University of Southern California Dissertations and Theses

Genetic and environmental risk factors for childhood cancer

(USC Thesis Other)

Genetic and environmental risk factors for childhood cancer

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content GENETIC AND ENVIRONMENTAL RISK FACTORS FOR CHILDHOOD CANCER by Jessica L. Barrington-Trimis A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements of the Degree DOCTOR OF PHILOSOPHY (EPIDEMIOLOGY) August 2014 Copyright 2014 Jessica L. Barrington-Trimis ii Epigraph “ I f y ou c a n be li e v e it , y o u c a n a c hieve it ” -Ed Trimis ii iii Acknowledgements I would like to thank Dr. Roberta McKean-Cowdin, for her guidance, leadership, and inspiration - for the endless hours she spent reading and editing my papers, for the many times she met with me to give me career and life advice, for always believing in me and inspiring me to apply for everything and never give up on anything. I would like to thank Dr. Jim Gauderman for all the opportunities he has given me, both in a nd out of the c lassr oom, for a lwa y s a nswe r ing that ‘sta ti sti c a l que sti on I ha ve fo r y o u…’, a nd fo r insti ll ing i n me a love of biost a ti sti c s. I would like to thank the rest of my dissertation committee, Dr. Duncan Thomas, Dr. Mariana Stern, and Dr. Shahab Asgharzadeh for their guidance and support in the completion of my dissertation work. I would like to thank Dr. Susan Searles Nielsen, Dr. Susan Preston-Martin, Dr. Jim Gauderman, Dr. Elizabeth A. Holly, Dr. Federico M. Farin, Dr. Beth A. Mueller, and Dr. Roberta McKean-Cowdin, the co- a uthors on m y manusc ript ti tl e d “ P arental Smoking and Risk of Childhood Brain Tumors by Functional Polymorphisms in Polycyclic Aromatic Hydrocarbon Metabolism Genes ”. I would like to thank Dr. Beth Mueller, Dr. Myles Cockburn, Dr. Joe Wiemels, and Dr. Roberta McKean-Cowdin, the co- a uthors on my manusc ript “ Rising Rates of Acute iii iv Lymphocytic Leukemia in Hispanic Children: a Review of Trends in Childhood Leukemia Incidence from 1973-2010 ”. I would like to thank Dr. Jim Gauderman, Dr. Catherine Metayer, Dr. Lisa Barcellos, Dr. Yang Wang, Dr. Joe Wiemels, Dr. Jacqueline Clavel, Dr. Laurent Orsi, Dr. Jeremie Rudant, Dr. Denise Anderson, Dr. Elizabeth Milne, and Dr. Roberta McKean-Cowdin, the co- a uthors on m y m a nusc ript “ Gene-Smoking Interactions and Risk of Childhood Acute Lymphocytic Leukemia among Hispanic children in a Genome-Wide Association Study ”. I would like to thank Dr. Myles Cockburn, Dr. Jim Gauderman, Dr. Catherine Metayer, and Dr. Roberta McKean-Cowdin, who helped draft our NIH R21 grant. I would like to thank Athena Foong for her help with formatting my dissertation, and for many years of friendship. I would like to thank my colleagues, especially Tara Kerin and Alice Lee for their continued support, and the many other students and faculty at USC who have been an inspiration to me throughout this process. I would like to thank my parents, Ed and Moira Trimis, for instilling in me the true value of education and for inspiring my love of learning, and my siblings, Nick, Stephanie, Olivia and Alexander Trimis, for their love and support. iv v Finally, I would like to thank my husband, Conrad Barrington, for his tremendous support over the many, many years I have spent as a student (with little complaint about the long hours I spent working), for always believing in me and never doubting my ability to achieve my goals, and for constantly grounding me and helping me to see those things most important in life. v vi Table of Contents Epigraph ............................................................................................................................ ii Acknowledgements .......................................................................................................... iii List of Tables .................................................................................................................... ix List of Figures ................................................................................................................... xi Abstract ............................................................................................................................ xii CHAPTER 1: Background and Review .......................................................................... 1 1.1 CHILDHOOD BRAIN TUMORS ............................................................................ 1 1.1.1 INCIDENCE AND SURVIVAL ....................................................................... 1 1.1.2 PHENOTYPES .................................................................................................. 3 1.1.3 GENETIC AND ENVIRONMENTAL RISK FACTORS ................................ 4 1.2 CHILDHOOD LEUKEMIA ..................................................................................... 6 1.2.1 INCIDENCE AND SURVIVAL ....................................................................... 6 1.2.2 PHENOTYPES .................................................................................................. 8 1.2.3 GENETIC AND ENVIRONMENTAL RISK FACTORS .............................. 10 1.2.4 GENETIC SUSCEPTIBILITY TO TOBACCO SMOKE AND PESTICIDES ............................................................................................................ 12 1.3 TABLES ................................................................................................................. 14 1.4 REFERENCES ....................................................................................................... 17 CHAPTER 2: Parental Smoking and Risk of Childhood Brain Tumors by Functional Polymorphisms in Polycyclic Aromatic Hydrocarbon Metabolism Genes ................................................................................................................................ 20 2.1 ABSTRACT ............................................................................................................ 20 2.2 INTRODUCTION .................................................................................................. 21 2.3 MATERIALS AND METHODS ............................................................................ 23 2.4 RESULTS ............................................................................................................... 27 vi vii 2.5 DISCUSSION ......................................................................................................... 31 2.6 TABLES AND FIGURES ...................................................................................... 35 2.7 SUPPLEMENTARY MATERIAL ......................................................................... 42 2.8 REFERENCES ....................................................................................................... 45 CHAPTER 3: Rising Rates of Acute Lymphocytic Leukemia in Hispanic Children: a Review of Trends in Childhood Leukemia Incidence from 1973-2010 ....................49 3.1 ABSTRACT ............................................................................................................ 49 3.2 INTRODUCTION .................................................................................................. 50 3.3 MATERIALS AND METHODS ............................................................................ 52 3.4 RESULTS ............................................................................................................... 55 3.5 DISCUSSION ......................................................................................................... 60 3.6 TABLES AND FIGURES ...................................................................................... 69 3.7 REFERENCES ....................................................................................................... 77 CHAPTER 4: Gene-Smoking Interactions and Risk of Childhood Acute Lymphocytic Leukemia among Hispanic children in a Genome-Wide Association Study................................................................................................................................. 80 4.1 ABSTRACT ............................................................................................................ 80 4.2 INTRODUCTION .................................................................................................. 81 4.3 MATERIALS AND METHODS ............................................................................ 84 4.4 RESULTS ............................................................................................................... 90 4.5 DISCUSSION ......................................................................................................... 97 4.6 TABLES AND FIGURES .................................................................................... 102 4.7 SUPPLEMENTARY MATERIAL ....................................................................... 115 4.8 REFERENCES ..................................................................................................... 124 vii viii Chapter 5: National Institute of Health R21 Grant “Genome-Wide Interaction Scan (GWIS) Analysis of Pesticide Exposure and Risk of Childhood Acute Lymphocytic Leukemia (ALL)” .................................................................................. 127 5.1 NARRATIVE ....................................................................................................... 127 5.2 ABSTRACT .......................................................................................................... 127 5.3 SPECIFIC AIMS .................................................................................................. 129 5.4 RESEARCH STRATEGY .................................................................................... 132 5.5 REFERENCES: .................................................................................................... 152 viii ix List of Tables TABLE 1.1. AGE ADJUSTED INCIDENCE RATES (2005-2009) AND AGE-ADJUSTED MORTALITY RATES (AAMR) FOR CHILDHOOD BRAIN TUMORS AND CHILDHOOD LEUKEMIA ................................................................................................................. 14 TABLE 1.2. SUSPECTED RISK FACTORS FOR CHILDHOOD BRAIN TUMORS AND CHILDHOOD LEUKEMIA ............................................................................................. 15 TABLE 2.1. CHARACTERISTICS OF CANDIDATE POLYMORPHISMS IN POLYCYCLIC AROMATIC HYDROCARBON (PAH) METABOLISM GENES .......................................... 35 TABLE 2.2. DEMOGRAPHIC CHARACTERISTICS OF CHILDREN WITH AND WITHOUT BRAIN TUMORS, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, BORN 1978-1990 ......... 36 TABLE 2.3. RISK OF CHILDHOOD BRAIN TUMORS IN RELATION TO EXPOSURE TO PARENTAL SMOKING DURING PREGNANCY, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, BORN 1978-1990.............................................................................. 37 TABLE 2.4. RISK OF CHILDHOOD BRAIN TUMORS IN RELATION TO PATERNAL SMOKING DURING PREGNANCY BY PAH METABOLISM GENOTYPE, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, BORN 1978-1990 .................................................................. 38 TABLE 2.5. RISK OF CHILDHOOD BRAIN TUMORS IN RELATION TO MATERNAL SMOKING DURING PREGNANCY BY PAH METABOLISM GENOTYPE, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, BORN 1978-1990 .................................................................. 39 TABLE 2.6. RISK OF CHILDHOOD BRAIN TUMORS IN RELATION TO PATERNAL SMOKING LEVEL DURING PREGNANCY BY POLYMORPHISMS IN SELECTED GENES, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, BORN 1978-1990 .................................. 40 SUPPLEMENTAL TABLE 2.1. RISK OF CHILDHOOD BRAIN TUMORS IN RELATION TO POLYCYCLIC AROMATIC HYDROCARBON (PAH) METABOLISM POLYMORPHISMS, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, N=479......................................... 42 SUPPLEMENTAL TABLE 2.2. ASSOCIATION BETWEEN EXPOSURE TO PRENATAL PARENTAL SMOKING AND SELECTED POLYMORPHISMS IN A CASE-ONLY ANALYSIS, WEST COAST CHILDHOOD BRAIN TUMOR STUDY, N=196 ................................................... 43 SUPPLEMENTAL TABLE 2.3. RISK OF CHILDHOOD BRAIN TUMORS IN RELATION TO MATERNAL SMOKING LEVEL DURING PREGNANCY BY POLYMORPHISMS IN SELECTED GENES, WEST COAST CHILDHOOD BRAIN TUMOR STUDY ......................................... 44 TABLE 3.1. AGE-ADJUSTED INCIDENCE RATES (AAIR), ANNUAL PERCENT CHANGE (APC), AND RATE DIFFERENCE FOR ALL CHILDHOOD LEUKEMIAS BY ETHNICITY AND SELECTED DEMOGRAPHIC CHARACTERISTICS ................................................... 69 ix x TABLE 3.2. AGE-ADJUSTED INCIDENCE RATES (AAIR), ANNUAL PERCENT CHANGE (APC), AND RATE DIFFERENCE FOR CHILDHOOD ALL AND AML BY ETHNICITY AND SELECTED DEMOGRAPHIC CHARACTERISTICS .................................................... 70 TABLE 4.1. DEMOGRAPHIC CHARACTERISTICS OF CASES AND CONTROLS IN THE PRIMARY ANALYSIS (CCLS), AND OF CASES IN THE REPLICATION ANALYSIS (ESCALE AND AUS-ALL) ..................................................................................... 102 TABLE 4.2. CHARACTERISTICS OF THE 20 MOST SIGNIFICANT SNPS IDENTIFIED IN THE CCLS ANALYSIS, USING TWO-STEP METHODS FOR ANY EXPOSURE (MATERNAL SMOKING PRIOR TO PREGNANCY, DURING PREGNANCY, OR IN EARLY CHILDHOOD), AND FOR ANY LEUKEMIA, WHEN RESTRICTING TO B-CELL ALL, OR WHEN RESTRICTING TO ALL DIAGNOSED BETWEEN 1-10 YEARS OF AGE. ...... 104 TABLE 4.3. INTERACTION EFFECTS FOR MATERNAL SMOKING DURING PREGNANCY BY LEUKEMIA SUBGROUP, CCLS .................................................................................. 105 TABLE 4.4. REPLICATION RESULTS FOR MATERNAL SMOKING DURING PREGNANCY BY LEUKEMIA SUBGROUP, CCLS, ESCALE, AUS-ALL .............................................. 106 TABLE 4.5. INTERACTION EFFECTS FOR MATERNAL SMOKING PRIOR TO PREGNANCY BY LEUKEMIA SUBGROUP, CCLS ............................................................................ 107 TABLE 4.6. REPLICATION RESULTS FOR MATERNAL SMOKING PRIOR TO PREGNANCY BY LEUKEMIA SUBGROUP, CCLS, ESCALE, AUS-ALL ........................................ 109 TABLE 4.7. INTERACTION EFFECTS FOR MATERNAL SMOKING IN EARLY CHILDHOOD BY LEUKEMIA SUBGROUP, CCLS ............................................................................ 110 SUPPLEMENTAL TABLE 4.1 CHARACTERISTICS OF TOP SNPS IDENTIFIED FROM TWO- STEP SCANS BY TOBACCO SMOKE EXPOSURE FOR DIAGNOSES WITH ALL .............. 