.41 to 263.91 >263.91 to 512.02 84 100 113 112 0.84 1.04 (0.56-1.25) (0.70-1.56) >512.02to 18898.1 121 112 1.42 (0.92-2.19) p,p'-DDT Not detected >3.24 to 1112.42 289 133 305 155 0.91 1.00 (0.68-1.20) 0.491 <11.45 83 108 1.00 0.001* HCB >11.45 to 16.11 84 115 1.04 (0.69-1.57) >16.11 to 22.78 103 113 1.36 (0.89-2.08) >22.78 to 1050.13 138 113 1.94 (1.25-3.03) Mirex Not detected 248 306 1.00 0.013 >1.43 to 60.46 174 154 1.44 (1.08-1.92) <6.07 79 112 1.00 O.001* Oxychlordane >6.07 to 9.76 >9.76 to 13.7 91 87 113 112 1.36 1.39 (0.88-2.08) (0.88-2.19) >13.7 to 58.21 151 112 2.68 (1.69-4.24) <8.97 91 117 1.00 0.009* trans-Nonachlor >8.97 to 14.08 >14.08 to 20.83 93 107 114 115 1.14 1.36 (0.76-1.73) (0.89-2.09) >20.83 to 106.23 131 114 1.70 (1.11-2.60) * statistically significant after controlling for false, discovery rate. -101 -Table 4.13: Statistically significant interactions between organochlorines and covariates. Europeans Other Ethnicities Organochlorine Exposure category Cases Controls Odds ratio (95% confidence interval) p for trend Cases Controls Odds ratio (95% confidence interval) p for trend PCB 105 Not Detected 231 246 1 0.537 70 54 1 0.045 >1.32-37.65 103 119 0.90 (0.66-1.25) 21 29 1.96 (0.93-4.14) p,p'-DDT Not Detected 279 247 1 0.403 42 26 1 0.013 >3.24 to 1112.42 90 92 1.15 (0.82-1.62) 41 65 0.40 (0.20-0.82) Table 4.14: Forward stepwise selection of organochlorine exposures and association with risk of NHL. Organochlorine Variables Entered in Stepwise Selection Procedure Organochlorine Variables Included in Model Exposure Categories (u.g/kg lipid) Cases Controls Odds Ratio (95% confidence interval) All Organochlorines including summary PCB variables Oxychlordane <6.07 77 111 >6.07to9.76 87 113 >9.76tol3.7 85 111 >13.7 to 58.21 140 109 Non dioxin-like <88.57 77 108 PCB summed >88.57 to 136.19 94 112 >136.19to 196.38 82 114 >196.38 to 6444.82 136 110 1.00 1.34 (0.84-2.14) 1.29 (0.78-2.14) 2.28 (1.33-3.91) 1.00 1.09 (0.68-1.76) 0.90 (0.53-1.52) 1.61 (0.94-2.73) PCBs Only [PCB 153 <5.93 85 109 >5.93to9.82 85 112 >9.82 to 15.46 104 114 >15.46 to 833.15 134 114 1.00 1.04 (0.67-1.62) 1.28 (0.82-2.01) 1.85 (1.18-2.92) Pesticides Only Oxychlordane <6.07 77 112 >6.07to9.76 87 113 >9.76tol3.7 85 112 >13.7 to 58.21 140 112 1.00 1.34 (0.87-2.08) 1.40 (0.88-2.22) 2.59 (1.62-4.15) -103-Table 4.15: Odds Ratios ( 95% CI) for the largest versus smallest organochlorine exposure categories by histologic subtype of NHL Organochlorine All NHL All B-cell Follicular Diffuse Large Cell Other B-cell T-Cell No. of Cases 422 375 141 67 167 47 Total PCBs 2.1 (1.4-3.3)f 2.2(1.4-3.5)1 2 .0 (1 .1 -3 .7 ) * 1.8(0.8-4.1) 3.0 (1.6-5.6)t 1.7 (0.7-4.3) Dioxin-like Summed 2.4(1.5-3.8)1 2 .6(1 .6-4 .2) f 2.5 (1.3-4.7)f 2.1 (0.9-4.9) 3.2(1.7-6.1)t 1.5 (0.6-3.8) PCB 1051 1.1 (0.8-1.4) 1.1 (0.8-1.5) 0.9 (0.6-1.4) 0.8(0.5-1.5) 1.4(1.0-2.0) 0 .7(0.4-1.5) PCB 118 1.8(1.2-2.7)1 1.9(1.2-3.0)1 2 .0 (1 .1 -3 .7 ) * 2.0 (0.9-4.7)* 1.8(1.0-3.3)* 1.2 (0.4-3.0) PCB 156 1.8(1.1-2.7)1 1.9(1.2-2.9)1 2 .4(1 .2-4 .5) f 1.3 (0.6-3.0) 1.8(1.0-3.4) 1.4 (0.6-3.4) Non dioxin-like Summed 2.2(1.4-3.4)1 2.3 (1.5-3.7)f 2.1 (1.1-3.9)* 1.8(0.8-4.1) 3.2(1.7-6.0)f 1.7(0.7-4.1) PCB 281 1.0 (0.7-1.3) 0.9 (0.7-1.3) 0.7(0.4-1.3) 1.3 (0.7-2.4) 1.0 (0.6-1.6) 1.1 (0.5-2.3) PCB 99 1.3 (0 .9-1.9)* 1.4 (0.9-2.1)* 1.3 (0.8-2.3) 1.0 (0.5-2.0) 1.7(1.0-2.9) 0.7 (0.3-1.7) PCB 138 1.5 (1 .0-2.2)* 1.5 (1.0-2.3)* 1.5 (0.9-2.7) 1.2 (0.6-2.6) 1.7(1.0-3.0)* 1.2 (0.5-2.8) PCB 153 1 .8(1.2-2.7)t 1.8(1.2-2.8)f 2 .0 (1 .1 -3 .7 ) * 1.3 (0.6-2.7) 2.1 (1.2-3.8)* 1.6(0.7-4.0) PCB 170 1.8(1.2-2.8) f 1.8(1.2-2.9)f 1.5 (0.8-2.8) 1.6 (0.7-3.6) 2 .4 (1 .3 -4 .5 ) * 1.9 (0.7-4.6) PCB 180 1.9(1.2-3.1) f 1.8(1.1-3.0)* 1.6 (0.8-3.1) 1.2 (0.5-2.9) 2 .6 (1 .3 -5 .3 ) * 3.5 (1.2-9.7)1 PCB 1832 1.2 (0.9-1.7) 1.2 (0.9-1.8) 1.6(1.0-2.7)* 0.8(0.4-1.6) 1.2 (0.8-1.8) 1.1 (0.5-2.4) PCB 187 1.9(1.2-3.0)1 2.0(1.3-3.2)1 1.8(1.0-3.3) 1.7 (0.7-4.0) 2.6(1.4-4.9)t 1.7 (0.7-4.3) P -HCCH 1.6(1.0-2.5)* 1.6(1.0-2.5)* 1.2 (0.6-2.3) 1.5 (0.6-3.5) 2.1 (1.2-3.8)t 1.4 (0.4-4.8) cis-nonachlor1 1.2 (0.9-1.6) 1.2 (0.9-1.6) 1.1 (0.7-1.6) 1.1 (0.6-1.9) 1.3 (0.9-1.9) 1.1 (0.6-2.2) p,p'-DDE 1.4 (0 .9-2.2)* 1.5 (1.0-2.4)* 1.8(0.9-3.3)* 0.6 (0.2-1.5) 1.8(1.0-3.2)* 1.0 (0.4-2.6) p,p'-DDT 0.9 (0.7-1.2) 0.9 (0.7-1.2) 0.7 (0.5-1.1) 1.0 (0.6-1.7) 1.0 (0.7-1.5) 1.1 (0.6-2.1) HCB 1.9(1.3-3.0)t 2.0 (1.3-3.2)f 2 .4 (1 .2 -4 .6 ) * 2.5 (1.0-6.0)* 1.6 (0.9-2.9)* 1.4 (0.5-4.0) Mirex1 1.4(1.1-1 .9)* 1.4(1.0-1.9)* 1.3 (0.9-2.0) 1.3 (0.8-2.3) 1.5 (1.0-2.1) 2 .2 (1 .2 -4 .3 ) * Oxychlordane 2 .7 (1 .7 -4 .2 ) t 2.8 (1.7-4.5)f 3.3 (1.7-6.4)f 2 .9(1.2-7.2)t 2.5 (1.3-4.7)f 2.4 (0.9-6.5)* Trans-nonachlor 1.7(1.1-2.6)1 1.8(1.2-2.8)t 1.7(0.9-3.1) 1.9(0.9-4.3)t 2 .0 (1 .1 -3 .6 ) * 1.1 (0.5-2.9) *p<0.05, t p O . 0 1 , W o categories, 2three categories Table 4.16: Characteristics of AHR gene variants included in study. SNP region and position* Flanking sequence Nucleotide change Codon change Amino acid change Observed minor allele frequency Observed minor allele frequency in European subjects Exon 1 -459 GATTT(G/A)GGAAG G/A n/a n/a 0.249 0.235 Intervening Sequence 1 +1204 C AGAA(C/G)T AT AC C/G n/a n/a 0.238 0.227 Intervening Sequence 1 +4640 ATGTT(G/A)GGATC G/A n/a n/a 0.150 0.176 Intervening Sequence 1 -3946 TTCAT(A/G)ACAAT A/G n/a n/a 0.336 0.366 Exon 2 +132 CTTAA(T/C)ACAGA T/C AAT/AAC none (Asp44Asp) 0.073 0.079 Intervening Sequence 7 +33 TATTG(G/T)ATGTA G/T n/a n/a 0.135 0.1.10 Exon 10 +1661 CATCA(G/A)ACACA G/A AGA/AAA Arg554Lys 0.142 0.113 * Relative to the sequence of accession #L19872. Figure 4.1: Pairwise r2 values of linkage disequilibrium between the AHR SNPs, based genotyping data. - 1 0 6 -Table 4.17: Single nucleotide polymorphisms (SNPs) of the AHR gene, genotype frequencies, and p-value for the test of Hardy-Weinberg equilibrium in European controls. Parent NHL study Organochlorine Study Variant Genotype Proportion HWE p-value Cases Controls Proportion Cases Controls Total 797 791 422 459 X1-459G/A G/G 0.560 0.575 453 436 0.572 245 259 G/A 0.352 281 278 0.353 148 163 A/A 0.068 54 54 0.062 25 30 Undetermined 0.020 9 23 0.012 4 7 IVS1+1205C/G C/C 0.578 0.818 467 451 0.590 253 267 C/G 0.338 264 272 0.337 138 159 G/G 0.065 53 50 0.060 25 28 Undetermined 0.020 13 18 0.012 6 5 IVS1+4640G/A G/G 0.707 0.096 546 577 0.722 298 338 G/A 0.229 203 161 0.217 100 91 A/A 0.033 28 24 0.034 16 14 Undetermined 0.031 20 29 0.027 8 16 IVS1-3946G/A G/G 0.441 0.652 342 358 0.440 184 204 G/A 0.408 327 321 0.411 168 194 A/A 0.123 106 89 0.125 59 51 Undetermined 0.028 22 23 0.024 11 10 Table 4.17 (continued): Single nucleotide polymorphisms (SNPs) of the AHR gene, genotype frequencies, and p-value for the test of Hardy-Weinberg equilibrium in European controls. Parent NH L study Organochlorine Study Variant Genotype Proportion HWE p-value Cases Controls Proportion Cases Controls X2+132T/C T/T 0.853 1 672 682 0.848 350 397 T/C 0.134 116 96 0.141 67 57 C/C 0.006 4 5 0.003 2 1 Undetermined 0.008 5 8 0.008 3 4 IVS7+33G/T G/G 0.741 0.843 598 579 0.746 314 343 G/T 0.213 167 172 0.218 94 98 T/T 0.025 18 22 0.024 9 12 Undetermined 0.020 14 18 0.012 5 6 X10+1661G/A G/G 0.735 0.082 594 573 0.742 314 340 G/A 0.206 175 177 0.208 95 102 A/A 0.045 22 25 0.043 11 13 Undetermined 0.014 6 16 0.007 2 4 Table 4.18: Percentage of genotype data determined from whole genome amplified samples and reliability of whole genome amplification. Proportion of Quality Error Quality Error genotypes derived control ratet (%) controls ratef (%) Variant from whole genome sample sample amplified samples, n pairs*, pairs§, (%) n n Cases Controls Total n=797 n=791 X1-459G/A 129 (16.2) 133 (16.8) 89 4.71 88 0 IVS1+1205C/G 133 (16.7) 134(16.9) 91 3.41 90 0 IVS1+4640G/A 130(16.3) 133 (16.8) 91 3.45 94 1.08 IVS1-3946G/A 130(16.3) 132(16.7) 88 3.53 93 0 X2+132T/C 127(15.9) 131 (1.6) 92 1.10 92 1.10 IVS7+33G/T 135 (16.9) 135(17.1) 89 2.30 94 0 XI0+1661 G/A 128 (16.1) 130(16.4) 94 2.17 93 0 * Quality control pairs include the genotype from a high quantity DNA sample and the genotype from the whole genome amplified sample of the same DNA. Only pairs where both genotypes were available are included. § Quality control pairs include the genotype from a whole genome amplified sample of a low DNA quantity sample and the genotype from a second higher quantity DNA sample from the same subject. Only pairs where both genotypes were available are included. | Error rate is the percentage of discordant genotypes between the quality control sample pairs where both genotypes are determined. - 1 0 9 -Table 4.19: Genotype distribution of SNPs in the AHR gene by categories of covariates, and chi-square tests. Age Sex Region Ethnicity Education SNP Genotype <60 60+ Male Female Vancouver Victoria European Other High school High school+ X1-459G/A G/G 414 475 511 378 702 187 719 170 435 440 G/A 241 318 291 268 452 107 419 140 271 275 A/A 49 59 67 41 87 21 76 32 55 52 p-value 0.437 0.052 0.668 0.004 0.944 IVS1+1205C/G C/C 425 493 524 394 727 191 737 181 451 453 C/G 231 305 281 255 433 103 405 131 254 270 G/G 45 58 63 40 81 22 73 30 52 50 0.477 0.117 0.739 0.021 0.833 IVS1+4640G/A G/G 518 605 618 505 918 205 822 301 537 567 G/A 165 199 206 158 273 91 330 34 184 173 A/A 19 33 35 17 37 15 46 6 32 19 0.395 0.207 0.006 O.001 0.108 IVS1-3946G/A G/G 318 382 387 313 569 131 496 204 345 341 G/A 296 352 378 270 511 137 537 111 316 322 A/A 86 109 102 93 151 44 172 23 96 96 0.927 0.268 0.37 O.001 0.962 Table 4.19 (continued): Genotype distribution of variants in the AHR gene by categories of covariates, and chi-square tests. Age Sex Region Ethnicity Education SNP Genotype <60 60+ Male Female Vancouver Victoria European Other High school High school+ X2+132T/C T/T 611 743 747 607 1085 269 1038 316 658 675 T/C 95 117 124 88 165 47 183 29 111 95 C/C 6 3 4 5 8 1 6 3 5 4 p-value 0.429 0.53 0.588 0.005 0.456 IVS7+33G/T G/G 518 659 648 529 936 241 962 215 576 583 G/T 162 177 194 145 270 69 237 102 166 166 T/T 23 17 26 14 35 5 15 25 18 22 p-value 0.133 0.383 0.466 <0.001 0.834 X10+1661G/A G/G 512 655 641 526 925 242- 963 204 572 577 G/A 169 183 201 151 284 68 241 111 176 168 A/A 26 21 29 18 41 6 17 30 20 26 p-value 0.143 0.537 0.077 <0.001 0.611 Table 4.20: Covariates included in logistic regression models measuring the association between AHR variants and NHL risk chosen by change in estimate criterion (5% threshold). SNP Confounders included in final model (age, sex, region , ethnicity, and education tested) X1-459G/A None IVS1+1205C/G None TVS1+4640G/A None IVS1-3946G/A None X2+132T/C None IVS7+33G/T None X10+1661G/A Ethnicity - 1 1 2 -Table 4.21: Variants the AHR gene and association with Non-Hodgkin lymphoma. SNP Model Genotype Odds ratio (95% confidence interval) p for trend FDR thres-hold LRT P value Xl-459 co-dominant G/A 0.97 (0.79-1.20) 0.777 G/A A/A 0.96 (0.65-1.44) additive nA 0.98 (0.83-1.15) 0.777 0.05 0.949 dominant G/A or A/A 0.97 (0.79-1.19) 0.775 0.959 recessive A/A 0.97 (0.66-1.44) 0.89 0.799 IVS1+1205 co-dominant C/G 0.94 (0.76-1.16) 0.766 C/G G/G 1.02 (0.68-1.54) additive nG 0.98 (0.83-1.15) 0.766 0.043 0.572 dominant C/G or G/G 0.95 (0.78-1.16) 0.624 0.682 recessive G/G 1.05 (0.70-1.56) 0.817 0.552 IVS1+4640 co-dominant G/A 1.33 (1.05-1.69) 0.029 G/A A/A 1.23 (0.71-2.15) additive nA 1.24(1.02-1.50) 0.029 0.007 0.304 dominant G/A or A/A 1.32(1.05-1.65) 0.016 0.794 recessive A/A 1.15 (0.66-2.00) 0.623 0.018 IVS1-3946 co-dominant G/A 1.07 (0.86-1.32) 0.192 G/A A/A 1.25 (0.91-1.71) additive nA 1.10(0.95-1.27) 0.192 0.014 0.684 dominant G/A or A/A 1.11 (0.90-1.35) 0.327 0.340 recessive A/A 1.21 (0.89-1.63) 0.218 0.556 X2+67 co-dominant T/C 1.23 (0.92-1.64) 0.256 T/C C/C 0.81 (0.22-3.04) additive nC 1.17(0.89-1.53) 0.256 0.021 0.395 dominant T/C or C/C 1.21 (0.91-1.60) 0.2 0.545 recessive C/C 0.79 (0.21-2.95) 0.727 0.168 IVS7+33 co-dominant G/T 0.94 (0.74-1.20) 0.425 G/T T/T 0.79 (0.42-1.49) additive nT 0.92 (0.76-1.13) 0.425 0.029 0.779 dominant G/T or T/T 0.92 (0.73-1.16) 0.5 0.610 recessive T/T 0.80 (0.43-1.51) 0.497 0.616 XI0+501 co-dominant A/G 0.97 (0.76-1.23) 0.714 A/G G/G 0.91 (0.49-1.67) additive nG 0.96 (0.79-1.18) 0.714 0.034 0.934 dominant A/G or G/G 0.96 (0.79-1.18) 0.748 0.844 recessive G/G 0.92 (0.50-1.68) 0.786 0.795 -113 -Table 4.22: Variants of the AHR gene and association with Non-Hodgkin lymphoma in Europeans using additive models of phenotypic expression. All ethnic groups Europeans Asians Indo-Pakistani n=1588 n=1236 n=170 n=66 SNP Odds ratio (95% CI) P-value Odds ratio (95% CI) P-value Odds ratio (95% CI) P-value Odds ratio (95% CI) P-value X1-459G/A 0.98 (0.83-1.15) 0.777 0.94 (0.78-1.13) 0.503 1.34 (0.84-2.14) 0.214 0.83 (0.39-1.78) 0.641 IVS1+1205C/G 0.98 (0.83-1.15) 0.766 0.92 (0.77-1.11) 0.405 1.37(0.86-2.19) 0.18 1.03 (0.50-2.13) 0.926 IVS1+4640G/A 1.24(1.02-1.50) 0.029 1.21 (0.98-1.48) 0.074 * * 2.39 (0.65-8.72) 0.187 IVS1-3946G/A 1.10(0.95-1.27) 0.192 1.15 (0.98-1.35) 0.097 0.84 (0.47-1.52) 0.564 1.25 (0.60-2.59) 0.546 X2+132T/C 1.17(0.89-1.53) 0.256 1.15 (0.86-1.55) 0.355 0.67 (0.21-2.17) 0.507 1.56 (0.43-5.58) 0.498 IVS7+33G/T 0.92 (0.76-1.13) 0.425 0.93 (0.72-1.20) 0.562 0.91 (0.59-1.40) 0.667 1.19(0.41-3.48) 0.753 X10+1661G/A 1.01 (0.84-1.21) 0.92 0.95 (0.74-1.23) 0.715 0.87 (0.57-1.33) 0.533 1.26 (0.45-3.47) 0.662 * odds ratio estimates could not be obtained because controls all had the major allele homozygote genotype. Table 4.23: Odds ratios (95% confidence intervals) from additive models of phenotypic expression within NHL subtypes. SNP All NHL cases All B-cell DLBC Follicular Other B-cell T-cell n 797 721 210 216 295 76 X1-459G/A 0.98 (0.83-1.15) 0.96 (0.82-1.14) 0.96 (0.75-1.23) 1.01 (0.79-1.28) 0.94 (0.75-1.17) 1.11 (0.77-1.61) IVS1+1205C/G 0.98 (0.83-1.15) 0.97 (0.82-1.15) 0.95 (0.74-1.23) 1.01 (0.79-1.30) 0.96 (0.77-1.19) 1.01 (0.69-1.49) IVS1+4640G/A 1.24(1.02-1.50)* 1.23 (1.02-1.50)* 1.25 (0.94-1.67) 1.29 (0.98-1.69)1 1.18(0.91-1.52) 1.25 (0.81-1.92) IVS1-3946G/A 1.10(0.95-1.27) 1.10(0.95-1.28) 0.99 (0.79-1.24) 1.21 (0.97-1.50)1 1.11 (0.91-1.36) 1.09 (0.77-1.54) X2+132T/C 1.17 (0.89-1.53) 1.15 (0.87-1.51) 1.11 (0.73-1.67) 1.03 (0.68-1.57) 1.25 (0.88-1.77) 1.38 (0.78-2.45) IVS7+33G/T 0.92 (0.76-1.13) 0.90 (0.73-1.10) 0.86 (0.62-1.18) 0.87 (0.63-1.18) 0.95 (0.72-1.24) 1.18 (0.76-1.84) X10+1661G/A 0.96 (0.79-1.18) 0.96 (0.78-1.18) 0.97 (0.71-1.32) 0.90 (0.65-1.24) 1.01 (0.76-1.33) 0.99 (0.62-1.59) *p<0.05, tpO. l Table 4.24: Statistically significant interactions between AHR SNPs and confounder variables. Confounder p-value for interaction Odds ratio (95% confidence interval) Sex SNP Male Female IVS1-3946G/A 0.042 0.96 (0.79-1.17) 1.30(1.05-1.61) IVS7+33G/T 0.037 1.09 (0.84-1.42) 0.71 (0.51-0.97) Region SNP Vancouver Victoria X1-459G/A 0.002 1.11 (0.92-1.32) 0.56 (0.38-0.82) IVS1+1205C/G 0.001 1.12(0.93-1.34) 0.55 (0.37-0.80) IVS7+33G/T 0.001 0.78 (0.62-0.97) 1.96(1.21-3.19) X10+1661G/A 0.001 0.82 (0.66-1.03) 1.96 (1.20-3.21) - 115 -Table 4.25: Estimated haplotypes and haplotype frequencies from seven AHR SNPs (0=major allele, l=minor allele). SNP XI IVS1 IVS1 IVS1 X2 IVS7 X10 Estimated frequency Estimated No. -459 +1205 +4640 -3946 +67 +33 +501A Haplotype frequency in G / A C / G G / A G / A T / C G/T / G Europeans 1 0 0 0 0 0 0 0 GCGGTGA 0.24678 0.24555 2 1 1 0 0 0 0 0 AGGGTGA 0.22034 0.21569 3 0 0 0 1 0 0 0 GCGATGA 0.18051 0.18474 4 0 0 1 1 0 0 0 GCAATGA 0.13225 0.15275 5 0 0 0 0 0 1 1 GCGGTTG 0.11623 0.10502 6 0 0 0 0 1 0 0 GCGGCGA 0.04645 0.05095 7 0 0 1 1 1 0 0 GCAACGA 0.01496 0.01630 8 1 1 0 0 0 1 1 AGGGTTG 0.01247 0.00499 9 0 0 0 0 0 0 1 GCGGTGG 0.00431 0.00176 10 1 0 1 1 1 0 0 ACGATGA 0.00400 0.00510 11 1 0 0 0 0 0 0 ACGGTGA 0.00393 0.00090 12 1 0 0 0 0 0 1 ACGGTGG 0.00239 0.00261 13 0 0 1 0 0 0 0 GCAGTGA 0.00211 0.00207 14 1 1 0 1 0 0 0 AGGATGA 0.00167 0.00140 15 0 1 0 0 1 0 0 GGGGCGA 0.00144 0.00135 16 0 0 0 1 0 1 1 GCGATTG 0.00138 0.00086 17 1 1 0 0 1 0 0 AGGGCGA 0.00117 0.00154 18 0 1 0 0 0 1 1 GGGGTTG 0.00104 0.00050 19 1 1 1 1 0 0 0 AGAATGA 0.00082 0.00113 20 1 1 0 1 0 1 0 AGGATTA 0.00066 0.00087 Haplotypes with frequencies <0.045 were pooled for generalized linear model. Table 4.25 (continued): Estimated haplotypes and haplotype frequencies from seven AHR SNPs (0=major allele, l=minor allele). SNP XI IVS1 IVS1 IVS1 X2 IVS7 XlO Estimated frequency Estimated No. -459 G / A +1205 C / G +4640 G / A -3946 G / A +67 T / C +33 G/T +501A / G Haplotype frequency in Europeans 21 1 0 0 0 1 0 0 ACGGCGA 0.00057 0 22 1 0 0 0 0 1 1 ACGGTTG 0.00052 0 23 0 1 0 0 0 0 0 GGGGTGA 0.00050 0.00054 24 0 0 0 1 0 0 1 GCGATGG 0.00049 0.00044 25 1 1 0 0 0 0 1 AGGGTGG 0.00045 0 26 0 0 1 0 1 0 0 GCAGCGA 0.00045 0.00062 27 0 1 0 1 0 1 0 GGGATTG 0.00035 0.00044 28 0 1 0 1 0 0 0 GGGATGA 0.00035 0.00044 29 0 0 0 0 1 1 0 GCGGCTA 0.00035 0.00044 30 1 0 0 1 0 1 1 ACGATTG 0.00034 0.00044 31 1 1 0 1 0 1 1 AGGATTG 0.00027 0 32 1 0 0 1 1 0 0 ACGACGA 0.00021 0.00056 33 1 0 0 1 0 0 0 ACGATGA 0.00020 0 34 0 0 0 1 0 1 0 GCGATTA 0.00003 0 *Haplotypes with frequencies <0.045 were pooled for generalized linear model. Table 4.26: Association between AHR haplotypes and risk of Non-Hodgkin lymphoma, in all ethnicities and in Europeans only. Haplotype Frequency in cases Frequency in controls Odds ratio p-value Main analysis (LRT p-value=0.791) Total n 797 791 GCGGTGA 0.237 0.256 1 AGGGTGA 0.217 0.224 1.06 (0.90-1.24) 0.489 GCGATGA 0.175 0.185 1.03 (0.82-1.28) 0.807 GCAATGA 0.148 0.118 1.31 (1.03-1.67) 0.030 GCGGTTG 0.114 0.119 1.03 (0.80-1.34) 0.812 GCGGCGA 0.052 0.041 1.35 (0.92-1.99) 0.126 pooled 0.057 0.057 1.07 (0.75-1.51) 0.719 Analysis in Europeans (LRT p-value=0.661) Total n 625 611 GCGGTGA 0.231 0.260 1 AGGGTGA 0.211 0.219 1.07 (0.84-1.37) 0.577 GCGATGA 0.183 0.185 1.11 (0.86-1.44) 0.413 GCAATGA 0.171 0.136 1.38 (1.05-1.81) 0.022 GCGGTTG 0.104 0.106 1.08(0.79-1.47) . 0.631 GCGGCGA 0.057 0.044 1.43 (0.93-2.19) 0.104 pooled 0.042 0.049 0.95 (0.61-1.46) 0.803 - 1 1 8 -Table 4.27: Log likelihood ratio test (LRT) p-values for models with gene-environment interactions between the rVS l+4640G/A SNP and organochlorine exposure. Main analysis European sub-group analysis Organochlorine LRT p-value LRT p-value FDR Threshold Total sum of PCBs 0.326 0.319 0.029 Summed dioxin-like PCBs 0.091 0.058 0.012 PCB 118 0.013 0.013 0.006 PCB 156 0.213 0.186 0.024 Summed non dioxin-like PCBs 0.586 0.484 0.038 PCB 99 0.264 0.368 0.026 PCB 138 0.127 0.130 0.015 PCB 153 0.564 0.501 0.035 PCB 170 0.506 0.419 0.032 PCB 180 0.782 0.719 0.047 PCB 187 0.749 0.697 0.044 Pesticides (3-HCCH 0.179 0.483 0.021 p, p'-DDE 0.940 0.655 0.05 HCB 0.718 0.537 0.041 Mirex 0.175 0.354 0.018 Oxychlordane 0.012 0.011 0.003 trans-Nonachlor 0.028 0.020 0.009 - 119-Table 4.28: Odds ratio estimates for interactions between the AHR 1YS1+4640G/A SNP and organochlorine exposure. IVS1+4640G/A Marginal Genotype^ =G/G Genotype=G/A or A/A Exposure Quartiles Odds ratio Cases/ Odds ratio p for Cases/ Odds ratio p for Organochlorine (US/kg) (95% CI) Controls (95% CI) trend Controls (95% CI) trend Marginal 298/338 1 116/105 1.25 (0.92-1.70) 0.149 PCB 118 <4.57 1.00 57/78 1 0.002 25/26 1 0.454 >4.57 to 7.78 1.12(0.74-1.69) 57/86 0.99 (0.60-1.63) 31/21 1.94 (0.84-4.52) >7.78 to 12.85 1.23 (0.81-1.88) 68/83 1.24 (0.75-2.06) 25/27 1.05 (0.46-2.43) >12.85 to 202.13 1.77(1.15-2.72) 98/80 1.95 (1.17-3.25) 25/31 0.97 (0.40-2.38) Oxychlordane <6.07 >6.07 to 9.76 1.00 1.36(0.88-2.08) 53/88 69/85 1 1.62 (0.98-2.69) <0.001 26/21 22/22 1 0.99 (0.41-2.39) 0.439 >9.76to 13.7 1.39(0.88-2.19) 64/79 1.67 (0.97-2.86) 23/27 0.92 (0.37-2.32) >13.7 to 58.21 2.68(1.69-4.24) 104/75 3.19(1.85-5.51) 39/35 1.32 (0.52-3.36) trans-Nonachlor <8.97 1.00 63/95 1 0.004 28/20 1 0.892 >8.97 to 14.08 1.14(0.76-1.73) 