115 SUPPLEMENTAL TABLE 4.2 CHARACTERISTICS OF TOP SNPS IDENTIFIED FROM TWO- STEP SCANS BY TOBACCO SMOKE EXPOSURE FOR DIAGNOSES WITH B-CELL ALL ......................................................................................................................... 118 SUPPLEMENTAL TABLE 4.3 CHARACTERISTICS OF TOP SNPS IDENTIFIED FROM TWO- STEP SCANS BY TOBACCO SMOKE EXPOSURE FOR DIAGNOSES WITH B-CELL ALL ......................................................................................................................... 121 TABLE 5.1. CARCINOGENIC POTENTIAL OF SELECTED PESTICIDES AND CHILDHOOD LEUKEMIA RISK ....................................................................................................... 146 x xi List of Figures FIGURE 2.1. RISK OF CHILDHOOD BRAIN TUMORS BY EPHX1 H139R GENOTYPE AND EXPOSURE TO PARENTAL SMOKING (MATERNAL/PATERNAL), WEST COAST CHILDHOOD BRAIN TUMOR STUDY ........................................................................... 41 FIGURE 3.1. AGE-ADJUSTED INCIDENCE RATES BY ETHNICITY AND SUBTYPE, SEER 13, 1992-2010 ................................................................................................................. 71 FIGURE 3.2. AGE-ADJUSTED INCIDENCE RATES FOR ALL BY PRECURSOR CELL SUBTYPE, SEER 18, 2000-2009................................................................................. 72 FIGURE 3.3. AGE-SPECIFIC INCIDENCE RATES BY SUBTYPE FOR A) HISPANIC CHILDREN, AND B) NON-HISPANIC CHILDREN, SEER 13, 1992-2010 ......................................... 73 FIGURE 3.4. AGE-SPECIFIC INCIDENCE RATES FOR CHILDHOOD ALL FOR A) HISPANIC AND B) NON-HISPANIC CHILDREN DIAGNOSED IN 1992-1993 AND 2009-2010, SEER 13 .................................................................................................................... 74 FIGURE 3.5. AGE-ADJUSTED INCIDENCE RATES BY AGE AT DIAGNOSIS FOR A) HISPANIC AND B) NON-HISPANIC CHILDREN, SEER 13, 1992-2010 ......................................... 75 FIGURE 3.6. AGE-ADJUSTED INCIDENCE RATES BY GENDER AND ETHNICITY, SEER 13, 1992-2010 ................................................................................................................. 76 FIGURE 3.7. AGE-ADJUSTED INCIDENCE RATES BY RACE/ETHNICITY, SEER 13, 1992-2010 ................................................................................................................. 76 FIGURE 4.1A. SNP EXCLUSION CRITERIA ........................................................................ 111 FIGURE 4.1B. SAMPLE EXCLUSION CRITERIA ................................................................... 112 FIGURE 4.2. SUMMARY EFFECT ESTIMATES FROM CASE-ONLY ANALYSES FOR RS7421154, BY SMOKING EXPOSURE AND SUBTYPE, FOR RESULTS FROM THE CCLS AND AUS-ALL STUDIES .............................................................................. 113 FIGURE 4.3. LOCUS ZOOM PLOT FOR RS7421154 ............................................................. 114 xi xii Abstract While epidemiologic studies of childhood cancer have been conducted for many decades, ionizing radiation is the only established environmental risk factor for childhood brain tumors and childhood leukemia. Other factors, such as pesticides, electromagnetic fields, smoking and immunologic challenge have been suggested, but not consistently demonstrated, as risk factors. Additionally, recent genome wide association studies (GWAS) of childhood leukemia have found variants in several genes that may increase risk of disease. However, it is likely that both genetic and environmental risk factors contribute to the incidence of childhood brain tumors and childhood leukemia. In my background chapter (chapter 1), I describe incidence, mortality, and 5-year survival, as well secular trends in incidence and mortality for both childhood brain tumors and childhood leukemias. I also review the various classifications of each disease, and suspected environmental and genetic risk factors, with an emphasis on the roles of parental smoking and exposure to pesticides prior to pregnancy, during pregnancy or in early childhood. In my second chapter, I describe my analysis of data from the West Coast Childhood Brain Tumor (WCCBT) study to evaluate gene-parental smoking interactions and risk of childhood brain tumors. The analysis focuses parental smoking during pregnancy and 9 selected candidate polymorphisms in 6 genes known to affect the activation or detoxification of polycyclic aromatic hydrocarbons, a suspected carcinogen found in tobacco smoke. The corresponding manuscript was published in PLoS One in October 2013. xii xiii In chapter 3, I describe my analysis of trends in the incidence of childhood leukemia from 1973-2010 using complete Surveillance, Epidemiology and End Results (SEER) registry data. I evaluated secular trends by ethnicity (Hispanic vs. non-Hispanic) for incidence of all childhood leukemias, acute lymphocytic leukemia, and acute myeloid leukemia by demographic characteristics, including age, gender, race, and ethnicity. Chapter 3 includes a draft of the corresponding manuscript. In my fourth chapter, I describe my analysis of Hispanic participants enrolled in the California Childhood Leukemia Study (CCLS). I used two-step approaches to scan the genome for gene-parental smoking interactions associated with risk of childhood acute lymphocytic leukemia (ALL). We replicated results in the ESCALE study of childhood leukemia in France, and in the AUS-ALL study of childhood leukemia in Australia. A draft of the manuscript can be found in chapter 4. Chapter 5 includes a National Institute of Health (NIH) R21 grant that I submitted with Dr. Roberta McKean-Cowdin (chair, PI), Dr. W. James Gauderman (chair, Co-I), and Dr. Myles Cockburn (Co-I) in October 2013. The aims of this grant were 1) to evaluate the association between ambient pesticide exposure early in life and childhood ALL risk, using advanced exposure models and data from the California Childhood Leukemia Study, and 2) to evaluate the role of genetic susceptibility in the association between ambient pesticide exposure and ALL, using a genome-wide interaction scan (GWIS) analysis to scan the genome for susceptibility loci. The complete grant application can be found in chapter 5. xiii 1 CHAPTER 1: Background and Review 1.1 CHILDHOOD BRAIN TUMORS 1.1.1 INCIDENCE AND SURVIVAL Overview Childhood brain tumors (CBT) are the most common type of solid tumor, and the second most common type of cancer in children, behind leukemias (all subtypes). In 2013, it was estimated that approximately 4,300 new cases of CBT would be diagnosed, with 3,050 of those expected to have occurred in children under the age of 15 (1). From 2005- 2009, the age-adjusted incidence rate of CBT in children under 15 years of age was 3.2 per 100,000, and in children under 20 years of age, 3.0 per 100,000 (see table 1.1). Incidence rates overall are higher in males than in females: among children under 15, age-adjusted incidence rates (AAIR) were 3.4 per 100,000 in males and 3.1 per 100,000 in females, and among children under 20, AAIRs were 3.1 per 100,000 in males, and 2.8 per 100,000 in females (SEER Cancer Statistics Review). The male to female ratio is similar for all subtypes of childhood brain tumor with the exception of medulloblastoma/PNET, which are about 50% more frequent in males. Mortality rates from 2005-2009 for CBT and childhood leukemia were approximately equal, according to data from the US Mortality Files at the National Center for Health Statistics, Centers for Disease Control and Prevention. In children under 15, the age-adjusted mortality rate (AAMR) for CBT was 0.7 per 100,000 for both males and females. Mortality rates have declined only slightly since 1975, when rates 1 2 were approximately 1.0 deaths per 100,000 children overall. However, 5-year survival rates for children diagnosed with CBTs have increased significantly over time. For children under 15 diagnosed between 1975-1977, 5-year survival was only 56.9%. Relative survival increased to 74.5% for children diagnosed from 2002-2008 and followed for survival through 2009 (P<0.05). Similar trends were observed when including diagnoses of all children under the age of 20 (SEER Cancer Statistics Review). Trends The incidence rates of childhood brain tumors have remained relatively stable since ~1986. Just prior to this, between 1983-1986, the rates of CBT increased substantially (annual percent change (APC) = 14.06, 95%CI: 4.05, 25.0). From 1973- 1983, rates increased only slightly (APC = 1.09; 95%CI: -1.34, 3.58), and from 1986- 2009, almost no increase in incidence rates of CBT was observed (APC = 0.10; 95%CI: - 0.39, 0.61)(2). The rise in rates in the mid 1980s may be attributable to improved diagnostic techniques, using magnetic resonance imaging (MRI) scans, which increased the ability to find otherwise undiagnosed or unconfirmed brain tumors in children. SEER first mandated reporting of benign CNS tumors on January 1, 2004. Prior to use of MRIs for diagnosis, when benign tumors were not reported to SEER registries, malignant CNS tumors may not have been confirmed by pathologic review. These tumors may therefore have been grouped with benign (unreported) tumors. The use of diagnostic MRIs to confirm malignant tumors may have led to the increase observed in the mid 1980s, and higher, yet stable incidence rate of brain tumors thereafter. 2 3 1.1.2 PHENOTYPES Childhood brain tumors constitute an extraordinarily heterogeneous group of neoplasms, of which more than 100 subtypes and histological types have been identified(3). CBT are classified using the World Health Organization International Classification of Diseases-Oncology (ICD-O-3) codes. The WHO broadly categorizes neoplasms based on a 4-tiered system: grade I tumors have low proliferative potential (tumors are well circumscribed and slowly progressing), and can often be cured with resection; grade II tumors have low proliferation but are often infiltrative with a higher likelihood of recurrence; grade III tumors are malignant and generally require adjuvant radiation and/or chemotherapy; grade IV tumors are highly malignant, with rapid disease evolution, and are usually fatal (3, 4). Childhood brain tumors are further classified by histological subtype: astrocytic tumors (subependymal giant cell astrocytoma, pilocytic astrocytoma, pilomyxoid astrocytoma, diffuse astrocytoma, pleomorphic xanthoastrocytoma, anplastic astrocytoma, glioblastoma, giant cell glioblastoma, gliosarcoma), oligodendroglial tumors (oligodendroglioma and anaplastic oligodendroglioma), oligoastrocytic tumors (oligoastrocytoma, anaplastic oligoastrocytoma), ependymal tumors (subependymoma, myxopapillary ependymoma, ependymoma, anaplastic ependymoma), choroid plexus tumors, other neuroepithelial tumors, neuronal and mixed neuronal-glial tumors, pineal tumors, and embryonal tumors (medulloblastoma, CNS primitive neuroectodermal tumor (PNET), atypical teratoid/rhabdoid tumor), or tumors of the cranial and perispinal nerves, meninges, or sellar region. More finite classifications within each of the above tumor subtypes are also 3 4 often utilized; for instance, medulloblastoma may be further characterized as classic, desmoplastic, large cell, anaplastic, or medulloblastoma with extensive nodularity(4). The most common subtype of CBT is pilocytic astrocytoma (approximately 26% of all cases diagnosed from 2000-2009 in the United States), followed by primitive neuroectodermal tumors (PNET)/medulloblastoma (22% of all CBT cases), mixed gliomas and malignant gliomas NOS (19%), anaplastic astrocytomas, glioblastomas, other astrocytomas, and astrocytomas not otherwise specified (NOS) (16%), ependymomas (8%), oligodendrogliomas (2%), and other CNS tumors (7%) (SEER Cancer Statistics Review). While the extensive variation in childhood brain tumor subtypes suggests potentially varying etiologies of disease, childhood brain tumors are often grouped for epidemiologic studies because of the rarity of CBT and resulting difficulty in obtaining sufficient sample sizes for each subtype within any given study. In chapter 2, I describe my research of childhood brain tumors, using data from the West Coast Childhood Brain Tumor (WCCBT) study. In this study, we group CBT as 1) astroglial tumors (48%), 2) PNET (27.2%), or 3) tumors of other histology (24.8%). 1.1.3 GENETIC AND ENVIRONMENTAL RISK FACTORS Several genetic syndromes have been associated with CBT risk, including neurofibromatosis 1 (NF1; associated with increased risk of glioma), neurofibromatosis 2 (NF2; associated with meningioma), tuberous sclerosis (TSC1, TSC2; associated with subependymal giant cell astrocytomas), Von Hippel-Lindau syndrome (VHL; associated with hemangioblastomas), Gorlin syndrome (PTCH1; associated with medulloblastomas), 4 5 Li-Fraumeni syndrome (TP53), and Turcot syndrome (APC, various mismatch repair genes)(3, 5-7). The only established environmental risk factor for childhood brain tumors is ionizing radiation, though there are many other suspected risk factors with less consistent evidence supporting an association, including nonionizing radiation (i.e. electromagnetic fields), occupational exposures, pesticides, and dietary factors (see table 1.2) (3, 5, 8, 9). Parental smoking has been investigated as a risk factor in more than 15 studies of CBT from 1982 to 2012. Both maternal and paternal smoking have been considered at different time periods (i.e. prior to pregnancy, during pregnancy, and in early childhood) with respect to CBT risk. Results have suggested that paternal smoking during pregnancy, but not maternal smoking, may increase risk of CBT. A meta-analysis, combining results from 10 case-control studies on paternal smoking during pregnancy published prior to 2000, found an odds ratio of 1.22 (95%CI: 1.05, 1.40) (10). In contrast, two meta- analyses evaluating the association between maternal smoking during pregnancy and risk of CBT found only a small, statistically non-significant 4-5% increase in risk (10, 11). The specific agent in environmental tobacco smoke that may explain the increased risk for paternal smoking and CBT is unknown. Polycyclic aromatic hydrocarbons (PAH) are one suspected agent found in tobacco smoke that are known to be carcinogenic in animal models, and which have been associated with cancer at other sites. Animal studies suggest that PAHs in tobacco smoke may possibly affect brain tumor risk (12, 13). The lack of consistency in finding an increased risk of CBT associated with exposure to maternal or paternal tobacco smoke could be related to one of several 5 6 different factors, including small sample size, errors in the measurement of parental smoking exposure, or differing genetic susceptibility to the effects of tobacco smoke. Studies of gene-environment interactions may help to elucidate the true effect of smoking, as such analyses allow for the consideration of differing genetic backgrounds, which may make some children better at metabolizing and detoxifying carcinogenic compounds in tobacco smoke. In chapter 2, I describe the results of our candidate GxE interaction analysis, which evaluates the risk of CBT associated with parental smoking during pregnancy by polymorphisms in six genes known to be associated with the activation or detoxification of PAHs. 1.2 CHILDHOOD LEUKEMIA 1.2.1 INCIDENCE AND SURVIVAL Overall The age adjusted incidence rate for leukemia from 2005-2009 was 4.5 per 100,000 among children diagnosed under 20 years of age; the incidence rate for acute lymphocytic leukemia (ALL) was 3.4 per 100,000 during this time (table 1.1). Among children diagnosed under the age of 15, the age adjusted incidence rate (AAIR) for all leukemias was 5.0 per 100,000, and 4.0 per 100,000 for ALL. Similar to CBT, incidence rates are higher for males than for females: for children diagnosed under the age of 20, the AAIR for males is 5.0 per 100,000, compared to 4.0 per 100,000 for females. The discrepancy in rates by gender is similar when including only children diagnosed under the age of 15 (AAIR males = 5.5 per 100,000; AAIR females = 4.5 per 100,000) (SEER 6 7 Cancer Statistics Review). A description of rates stratified by other demographic characteristics (age, race) can be found in chapter 3. The incidence of ALL peaks around age 2-3 and decreases thereafter until age 10, when rates remain relatively constant through age 16, and decrease slightly from 17-19 years. The incidence of AML is highest among infants (0-1 years of age), followed by a decrease in rates from age 2-5, and a small, gradual increase in incidence through age 19. The age-adjusted mortality rates (AAMR) are similar for all leukemias and CBT (0.7 per 100,000 for children under 20), with lower mortality rates observed for ALL (0.3 per 100,000). Rates are similar among males and females, and for children diagnosed under the age of 15 (SEER Cancer Statistics Review). Trends From 1973-2009, the incidence rates for all leukemias increased significantly (APC = 0.80; 95%CI: 0.30, 0.75). This increase in rates was observed for both ALL (APC = 0.78; 95%CI: 0.51, 1.06) and AML (APC = 0.81; 95%CI: 0.33, 1.28). An analysis of trends in incidence rates, and review of differences in trends by ethnicity, is presented in chapter 2. During this same time period (1973-2009), survival rates for childhood leukemia increased. Among children diagnosed from 2002-2008, under the age of 15, five-year survival for all leukemias was 86.6%, up from 50.3% observed for children diagnosed from 1975-1977. Similar increases in survival are observed for both ALL and AML: 5-year survival for those diagnosed from 2002-2008 is 91.2% for ALL, and 64.2% for AML, up from 57.5% (ALL) and 18.8% (AML) for diagnoses between 1975-1977. 7 8 Mortality decreased most appreciably from 1975-1990, when mortality rates dropped from 2.0 per 100,000 to 1.1 per 100,000, followed by a steady decrease from 1990 to 2009, when mortality rates were 0.6 per 100,000. A similar pattern was observed for ALL mortality: rates in 1975 were 1.2 per 100,000, which decreased to 0.5 per 100,000 in 1990, and 0.3 per 100,000 in 2009. 1.2.2 PHENOTYPES Childhood leukemias are classified into four major groups. These classifications incorporate the rate at which disease develops (acute or chronic), and the type of white blood cell affected (lymphocytic or myeloid). The 4 major classifications include: acute lymphocytic leukemia (ALL) (approximately 78-80% of childhood leukemias; World Health Organization International Classification of Diseases-Oncology (ICD-O-3) codes 9828, 9832-7), acute myeloid leukemia (AML) sometimes called acute non-lymphocytic leukemia (ANLL) (approximately 16%; ICD-O-3: 9840, 9860-1, 9866-7, 9870-4, 9891, 9895-6, 9910), chronic myeloid leukemia (approximately 2%; ICD-O-3: 9863, 9875-6, 9945), and chronic lymphocytic leukemia (less than 1% of all childhood leukemias; ICD- O-3: 9823)(14). Our GxE analysis described in chapter 4 includes cases of ALL, the most common type of childhood leukemia. Subtypes of leukemia can be further classified according to affected cell type, or type of chromosomal translocation involved. ALL is broadly categorized by type of lymphocyte affected: B-cell precursor ALL (majority of ALL cases), T-cell precursor ALL or other ALL. Childhood B-cell ALL can be further classified by chromosomal translocation: 1) t(12;21)(p13;q22) leading to a gene fusion of TEL with AML1 (TEL- 8 9 AML1) (20-25% of B-cell ALL cases), 2) t(1;19)(q23;p13) leading to an E2A-PBX1 gene fusion (5% of B-cell ALL cases), 3) t(9;22)(q34;q11) resulting in a BCR-ABL fusion (3-5% of B-cell ALL cases), or 4) hyperdiploidy, leading to increased gene dosage, approximately 35% of all B-cell ALL cases. Translocations at t(4;11), t(9;11), t(11;19) leading to MLL fusions are observed in a only small number of childhood ALL cases, however, B-cell progenitor-monocytic MLL-AF4; 11q23 translocations are observed in the majority (approximately 85%) of infant B-cell precursor ALL cases. TEL-AML1 and BCR-ABL translocations are also found in a small number of infant ALL cases. Within T-cell precursor ALL, a 1q deletion; t(1;14)(p32;q11), leading to an SIL-SCL fusion is observed in approximately 25% of cases(14, 15). The majority of chromosomal translocations observed in ALL can be backtracked to the in-utero time period, which suggests that these translocations may represent an early initial event in the etiology of disease. Exposures around this time period, such as paternal or maternal smoking prior to or during pregnancy, may also be important in the etiology of childhood ALL. We will focus on these early life exposures and early age cases in our GxE analyses (described in chapter 4). Chromosomal translocations are observed less frequently in the other subtypes of leukemia. Among cases of infant AML, 11q23 translocations resulting in MLL-AF6, MLL-AF9, MLL-AF10 or other fusions are found in approximately 50% of all infant AML cases. An AML1-ETO fusion, resulting from a translocation at t(8;21)(q22;q22) is observed in about 15% of all infant and childhood AML cases. 9 10 1.2.3 GENETIC AND ENVIRONMENTAL RISK FACTORS Several genetic risk factors have been identified for childhood leukemia, including Dow n s y ndr ome, ne urof ibroma tosi s , F a nc oni’s a ne mi a , a nd B loom ’s syndrome(15). Recent genome-wide association studies (GWAS) of childhood acute lymphocytic leukemia in non-Hispanic White children (sample sizes ranging from 284- 1384 cases and 270-17,958 controls) have additionally identified and validated three potential genes that may be associated with increased risk of disease: ARID5B, IKZF1, and CEBPE (16-20). Other GWAS studies have identified additional single nucleotide polymorphisms that may be associated with risk of childhood leukemia, but results have not yet been validated (21, 22). Additional known risk factors for childhood leukemia include male sex, age (between 2-6 for ALL, younger than 1 year for AML), and race (Hispanic>White>Black). Ionizing radiation is the only established environmental risk factor, though there is some evidence to implicate a long list of unconfirmed risk factors, such as exposure to non- ionizing radiation, pesticides, alcohol or tobacco use, or early infection/allergy (14, 23, 24). A list of suspected risk factors is presented in table 1.2. Exposure to parental smoking early in life (i.e. prior to conception, during pregnancy, in early childhood, or while breastfeeding) is one suspected risk factor for childhood leukemia. A recent meta-analysis, using data from 18 published studies, observed an increased risk of childhood acute lymphocytic leukemia (ALL) with exposure to paternal smoking overall, prior to conception, and during pregnancy (OR overall =1.11, 95%CI: 1.05, 1.18; OR prior to conception =1.24, 95%CI: 1.07, 1.43; OR during 10 11 pregnancy =1.25, 95%CI: 1.08, 1.46)(25). A second meta-analysis, using 10 studies, observed similar results (OR around conception =1.15, 95%CI: 1.06, 1.24; OR >20cig/day around conception =1.44, 95%CI: 1.24, 1.68)(26). A meta-analysis evaluating the effect of maternal smoking during pregnancy found no association, using 20 published studies (OR = 1.03, 95%CI: 0.96, 1.12)(27). Exposure to pesticides has also been suggested as a risk factor for childhood leukemia. Meta-analyses of epidemiologic studies of residential exposure during pregnancy found elevated, statistically significant risks of childhood ALL with exposure to unspecified pesticides (OR=1.54; 95%CI: 1.13, 2.11; 11 studies), insecticides (OR=2.05; 95%CI: 1.80, 2.32; 8 studies), and herbicides (OR=1.61; 95%CI: 1.20, 2.16; 5 studies)(28). For residential exposure in childhood, increased risk estimates were observed both for exposure to unspecified pesticides (OR=1.38; 95%CI: 1.12, 1.70; 9 studies), and insecticides (OR=1.61; 95%CI: 1.33, 1.95; 7 studies)(28). Meta-analyses of epidemiologic studies of maternal occupational exposure to pesticides and childhood leukemia observed increased risks for maternal prenatal exposure overall (OR=2.09; 95%CI: 1.51, 2.88; 16 studies), with exposure to insecticides (OR=2.72; 95%CI: 1.47, 5.04; 6 studies) and herbicides (OR=3.62; 95%CI: 1.28, 10.3; 2 studies), and with exposure to any pesticides when restricting to ALL only (OR=2.64; 95%CI: 1.40, 5.00; 5 studies)(29). No association was observed for paternal occupational exposure to pesticides and childhood leukemia (OR=1.09; 95%CI: 0.88, 1.34; 30 studies)(29). Several positive, but statistically non-significant associations were observed for ambient exposure to specific pesticides and childhood ALL based on Pesticide Use Reporting 11 12 (PUR) data and residential address, using a subset of participants from phases I and II (1995-2002) of the California Childhood Leukemia Study (CCLS)(30). 1.2.4 GENETIC SUSCEPTIBILITY TO TOBACCO SMOKE AND PESTICIDES Tobacco Smoke Several studies have begun to explore the role of variants in genes that increase susceptibility to tobacco smoke, when evaluating the association between parental smoking and risk of childhood leukemia. Using a candidate gene approach, one case- control study found that maternal smoking was associated with ALL among children with genetic variants in CYP1A1(*2A) and GSTM1(null) (31). Another case-control study found an interaction between a high-risk haplotype for 5 CYP1A1 polymorphisms and paternal smoking during pregnancy; children with the high-risk haplotype exposed to tobacco smoke were 2.8 times as likely to be diagnosed with ALL (95%CI: 1.5, 5.3)(32). Increasingly, we can expect to see interaction analyses conducted on a larger scale using genome-wide data to identify susceptibility loci. In chapter 4, we extend the work of the initial GWAS projects and candidate gene GxE analyses by using two-step methods to scan the genome for gene-parental smoking interactions in a population based case-control study of Hispanic children younger than 15 years of age at diagnosis. Pesticides Studies have also evaluated genetic susceptibility to pesticide exposure early in life. A recent candidate gene study of MDR1 variants, indoor pesticide exposure and 12 13 childhood ALL found that children with a CGC haplotype for 3 MDR1 SNPs were less susceptible to the effects of pesticides that children with a variant haplotype(33). One additional study evaluated gene-pesticide interactions for 4 genes and residential pesticide exposure, and found increased interaction ORs for two variants in CYP1A1(34). No studies have evaluated susceptibility to ambient pesticides and childhood ALL risk using genome-wide data. In chapter 5, we describe our proposal to assess genetic susceptibility to ambient pesticides (GxE interactions) and risk of childhood ALL using state-of-the-art ambient pesticide exposure models and novel, two-step statistical methods for GxE interaction analysis. 13 14 1.3 TABLES Table 1.1. Age Adjusted Incidence Rates (2005-2009) and Age-Adjusted Mortality Rates (AAMR) for Childhood Brain Tumors and Childhood Leukemia Childhood brain tumors (CBT) Childhood Leukemia Acute lymphocytic leukemia (ALL) Age-adjusted Incidence Rate Age 0-14 3.2 5.0 4.0 Males 3.4 5.5 4.4 Females 3.1 4.5 3.6 Age 0-19 3.0 4.5 3.4 Males 3.1 5.0 3.8 Females 2.8 4.0 3.0 Age-adjusted Mortality Rate Age 0-14 0.7 0.7 0.3 Males 0.7 0.7 0.3 Females 0.7 0.6 0.3 Age 0-19 0.6 0.7 0.3 Males 0.7 0.8 0.4 Females 0.6 0.6 0.3 14 15 Table 1.2. Suspected Risk Factors for Childhood Brain Tumors and Childhood Leukemia Risk Factor Childhood Brain Tumors Childhood Leukemia Parental Exposures Occupational Plastic materials (i.e. polystyrene) x Hydrocarbons (freon, gasoline) x Paints or paint thinners x Pesticides x x Organic dust (i.e. PCB) x Lifestyle Characteristics Paternal smoking x x Alcohol consumption x Maternal Drug Use Amphetamines or diet pills x Mind-altering drugs (i.e. marijuana) x Vitamins* x Iron* x Folate* x Oral contraceptives x Antihistamines x Pregnancy maintaining drugs x Teratogenic drugs (i.e. CNS depressants) x Antibiotics x Residential Pesticides x Paints or paint thinners x Air pollution x Residential EMF x Food Consumption Protein sources x Vegetables x Fish/seafood x Sugars and syrups x Meat and meat products (dietary nitrites) x x Artificial sweeteners x Coffee/Tea x Viruses Epstein-Barr virus x JC Virus x SV40 x Other virus x Other Maternal history of fetal loss x Maternal age (younger than 20, or older than 35) x Ionizing radiation x x Non-ionizing radiation, including EMF and RF x x Low frequency microwave irradiation x Paternal age (older than 40) x Parental heat exposure prior to pregnancy x Maternal diseases during pregnancy and delivery x *Protective factor 15 16 Table 1.2. continued. Risk Factor Childhood Brain Tumors Childhood Leukemia Childhood Exposures Residential Residence near gas stations/repair garages (i.e. benzene exposure) x Radon x Pesticides x Artwork (source of organic solvents) x Residence in high population mixing/high growth/high density area* x Residence in a high density household (i.e. more than 1 person per room)* x Infection/Allergy History of infection* x Hay fever, eczema, hives* x Allergic disorder (i.e. asthma)* x Vaccination (HiB, BCG, Measles)* x Daycare attendance* x Other Electrical appliance usage (i.e. electric blanket, hair dryer, video games) x Chloramphenicol drug use x Ionizing radiation x x Radiation/X-ray exposure x x CT scans x Breastfeeding x High birth order x Long interval between proband and proceeding sibling x High birthweight x Family history of disease (cancer, autoimmune disease) x Genetic Risk Factors Down Syndrome x Neurofibromatosis x x Fanconi's anemia x Bloom's syndrome x Tuberous sclerosis x Von Hippel-Lindau syndrome x Li-Fraumeni syndrome x Turcot syndrome x Gorlin syndrome x *Protective factor 16 17 1.4 REFERENCES 1. Dolecek TA, Propp JM, Stroup NE, et al. CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2005- 2009. Neuro Oncol 2012;14 Suppl 5:v1-49. 