67/83 1.32 (0.81-2.14) 26/26 0.85 (0.36-2.00) >14.08 to 20.83 1.36(0.89-2.09) 73/81 1.50(0.91-2.47) 31/28 0.98 (0.39-2.46) >20.83 to 106.23 1.70(1.11-2.60) 95/79 2.04(1.25-3.33) 31/31 0.89 (0.35-2.29) Table 4.29: Odds ratio estimates for interactions between the AHR IVS1+4640G/A SNP and organochlorine exposure in Europeans. IVS1+4640G/A Marginal Genotype=G/G Genotype= =G/A or A/A Exposure Quartiles Odds ratio Cases/ Odds ratio p for Cases/ Odds ratio p for Organochlorine (95% CI) Controls (95% CI) trend Controls (95% CI) trend Marginal 225/258 1 106/96 1.27 (0.91-1.76) 0.160 PCB 118 <4.57 1 39/58 1 0.031 23/22 1 0.273 >4.57 to 7.78 1.29 (0.81-2.05) 48/63 1.16(0.65-2.05) 30/17 2.12 (0.85-5.33) >7.78 to 12.85 1.10(0.68-1.77) 50/68 1.09 (0.60-1.98) 23/26 0.91 (0.37-2.24) >12.85 to 202.13 1.58 (0.97-2.56) 76/64 1.79 (0.98-3.25) 24/31 0.89 (0.34-2.29) Oxychlordane <6.07 >6.07 to 9.76 1 1.25 (0.77-2.03) 34/61 50/67 1 1.41 (0.78-2.56) 0.001 22/18 22/20 1 1.12(0.44-2.86) 0.491 >9.76 to 13.7 1.42 (0.85-2.40) 53/62 1.60 (0.85-3.02) 23/24 1.10(0.41-2.95) >13.7 to 58.21 2.38(1.42-4.00) 84/63 2.72(1.44-5.12) 37/34 1.38 (0.51-3.72) trans- 1 Nonachlor <8.97 42/66 1 0.013 23/17 1 0.839 >8.97 to 14.08 1.09 (0.68-1.75) 54/71 1.16(0.66-2.06) 26/24 0.98 (0.39-2.47) >14.08 to 20.83 1.32 (0.80-2.18) 63/69 1.39 (0.77-2.53) 28/25 1.05 (0.38-2.91) >20.83 to 106.23 1.71 (1.02-2.87) 66/52 2.02 (1.09-3.74) 29/30 0.92 (0.33-2.61) Table 4.30: Global log likelihood ratio test (LRT) p-values for interaction between haplotypes and organochlorine exposure. Main analysis European sub-group analysis Organochlorine LRT p-value LRT p-value Total sum of PCBs 0.634 0.829 Summed dioxin-like PCBs 0.335 0.448 PCB 118 0.374 0.412 PCB 156 0.483 0.836 Summed non dioxin-like PCBs 1.000 0.817 PCB 99 0.766 0.912 PCB 138 0.725 0.860 PCB 153 0.723 0.917 PCB 170 0.603 0.886 PCB 180 0.602 0.858 PCB 187 0.439 0.776 Pesticides 3-HCCH 0.663 0.819 p, p'-DDE 0.952 0.903 HCB 0.883 0.784 Mirex 0.709 0.774 Oxychlordane 0.157 0.380 trans-Nonachlor 0.176 0.282 - 1 2 2 -Chapter 5: Discussion In this case-control study, significant associations were found between several PCB congeners, organochlorine pesticides, and risk of non-Hodgkin lymphoma. Associations were significant even after adjusting for multiple comparisons. There is substantial support for the study hypothesis that plasma organochlorine concentrations are associated with increased NHL risk. In particular, the dioxin-like PCBs 118 and 156, and non dioxin-like PCBs 138, 153, 170, 180 and 187 were found to increase risk of NHL. The results indicate a 2-fold increase in risk for subjects with the highest plasma PCB concentrations versus those with the lowest concentrations, after controlling for confounders such as age. Odds ratio estimates for individual PCB congeners were similar, and increased NHL risk by approximately 50-90%. The pesticides hexachlorobenzene, mirex, and the derivatives of chlordane, oxychlordane and trans-nonachlor, were also found to increase NHL risk. There were significant odds ratios comparing the highest to lowest exposure categories, ranging from 1.4 for mirex up to 2.7 for oxychlordane. There were also significant dose-response trends for these organochlorines. The pesticide p,p'-DDE also had a significant dose-response trend. However, no association was found between PCBs 28, 99, 138, and 183, P-hexachlorocyclohexane, cis-nonachlor, and p,p'-DDT and risk of NHL, after controlling for multiple comparisons. The association between 7 SNPs of the AHR gene and risk of NHL was also measured, under the hypothesis that the carcinogenicity of organochlorines may act through the AhR pathway, and so genetic susceptibility within this pathway may also exist. An increased risk of NHL was found for subjects with the IVS1+4640G/A minor allele. The -123 -odds ratio comparing those with the G/A or A/A genotype to those with the G/G genotype was 1.32 (95% 0=1.05-1.65). Similar results were seen when analyses were restricted to ethnic groups, and NHL subtypes. Effect modification was observed for PCB 118 and chlordane-related organochlorines oxychlordane and trans-nonachlor by the IVS+4640G/A SNP on NHL risk. PCB 118 is a dioxin-like congener that is known to bind and activate AhR. Increasing plasma concentrations of these organochlorines increased risk of NHL in subjects with the common G/G genotype, but the increase in risk was not observed for subjects with the G/A or A/A genotypes. These results also give support to the study hypothesis that organochlorines increase the risk of NHL, and the biological mechanism of carcinogenicity may involve the AhR pathway. The effect modifications observed were unchanged when analysis was done only in Europeans. However, the genetic associations and interactions observed were not significant based on the FDR. Haplotype analyses using the seven genotyped SNPs in this study revealed no association with NHL or interactions with organochlorines. This study had several strengths. This is the largest study to date to investigate the relationship between plasma organochlorine concentrations and risk of NHL. The large number of samples allowed for detection of weak associations that earlier studies were not powered to detect. This is also the first study to date to examine whether gene-environment interactions modify the risk of NHL. The study used a biological measure of exposure, organochlorines concentrations in blood plasma, to avoid recall bias and misclassification associated with questionnaire data. Furthermore, quality control samples were included in the organochlorine analysis to ensure reliability of measurements. The laboratory methods to measure organochlorine concentrations were very reliable, as evidenced by high - 124-repeatability between duplicate samples. Reliability allows for detection of observed associations that may be diluted with measurement error. For the genetic data, reliability between genotyping results for samples that were whole genome amplified and samples that were not whole genome amplified was also high, ensuring that whole genome amplification supplied high quality DNA. As well, only 15% of genotypes were derived from whole genome amplified samples, so the chances of bias due to amplification, if any, will be low. The plasma organochlorine concentrations measured among controls in the study can be assumed to reflect concentrations found in the general population of a similar age. The concentrations detected in this study are lower than those reported in previous US studies (94-96). Comparison to a recent Canadian study by Tsuji et al (136), where plasma organochlorine levels were measured in a sample of adult residents from Hamilton, Ontario, found similar levels to those reported in this study. The only noted difference was that levels of DDT and DDE were much higher in our study population, with mean DDT concentrations at 48.03±120.23 ug/kg in males and 33.27±73.24 ug/kg in females, compared to 8.8±7.4 ug/kg in males and 9.2±9.1 in females in the Hamilton population. Mean concentrations of DDE in the control population of this study were 595.16±1401.26 ug/kg for males and 617.61±1111.54 pg/kg for females compared to 245.3±293.9 ug/kg for males and 245.8±370.4 ug/kg for females in Hamilton. The age range of the participants in this study was not reported, so it is difficult to know whether differences in organochlorine concentrations are attributed to age differences. But there were only differences in concentrations for DDT/DDE, and it would be expected that there would be differences in concentrations of all the organochlorines if age had an effect. The variation in measured DDT/ DDE concentrations between the two studies may be due to differences in food - 125 -contamination levels in the two regions, or may be due to differences in dietary habits of the two populations. In comparison of pesticide contaminant levels in foods from Vancouver (137) and Toronto (138), which is in close proximity to Hamilton, there were a higher number of food items with detectable concentrations of p,p'-DDT and p,p'-DDE in Vancouver. This may be an indication of higher DDT/ DDE contaminants present in foods from Vancouver. Another Canadian study measured organochlorines in blood plasma of a group of Great Lakes fish eaters (139). The participants were much younger, with a median age of 37, whereas about half of the controls in this study were over the age of 60. It would be expected that the control population of this study would have higher plasma organochlorine levels, except that the lipid-adjusted plasma concentrations of DDE, mirex, and total PCBs were instead lower. However, concentrations of other pesticides such as P-HCCH, oxychlordane, and trans-nonachlor were similar. The population included in the Great Lakes fish eaters study likely had higher exposure to certain organochlorines due to frequent consumption of polluted fish. The relationship between Great Lakes fish consumption and increased levels of organochlorine contaminants in the body has been documented in other studies (140-142). In particular, PCBs, mirex, and DDT/ DDE are documented as some of the most common contaminants of Lake Ontario and other Great Lakes fish (143). The pesticides aldrin, a-chlordane, and y-chlordane were not detected in samples in this study, while PCBs 52, 101, and 128 were detected at very low frequencies. PCB congeners 28 and 105 were also detected at low frequencies. Aldrin is metabolized by mammals and in the environment and so is not normally detected. The metabolite of aldrin, dieldrin, was not measured in this study. Chlordane is also metabolized, and the metabolite - 1 2 6 -oxychlordane is more readily detected. Previous studies measuring chlordane also found lower frequencies of detection (19). In similarity, PCB congeners 28, 52, 101, 105, and 128 were frequently non-detectable in another Canadian study of Ontario anglers (140). In general, lower chlorinated PCBs in this study were less frequently detected than higher chlorinated PCBs, and were also less correlated than higher chlorinated PCBs. These results are consistent with the fact that lower chlorinated PCBs are more readily metabolized in the environment by bacteria and are considered less persistent in the food chain (144, 145). They are also more efficiently metabolized in humans and appear only transiently in tissues (145). In this study, precautions were taken to reduce bias in exposure assessment. Blood samples from cases were collected before chemotherapy treatment to avoid bias due to changes in plasma organochlorine levels from the treatment. Thirty-two cases were also excluded from analyses because of possible fluctuations in plasma organochlorine and lipid levels that might occur from body weight change due to cancer. Increases in plasma organochlorines have been documented after a weight loss program, suggesting that the compounds are released during lipid mobilization and decrease in fat mass (114). Despite the exclusions of the cases with weight loss, the remaining cases left in analysis differed from controls in their plasma lipid levels. Lipid-adjusted organochlorine concentrations were used in analyses, and differences in organochlorine concentrations observed between cases and control might solely be a manifestation of varying lipid levels between the groups. Comparison of lipid-adjusted concentrations between these groups would only be reasonable if the differences in lipid levels were not related to rate of release of organochlorines into the bloodstream. The biological reason for the difference in lipid levels between cases and controls is unknown, so it is also unknown whether bias has affected results. However, the - 1 2 7 -difference in lipid levels was small (~5%). Furthermore, if bias was introduced, it would be expected that results for every organochlorine compound would be affected. Not all organochlorine compounds were found to be associated with NHL risk, providing assurance that results were valid. In contrast, lipid levels of cases excluded from analyses did not differ from levels of cases included in analyses. It was expected that there would be a difference due to the weight loss in the cases that were excluded. However, excluded cases had weight loss any time within the year prior to blood collection. For some of them, the period of weight loss occurred months before, so that lipid levels would have been stable at the time of blood collection. Even though lipid levels were similar in the two groups, cases that were excluded had consistently higher plasma organochlorine concentrations than cases that were included. The concentrations were nearly two-fold higher in a few compounds. Such large differences give support to the exclusion of these cases, as inclusion of their data could bias the results away from the null, when in fact the differences in organochlorine concentrations could be related to the disease rather than a cause of the disease. There is a possible biological explanation for the higher organochlorine concentrations observed in the cases with prior weight loss. During rapid or significant weight loss, the release of organochlorines is small relative to the decrease in fat content (114). This would cause the overall concentration of organochlorines to increase in every fatty tissue to reach a new equilibrium. In other words, total body fat decreases after weight loss, but organochlorine concentrations have not decreased to the same extent, thus elevating the overall body burden. -128-For many organochlorine compounds, median concentrations differed by age and ethnicity. Subjects over 60 had higher concentrations than those under 60, which is consistent with the bioaccumulative potential of these chemicals in human fatty tissue. Organochlorine concentrations were also higher among European subjects compared to subjects of other ethnicities. The difference may again be attributed to genetics, but may also be due to dietary differences since North American diets are known to be higher in fat than diets of other regions. In contrast, higher levels of DDE and DDT were measured in subjects of other ethnicities. DDT, despite a ban in North America in the 1970's, is still used in some developing countries worldwide for malaria control, which may explain the higher exposure observed in subjects of other ethnicities. The results of this study confirm previously reported associations between organochlorine exposure and increased risk of NHL. However, comparison of results from this study with those other studies is difficult due to differences in methods of organochlorine measurement and varying population risks. Also, the organochlorines measured differ between studies. For example, the panel of PCB congeners measured in each study varies widely, and for the study by Rothman et al. (94), only the association between total PCB exposure and NHL risk is reported. Also, confounders adjusted for in the logistic regression models also vary. Despite the differences, some accordance is seen between results of this study and previous studies. Similar case-control studies in the US have also found that plasma PCB concentrations increase the risk of NHL (94, 96, 97). In particular, De Roos et al (96) also found significant associations with PCBs 156 and 180. Results from another study observed that PCB exposure, as assessed by levels found in carpet dust in homes of participants, was associated with NHL risk (146). Similarly, the highest exposure -129-category for PCB 180 had significantly elevated NHL risk, and there was a positive trend in risk with increasing PCB 180 levels. A study combining results from three cohort studies, which includes the CLUE I cohort used in the Rothman et al study (94), found dose-response trends for PCB congeners 118, 138, and 153 (91). Significant dose-response trends for these congeners were also observed in this study. In contrast, a case-control study of past occupational exposure to various solvents and chemicals and NHL did not find an association with PCB exposure (89). However, exposure assessment was done through questionnaire data on job history and may be subject to recall bias and misclassification. As well, there were very small numbers of individuals with reported PCB exposure. Another retrospective cohort study of mortality in workers exposed to PCBs in an electrical capacitor manufacturing plant did not find an increase in mortality from NHL with increasing levels of exposure (147). But the number of deaths from NHL was again too few warrant any definite conclusions. Another mortality study of workers employed at a Canadian transformer manufacturing plant, where transformer fluid containing PCBs were used, also did not find an excess of deaths from NHL (148). However, exposure to PCBs at the plant was very limited, as only 85 of the over 51 000 transformers at the plant contained fluid with PCBs. Health concern was emphasized on the considerably higher exposure to mineral oil, which is refined from naphthenic base crudes. In this study, an interaction was observed between ethnicity and PCB congener 105. Detectable levels of PCB 105 in the blood were not associated with NHL risk in Europeans, but there was an increased risk for those not identified as Europeans. The group not identified as Europeans consists of various ethnic groups that were categorized as one for analysis purposes, so it is difficult to assess the role of ethnicity for this increased risk. But - 130-PCB 105 is a dioxin-like PCB, and two other dioxin-like PCBs, 118 and 156, were also found to increase risk of NHL in this study. TCDD, the most toxic dioxin and organochlorine known, is the only organochlorine compound classified as carcinogenic to humans (group 1) (149). Dioxin-like PCBs are structurally similar to dioxins and TCDD, so it is possible that PCB 105 also increases risk of NHL, although only in particular subsets of the population. The International Agency for Research on Cancer (IARC) classifies PCBs as probable human carcinogens (2A), with sufficient evidence of carcinogenicity in animals but limited evidence from human studies (150). Rats and mice administered with PCBs were shown to produce cancers of the liver and lung. Most of the human studies reviewed were occupational studies, including some of the ones just mentioned, and studies of victims of the Yusho incident. Any evidence of increase in cancers involves cancers of the digestive system or hematopoietic neoplasms. The results of this study provide further epidemiological evidence that PCBs are possible human carcinogens. All individual PCB congeners with significant associations with NHL had similar odds ratios comparing groups with highest exposure to those with the lowest exposure. These results likely reflect the correlation between PCB congeners, particularly for congeners with a higher degree of chlorination. The high correlation between PCB congeners, in addition to the fact that exposure to these compounds in the general population all come from dietary intake, makes it very difficult to determine individual congener effects and risk profiles. Attempts to tease out effects due to any single PCB congener would be very complex, considering the mixture in exposure and multiple other factors that may be involved, such as interactions between different compounds and genetic susceptibility. It - 1 3 1 -may be useful to look at summary exposure variables to assess the overall risk rather than focusing on individual PCB congeners. In this study, a total summed PCB variable was analyzed to represent overall PCB exposure. Summary variables based on functional grouping, summed dioxin-like and summed non dioxin-like PCBs, were also analyzed. Analyzing these summary PCB exposure variables gave higher estimates than analyzing individual PCB congeners. This may indicate that having high body burden of PCBs overall may have a more detrimental effect than having a high body burden of any one PCB in particular. This study found significant associations with the pesticides P-HCCH, HCB, mirex, the chlordane-related products oxychlordane and trans-nonachlor, p,p'-DDE, and risk of NHL. The other chlordane-related compound, cis-nonachlor, and the parent compound of p,p'-DDE, p,p'-DDT, were not found to be associated with NHL risk. The study by Hardell et al (93), which measured organochlorine concentrations in adipose tissues of participants, also observed higher concentrations of chlordane and chlordane-related products in cases compared to controls. The odds ratio for trans-nonachlor comparing those with concentrations higher than the median versus those lower than the median was significant, while the odds ratios for cis-nonachlor and oxychlordane were not significant. Cis-nonachlor was also not found to be associated to NHL risk in this study. A second study in which adipose tissue levels of organochlorines were measured found oxychlordane to significantly increase the risk of NHL, but not trans-nonachlor (98). Significant trends for a-chlordane and y-chlordane and risk of NHL were found in a study of carpet dust in homes, indicating that residential use of the insecticide is associated to the disease (151). However, other studies which measured chlordane or chlordane-related - 132-product levels in blood did not find an association with NHL (95-97). Population-based studies using questionnaire data to obtain information on pesticide exposure have also found that chlordane does not affect NHL risk (84, 152), although one case-control study of pesticide use in the Midwestern US did find a link between chlordane and NHL (79). Interpretation from these studies using occupational data is difficult, as workers using pesticides are often exposed to multiple types. IARC classifies chlordane as possibly carcinogenic to humans (group 2b), with inadequate evidence in humans but sufficient evidence in experimental animals for carcinogenicity (153). In similarity with oxychlordane and trans-nonachlor, results of previous studies which measured HCB exposure did not entirely agree with the results of this study. A case-control study in Sweden measuring HCB concentrations in blood and adipose tissue samples found a significant elevated risk of NHL (97), while the two American studies measuring HCB in blood did not find an elevated risk (95, 96). Another study which measured adipose tissue HCB concentrations in cadavers by Quintana et al did not find a link with NHL (98). They showed a near-significant association with HCB, but found that the association weakened when adjusting for heptachlor epoxide, a metabolite of heptachlor. Like chlordane, HCB is also classified as possibly carcinogenic to humans (group 2b) by IARC (154). P-HCCH was found to be significantly associated with NHL risk, but was not considered significant when multiple comparisons were taken into account, under the FDR method. This organochlorine was not found to be associated with NHL risk in two other studies that measured serum organochlorine concentrations (95, 96). The study by Quintana et al (98) did find an increased odds of NHL for increasing P-HCCH exposure. However, -133 -methodological flaws in this study from using post-diagnosis and post-mortem adipose tissue samples preclude useful comparison of results. Results from this study confirm that p,p'-DDT does not modify the risk of NHL. However, p,p'-DDE was found to increase risk of NHL. Two other studies in the US by Rothman et al (95) and De Roos et al (96) found that blood concentrations of DDT and DDE were not associated with NHL risk. The study by Rothman et al (94) measured the compounds o,p'-DDT, p,p'-DDT, o,p'-DDE, and p,p'-DDE, and analyzed them together as DDT-related compounds. The study by De Roos et al (96), similar to this study, measured p,p'-DDT and p,p'-DDE and neither was found to be associated. Quintana et al (98) also showed an association with DDT that weakened when adjusting for heptachlor epoxide. A Canadian case-control study on pesticide exposure found that DDT exposure was significantly linked to NHL (84), while a Swedish case-control study did not (81). A study measuring organochlorines in carpet dust as an indicator for exposure (146) found an elevated risk of NHL for DDE exposure in men, as well as a significant dose-response for DDE. Limitations in the assessments of exposure in theses studies and the finding of small and inconsistent excesses across all the studies measuring DDT/ DDE preclude an assessment for risk. The use of biological samples to assess DDT/ DDE exposure is presumably unbiased and more accurate than use of questionnaire data. Even though previous studies using biological markers of exposure did not detect an association with DDE, the dose-response observed in this study was significant even after adjusting for multiple comparisons. However, individual odds ratio estimates were not significant. It is possible that p,p'-DDE has a very small effect on NHL risk but the studies thus far have not had the power to detect it. - 134 -Even though p,p'-DDT was not associated with NHL risk, a significant interaction with ethnicity was observed in this study. Subjects who did not classify themselves as European had a decreased odds of NHL with detectable levels of p,p'-DDT, while the risk was not increased in Europeans. This protective effect is in contrast to the study hypothesis and to the significant linear dose-response observed for p,p'-DDE, which increased NHL risk. The apparent protective effect of p,p'-DDT is most likely a false positive finding because the group of individuals came from different ethnic backgrounds, including those of mixed ethnicities. They were grouped for analysis because of small numbers, but may not have similar cultural background to offer explanation for a protective effect of p,p'-DDT. As well, a number of tests of interaction between confounder variables and organochlorine variables were conducted, increasing the likelihood of false-positives. P-HCCH, DDT, and DDE are all classified as possibly carcinogenic to humans (group 2B) by IARC, with inadequate evidence for carcinogenicity in humans but sufficient evidence in animals. There was little human data available for determining carcinogenicity of P-HCCH in the IARC report, and the information that was available was based on small numbers of cases (150). For DDT/ DDE, IARC notes that there is some epidemiological evidence for a slight increase in lung cancer, and studies of pesticide exposure have also suggested a link between DDT use and specific cancers, including NHL. Inconsistencies between results of this study and previous studies may be due to differences in study design, differences in population, and previous studies being underpowered. The chance of false positive results in this study is low, since many of the odds ratios and tests forHrend would have been significant even at the p=0.01 level. Also, - 135-many of the significant results remained the same even after application of the FDR procedure for multiple testing. Plasma concentrations of oxychlordane had the strongest effect on NHL risk than any other organochlorine compound in this population. When adjusting for oxychlordane, only one other summary PCB variable in this study contributed to NHL risk, suggesting that plasma oxychlordane concentrations represent a large proportion of the effect on NHL risk. Oxychlordane was also frequently detected in samples and correlated with other organochlorines. It is well established that chlordane and chlordane-related products are carcinogenic in animals (153). Interestingly, oxychlordane is the most toxic of chlordane and chlordane-related compounds in rats (155). This means that consumption of chlordane residues in food result in formation of a metabolite in the body that is even more potent than the parent compound. Studies in rats have also shown that the relative order of toxicity of chlordane-related compounds begins with oxychlordane as having the highest toxicity, followed by trans-nonachlor, then technical chlordane, and lastly cis-nonachlor (155, 156). The results in this study somewhat approximates this order, oxychlordane had the highest effect, followed by trans-nonachlor, and finally cis-nonachlor, which did not have a statistically significant effect. This consistency provides additional support that the associations observed in this study are not false positives and has a biological basis. The order of toxicity observed in rats and the order of increasing effect on NHL risk observed in this study may relate to the degradative properties and bioaccumulative potential of the compounds. Oxychlordane and trans-nonachlor have the highest bioaccumulative potential of chlordane products in both animals and humans, and constitute a large percentage of chlordane products measured in biological samples (155). In this study, - 1 3 6 -oxychlordane and trans-nonachlor were detected in over 95% of samples, while cis-nonachlor was detected in only about 25% of samples. This study is the only one to date which investigated the relationship between plasma concentrations of the pesticide mirex and NHL. The US studies by Cantor et al (95) and De Roos et al (96) measured mirex, but it was only detected in a small number of samples and not included in analyses. A significant relationship was