2. McKean-Cowdin R, Razavi P, Barrington-Trimis J, et al. Trends in childhood brain tumor incidence, 1973-2009. J Neurooncol 2013;115(2):153-60. 3. Tihan T, Wiencke JK, Razavi P, et al. Epidemiology of Childhood Brain Tumors. In: Mehta M, Chang SM, Guha A, et al., eds. Principles and Practice of Neuro- Oncology: A Multidisciplinary Approach: Demos Medical Publishing, 2011. 4. Louis DN, Ohgaki H, Wiestler OD, et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 2007;114(2):97-109. 5. Pollack IF, Jakacki RI. Childhood brain tumors: epidemiology, current management and future directions. Nat Rev Neurol 2011;7(9):495-506. 6. Ohgaki H. Epidemiology of brain tumors. Methods Mol Biol 2009;472:323-42. 7. Fleming AJ, Chi SN. Brain tumors in children. Curr Probl Pediatr Adolesc Health Care 2012;42(4):80-103. 8. Infante-Rivard C, Weichenthal S. Pesticides and childhood cancer: an update of Zahm and Ward's 1998 review. J Toxicol Environ Health B Crit Rev 2007;10(1- 2):81-99. 9. Kheifets LI. Electric and magnetic field exposure and brain cancer: a review. Bioelectromagnetics 2001;Suppl 5:S120-31. 10. Boffetta P, Trédaniel J, Greco A. Risk of childhood cancer and adult lung cancer after childhood exposure to passive smoke: A meta-analysis. Environmental Health Perspectives 2000;108(1):73-82. 11. Huncharek M, Kupelnick B, Klassen H. Maternal smoking during pregnancy and the risk of childhood brain tumors: a meta-analysis of 6566 subjects from twelve epidemiological studies. Journal of neuro-oncology 2002;57(1):51-7. 12. Rice JM, Ward JM. Age dependence of susceptibility to carcinogenesis in the nervous system. Ann N Y Acad Sci 1982;381:274-89. 13. Markovits P, Maunoury R, Tripier MF, et al. Normal and benzo(a)pyrene- transformed fetal mouse brain cell. I. Tumorigenicity and immunochemical detection of glial fibrillary acidic protein. Acta Neuropathol 1979;47(3):197-203. 14. Kim AS, Eastmond DA, Preston RJ. Childhood acute lymphocytic leukemia and perspectives on risk assessment of early-life stage exposures. Mutat Res 2006;613(2-3):138-60. 15. Wiemels J. Perspectives on the causes of childhood leukemia. Chem Biol Interact 2012;196(3):59-67. 16. Papaemmanuil E, Hosking FJ, Vijayakrishnan J, et al. Loci on 7p12.2, 10q21.2 and 14q11.2 are associated with risk of childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1006-10. 17. Prasad RB, Hosking FJ, Vijayakrishnan J, et al. Verification of the susceptibility loci on 7p12.2, 10q21.2, and 14q11.2 in precursor B-cell acute lymphoblastic leukemia of childhood. Blood 2010;115(9):1765-7. 17 18 18. Treviño LR, Yang W, French D, et al. Germline genomic variants associated with childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1001-5. 19. Healy J, Richer C, Bourgey M, et al. Replication analysis confirms the association of ARID5B with childhood B-cell acute lymphoblastic leukemia. Haematologica 2010;95(9):1608-11. 20. Bonaventure A, Goujon-Bellec S, Rudant J, et al. Maternal smoking during pregnancy, genetic polymorphisms of metabolic enzymes, and childhood acute leukemia: the ESCALE Study (SFCE). Cancer causes & control : CCC 2011;23(2):329-45. 21. Han S, Lee K-M, Park SK, et al. Genome-wide association study of childhood acute lymphoblastic leukemia in Korea. Leukemia Research 2010;34(10):1271-4. 22. Sherborne AL, Hosking FJ, Prasad RB, et al. Variation in CDKN2A at 9p21.3 influences childhood acute lymphoblastic leukemia risk. Nature Genetics 2010;42(6):492-4. 23. Belson M, Kingsley B, Holmes A. Risk Factors for Acute Leukemia in Children: A Review. Environmental Health Perspectives 2006;115(1):138-45. 24. Infante-Rivard C. Chemical risk factors and childhood leukaemia: a review of recent studies. Radiation Protection Dosimetry 2008;132(2):220-7. 25. Liu R, Zhang L, McHale CM, et al. Paternal smoking and risk of childhood acute lymphoblastic leukemia: systematic review and meta-analysis. J Oncol 2011;2011:854584. 26. Milne E, Greenop KR, Scott RJ, et al. Parental Prenatal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia. American journal of epidemiology 2011;175(1):43-53. 27. Klimentopoulou A, Antonopoulos CN, Papadopoulou C, et al. Maternal smoking during pregnancy and risk for childhood leukemia: A nationwide case-control study in Greece and meta-analysis. Pediatric Blood & Cancer 2011;58(3):344-51. 28. Turner MC, Wigle DT, Krewski D. Residential Pesticides and Childhood Leukemia: A Systematic Review   and Meta-Analysis. Environmental Health Perspectives 2009. 29. Wigle DT, Turner MC, Krewski D. A Systematic Review and Meta-analysis of Childhood Leukemia and Parental Occupational Pesticide Exposure. Environmental Health Perspectives 2009;117(10):1505-13. 30. Rull RP, Gunier R, Von Behren J, et al. Residential proximity to agricultural pesticide applications and childhood acute lymphoblastic leukemia. Environ Res 2009;109(7):891-9. 31. Clavel J, Bellec S, Rebouissou S, et al. Childhood leukaemia, polymorphisms of metabolism enzyme genes, and interactions with maternal tobacco, coffee and alcohol consumption during pregnancy. Eur J Cancer Prev 2005;14(6):531-40. 32. Lee KM, Ward MH, Han S, et al. Paternal smoking, genetic polymorphisms in CYP1A1 and childhood leukemia risk. Leuk Res 2009;33(2):250-8. 33. Urayama KY, Wiencke JK, Buffler PA, et al. MDR1 gene variants, indoor insecticide exposure, and the risk of childhood acute lymphoblastic leukemia. Cancer Epidemiol Biomarkers Prev 2007;16(6):1172-7. 18 19 34. Infante-Rivard C, Labuda D, Krajinovic M, et al. Risk of childhood leukemia associated with exposure to pesticides and with gene polymorphisms. Epidemiology 1999;10(5):481-7. 19 20 CHAPTER 2: Parental Smoking and Risk of Childhood Brain Tumors by Functional Polymorphisms in Polycyclic Aromatic Hydrocarbon Metabolism Genes Barrington-Trimis JL, Searles Nielsen S, Preston-Martin S, Gauderman WJ, Holly EA, Farin FM, Mueller BA, McKean-Cowdin R (2013) Parental Smoking and Risk of Childhood Brain Tumors by Functional Polymorphisms in Polycyclic Aromatic Hydrocarbon Metabolism Genes. PLoS ONE 8(11): e79110. doi:10.1371/journal.pone.0079110 (1) 2.1 ABSTRACT A recent meta-analysis suggested an association between exposure to paternal smoking during pregnancy and childhood brain tumor risk, but no studies have evaluated whether this association differs by polymorphisms in genes that metabolize tobacco- smoke chemicals. We assessed 9 functional polymorphisms in 6 genes that affect the metabolism of polycyclic aromatic hydrocarbons (PAH) to evaluate potential interactions with parental smoking during pregnancy in a population-based case-control study of childhood brain tumors. C a se s (N = 202) we r e ≤1 0 y e a rs old, dia g nose d fr om 1984 -1991 and identified in three Surveillance, Epidemiology, and End Results (SEER) registries in the western U.S. Controls in the same regions (N=286) were frequency matched by age, sex, and study center. DNA for genotyping was obtained from archived newborn dried blood spots. We found positive interaction odds ratios (ORs) for both maternal and paternal smoking during pregnancy, EPHX1 H139R, and childhood brain tumors (P interaction =0.02; 0.10), such that children with the high-risk (greater PAH activation) genotype were at a higher risk of brain tumors relative to children with the low-risk 20 21 genotype when exposed to tobacco smoke during pregnancy. A dose-response pattern for paternal smoking was observed among children with the EPHX1 H139R high-risk genotype only (OR no exposure =1.0; OR ≤3 hours/day =1.32, 95%CI:0.52, 3.34; OR >3hours/day =3.18, 95%CI:0.92, 11.0; P trend =0.07). Parental smoking during pregnancy may be a risk factor for childhood brain tumors among genetically susceptible children who more rapidly activate PAH in tobacco smoke. 2.2 INTRODUCTION The association between parental smoking during pregnancy and risk of childhood brain tumors is inconsistent in the literature. Most studies have reported positive associations between paternal smoking during pregnancy and childhood brain tumor risk, although the findings from only three studies were statistically significant (2- 4). Seven studies reported positive, but non-statistically significant associations (5-11), and two reported no association (12, 13). A meta-analysis, combining ten studies published prior to 2000, estimated a 22% increase in risk of childhood brain tumors with exposure to paternal tobacco smoke during pregnancy (95%CI:1.05, 1.40) (14). Studies examining the association between maternal smoking during pregnancy and childhood brain tumors generally suggest little to no increased risk. Ten studies reported no association (2, 3, 6, 9, 11, 12, 15-18), and six studies reported a positive, but statistically non-significant association (5, 7, 10, 19-21). Two meta-analyses estimated a statistically non-significant 4-5% increase in childhood brain tumor risk with maternal smoking during pregnancy using 12 of the above studies (14, 22). However, a more recent prospective study reported a statistically significant 24% increase in childhood 21 22 brain tumor risk with maternal smoking during pregnancy (23). Although many studies have evaluated parental smoking and childhood brain tumors, none have evaluated potential interactions with functional polymorphisms in genes whose enzyme products metabolize tobacco smoke carcinogens, such as polycyclic aromatic hydrocarbons (PAH). Animal studies suggest this class of chemicals may possibly affect brain tumor risk (24, 25). Several genes are associated with the activation (transformation to more carcinogenic intermediates) or detoxification of PAH. We focused on 6 genes of potential importance to our analysis of parental smoking (PAH exposure) and childhood brain tumors (Table 2.1). Microsomal epoxide hydrolase (mEH), coded by EPHX1, detoxifies selected substances (by catalyzing the hydrolysis of epoxide intermediates for excretion), and activates others, including PAH (26, 27). Single nucleotide polymorphisms (SNPs) in exon 3 (Y113H) and exon 4 (H139R) of EPHX1 alter enzyme activity through amino acid changes (26, 28). A variant leading to a histidine (H) replacement of tyrosine (Y) at EPHX1 Y113H results in decreased mEH activity, whereas a variant leading to an arginine (R) substitution of a histidine (H) at H139R results in increased mEH activity (28). Myeloperoxidase (MPO) and sulfotransferase (SULT1A1) also activate carcinogens found in tobacco smoke, including PAHs. Variations in genotype at MPO G- 463A(29), or SULT1A1 R213H (30) result in greater enzyme activity leading to faster PAH activation. NAD(P)H: quinone oxidoreductase (NQO1), and glutathione S- transferases (including GSTM1 and GSTP1) detoxify PAHs. Variant alleles at NQO1 22 23 (P187S) (31, 32), GSTP1 I105V and GSTP1 A114V (33-35), or a null genotype at GSTM1(35) result in decreased enzyme activity (detoxification) of at least some PAHs. We analyzed the interaction between childhood brain tumors, exposure to parental smoking during pregnancy, an d the c hil d’s ge not y p e for the a bove 9 func ti ona l polymorphisms to evaluate whether the association between childhood brain tumors and parental smoking during pregnancy varies by genetic polymorphisms in the child. 2.3 MATERIALS AND METHODS Participants Participants were cases and controls enrolled in the West Coast Childhood Brain Tumor study (36) for whom a dried blood spot was located in newborn screening archives in California or Washington state (202 cases/286 controls) (37). Cases were identified through the Surveillance, Epidemiology and End Results (SEER) registries in the Los Angeles, San Francisco-Oakland, and Seattle regions, and include children diagnosed with a tumor of the brain, cranial nerves, or meninges [International Classification of Diseases-Oncology (ICD-O) (World Health Organization 1976) codes 191.0-192.1] between 1984-1991. Controls living in the same regions were identified using random digit dialing, and were frequency matched to cases by age, sex, and study center. This analysis includes children born in Washington State in 1978 or later, or in California in 1982 or later, the birth years for which a specimen could still remain in the state archives. C hil dre n mee ti ng these c rite ria we re ≤10 y e a rs old. S p e c im e ns we r e obtained for 93% of eligible cases and 83% of eligible controls, as detailed elsewhere(37). 23 24 Cases and controls in this sample were similar to those in the larger study with respect to race/ethnicity and maternal education, but were born more recently and were therefore younger at diagnosis/reference date. Fewer astroglial cases and more medulloblastoma/primitive neuroectodermal tumor (PNET) cases were included in the present sample, consistent with a younger age at diagnosis (37). Fewer case and control mothers and fathers smoked during pregnancy in more recent years than during earlier years. Exposure to parental smoking Parental smoking was assessed by in-person interview with the subj e c ts’ mot he rs. Mothers were asked if they ever smoked tobacco during their pregnancy with the enrolled child (yes/no), and the number of cigarettes smoked per day or week. They also were asked whether there was regular tobacco smoke exposure during pregnancy (yes/no, and hours pe r d a y ) fr om the c hil d’s fa the r in the hom e , fr om a n y other house h old re sident, or at work. Ma ter na l e x posure to tobac c o smok e fr om the c hil d’s fa ther du ring pre g n a nc y will be he re a fte r re fe r re d to a s “ pa te rna l smok ing . ” Mothers and fathers also were asked if they ever smoked at least once a day for 3 months or longer prior to the pregnancy with the participating child (yes/no). Maternal smoking during pregnancy was categorized by the typical number of cigarettes smoked per day: never smoked, 1-10, or 11+ cigarettes. Paternal smoking during pregnancy was categorized by the median number of hours per day the mother wa s e x pose d to t oba c c o s moke f rom th e f a ther (no ne , ≤3 hours pe r da y , > 3 hours pe r da y ). 24 25 Genotyping Subject s’ DN A wa s e x t ra c ted fr om drie d blood spot spe c im e ns fr om n e ona tal screening archives in California and Washington using the QIAamp DNA Mini Kit (QIAGEN, Valencia, CA) at the Center for Ecogenetics and Environmental Health Functional Genomics Laboratory at the University of Washington (Seattle, WA). Custom TaqMan Detection System-based assays-by-Design Service (Applied Biosystems, Inc., Foster City, CA) were used to assess EPHX1 H139R (rs2234922), EPHX1 Y113H (rs1051740), and EPHX1 C-613T (rs2854448), SULT1A1 R213H (rs9282861), NQO1 P187S (rs1800566), GSTP1 I105V (rs1695), GSTP1 A114V (rs1138272), and rs2243828 (in complete linkage disequilibrium with MPO G-463A (rs2333227)). Microsomal epoxide hydrolase (mEH) activity was computed using EPHX1 H139R and Y113H polymorphisms: low activity —0,1, or 2 stable alleles at H139R/Y113H (HH/HH, HH/HR, HY/HH, HH/RR, HY/HR, YY/HH), or high activity —3 or 4 stable alleles (HY/RR, YY/HR, YY/RR). One multiplex PCR-based assay (38) assessed GSTM1 null status. Complete genotyping data for all 9 polymorphisms was available for 200 (99.0%) cases and 284 (99.6%) controls. For 6% of cases and controls, duplicate and quadruplicate specimens were analyzed, blinded to initial results; analyses demonstrated complete concordance. Hardy Weinberg equilibrium was met (P>0.01) for all genotype frequencies for controls when stratified by race/ethnicity, with the exception of EPHX1 Y113H for Los Angeles non-Hispanic Whites (P<0.0001), and for NQ01 P187S for the he ter o g e n e ous ‘O the r’ e t hnicit y ( P=0.0003). Statistical analysis 25 26 We used unconditional logistic regression to evaluate the primary associations and potential interaction of genotype at each locus with maternal and/or paternal smoking during pregnancy. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed to estimate relative risks. For main associations and interaction analyses, genotypes were dichotomized and classified as low- or high-risk based on the ability of each variant to increase or decrease the activation or detoxification of PAHs (Table 2.1). All models were adjusted for frequency matching factors (age at diagnosis/reference age (<5, 5-10 years), sex, region (Los Angeles, San Francisco, Seattle), race/ethnicity (African- American, Non-Hispanic White, Hispanic, Asian/Other), and birth year (1978-84, 1985- 90)). Mode ls we r e a lso a djust e d fo r mot he r ’s e duc a ti on (no c oll e ge , some c oll e g e , college or graduate degree) a priori with the expectation that maternal education is associated both with maternal or paternal smoking and childhood brain tumors. A parallel set of models were additionally adjusted for spousal smoking. Formal tests of interaction were conducted using a product term in each model. Case-only analyses were conducted after confirming independence of each gene-smoking association among controls. Consistencies of all associations were further evaluated by race/ethnicity (non-Hispanic White or Hispanic). Polytomous logistic regression was used to evaluate whether gene- environment interactions differed by histological tumor type (astroglial, medulloblastoma/PNET, or ependymoma/other); formal tests of heterogeneity were conducted. Tests for trend in dose analyses were evaluated using a 1df test for the categorized dose variable. Due to a priori hypotheses regarding the suspected functionality of the tested polymorphisms in the metabolism of tobacco smoke, no corrections for multiple comparisons were made. All reported P-values are two-sided. 26 27 Ethics Statement Institutional Review Board approvals were obtained in California from the University of Southern California Institutional Review Board and the Committee for the Protection of Human Subjects at the Health and Human Services Agency of the State of California, and in Washington from the Fred Hutchinson Cancer Research Center and the Washington State Department of Health. Written informed consent for all participants was obtained prior to interview. Before release from neonatal archives in both California and Washington, all dried blood-spot specimens were anonymized by the assignment of a random specimen identification number that could not be linked to identifying information. 2.4 RESULTS Cases and controls were similar with regard to frequency-matched variables (Table 2.2). A higher proportion of controls were white (67.8% v. 53.6%, P=0.02), and control mothers were more likely to have a college or graduate degree (29.8% vs. 20.8%, P=0.02). The ORs for childhood brain tumors in relation to maternal smoking during pregnancy were less than one, but not statistically significant (Table 2.3). One exception was maternal smoking at the lowest smoking level (OR=0.23; 95%CI:0.08, 0.65) relative to never smoking. We observed a statistically non-significant increased OR associated with paternal smoking during pregnancy (OR=1.24; 95%CI:0.66, 2.35). Exposure to paternal smoking 27 28 for >3 hours per day, vs. no exposure, was positively associated with childhood brain tumors (OR=1.30; 95%CI:0.65, 2.59). The OR for smoking by both parents during pregnancy was consistent with no association (data not shown). Results were similar when examined by histology (data not shown). No association was observed for maternal exposure to tobacco smoke from other household residents. However, the number of mothers reporting exposure from other household members during pregnancy was small (10.4% of cases, 7.4% of controls; data not shown). We modeled the direct genotype- c hil dhood b ra i n tum or a ssocia ti on usin g ‘low - risk’ or ‘hi g h - risk’ ge no t y pe s (s e e S uppleme ntal Ta ble 2.1). No polymorphisms were associated with childhood brain tumors. When we examined the association between maternal and paternal smoking (ne ve r/ e ve r durin g pre g n a nc y ) a nd c hil dhood bra i n tum or risk, b y ‘low - ris k’ or ‘hig h - risk’ genotype, we found a positive interaction OR for paternal smoking and EPHX1 H139R (OR interaction =2.21; P interaction =0.10, Table 2.4). In children with a high-risk genotype (HR/RR) for EPHX1 H139R, exposure to paternal tobacco smoke during pregnancy was associated with increased risk of childhood brain tumors (OR=1.78; 95%CI:0.81, 3.91), whereas there was little observed association in children with a low-risk genotype (HH) (OR=0.83; 95%CI:0.45, 1.54). The case-only analysis showed a similar association (OR=1.99; 95%CI:0.96, 4.20; see Supplemental Table 2.2). Effect estimates changed minimally after adjustment for maternal smoking, with the exception of SULT1A1 R213H: we found a statistically significantly increased OR for children with the high-risk genotype after adjustment (OR high-risk =2.19; 95%CI:1.03, 4.65). Results were comparable when log-additive models were evaluated (data not shown). Other potential interactions 28 29 were either statistically non-significant (e.g. mEH activity, SULT1A1, GSTM1) or did not manifest in a biologically plausible manner (e.g. GSTP1 A114V) (see Table 2.4). We observed similar results for paternal smoking prior to pregnancy (never/ever) for all polymorphisms, with a positive interaction OR of a similar magnitude for EPHX1 H139R (OR interaction =1.91; P interaction =0.13; data not shown). Results were similar when examined by histology (data not shown). As with paternal smoking, we observed an interaction between maternal smoking and EPHX1 H139R (OR interaction =4.18; P interaction =0.02; Table 2.5). Although shifted downward relative to paternal smoking ORs, the OR for children with a high-risk variant was again greater than that for children with a low-risk variant (EPHX1 H139R: OR high- risk =1.09; 95%CI:0.44, 2.71; OR low-risk =0.28; 95%CI:0.12, 0.68). A similar interaction was observed for mEH activity (OR high-risk =0.87; 95%CI:0.42, 1.79; OR low-risk =0.25; 95%CI:0.07, 0.85; OR interaction =4.49; P interaction =0.03). The findings were supported by case-only analyses (EPHX1 H139R: OR=3.07; 95%CI:1.14, 8.28; mEH activity: OR=3.29; 95%CI:1.01, 10.8; see Supplemental Table 2.2). Results were similar after adjustment for paternal smoking. Results did not differ by state (CA or WA) or histology (data not shown). Smaller and statistically non-significant positive interaction ORs were observed for EPHX1 H139R and mEH activity for maternal smoking prior to pregnancy (never/ever). A positive association between hours per day of exposure to paternal smoking during pregnancy and childhood brain tumor risk was observed only among children with a high-risk genotype (HR or RR) for EPHX1 H139R (P interaction =0.07; Table 2.6). For children with the high-risk genotype, those exposed to paternal smoking for >3 hours per 29 30 day were 3.18 times as likely as unexposed children to develop a childhood brain tumor (95%CI:0.92, 11.0). In contrast, among children with a low-risk genotype (HH), there was no childhood brain tumor-paternal smoking association (OR >3 hrs/day =0.96; 95%CI:0.42, 2.20). A similar association was seen for SULT1A1 R213H, although the interaction did not reach statistical significance. Among children with a high-risk genotype (RR), children exposed to >3 hours per day of smoke from the father were 2.57 times as likely as unexposed children to develop a childhood brain tumor (95%CI:0.94, 7.01). This association was greater after adjusting for maternal smoking during pregnancy (OR >3 hrs/day =4.91; 95%CI:1.55, 15.6; P trend =0.01). No increased risk was observed among children with a low-risk genotype (OR >3 hrs/day =0.75; 95%CI:0.28, 1.96). Adjustment for maternal smoking had minimal effects on remaining polymorphisms. A suggestion of increasing ORs among carriers of high-risk genotypes also was observed by duration of exposure for EPHX1 Y113H and mEH activity (see Table 5), and for NQO1 P187S (P interaction = 0.54, data not shown). Similar to the paternal smoking data, a statistically significant interaction was observed for level of maternal smoking during pregnancy and EPHX1 H139R genotype (P interaction =0.003; see Supplemental Table 2.3). An interaction also was observed for mEH activity (P interaction =0.03). Among children with a high-risk variant (RR or HR) for EPHX1 H139R, children whose mothers smoked 11 or more cigarettes per day were twice as likely to develop a childhood brain tumor as children of mothers who did not smoke (OR=2.19; 95%CI:0.72, 6.63), however, the number of children exposed to high levels of maternal smoking was low. Among children with a low-risk genotype, there was no increase in childhood brain tumor risk observed in relation to smoking. Results were 30 31 similar for high mEH activity. Effect estimates decreased slightly after adjustment for paternal smoking. Figure 1 shows trends in childhood brain tumor risk by EPHX1 H139R genotype for children exposed to paternal or maternal tobacco smoke during pregnancy. The pattern of increased risk associated with exposure to tobacco smoke for children with a high-risk genotype, in contrast to those with a low-risk genotype, persists for children exposed to either maternal or paternal smoking, as evidenced by the parallel interactions presented. Similar patterns were observed for mEH activity. 2.5 DISCUSSION Our study expands on previous studies by evaluating the modifying effect of selected genetic polymorphisms involved in the metabolism of carcinogens present in tobacco smoke. We identified biologically plausible interactions between EPHX1 H139R and both maternal and paternal smoking overall (never/ever) and by level of exposure. Our results suggest that childhood brain tumor risk may be associated with exposure to parental smoking during pregnancy for children with genetic susceptibility to carcinogenic PAHs present in tobacco smoke. mEH is considered a detoxification enzyme for many substrates. However, in the process of PAH detoxification, carcinogenic highly-activated intermediates may be generated. mEH metabolizes PAHs to bay region diol-epoxides (27) that have potential to bind to DNA and cause mutations. A variant at exon 4 in EPHX1 results in increased mEH activity (28), and presumably greater levels of activated PAHs. Further, EPHX1 is expressed in the brain and during the fetal period (39). 31 32 The parental smoking-childhood brain tumor association was quite different for maternal vs. paternal smoking, overall and by strata of genotype for EPHX1 H139R and mEH activity. In our primary associations analysis, exposure to maternal smoking during pregnancy resulted in an OR <1, whereas exposure to paternal smoking was positively associated with childhood brain tumors. However, interaction ORs for childhood brain tumors, EPHX1 H139R and parental smoking were above null for both maternal smoking and paternal smoking (Figure 1). Similar results were observed for mEH activity. Potential reasons for the observed protective association between maternal smoking (disregarding genotype) and childhood brain tumors are likely related to one of two different explanations. First, the data on maternal smoking during pregnancy may be subject to maternal reporting bias. If mothers of cases were more likely than mothers of controls to underreport smoking, an artificially low association could result. Second, a similar bias could have occurred if among smokers we contacted, mothers of cases were less likely than mothers of controls to participate in the study. The occurrence of ORs <1 for maternal smoking during pregnancy, especially more recently when smoking has become less socially acceptable, is consistent with either possible source of bias. Although these factors may have biased the maternal smoking-childhood brain tumor association downward, they are unlikely to account for the observed interactions. Gene- environment interactions are largely unaffected by selection bias (40) and biased conservatively by any reporting/recall that may differ by case status (41). Confirmation of the interactions in the case-only analysis suggests the finding is not due to control selection or differential reporting. 