found in this study. Subjects with detectable amounts of mirex in their blood had a 39% increased odds of developing NHL than those without detectable levels in their blood. Mirex is classified as a possible carcinogen to humans (group 2b) by IARC, with sufficient evidence as a carcinogen in experimental animals but there is a lack of data available to determine carcinogenicity in humans (157). Although this evaluation was made in 1987, there is still presently an absence of data to assess the effects of mirex in humans. Population studies of pesticide exposure and human health effects have not included mirex as a pesticide of interest, despite it being a major contaminant in human tissues for those living in the Great Lakes region (143), which constitute a large proportion of the population in both Canada and the US. One Canadian study looking at the association between organochlorines and risk of breast cancer did measure adipose tissue concentrations of mirex (158). They found that mirex was not significantly associated with breast cancer risk in pre- or post-menopausal women. However, an interaction was seen with lactation status in post-menopausal women, where women who never lactated and had the highest exposure to mirex were at an increased risk of breast cancer. These results along with those of the present study suggest that mirex may play a role in carcinogenesis, although this role would have to be confirmed in other studies. - 137-Combinations of organochlorines may not provide additional information on NHL risk, due to the high correlation between organochlorine concentrations. In a stepwise selection procedure, a combination of oxychlordane and summed non dioxin-like PCBs both contributed to increased NHL risk. Few variables were included in the final model due to the high correlation of the concentrations. Oxychlordane had the strongest effect of an individual organochlorine, and along with non dioxin-like PCB exposure, may be representative of overall organochlorine exposure and its effect on NHL risk. Oxychlordane was also representative of overall pesticide exposure. In a stepwise selection procedure with the PCB exposure variables, the final model included PCB 153. PCB 153 was the most abundant PCB congener detected in this study, and concentrations were highly correlated with other PCBs. Associations between organochlorines and subgroups of NHL were determined in separate analyses. These analyses are limited by the small numbers of cases within subgroups, so results should be regarded as exploratory and not conclusive. As well, multiple comparisons may increase the number of false positives detected. Overall, there were similar results for the subgroup analyses. However, notable observations are that some organochlorines had a stronger association with other B-cell types and follicular lymphoma. For the other B-cell subgroup, there were stronger associations observed for the summary PCB measures, as well as for individual non dioxin-like PCB congeners 170, 180 and 187. The other B-cell category contains a number of different subtypes, including MALT, CLL/SLL, mantle cell and lymphoplasmacytic lymphomas. These subtypes were grouped together for analysis due to small numbers of cases within each subtype. It is difficult to assess whether the increased risks observed for this group have a biological basis or is due to - 138-random chance. These results will have to be confirmed by studies powered to look at risks within NHL subtypes. In the follicular subtype, there was an increase in the odds ratios for the highest exposure category versus the lowest exposure category by at least 10% for PCB 156, HCB, and oxychlordane. For oxychlordane, there was an over 3-fold risk of NHL for those with the highest plasma concentrations. One of the hallmarks of follicular lymphomas is the 14; 18 translocation, which has been found in up to 100% of follicular lymphoma patients (159). This translocation is believed to be an important part of carcinogenesis for the subtype. In t(14;18) follicular lymphomas, the B-cell leukemia/lymphoma 2 (BCL2) gene on chromosome 18q21 is juxtaposed near chromosome 14q32, the location of the immunoglobulin heavy chain (IGH) locus. The translocation mechanism is believed to involve V(D)J recombination (160). The translocated BCL2 gene then becomes under control of the IGH-Ep enhancer and becomes overexpressed (161). Overexpression of the anti-apoptotic BCL2 gene then results in increased cell survival, a hallmark of cancer cells. Epidemiological studies have found that the 14; 18 translocation is associated with occupational exposure to pesticides. In a study by Roulland et al (162), t(14;18) lymphocytes characterized by the presence of BCL2-IGH DNA was more frequently found in farmers with higher pesticide use. Another study found that organochlorine insecticides, lindane, dieldrin, and toxaphene in particular, were significantly associated with risk of t(14;18) positive NHL, but no association was seen in t(14;18) negative NHL cases (163). Chlordane exposure was also assessed in the study but no relationship was found. Nevertheless, these studies suggest a link between organochlorine exposure and development of 14; 18 translocations and the possibility that organochlorines are responsible for follicular - 1 3 9 -lymphomagenesis in particular. The increase in risk estimates for some organochlorines in the follicular lymphoma subgroup provide some support for this hypothesis, although it is unclear why the increase was only observed for specific organochlorines. Analyzing AHR variants provided additional support for the association between organochlorine exposure and NHL risk in this study. Measuring genetic association in epidemiological studies can provide many benefits: Accounting for gene-environment interactions can give better risk estimates for disease; genetic associations can provide insight into etiology of disease; and genetic associations are not affected by biases that plague epidemiological studies involving environmental or behavioural risk factors. Epidemiological studies involving genetic risk factors can help to confirm studies of environmental risk factors if the environmental exposure is influenced by a genetic trait. For example, organophosphate pesticides such as sheep dip, were suspected in causing neurological damage in farmers from chronic use. However, a report from the Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (COT) in the UK in 1999 determined that there was insufficient epidemiological evidence to support the link between prolonged low dose exposure and neurological effects (164). Their critiques of the epidemiological literature included confounding bias from other behavioral or environmental factors, and exposure measurement bias. Subsequently, polymorphisms in the paraoxonase (PON1) gene responsible for metabolism and detoxification of organophosphates were found to be associated with health effects that were reported to be from organophosphate exposure. Farmers who were ill were more likely to carry a genotype that conferred a less efficient paraoxonase (165). It is unlikely that the genotype is subject to the same confounding and exposure assessment biases that previous epidemiological studies were, and the findings - 1 4 0 -provide evidence that chronic exposure to organophosphates does have a causal effect on health. Similarly, polymorphism in genes metabolizing organochlorines, including AHR, can help to determine the validity of the link of organochlorine exposure to increased NHL risk. Seven AHR variants were genotyped in this study. For all variants, the genotype frequencies observed in control samples were similar to those expected under Hardy-Weinberg equilibrium. It is unlikely that population substructure, biased sampling of individuals or genotyping error affected results. Genotype frequencies for the 7 variants were found to differ by ethnicity, but did not differ by age, sex, region of residence or education level. To account for this difference by ethnicity, interaction between the SNPs and ethnicity were tested, and all analyses were conducted in the full population and only within Europeans for comparison. No significant interactions were observed. Results were consistent in the two analyses for all variants, providing assurance that significant associations observed were not due to ethnic differences in genotype frequencies. There were significant interactions found between AHR SNPs and sex and region of residence. Background genetic variation may account for differences in NHL risk between males and females for the IVS1-3946G/A and IVS7+33G/T SNPs. These SNPs were not found to be associated with NHL risk, but significant associations were observed for females. It is possible that these interactions may be chance findings because multiple tests of interaction were conducted, and the p-values were close to 0.05. Interaction was also found between the two pairs of SNPs in LD and region of residence. The X1-459G/A and IVS1+1205C/G SNPs were associated with a decreased risk of NHL in the CRD. The IVS7+33G/T and the XI0+1661 G/A SNPs were associated with decreased NHL risk in the GVRD, but an increased risk in the CRD. However, no interactions were observed between -141 -these SNPs and age or ethnicity, so the differences in risk between the two regions cannot be due to these factors. Other demographic and lifestyle factors are assumed to be very similar between the two regions. So there is no clear explanation for why these AHR SNPs would give different risks for different places of residences. The intronic IVS1+4640G/A SNP was the only variant found to be significantly associated with NHL risk in this study. The minor allele was linked to an increase in NHL