32 33 The differences in maternal vs. paternal smoking ORs may be due to true biological differences in these associations with childhood brain tumor risk. However, if this were the case, we might have expected to observe dissimilar interaction ORs for maternal and paternal smoking with respect to EPHX1 H139R genotype. Our data suggest that children with a high-risk genotype are at a greater risk of childhood brain tumors if exposed to either maternal or paternal smoking during pregnancy, relative to children with a low-risk genotype and similar exposures. This may indicate that PAH activation increases risk regardless of the source of parental exposure. The carcinogenic process may be initiated through maternal exposure to environmental tobacco smoke from the father, or through the sperm, as a result of pa ter na l smok in g sho rtly b e for e the c hil d’s c o nc e pti on. Althou g h our prima r y r e sult s focused on paternal smoking during pregnancy, we observed similar interaction ORs for paternal smoking prior to pregnancy. Paternal smoking may induce genotoxic effects on sperm; studies of male smokers have demonstrated greater levels of oxo 8 dG (an oxidative product of DNA damage)(42), 8-hydroxydeoxyguanosine(43), and benzo(a)pyrene diol epoxide-DNA adducts(44, 45) in sperm DNA, and an increased risk of aneuploidy(46). However, the potential role of these in the etiology of brain tumors has not been established. Both strengths and limitations of this analysis need to be considered in the interpretation of the data. Although this is a relatively large population-based study of childhood brain tumors with comprehensive ascertainment of cases and highly comparable population-based controls, our sample is small for gene-environment interaction analyses. Therefore, these findings could be explained by chance. We also 33 34 focused on polymorphisms from a small number of candidate genes relevant to PAH specifically. We did not explore other genes associated with metabolism of other potential carcinogens in tobacco smoke and therefore may have missed some important interactions. We did not have DNA or genotype data for mothers, which during the pre g na nc y c ould influe n c e P AH meta boli sm in c ombi na ti on with the c h il d’s g e not y p e . However, to our knowledge this is the first assessment of these interactions. Moreover, use of archival dried blood spots allowed inclusion of all cases regardless of survival status, therefore minimizing survival bias that may be problematic in case-control studies of highly fatal diseases. Our study supports previous findings that parental smoking may be a risk factor for childhood brain tumors, and provides new information that risk may vary by genetic susceptibility. Studies that have reported no association may have been limited by inaccurate self-report of maternal smoking, and a lack of data on the genetic susceptibility of children in the study. Future studies of childhood brain tumors and parental smoking should include biological markers of smoking, in addition to data on the genetic susceptibility of children to tobacco smoke, to confirm and extend the results reported here. 34 35 2.6 TABLES AND FIGURES Table 2.1. Characteristics of Candidate Polymorphisms in Polycyclic Aromatic Hydrocarbon (PAH) Metabolism Genes Enzyme Expression Gene Polymor. ID Polymor. Chr. Enzyme Effect Effect of High-Risk Allele Ref. Microsomal Epoxide Hydrolase (mEH) Fetus: Yes Brain: Yes (39) EPHX1 rs2234922 H139R 1 Activates PAHs R: Faster PAH activation (26-28) EPHX1 rs1051740 Y113H 1 Activates PAHs Y: Faster PAH activation EPHX1 rs2854448 C-613T 1 Activates PAHs T: More mEH (faster PAH activation) Myeloperoxidase Brain: Yes (47) MPO rs2333227 G-463A 17 Activates PAHs G: Greater activity (faster PAH activation) (29) Sulfotransferase 1A1 Fetus: Yes Brain: Yes (48) SULT1A1 rs9282861 R213H 16 Activates PAHs R: Greater activity (faster PAH activation) (30) NAD(P)H: Quinone Oxireductase Brain: Yes (49) NQO1 rs1800566 P187S 16 Catalyzes detoxification of PAH quinines S: Reduced enzyme activity (reduced PAH detoxification) (31, 32) Glutathione S- Transferase Pi 1 Fetus: Yes Brain: Yes (50) GSTP1 rs1695 I105V 11 Detoxifies PAH intermediates V: Reduced PAH detoxification (33-35) GSTP1 rs1138272 A114V 11 Detoxifies PAH intermediates V: Reduced PAH detoxification Glutathione S- Transferase Mu 1 Brain: Yes (50) GSTM1 Null 1 PAH detoxification Null: No enzyme activity (reduced PAH detoxification) (33-35) 35 36 Table 2.2. Demographic Characteristics of Children With and Without Brain Tumors, West Coast Childhood Brain Tumor Study, Born 1978-1990 Cases n(%) Controls n(%) N=202 N=285 Race / Ethnicity White 105 (53.6) 192 (67.8) Hispanic 62 (31.6) 61 (21.6) African American 14 (7.1) 13 (4.6) Asian/other 15 (7.7) 17 (6.0) Unknown 6 2 Male 121 (59.9) 168 (58.9) Birth year 1978-1980 10 (5.0) 27 (9.5) 1981-1983 52 (25.7) 80 (28.1) 1984-1986 93 (46.0) 107 (37.5) 1987-1990 47 (23.3) 71 (24.9) Age at diagnosis (years) a <5 168 (83.2) 222 (77.9) 5-10 34 (16.8) 63 (22.1) Mother's Education No college b 103 (51.0) 112 (39.3) Some college (no degree) 57 (28.2) 88 (31.9) College or graduate degree 42 (20.8) 85 (29.8) Histologic tumor type Astroglial 97 (48.0) PNET c 55 (27.2) Other 50 (24.8) a Reference age for controls b <High school degree, high school degree, or basic or technical training only c Primitive neuroectodermal tumor 36 37 Table 2.3. Risk of Childhood Brain Tumors in Relation to Exposure to Parental Smoking during pregnancy, West Coast Childhood Brain Tumor Study, Born 1978-1990 Exposure Cases N=202 n (%) Controls N=285 n (%) Adj a OR 95% CI Maternal smoking (N=125 cases; 200 controls b ) No exposure to tobacco smoke during pregnancy 104 (83.2) 153 (76.5) 1.00 Mother smoked during pregnancy 21 (16.8) 47 (23.5) 0.55 0.29, 1.05 Mother only 4 (3.2) 12 (6.0) 0.41 0.12, 1.42 Mother and other passive/father c 17 (13.6) 35 (17.5) 0.60 0.30, 1.21 1-10 cigarettes/day 5 (4.0) 26 (13.0) 0.23 0.08, 0.65 11+ cigarettes/day 16 (12.8) 21 (10.5) 1.00 0.46, 2.17 P for trend 0.42 Paternal smoking (N=149 cases; 210 controls d ) No exposure to tobacco smoke during pregnancy 104 (69.8) 153 (72.9) 1.00 Father smoked during pregnancy 45 (30.2) 57 (27.1) 1.03 0.62, 1.71 Father only 25 (16.8) 27 (12.9) 1.24 0.66, 2.35 Father and other passive/mother c 20 (13.4) 30 (14.3) 0.82 0.41, 1.63 ≤ 3 h o u r s / d a y e 24 (16.1) 33 (15.7) 0.86 0.46, 1.61 >3 hours/day 21 (14.1) 24 (11.4) 1.30 0.65, 2.59 P for trend 0.64 a Odds ratio and 95% CI, adjusted for race, sex, age at diagnosis/reference, mother's education, birth year and center b Excludes children exposed to only paternal or other passive smoking c Other passive is exposure to tobacco smoke from a household resident other than the father, or at the workplace d Excludes children exposed to only maternal or other passive smoking e Hours per day of exposure from the father only, or from the father and another source 37 38 Table 2.4. Risk of Childhood Brain Tumors in Relation to Paternal Smoking during pregnancy by PAH Metabolism Genotype, West Coast Childhood Brain Tumor Study, Born 1978-1990 Polymorphism Low-risk genotype High-risk genotype Interaction OR a P-value for interaction a No/Yes Adj. b OR Adj. c OR No/Yes Adj. b OR Adj. c OR 95%CI 95%CI 95%CI 95%CI b c b c EPHX1 H139R Cases 107/24 44/21 Controls 144/37 0.83 0.45, 1.54 1.10 0.57, 2.11 82/20 1.78 0.81, 3.91 1.84 0.81, 4.21 2.21 2.26 0.10 0.10 Y113H Cases 67/21 84/24 Controls 113/27 0.99 0.52, 1.89 1.17 0.60, 2.33 113/30 1.42 0.69, 2.94 1.54 0.72, 3.32 1.19 1.27 0.71 0.62 C-613T Cases 70/26 81/19 Controls 126/35 1.20 0.64, 2.26 1.47 0.75, 2.90 100/22 1.02 0.49, 2.12 1.16 0.55, 2.49 0.83 0.78 0.69 0.61 mEH Activity d Cases 66/15 85/30 Controls 84/23 0.84 0.38, 1.84 1.01 0.45, 2.33 142/34 1.43 0.78, 2.64 1.56 0.82, 2.98 1.67 1.82 0.29 0.23 MPO G-463A e Cases 104/32 47/13 Controls 150/35 1.29 0.71, 2.32 1.48 0.80, 2.77 76/21 0.81 0.35, 1.89 1.07 0.44, 2.64 0.67 0.64 0.43 0.39 SULT1A1 R213H Cases 74/19 77/26 Controls 121/30 0.76 0.38, 1.54 0.85 0.41, 1.75 105/27 1.52 0.77, 3.00 2.19 1.03, 4.65 1.61 1.75 0.31 0.25 NQO1 P187S Cases 82/29 69/16 Controls 135/41 0.93 0.51, 1.69 1.08 0.57, 2.03 91/16 1.22 0.54, 2.75 1.50 0.63, 3.58 1.28 1.25 0.62 0.66 GSTP1 I105V Cases 66/17 85/28 Controls 74/18 1.01 0.46, 2.24 1.26 0.54, 2.98 152/39 1.10 0.60, 2.01 1.26 0.67, 2.37 1.16 1.10 0.77 0.85 A114V e Cases 132/42 19/3 Controls 188/41 1.30 0.78, 2.18 1.66 0.95, 2.89 38/15 0.33 0.07, 1.57 0.32 0.06, 1.66 0.27 0.25 0.08 0.07 GSTM1 f Cases 82/20 68/25 Controls 109/30 0.74 0.37, 1.50 0.82 0.39, 1.72 117/27 1.46 0.75, 2.87 1.86 0.90, 3.83 1.88 1.81 0.18 0.22 a Interaction between genotype and smoking, using dichotomous genotype and exposure levels never and ever b Adjusted for race, sex, age at diagnosis, mother's education, birth year and center c Additionally adjusted for maternal smoking d Microsomal epoxide hydrolase (mEH) activity: low —0,1 or 2 stable alleles (HH/HH, HH/HR, HY/HH, HH/RR, HY/HR, YY/HH); high —3 or 4 stable alleles (HY/RR, YY/HR, YY/RR) e Missing gene information for 1 control f Missing gene information for 1 case 38 39 Table 2.5. Risk of Childhood Brain Tumors in Relation to Maternal Smoking during pregnancy by PAH Metabolism Genotype, West Coast Childhood Brain Tumor Study, Born 1978-1990 Polymorphism Low-risk genotype High-risk genotype Interaction OR a P-value for interaction a No/Yes Adj. b OR Adj. c OR No/Yes Adj. b OR Adj. c OR 95%CI 95%CI 95%CI 95%CI b c b c EPHX1 H139R Cases 123/8 52/13 Controls 149/32 0.28 0.12, 0.68 0.27 0.11, 0.68 87/15 1.09 0.44, 2.71 0.88 0.34, 2.29 4.18 4.20 0.02 0.02 Y113H Cases 99/9 76/12 Controls 119/24 0.46 0.19, 1.11 0.43 0.17, 1.09 117/23 0.85 0.36, 2.02 0.73 0.30, 1.81 1.96 1.98 0.26 0.25 C-613T Cases 84/12 91/9 Controls 134/27 0.58 0.26, 1.28 0.49 0.21, 1.15 102/20 0.50 0.20, 1.30 0.48 0.18, 1.27 0.73 0.75 0.59 0.63 mEH Activity d Cases 77/4 98/17 Controls 87/20 0.25 0.07, 0.85 0.25 0.07, 0.86 149/27 0.87 0.42, 1.79 0.74 0.34, 1.58 4.49 4.44 0.03 0.03 MPO G-463A e Cases 120/16 55/5 Controls 156/29 0.66 0.32, 1.36 0.57 0.26, 1.22 80/17 0.25 0.08, 0.85 0.25 0.07, 0.87 0.59 0.60 0.42 0.44 SULT1A1 R213H Cases 82/11 93/10 Controls 131/20 0.56 0.22, 1.38 0.59 0.23, 1.50 105/27 0.41 0.17, 0.96 0.29 0.11, 0.74 0.48 0.47 0.21 0.20 NQO1 P187S Cases 96/15 79/6 Controls 144/32 0.56 0.27, 1.18 0.54 0.25, 1.19 92/15 0.47 0.16, 1.38 0.41 0.13, 1.25 0.74 0.74 0.63 0.64 GSTP1 I105V Cases 75/8 100/13 Controls 76/16 0.48 0.17, 1.40 0.43 0.14, 1.35 160/31 0.59 0.28, 1.25 0.54 0.25, 1.19 1.52 1.53 0.49 0.48 A114V e Cases 155/19 20/2 Controls 188/41 0.49 0.26, 0.93 0.41 0.21, 0.80 47/6 0.70 0.09, 5.39 1.09 0.12, 10.1 1.19 1.14 0.85 0.89 GSTM1 f Cases 90/12 84/9 Controls 117/22 0.63 0.27, 1.45 0.68 0.28, 1.62 119/25 0.39 0.16, 0.98 0.32 0.13, 0.83 0.70 0.72 0.54 0.57 a Interaction between genotype and smoking, using dichotomous genotype and exposure levels never and ever b Adjusted for race, sex, age at diagnosis, mother's education, birth year and center c Additionally adjusted for paternal smoking d Microsomal epoxide hydrolase (mEH) activity: low —0,1 or 2 stable alleles (HH/HH, HH/HR, HY/HH, HH/RR, HY/HR, YY/HH); high —3 or 4 stable alleles (HY/RR, YY/HR, YY/RR) e Missing gene information for 1 control f Missing gene information for 1 case 39 40 Table 2.6. Risk of Childhood Brain Tumors in Relation to Paternal Smoking Level during pregnancy by Polymorphisms in Selected Genes, West Coast Childhood Brain Tumor Study, Born 1978-1990 Polymorphism Exposure a Low-risk genotype High-risk genotype Cases/ Controls Adj. b OR 95%CI Adj. c OR 95%CI Cases/ Controls Adj. b OR 95%CI Adj. c OR 95%CI P-value for interaction d b c EPHX1 H139R Never 107/144 1.00 1.00 44/82 1.00 1.00 0.07 0.07 ≤ 3 h o u r s 12/18 0.74 0.33, 1.65 0.83 0.36, 1.93 12/15 1.32 0.52, 3.34 1.37 0.53, 3.57 >3 hours 12/19 0.96 0.42, 2.20 1.54 0.62, 3.83 9/5 3.18 0.92, 11.0 3.32 0.93, 11.9 P for trend 0.71 0.52 0.07 0.07 EPHX1 Y113H Never 84/113 1.00 1.00 67/113 1.00 1.00 0.47 0.45 ≤ 3 h o u r s 17/19 1.01 0.48, 2.14 1.23 0.80, 1.89 7/14 0.87 0.31, 2.48 0.93 0.32, 2.70 >3 hours 7/11 0.93 0.32, 2.73 1.4 0.91, 2.15 14/13 2.03 0.83, 4.99 2.31 0.89, 6.01 P for trend 0.93 0.64 0.18 0.12 mEH Activity e Never 66/84 1.00 1.00 85/142 1.00 1.00 0.16 0.12 ≤ 3 h o u r s 10/12 0.94 0.36, 2.44 1.11 0.42, 2.96 14/21 1.05 0.48, 2.29 1.13 0.51, 2.50 >3 hours 5/11 0.69 0.21, 2.31 0.88 0.25, 3.15 16/13 2.08 0.90, 4.79 2.41 0.98, 5.89 P for trend 0.58 0.94 0.12 0.08 SULT1A1 R213H Never 74/121 1.00 1.00 77/105 1.00 1.00 0.23 0.16 ≤ 3 h o u r s 10/16 0.78 0.32, 1.92 0.84 0.34, 2.10 14/17 1.09 0.48, 2.48 1.44 0.60, 3.43 >3 hours 9/14 0.75 0.28, 1.96 0.86 0.32, 2.30 12/10 2.57 0.94, 7.01 4.91 1.55, 15.6 P for trend 0.47 0.69 0.10 0.01 a Hours of exposure per day b Adjusted for race, sex, age at diagnosis/reference, mother's education, birth year and center c Additionally adjusted for maternal smoking d Interaction between genotype and smoking, using hours of exposure per day (interaction for trend) e Microsomal epoxide hydrolase (mEH) activity: low —0,1 or 2 stable alleles (HH/HH, HH/HR, HY/HH, HH/RR, HY/HR, YY/HH); high —3 or 4 stable alleles (HY/RR, YY/HR, YY/RR) 40 41 Figure 2.1. Risk of childhood brain tumors by EPHX1 H139R genotype and exposure to parental smoking (maternal/paternal), West Coast Childhood Brain Tumor Study 41 42 2.7 SUPPLEMENTARY MATERIAL Supplemental Table 2.1. Risk of childhood brain tumors in relation to polycyclic aromatic hydrocarbon (PAH) metabolism polymorphisms, West Coast Childhood Brain Tumor Study, N=479 Gene Polymorphism Genotype Cases Controls Adj . OR a N(%) N(%) N=196 N=283 95%CI Microsomal expoxide hydrolase (mEH) (EPHX1) H139R HH (low-risk) 131 (66.8) 181 (64.0) 1.00 HR/RR (high-risk) 65 (33.2) 102 (36.0) 1.15 0.64, 2.06 Y113H HH/YH (low-risk) 108 (55.1) 143 (50.5) 1.00 YY (high-risk) 88 (44.9) 140 (49.5) 1.15 0.65, 2.02 C-613T CC (low-risk) 96 (49.0) 161 (56.9) 1.00 CT/TT (high-risk) 100 (51.0) 122 (43.1) 0.74 0.42, 1.30 mEH Activity b Low (low-risk) 81 (41.3) 107 (37.8) 1.00 High (high-risk) 115 (58.7) 176 (62.2) 0.97 0.55, 1.72 Myeloperoxidase (MPO) G-463A c AA (low-risk) 136 (69.4) 185 (65.6) 1.00 GA/GG (high-risk) 60 (30.6) 97 (34.4) 1.11 0.61, 2.04 Sulfotransferase (SULT1A1) R213H RH/HH (low-risk) 93 (47.5) 151 (53.4) 1.00 RR (high-risk) 103 (52.5) 132 (46.7) 0.63 0.35, 1.12 NAD(P)H: quinone oxidoreductase (NQO1) P187S PP (low-risk) 111 (56.6) 176 (62.2) 1.00 PS/SS (high-risk) 85 (43.4) 107 (37.8) 0.67 0.37, 1.23 Glutathione S, Transferase Pi 1 (GSTP1) I105V II (low-risk) 83 (42.3) 92 (32.5) 1.00 IV/VV (high-risk) 113 (57.7) 191 (67.5) 0.76 0.42, 1.36 A114V c AA (low-risk) 174 (88.8) 229 (81.2) 1.00 AV/VV (high-risk) 22 (11.2) 53 (18.8) 0.66 0.31, 1.41 Glutathione S-Transferase Mu 1 (GSTM1) Null d No (low-risk) 102 (52.3) 139 (49.1) 1.00 Yes (high-risk) 93 (47.8) 144 (50.9) 0.64 0.37, 1.12 a Adjusted for race, sex, age at diagnosis/reference, mother's education, birth year and center b Microsomal epoxide hydrolase (mEH) activity: low —0,1 or 2 stable alleles (HH/HH, HH/HR, HY/HH, HH/RR, HY/HR, YY/HH); high —3 or 4 stable alleles (HY/RR, YY/HR, YY/RR) c Missing gene information for 1 control d Missing gene information for 1 case 42 43 Supplemental Table 2.2. Association between exposure to prenatal parental smoking and selected polymorphisms in a case-only analysis, West Coast Childhood Brain Tumor Study, N=196 Polymorphism Genotype Exposed N(%) Unexposed N(%) Adj. OR a 95% CI Adj. OR b 95% CI N = 21 N = 175 Paternal Smoking EPHX1 H139R HH 24 (51.1) 110 (71.0) 1.00 1.00 HR/RR 23 (48.9) 45 (29.0) 1.99 0.96, 4.20 1.60 0.74, 3.47 EPHX1 Y113H YY 21 (46.7) 67 (44.4) 1.00 1.00 YH/HH 24 (53.3) 84 (55.6) 1.18 0.58, 2.43 1.01 0.47, 2.19 mEH Activity Slow 15 (31.9) 68 (43.9) 1.00 1.00 Intermediate/Fast 32 (68.1) 87 (56.1) 1.55 0.74, 3.25 1.24 0.57, 2.71 SULT1A1 R213H RH/HH 19 (42.2) 74 (49.0) 1.00 1.00 RR 26 (57.8) 77 (51.0) 1.51 0.73, 3.11 1.56 0.72, 3.35 Maternal Smoking EPHX1 H139R HH 8 (38.1) 123 (70.3) 1.00 1.00 HR/RR 13 (61.9) 52 (29.7) 3.07 1.14, 8.28 2.51 0.88, 7.16 EPHX1 Y113H YY 12 (57.1) 76 (43.4) 1.00 1.00 YH/HH 9 (42.9) 99 (56.6) 1.88 0.70, 5.15 1.87 0.64, 5.44 mEH Activity Slow 4 (19.1) 77 (44.0) 1.00 1.00 Intermediate/Fast 17 (80.9) 98 (56.0) 3.29 1.01, 10.8 2.99 0.87, 10.3 SULT1A1 R213H RH/HH 11 (52.4) 82 (46.9) 1.00 1.00 RR 10 (47.6) 93 (53.1) 1.08 0.41, 2.85 0.89 0.31, 2.50 a Adjusted for race, sex, age at diagnosis, mother's education, birth year and center b Additionally adjusted for spousal smoking (maternal or paternal) 43 44 Supplemental Table 2.3. Risk of childhood brain tumors in relation to maternal smoking level during pregnancy by polymorphisms in selected genes, West Coast Childhood Brain Tumor Study Polymorphism Exposure a Low-risk genotype High-risk genotype Cases/ Controls Adj. b OR 95%CI Adj. c OR 95%CI Cases/ Controls Adj. b OR 95%CI Adj. c OR 95%CI p-value for interaction d b c EPHX1 Never 123/149 1.00 1.00 52/87 1.00 1.00 0.003 0.01 H139R 1-10 /day 3/18 0.17 0.05, 0.62 0.17 0.04, 0.62 2/8 0.28 0.05, 1.50 0.24 0.04, 1.29 11+ / day 5/14 0.46 0.15, 1.41 0.45 0.14, 1.44 11/7 2.19 0.72, 6.63 1.74 0.55, 5.51 P for trend 0.02 0.02 0.38 0.63 EPHX1 Never 99/119 1.00 1.00 76/117 1.00 1.00 0.83 0.34 Y113H 1-10 /day 2/14 0.16 0.03, 0.77 0.16 0.03, 0.76 3/12 0.41 0.10, 1.65 0.37 0.09, 1.52 11+ / day 5/14 0.95 0.31, 2.88 0.91 0.29, 2.89 9/11 1.39 0.48, 4.03 1.17 0.38, 3.55 P for trend 0.37 0.32 0.87 0.89 mEH Activity e Never 77/87 1.00 1.00 98/149 1.00 1.00 0.03 0.04 1-10 /day 1/11 0.09 0.01, 0.77 0.09 0.01, 0.77 4/15 0.36 0.11, 1.18 0.33 0.10, 1.08 11+ / day 3/9 0.59 0.13, 2.65 0.59 0.13, 2.76 13/12 1.57 0.63, 3.87 1.31 0.51, 3.38 P for trend 0.11 0.12 0.74 0.96 SULT1A1 Never 82/131 1.00 1.00 93/105 1.00 1.00 0.18 0.17 R213H 1-10 /day 2/13 0.21 0.04, 1.08 0.23 0.04, 1.15 3/13 0.22 0.06, 0.85 0.18 0.05, 0.72 11+ / day 9/7 1.05 0.33, 3.37 1.13 0.34, 3.73 7/14 0.62 0.22, 1.73 0.42 0.13, 1.29 P for trend 0.51 0.63 0.13 0.04 a Number of cigarettes smoked per day b Adjusted for race, sex, age at diagnosis/reference, mother's education, birth year and center c Additionally adjusted for paternal smoking d Interaction between genotype and smoking, using cigarettes per day (interaction for trend) e Microsomal epoxide hydrolase (mEH) activity: low —0,1 or 2 stable alleles (HH/HH, HH/HR, HY/HH, HH/RR, HY/HR, YY/HH); high —3 or 4 stable alleles (HY/RR, YY/HR, YY/RR) 44 45 2.8 REFERENCES 1. Barrington-Trimis JL, Searles Nielsen S, Preston-Martin S, et al. Parental smoking and risk of childhood brain tumors by functional polymorphisms in polycyclic aromatic hydrocarbon metabolism genes. PLoS One 2013;8(11):e79110. 2. Preston-Martin S, Yu MC, Benton B, et al. N-Nitroso compounds and childhood brain tumors: a case-control study. Cancer research 1982;42(12):5240-5. 3. McCredie M, Maisonneuve P, Boyle P. Antenatal risk factors for malignant brain tumours in New South Wales children. Int J Cancer 1994;56(1):6-10. 4. Sorahan T, Lancashire RJ, Hultén MA, et al. Childhood cancer and parental use of tobacco: deaths from 1953 to 1955. Br J Cancer 1997;75(1):134-8. 5. Howe GR, Burch JD, Chiarelli AM, et al. An exploratory case-control study of brain tumors in children. Cancer research 1989;49(15):4349-52. 6. Norman MA, Holly EA, Ahn DK, et al. Prenatal exposure to tobacco smoke and childhood brain tumors: results from the United States West Coast childhood brain tumor study. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology 1996;5(2):127-33. 7. Hu J, Mao Y. Parental Cigarette Smoking, Hard Liquor Consumption and the Risk of Childhood Brain Tumors-A Case Study in Northeast China. Acta Oncologica 2000. 8. Cordier S, Monfort C, Filippini G, et al. Parental exposure to polycyclic aromatic hydrocarbons and the risk of childhood brain tumors: The SEARCH International Childhood Brain Tumor Study. American journal of epidemiology 2004;159(12):1109-16. 9. John EM, Savitz DA, Sandler DP. Prenatal exposure to parents' smoking and childhood cancer. Am J Epidemiol 1991;133(2):123-32. 10. Filippini G, Farinotti M, Lovicu G, et al. Mothers' active and passive smoking during pregnancy and risk of brain tumours in children. Int J Cancer 1994;57(6):769-74. 11. Gold EB, Leviton A, Lopez R, et al. Parental smoking and risk of childhood brain tumors. Am J Epidemiol 1993;137(6):620-8. 12. Bunin GR, Buckley JD, Boesel CP, et al. Risk factors for astrocytic glioma and primitive neuroectodermal tumor of the brain in young children: a report from the Children's Cancer Group. Cancer Epidemiology … 1994. 13. Filippini G, Maisonneuve P, McCredie M, et al. Relation of childhood brain tumors to exposure of parents and children to tobacco smoke: The Search international case-control study. International Journal of Cancer 2002;100(2):206-13. 14. Boffetta P, Trédaniel J, Greco A. Risk of childhood cancer and adult lung cancer after childhood exposure to passive smoke: A meta-analysis. Environmental Health Perspectives 2000;108(1):73-82. 45 46 15. Kuijten RR, Bunin GR, Nass CC, et al. Gestational and familial risk factors for childhood astrocytoma: results of a case-control study. Cancer research 1990;50(9):2608-12. 16. Stjernfeldt M, Lindsten J, Berglund K. Maternal smoking during pregnancy and risk of childhood cancer. The Lancet 1986. 17. Sorahan T, Prior P, Lancashire RJ, et al. Childhood cancer and parental use of tobacco: deaths from 1971 to 1976. British Journal of Cancer 1997;76(11):1525- 31. 18. Pershagen G, Ericson A, Otterblad-Olausson P. Maternal smoking in pregnancy: does it increase the risk of childhood cancer? Int J Epidemiol 1992;21(1):1-5. 19. Cordier S, Iglesias MJ, Le Goaster C. Incidence and risk factors for childhood brain tumors in the Ile de France. … journal of cancer 1994. 20. Kramer S, Ward E, Meadows AT, et al. Medical and drug risk factors associated with neuroblastoma: a case-control study. Journal of the National Cancer Institute 1987;78(5):797-804. 21. Schwartzbaum JA. Influence of the mother's prenatal drug consumption on risk of neuroblastoma in the child. Am J Epidemiol 1992;135(12):1358-67. 22. Huncharek M, Kupelnick B, Klassen H. Maternal smoking during pregnancy and the risk of childhood brain tumors: a meta-analysis of 6566 subjects from twelve epidemiological studies. Journal of neuro-oncology 2002;57(1):51-7. 23. Brooks DR, Mucci LA, Hatch EE, et al. Maternal smoking during pregnancy and risk of brain tumors in the offspring. A prospective study of 1.4 million Swedish births. Cancer causes & control : CCC 2004;15(10):997-1005. 24. Rice JM, Ward JM. Age dependence of susceptibility to carcinogenesis in the nervous system. Ann N Y Acad Sci 1982;381:274-89. 25. Markovits P, Maunoury R, Tripier MF, et al. Normal and benzo(a)pyrene- transformed fetal mouse brain cell. I. Tumorigenicity and immunochemical detection of glial fibrillary acidic protein. Acta Neuropathol 1979;47(3):197-203. 26. Lacko M, Oude Ophuis MB, Peters WH, et al. Genetic polymorphisms of smoking-related carcinogen detoxifying enzymes and head and neck cancer susceptibility. Anticancer Res 2009;29(2):753-61. 27. Hulla JE, Miller MS, Taylor JA, et al. Symposium overview: the role of genetic polymorphism and repair deficiencies in environmental disease. Toxicol Sci 1999;47(2):135-43. 28. Hassett C, Aicher L, Sidhu JS, et al. Human microsomal epoxide hydrolase: genetic polymorphism and functional expression in vitro of amino acid variants. Hum Mol Genet 1994;3(3):421-8. 29. Taioli E, Benhamou S, Bouchardy C, et al. Myeloperoxidase G-463A polymorphism and lung cancer: a HuGE genetic susceptibility to environmental carcinogens pooled analysis. Genet Med 2007;9(2):67-73. 30. Kotnis A, Kannan S, Sarin R, et al. Case-control study and meta-analysis of SULT1A1 Arg213His polymorphism for gene, ethnicity and environment interaction for cancer risk. Br J Cancer 2008;99(8):1340-7. 31. Nisa H, Kono S, Yin G, et al. Cigarette smoking, genetic polymorphisms and colorectal cancer risk: the Fukuoka Colorectal Cancer Study. BMC Cancer 2010;10:274. 46 47 32. Kim HN, Kim NY, Yu L, et al. Polymorphisms of drug-metabolizing genes and risk of non-Hodgkin lymphoma. Am J Hematol 2009;84(12):821-5. 33. Lavender NA, Benford ML, VanCleave TT, et al. Examination of polymorphic glutathione S-transferase (GST) genes, tobacco smoking and prostate cancer risk among men of African descent: a case-control study. BMC Cancer 2009;9:397. 34. Koh WP, Nelson HH, Yuan JM, et al. Glutathione S-transferase (GST) gene polymorphisms, cigarette smoking and colorectal cancer risk among Chinese in Singapore. Carcinogenesis 2011;32(10):1507-11. 35. Kukkonen MK, Hämäläinen S, Kaleva S, et al. Genetic polymorphisms of xenobiotic-metabolizing enzymes influence the risk of pulmonary emphysema. Pharmacogenet Genomics 2011;21(12):876-83. 36. Preston-Martin S, Gurney JG, Pogoda JM, et al. Brain tumor risk in children in relation to use of electric blankets and water bed heaters. Results from the United States West Coast Childhood Brain Tumor Study. Am J Epidemiol 1996;143(11):1116-22. 37. Searles Nielsen S, Mueller BA, Preston-Martin S, et al. Childhood brain tumors and maternal cured meat consumption in pregnancy: differential effect by glutathione S-transferases. Cancer Epidemiol Biomarkers Prev 2011;20(11):2413-9. 38. Chen CL, Liu Q, Relling MV. Simultaneous characterization of glutathione S- transferase M1 and T1 polymorphisms by polymerase chain reaction in American whites and blacks. Pharmacogenetics 1996;6(2):187-91. 39. Farin FM, Omiecinski CJ. Regiospecific expression of cytochrome P-450s and microsomal epoxide hydrolase in human brain tissue. J Toxicol Environ Health 1993;40(2-3):317-35. 40. Morimoto LM, White E, Newcomb PA. Selection bias in the assessment of gene- environment interaction in case-control studies. Am J Epidemiol 2003;158(3):259- 63. 41. Garcia-Closas M, Rothman N, Lubin J. Misclassification in case-control studies of gene-environment interactions: assessment of bias and sample size. Cancer Epidemiol Biomarkers Prev 1999;8(12):1043-50. 42. Fraga CG, Motchnik PA, Wyrobek AJ, et al. Smoking and low antioxidant levels increase oxidative damage to sperm DNA. Mutat Res 1996;351(2):199-203. 43. Shen HM, Chia SE, Ni ZY, et al. Detection of oxidative DNA damage in human sperm and the association with cigarette smoking. Reprod Toxicol 1997;11(5):675-80. 44. Zenzes MT, Bielecki R, Reed TE. Detection of benzo(a)pyrene diol epoxide- DNA adducts in sperm of men exposed to cigarette smoke. Fertil Steril 1999;72(2):330-5. 45. Zenzes MT, Puy LA, Bielecki R, et al. Detection of benzo[a]pyrene diol epoxide- DNA adducts in embryos from smoking couples: evidence for transmission by spermatozoa. Mol Hum Reprod 1999;5(2):125-31. 46. Shi Q, Ko E, Barclay L, et al. Cigarette smoking and aneuploidy in human sperm. Mol Reprod Dev 2001;59(4):417-21. 47. Green PS, Mendez AJ, Jacob JS, et al. Neuronal expression of myeloperoxidase is increased in Alzheimer's disease. J Neurochem 2004;90(3):724-33. 47 48 48. Richard K, Hume R, Kaptein E, et al. Sulfation of thyroid hormone and dopamine during human development: ontogeny of phenol sulfotransferases and arylsulfatase in liver, lung, and brain. J Clin Endocrinol Metab 2001;86(6):2734- 42. 49. van Muiswinkel FL, de Vos RA, Bol JG, et al. Expression of NAD(P)H:quinone oxidoreductase in the normal and Parkinsonian substantia nigra. Neurobiol Aging 2004;25(9):1253-62. 50. Hayes JD, Pulford DJ. The glutathione S-transferase supergene family: regulation of GST and the contribution of the isoenzymes to cancer chemoprotection and drug resistance. Crit Rev Biochem Mol Biol 1995;30(6):445-600. 48 49 CHAPTER 3: Rising Rates of Acute Lymphocytic Leukemia in Hispanic Children: a Review of Trends in Childhood Leukemia Incidence from 1973-2010 3.1 ABSTRACT Childhood leukemia is the most common cancer in children under 20 years of age in the United States, and incidence rates have steadily increased over the last several decades. The incidence rate for all childhood leukemias is higher among Hispanic children (2006-2010 age-adjusted incidence rate (AAIR)= 5.71; 95%CI:5.47, 5.97) than among non-Hispanic children (2006-2010 AAIR=3.95; 95%CI:3.81, 4.09). Surveillance, Epidemiology and End Results (SEER) data were used to evaluate trends in the incidence of childhood leukemia diagnosed among children less than 20 years of age from 1992- 2010 overall and by age, ethnicity, gender, and histologic subtype. Tests for trend differences were computed by including an interaction term in linear regression models. Hispanic children were more likely than non-Hispanic children to be diagnosed with leukemia and with ALL. Further, a greater increase in AAIRs was observed from 1992- 2010 for Hispanic children as compared to non-Hispanic children for all types of leukemia (Annual percent change (APC) Hispanic =1.16, 95%CI:0.55, 1.76; APC non- Hispanic =0.52, 95%CI:0.01, 1.03) and when restricting to ALL (APC Hispanic =1.31, 95%CI:0.68, 1.94; APC=0.50 non-Hispanic ; 95%CI:-0.19, 1.19). By age group, statistically significant increases in ALL AAIRs were observed among older Hispanic children (diagnosed from 15-19 years of age), but not among non-Hispanic children 49 50 (P interaction =0.009). The increase in incidence rates of childhood ALL appears to be driven by increases in older Hispanic children (10-19 years of age). Future studies are needed to evaluate characteristics and risk factors that may be particularly prevalent and increasing among this demographic group. 