risk. Different risk estimates for dioxin-like PCB congener 118, and chlordane-related products oxychlordane and trans-nonachlor exposure were found for categories of this SNP. Major allele homozygotes (G/G genotype) with the highest exposure to these compounds had a higher risk of NHL compared to those with the lowest exposure. But no increased risk was seen in the heterozygote G/A and minor allele homozygote A/A genotypes. In other words, the G/A and A/A genotypes were associated with an increased risk compared to the G/G genotype at lower organochlorine exposure, but the G/G genotype conferred a higher risk at higher organochlorine exposures. PCB 118 is known to bind AhR and was hypothesized a priori to interact with AHR. It is unknown whether chlordane-related products bind AhR and activate the pathway. According to the Comparative Toxigenomics Database (166), no studies have looked at the binding or activation of AhR by oxychlordane and nonachlor. The only genetic study done with oxychlordane involves activation of the androgen receptor, and oxychlordane was found to have no effect (167). Nonachlor has been documented to bind or affect the activity of several proteins, including estrogen and progesterone receptors (168), the pregnane X receptor (169), and cytochrome P450 enzymes CYP2B6 and CYP3A3 (169). Results from this study suggest that these chlordane-related products may also be involved with binding or activation of the AhR. - 142 -This is the first epidemiological study to look at disease associations to the IVS1+4640G/A SNP, so no information is available on its biological effects. Also, because the SNP is not part of the coding sequence of the gene, any possible functional changes in the AhR protein cannot be predicted. Intronic SNPs can be causally linked to disease if they are involved in regulation of the gene or if they affect splicing of the gene. The IVS1+4640G/A SNP may be involved in regulation of the gene, but is unlikely to affect splicing due to its distant location from intron/ exon junctions where splice sites are found. It is also possible that the F/S1+4640G/A SNP is not the causal variant and is in linkage disequilibrium with a causal SNP, particularly because it is a tag-SNP. From the FfapMap website (124), the IVS1+4640G/A SNP is not in LD (based on 0.5C) of DNA repair gene XRCC1 is associated with risk of lung cancer in relation to tobacco smoking. Pharmacogenet Genomics 2005;15:457-63. 107. Ahmed FE. Gene-gene, gene-environment & multiple interactions in colorectal cancer. J Environ Sci Health C Environ Carcinog Ecotoxicol Rev 2006;24:1-101. 108. Pachkowski BF, Winkel S, Kubota Y, Swenberg JA, Millikan RC, Nakamura J. XRCC1 genotype and breast cancer: functional studies and epidemiologic data show interactions between XRCC1 codon 280 His and smoking. Cancer Res 2006;66:2860-8. 109. Hestermann EV, Stegeman JJ, Hahn ME. Relative contributions of affinity and intrinsic efficacy to aryl hydrocarbon receptor ligand potency. Toxicol Appl Pharmacol 2000;168:160-72. 110. Vezina CM, Walker NJ, Olson JR. Subchronic exposure to TCDD, PeCDF, PCB 126, and PCB 153: effect on hepatic gene expression. Environ Health Perspect 2004;112:1636-44. 111. Hestermann EV, Stegeman JJ, Hahn ME. Relationships among the cell cycle, cell proliferation, and aryl hydrocarbon receptor expression in PLHC-1 cells. Aquat Toxicol 2002;58:201-13. 112. De Roos AJ, Gold LS, Wang S, et al. Metabolic Gene Variants and Risk of Non-Hodgkin's Lymphoma. Cancer Epidemiol Biomarkers Prev 2006;15:1647-1653. 113. Jaffe ES, Harris, N. L., Stein, H., and Vardiman, J. W. World Health Organization Classification of Tumours: Pathology and Genetics of Tumours of Haematopoietic and Lymphoid Tissues. Lyon: IARC Press, 2001. 114. Chevrier J, Dewailly E, Ayotte P, Mauriege P, Despres JP, Tremblay A. Body weight loss increases plasma and adipose tissue concentrations of potentially toxic pollutants in obese individuals. Int J Obes Relat Metab Disord 2000;24:1272-8. 115. Patterson DG, Jr., Isaacs SG, Alexander LR, et al. Determination of specific polychlorinated dibenzo-p-dioxins and dibenzofurans in blood and adipose tissue by isotope dilution-high-resolution mass spectrometry. IARC Sci Publ 1991:299-342. 116. Fleiss JL. The Design and Analysis of Clinical Experiments. Toronto: John Wiley & Sons, Inc., 1986:9-11. 117. Ewing B, Green P. Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Res 1998;8:186-94. 118. Ewing B, Hillier L, Wendl MC, Green P. Base-calling of automated sequencer traces using phred. I. Accuracy assessment. Genome Res 1998;8:175-85. - 1 6 4 -119. Gordon D, Abajian C, Green P. Consed: a graphical tool for sequence finishing. Genome Res 1998;8:195-202. 120. Nickerson DA, Tobe VO, Taylor SL. PolyPhred: automating the detection and genotyping of single nucleotide substitutions using fluorescence-based resequencing. Nucleic Acids Res 1997;25:2745-51. 121. Ng PC, Henikoff S. Predicting deleterious amino acid substitutions. Genome Res 2001;11:863-74. 122. Ng PC, Henikoff S. SIFT: Predicting amino acid changes that affect protein function. Nucleic Acids Res 2003;31:3812-4. 123. Ramensky V, Bork P, Sunyaev S. Human non-synonymous SNPs: server and survey. Nucleic Acids Res 2002;30:3894-900. 124. Thorisson GA, Smith AV, Krishnan L, Stein LD. The International HapMap Project Web site. Genome Res 2005;15:1592-3. 125. SPSS Inc. Statistical Package for the Social Sciences Version 13.0 (SPSS 13.0). Chicago: SPSS Inc., 2004. 126. SAS Institute Inc. Statistical Analysis System 9.1.2 (SAS 9.1.2). Cary, NC, USA, 2004. 127. Hornung RW, Reed LD. Estimation of average concentration in the presence of non-detectable values. Applied Occupational and Environmental Hygiene 1990;5:46-51. 128. Rothman KJ, Greenland S. Modern Epidemiology. Philadelphia: Lippincott-Raven, 1998:256-257. 129. Greenland S, Finkle WD. A critical look at methods for handling missing covariates in epidemiologic regression analyses. Am J Epidemiol 1995;142:1255-64. 130. Davey Smith G, Hart C, Hole D, et al. Education and occupational social class: which is the more important indicator of mortality risk? J Epidemiol Community Health 1998;52:153-60. 131. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological) 1995;57:289-300. 132. Barrett JC, Fry B, Mailer J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005;21:263-5. 133. Burkett K, Graham J, McNeney B. Hapassoc: Software for likelihood inference of trait associations with SNP haplotypes and other attributes. Journal of Statistical Software 2006;16:1-19. - 165 -134. Burkett K, McNeney B, Graham J. A note on inference of trait associations with SNP haplotypes and other attributes in generalized linear models. Hum Hered 2004;57:200-6. 135. R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2006. 136. Tsuji LJ, Wainman BC, Martin ID, Weber IP, Sutherland C, Nieboer E. Abandoned Mid-Canada Radar Line sites in the Western James region of Northern Ontario, Canada: A source of organochlorines for First Nations people? Sci Total Environ 2006;370:452-66. 137. Health Canada. Concentrations (ppb) of pesticide residues in foods from Total Diet Study in Vancouver, 1995, 1995. 138. Health Canada. Concentrations (ppb) of pesticide residues in foods from Total Diet Study in Toronto, 1996, 1996. 139. Cole DC, Sheeshka J, Murkin EJ, et al. Dietary intakes and plasma organochlorine contaminant levels among Great Lakes fish eaters. Arch Environ Health 2002;57:496-509. 140. Kearney JP, Cole DC, Ferron LA, Weber JP. Blood PCB, p,p'-DDE, and mirex levels in Great Lakes fish and waterfowl consumers in two Ontario communities. Environ Res 1999;80:S138-S149. 141. Turyk M, Anderson HA, Hanrahan LP, et al. Relationship of serum levels of individual PCB, dioxin, and furan congeners and DDE with Great Lakes sport-caught fish consumption. Environ Res 2006;100:173-83. 142. Fiore BJ, Anderson HA, Hanrahan LP, Olson LJ, Sonzogni WC. Sport fish consumption and body burden levels of chlorinated hydrocarbons: a study of Wisconsin anglers. Arch Environ Health 1989;44:82-8. 143. Ontario Ministry of the Environment. Guide to Eating Ontario Sport Fish 2005-2006: Twenty-third Edition, Revised: Queen's Printer for Ontario, 2005:15-17. 144. Committee on Remediation of PCB-Contaminated Sediments, Board on Environmental Studies and Toxicology, Division on Life and Earth Studies, National Research Council. Appendix E. PCB Biodegradation. A risk-management strategy for PCB-contaminated sediments. Washington, D.C: The National Academy Press, 2001:329. 145. Hansen LG. Stepping backward to improve assessment of PCB congener toxicities. Environ Health Perspect 1998;106 Suppl 1:171-89. 146. Colt JS, Severson RK, Lubin J, et al. Organochlorines in carpet dust and non-Hodgkin lymphoma. Epidemiology 2005;16:516-25. - 166 -147. Ruder AM, Hein MJ, Nilsen N, et al. Mortality among workers exposed to polychlorinated biphenyls (PCBs) in an electrical capacitor manufacturing plant in Indiana: an update. Environ Health Perspect 2006;114:18-23. 148. Yassi A, Tate R, Fish D. Cancer mortality in workers employed at a transformer manufacturing plant. Am J Ind Med 1994;25:425-37. 149. International Agency for Research on Cancer. Volume 69: Polychlorinated Dibenzo-para-Dioxins and Polychlorinated Dibenzofurans. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans: World Health Organization, 1997. 150. International Agency for Research on Cancer. Overall Evaluations of Carcinogenicity: An Updating of IARC Monographs Volumes 1 to 42. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, 1987:104. 151. Colt JS, Davis S, Severson RK, et al. Residential insecticide use and risk of non-Hodgkin's lymphoma. Cancer Epidemiol Biomarkers Prev 2006;15:251-7. 152. De Roos AJ, Zahm SH, Cantor KP, et al. Integrative assessment of multiple pesticides as risk factors for non-Hodgkin's lymphoma among men. Occup Environ Med 2003;60:E11. 153. International Agency for Research on Cancer. Volume 79: Some Thyrotropic Agents, Summary of Data Reported and Evaluation. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, 2001:27-30. 154. International Agency for Research on Cancer. Volume 79: Some Thyrotropic Agents, Summary of Data Reported and Evaluation. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, 2001:31-33. 155. Bondy G, Armstrong C, Coady L, et al. Toxicity of the chlordane metabolite oxychlordane in female rats: clinical and histopathological changes. Food Chem Toxicol 2003;41:291-301. 156. Bondy GS, Newsome WH, Armstrong CL, et al. trans-Nonachlor and cis-nonachlor toxicity in Sprague-Dawley rats: comparison with technical chlordane. Toxicol Sci 2000;58:386-98. 157. International Agency for Research on Cancer. Volume 20: Some Halogenated Hydrocarbons, Summary of Data Reported and Evaluation. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, 1979:11-12. 158. Aronson KJ, Miller AB, Woolcott CG, et al. Breast adipose tissue concentrations of polychlorinated biphenyls and other organochlorines and breast cancer risk. Cancer Epidemiol Biomarkers Prev 2000;9:55-63. - 167-159. Godon A, Moreau A, Talmant P, et al. Is t(14;18)(q32;q21) a constant finding in follicular lymphoma? An interphase FISH study on 63 patients. Leukemia 2003;17:255-9. 160. Tsujimoto Y, Gorham J, Cossman J, Jaffe E, Croce CM. The t(14;18) chromosome translocations involved in B-cell neoplasms result from mistakes in VDJ joining. Science 1985;229:1390-3. 161. Kuppers R, Dalla-Favera R. Mechanisms of chromosomal translocations in B cell lymphomas. Oncogene 2001;20:5580-94. 162. Roulland S, Lebailly P, Lecluse Y, Briand M, Pottier D, Gauduchon P. Characterization of the t(14;18) BCL2-IGH translocation in farmers occupationally exposed to pesticides. Cancer Res 2004;64:2264-9. 163. Schroeder JC, Olshan AF, Baric R, et al. Agricultural risk factors for t(14;18) subtypes of non-Hodgkin's lymphoma. Epidemiology 2001;12:701-9. 164. The Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (COT). Organophosphates, 1999:Appendix 4: Summaries and critiques of the epidemiological literature. 165. Cherry N, Mackness M, Durrington P, et al. Paraoxonase (PON1) polymorphisms in farmers attributing ill health to sheep dip. Lancet 2002;359:763-4. 166. Mattingly CJ, Rosenstein MC, Davis AP, Colby GT, Forrest IN, Jr., Boyer JL. The comparative toxicogenomics database: a cross-species resource for building chemical-gene interaction networks. Toxicol Sci 2006;92:587-95. 167. Schrader TJ, Cooke GM. Examination of selected food additives and organochlorine food contaminants for androgenic activity in vitro. Toxicol Sci 2000;53:278-88. 168. Scippo ML, Argiris C, Van De Weerdt C, et al. Recombinant human estrogen, androgen and progesterone receptors for detection of potential endocrine disruptors. Anal Bioanal Chem 2004;378:664-9. 169. Lemaire G, de Sousa G, Rahmani R. A PXR reporter gene assay in a stable cell culture system: CYP3A4 and CYP2B6 induction by pesticides. Biochem Pharmacol 2004;68:2347-58. 170. Ronmark E, Lundqvist A, Lundback B, Nystrom L. Non-responders to a postal questionnaire on respiratory symptoms and diseases. Eur J Epidemiol 1999;15:293-9. 171. Pinard R, de Winter A, Sarkis GJ, et al. Assessment of whole genome amplification-induced bias through high-throughput, massively parallel whole genome sequencing. BMC Genomics 2006;7:216. - 1 6 8 -172. Keohavong P, Thilly WG. Fidelity of DNA polymerases in DNA amplification. Proc Natl Acad Sci U S A 1989;86:9253-7. 173. Szklo M, Nieto FJ. Epidemiology: Beyond the Basics. Sudbury: Jones and Bartlett Publishers, 2004. 174. Dallaire F, Dewailly E, Muckle G, Ayotte P. Time trends of persistent organic pollutants and heavy metals in umbilical cord blood of Inuit infants born in Nunavik (Quebec, Canada) between 1994 and 2001. Environ Health Perspect 2003; 111:1660-4. 175. Furst P. Dioxins, polychlorinated biphenyls and other organohalogen compounds in human milk. Levels, correlations, trends and exposure through breastfeeding. Mol Nutr Food Res 2006;50:922-33. 176. Sandin S, Hjalgrim H, Glimelius B, Rostgaard K, Pukkala E, Askling J. Incidence of non-Hodgkin's lymphoma in Sweden, Denmark, and Finland from 1960 through 2003: an epidemic that was. Cancer Epidemiol Biomarkers Prev 2006;15:1295-300. 177. Hardell L, Eriksson M. Is the decline of the increasing incidence of non-Hodgkin lymphoma in Sweden and other countries a result of cancer preventive measures? Environ Health Perspect 2003; 111:1704-6. 178. BC Cancer Agency. BC Incidence Trends By Cancer Type, Year of Diagnosis and Gender, Diagnosis Years 1970 to 2004, 2006. 179. Wolff MS, Britton JA, Teitelbaum SL, et al. Improving organochlorine biomarker models for cancer research. Cancer Epidemiol Biomarkers Prev 2005;14:2224-36. 180. Juan CY, Thomas GO, Sweetman AJ, Jones KC. An input-output balance study for PCBs in humans. Environ Int 2002;28:203-14. 181. Patterson DG, Jr., Todd GD, Turner WE, Maggio V, Alexander LR, Needham LL. Levels of non-ortho-substituted (coplanar), mono- and di-ortho-substituted polychlorinated biphenyls, dibenzo-p-dioxins, and dibenzofurans in human serum and adipose tissue. Environ Health Perspect 1994;102 Suppl 1:195-204. 182. Covaci A, Koppen G, Van Cleuvenbergen R, et al. Persistent organochlorine pollutants in human serum of 50-65 years old women in the Flanders Environmental and Health Study (FLEHS). Part 2: Correlations among PCBs, PCDD/PCDFs and the use of predictive markers. Chemosphere 2002;48:827-32. 183. Puga A, Xia Y, Elferink C. Role of the aryl hydrocarbon receptor in cell cycle regulation. Chem Biol Interact 2002;141:117-30. 184. Marlowe JL, Puga A. Aryl hydrocarbon receptor, cell cycle regulation, toxicity, and tumorigenesis. J Cell Biochem 2005;96:1174-84. -169-185. Koul HK, Maroni PD, Meacham RB, Crawford D, Koul S. p42/p44 Mitogen-activated protein kinase signal transduction pathway: a novel target for the treatment of hormone-resistant prostate cancer? Ann N Y Acad Sci 2004;1030:243-52. 186. Santen RJ, Song RX, McPherson R, et al. The role of mitogen-activated protein (MAP) kinase in breast cancer. J Steroid Biochem Mol Biol 2002;80:239-56. 187. Lengyel E, Sawada K, Salgia R. Tyrosine kinase mutations in human cancer. Curr Mol Med 2007;7:77-84. 188. Esfandiary H, Chakravarthy U, Patterson C, Young I, Hughes AE. Association study of detoxification genes in age related macular degeneration. Br J Ophthalmol 2005;89:470-4. 189. van der Spuy J, Kim JH, Yu YS, et al. The expression of the Leber congenital amaurosis protein AIPL1 coincides with rod and cone photoreceptor development. Invest Ophthalmol Vis Sci 2003;44:5396-403. 190. Allan LL, Sherr DH. Constitutive activation and environmental chemical induction of the aryl hydrocarbon receptor/transcription factor in activated human B lymphocytes. Mol Pharmacol 2005;67:1740-50. 191. Denison MS, Nagy SR. Activation of the aryl hydrocarbon receptor by structurally diverse exogenous and endogenous chemicals. Annu Rev Pharmacol Toxicol 2003;43:309-34. 192. Ross P, De Swart R, Addison R, Van Loveren H, Vos J, Osterhaus A. Contaminant-induced immunotoxicity in harbour seals: wildlife at risk? Toxicology 1996;112:157-69. 193. De Guise S, Martineau D, Beland P, Fournier M. Possible mechanisms of action of environmental contaminants on St. Lawrence beluga whales (Delphinapterus leucas). Environ Health Perspect 1995;103 Suppl 4:73-7. 194. De Krey GK, Baecher-Steppan L, Fowles JR, Kerkvliet NI. Polychlorinated biphenyl-induced suppression of cytotoxic T lymphocyte activity: role of prostaglandin .^ Toxicol Lett 1994;74:211-20. 195. Kerkvliet NI, Baecher-Steppan L, Smith BB, Youngberg JA, Henderson MC, Buhler DR. Role of the Ah locus in suppression of cytotoxic T lymphocyte activity by halogenated aromatic hydrocarbons (PCBs and TCDD): structure-activity relationships and effects in C57B1/6 mice congenic at the Ah locus. Fundam Appl Toxicol 1990;14:532-41. 196. Van Den Heuvel RL, Koppen G, Staessen JA, et al. Immunologic biomarkers in relation to exposure markers of PCBs and dioxins in Flemish adolescents (Belgium). Environ Health Perspect 2002; 110:595-600. - 170-197. Maher J, Davies ET. Targeting cytotoxic T lymphocytes for cancer immunotherapy. Br J Cancer 2004;91:817-21. 198. Oldham RK. Natural killer cells: artifact to reality: an odyssey in biology. Cancer Metastasis Rev 1983;2:323-36. 199. Kerkvliet NI, Shepherd DM, Baecher-Steppan L. T lymphocytes are direct, aryl hydrocarbon receptor (AhR)-dependent targets of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD): AhR expression in both CD4+ and CD8+ T cells is necessary for full suppression of a cytotoxic T lymphocyte response by TCDD. Toxicol Appl Pharmacol 2002;185:146-52. 200. Silkworth JB, Antrim LA, Sack G. Ah receptor mediated suppression of the antibody response in mice is primarily dependent on the Ah phenotype of lymphoid tissue. Toxicol Appl Pharmacol 1986;86:380-90. 201. Silkworth JB, Antrim L, Kaminsky LS. Correlations between polychlorinated biphenyl immunotoxicity, the aromatic hydrocarbon locus, and liver microsomal enzyme induction in C57BL/6 and DBA/2 mice. Toxicol Appl Pharmacol 1984;75:156-65. 202. Kerkvliet NI, Steppan LB, Brauner JA, et al. Influence of the Ah locus on the humoral immunotoxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin: evidence for Ah-receptor-dependent and Ah-receptor-independent mechanisms of immunosuppression. Toxicol Appl Pharmacol 1990;105:26-36. 203. Johnson CD, Balagurunathan Y, Tadesse MG, et al. Unraveling gene-gene interactions regulated by ligands of the aryl hydrocarbon receptor. Environ Health Perspect 2004;112:403-12. -171 -