3.2 INTRODUCTION Childhood leukemia is the most common type of cancer in children under 20 years of age, with an age adjusted incidence rate of 4.6 cases per 100,000 children for all leukemia types diagnosed between 2006-2010(1). Several characteristics, including male gender, younger age [2-6 years for acute lymphocytic leukemia (ALL) and <1 year of age for acute myeloid leukemia (AML)], race/ethnicity (highest rates in Hispanic children), a nd fa mi li a l s y ndrom e s ( e .g . ne u rof ibroma tosi s, F a nc oni’s a ne mi a , Do wns a nd B loom ’s syndrome) are associated with increased risk of childhood leukemia(2). Incidence rates for childhood ALL in all races/ethnicities have increased approximately 1% per year since 1973, suggesting a potential increase in the prevalence of causal risk factors or a change in the confluence of environmental or lifestyle factors(2). In the past 20 years, numerous studies have been conducted to evaluate the association between childhood leukemia and such factors. Ionizing radiation remains the only established causal environmental risk factor for childhood leukemia. Other lifestyle or environmental factors have been suggested as causal agents, with the strongest evidence for residential or occupational exposure to pesticides(3, 4). A series of meta- analyses also suggest ALL risk may be higher for children with exposure to non-ionizing radiation(5-7), paternal tobacco use(8, 9), and high birth weight(10, 11), and may be 50 51 protected by exposure to atopic conditions (e.g. allergy, asthma, hay fever)(12, 13) and day care attendance (as a surrogate for early life immune challenge)(14). However, few studies have evaluated associations between these risk factors and childhood leukemia by Hispanic ethnicity. Data on temporal trends in relevant exposures by ethnicity, as well as data on the ethnic-specific risks associated with each respective environmental factor may help to explain the observed patterns of childhood leukemia incidence in Hispanic and non-Hispanic populations. Studies of genetic susceptibility for childhood leukemia have been implemented using both directed, candidate gene approaches, and, more recently, agnostic genome- wide approaches. As of April 2008, pathway or candidate gene studies (59 total studies of gene main effects) of childhood ALL have identified probable susceptibility loci in a number of genes, including MTHFR (14 studies), GSTM1 and GSTT1 (13 studies each), GSTP1, CYP1A1, and NQO1 (7 studies each), and other genes reported in fewer than 5 studies (15). A recent meta-analysis of SNPs in MTHFR using a total of 21 studies published through September 2010, found a reduced risk of ALL associated with the variant allele at 677T(16). Studies have been less successful in identifying susceptibility loci for childhood AML, due in part to the rarity of this subtype, though 3 studies have found associations with GSTM1, GSTT1, and CYP1A1 (15). Recently, studies utilizing genome-wide data have been conducted to evaluate whether single nucleotide polymorphisms (SNPs) may be associated with increased risk of childhood leukemia. Genome-wide association studies (GWAS) of childhood ALL have identified several potential genetic risk factors for ALL. The most promising results from GWAS suggest the association between SNPs in ARID5B, IKZF1, CEBPE, and 51 52 CDKN2A/B and childhood ALL risk(17-21). In a replication study of these SNPs that included both Hispanic and non-Hispanic children, similar results were observed for ARID5B (all subtypes) and CEBPE (when restricting to B-cell ALL), with higher risk allele frequencies and stronger associations observed among Hispanic children for variants in both genes. Additional associations were observed for two IKZF1 variants and ALL risk (all subtypes) among non-Hispanic Whites but not among Hispanic children. Results were mixed for CDKN2A variants(22). Other GWAS and candidate gene studies have found additional variants that may be associated with increased risk of childhood ALL, but results have not yet been validated (21, 23, 24). A recent review of childhood leukemia suggests that less than 10% of all cases can be explained by recognized genetic or environmental risk factors(2). In this analysis of Surveillance, Epidemiology and End Results (SEER) registry data, we evaluate incidence rate patterns in the United States from 1992-2010 and examine potential differences in patterns by race and ethnicity. An analysis of recent and long-term trends in incidence may inform our search for preventable risk factors for childhood leukemia, particularly in demographic groups experiencing high and increasing rates of childhood leukemia. 3.3 MATERIALS AND METHODS Data Trends in childhood leukemia incidence and survival from 1992-2010 were a na l y z e d using d a ta f rom the Na ti ona l C a n c e r I nsti tut e ’s S urve il lanc e , Epid e mi olog y a nd End Results (SEER) population-based registries. Overall trends in leukemia incidence 52 53 from 1992-2010 were analyzed using SEER 13 registries (Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco/Oakland, Seattle/Puget Sound, Utah, Alaska, Los Angeles, San Jose-Monterey, and Rural Georgia), with complete data from these years. Age-adjusted incidence rates (AAIR) for recent years (2006-2010) by race/ethnicity are shown using data available from all 18 SEER registries (Alaska, Atlanta, Connecticut, Detroit, Greater California, Greater Georgia, Hawaii, Iowa, Kentucky, Los Angeles, Louisiana, New Jersey New Mexico, Rural Georgia, San Francisco-Oakland, San Jose-Monterey, Seattle/Puget Sound, and Utah). All cases of childhood leukemia diagnosed in children at age 19 or younger were included in analyses. Childhood leukemia was defined according to the International Classification of Childhood Cancer (ICCC) which uses the World Health Organization International Classification of Diseases for Oncology (ICDO-3) histology codes to define diagnosis with childhood leukemia: 9811-9818, 9835-9837 (precursor cell acute lymphoid leukemias); 9823, 9826, 9832-9833, 9940 (mature B-cell lymphoid leukemias); 9827, 9831, 9834, 9948 (Mature T-cell and NK cell lymphoid leukemias); 9820 (lymphoid leukemia, not otherwise specified (NOS)); 9840, 9861, 9865-9867, 9869-9874, 9891, 9895-9898, 9910-9911, 9920, 9931 (acute myeloid leukemias); 9800-9801, 9805- 9809, 9860, 9930, 9965-9967, 9971 (unspecified and other specified leukemias). From 1973-2009, precursor cell acute lymphocytic leukemias were classified by molecular subtype: B-cell (9836); T-cell (9837); NOS (9835); Burkitt cell leukemia (9826). For childhood leukemia cases diagnosed beginning January 1, 2010, classifications of precursor B-cell lymphoblastic leukemia utilized additional genetic data: 9811 (NOS), 9812 (BCR-ABL1), 9813 (MLL rearranged), 9814 (TEL-AML1), 9815 (hyperdiploidy), 53 54 9816 (hypodiploidy ALL), 9817 (IL3-IGH), and 9818 (E2A-PBX1 (TCF3-PBX1)). Details on these new classifications, and changes to classifications for acute myeloid leukemias, are provided in the 2010 Hematopoietic and Lymphoid Database (http://seer.cancer.gov/seertools/hemelymph/2010/). Individual attributes were available through SEER registries. SEER 13 (1992- 2010) and SEER 18 (2000-2010) data includes standard demographic variables (such as age, sex, and 4 categories of race), but also includes an additional two categories of race (American Indian, Asian/Pacific Islander), and a variable describing origin (Hispanic, non-Hispanic). Complete descriptions of the variables available through SEER can be accessed at (http://www.seer.cancer.gov/data/seerstat/nov2012/). Statistical analysis Age adjusted incidence rates (AAIRs) were calculated using SEER*stat 8.0.4 (www.seer.cancer.gov/seerstat), using the 2000 U.S. Standard Population for age standardization, based on recent data from 2006-2010. Trends in incidence rates from 1992-2010 by ethnicity were evaluated 1) using estimates of annual percent change (APC) and 2) by calculating the difference between the average rate from 1992-1993 and the average rate from 2009-2010. APC estimates were calculated by fitting a least squares regression line using the natural log of the rates as the outcome variable, with year of diagnosis as the primary predictor variable. JoinPoint was used to determine whether a piecewise linear spline was a better fit for the data than a simple linear estimator for the overall trends. 95% confidence intervals and p-values estimates for AAIR and APC estimates were calculated using SEER*stat and the JoinPoint regression program 4.0.4. 54 55 (http://surveillance.cancer.gov/joinpoint/). Tests for differences in APC by year of diagnosis and other demographic characteristics were completed by inclusion of an interaction term (characteristic x year of diagnosis) in a linear regression model with the natural log of rates as the outcome variable. Similar models were fit to test for differences in the absolute change in incidence rates, using annual incidence rate as the outcome variable in these analyses. Tests of significance for each interaction term were computed using outcome values weighted by the inverse variance for each data point. All interaction tests were completed using STATA v.13.1, with a two-sided level of significance of 0.05. 3.4 RESULTS Overall trends The AAIR for all leukemia subtypes combined from 2006-2010 was 5.71 per 100,000 persons for Hispanic children (95%CI: 5.47, 5.97) and 3.95 per 100,000 persons for non-Hispanic children (95%CI: 3.81, 4.09) (Table 3.1). Hispanic children were 1.45 times as likely as non-Hispanic children to be diagnosed with leukemia (95%CI: 1.37, 1.53). Incidence rates were highest in Hispanic White children (AAIR=6.07 per 100,000; 95%CI: 5.80, 6.35), as compared to other racial/ethnic groups (Hispanic Black, Non- Hispanic White, Non-Hispanic Black, Asian/Pacific Islander, or American Indian). Males had higher incidence rates than females in both Hispanic and non-Hispanic children. Incidence rates were highest in children diagnosed between 1-4 years of age in both ethnic groups (AAIR Hispanic =10.55 per 100,000; 95%CI: 9.83, 11.32; AAIR non-Hispanic =8.34 per 100,000; 95%CI: 7.89, 8.80). The incidence rates were higher for Hispanic children 55 56 both for ALL (AAIR hispanic =4.68 per 100,000, 95%CI: 4.46, 4.91; AAIR non-hispanic 3.04 per 100,000, 95%CI: 2.92, 3.16), and for AML (AAIR hispanic =0.89 per 100,000, 95%CI: 0.80, 0.99; AAIR non-hispanic 0.81 per 100,000, 95%CI: 0.75, 0.87). From 1992-2010, the incidence rate for all leukemias increased significantly among both Hispanic children (APC=1.16; 95%CI:0.55, 1.76) and non-Hispanic children (APC=0.52; 95%CI: 0.01, 1.04), with the average annual increase in incidence rates in Hispanic children more than twice that observed among non-Hispanic children (Table 3.1). Among Hispanic children, statistically significant increases were observed for White Hispanic children (APC=1.24; 95%CI: 0.67, 1.82), females (APC=1.24; 95%CI: 0.51, 1.99), children diagnosed at age 10-14 years (APC=1.72; 95%CI: 0.17, 3.31) or 15- 19 years (APC=2.65; 95%CI: 0.56, 4.79), and in the ALL subtype (APC=1.31; 95%CI: 0.68, 1.94). Among non-Hispanic children, statistically significant increases were only observed for females (APC=0.59; 95%CI: 0.01, 1.18), and overall for AML (APC=0.84; 95%CI: 0.09, 1.60). Rate differences between 1992-1993 and 2009-2010 were generally greater in Hispanic children than in non-Hispanic children, particularly among the oldest age group of children (age 15-19 years). When restricting analyses to ALL, Hispanic children experienced statistically significantly higher AAIRs than non-Hispanic children from 2006-2010 for all demographic groups (by race, gender and age, except in children <1yr) (Table 3.2). The difference in rates between Hispanic and non-Hispanic children was greatest among those in the oldest age group —Hispanic children aged 15-19 were 2.34 times as likely to be diagnosed with ALL as non-Hispanic children (95%CI: 1.98, 2.77). 56 57 Differences in APC between Hispanic and non-Hispanic children also were more pronounced among children diagnosed with ALL (Table 3.2). There was a statistically significant increase in the APC of ALL for Hispanic children (APC=1.31; 95%CI: 0.68, 1.94), but not for non-Hispanic children (APC=0.50, 95%CI -0.19, 1.19), although the difference did not reach statistical significance (P interaction =0.18). These patterns were also observed among Hispanic children for both genders, and among older children (aged 10- 14, or 15-19 years). Rate differences for ALL between the 1992-1993 and 2009-2010 average rates (Table 3.2) were greater in Hispanic children than in non-Hispanic children overall (P interaction = 0.032), and for children of a White race (P interaction = 0.036), among children 10-14 years of age (P interaction =0.025) and among children 15-19 years of age (P interaction = 0.001). The AAIRs for childhood AML from 2006-2010 overall were similar by ethnicity (Table 3.2). The incidence rate of AML was slightly higher for Hispanic White children than for non-Hispanic White children (RR=1.16; 95%CI:1.00, 1.35), and for Hispanic children aged 1-4, relative to non-Hispanic children in the same age group (RR=1.37; 95%CI:1.04, 1.79). The incidence rates were highest in children under 1 year of age for both ethnic groups (AAIR Hispanic =2.10; 95%CI:1.51, 2.85; AAIR non-Hispanic =1.93; 95%CI: 1.52, 2.42). When restricting analyses to AML diagnoses, no statistically significant differences in APC or rate difference between Hispanic children and non-Hispanic children were observed for any demographic group (P interaction >0.05) (Table 3.2). Among Hispanic children, a statistically significant increase in the incidence of AML was 57 58 observed only among children aged 1-4 (APC=4.38; 95%CI:1.39, 7.47). Among non- Hispanic children, a statistically significant increase in age-adjusted incidence rates was observed in females (APC=1.22; 95%CI:0.10, 2.35), and among children in the youngest age group (<1) from 1992-2010 (APC=4.41; 95%CI:1.13, 7.80). Trends in incidence rates of childhood leukemia from 1992-2010 by subtype and ethnicity are presented in Figure 3.1. Overall, the incidence of childhood leukemia increased at a greater rate for Hispanic than for non-Hispanic children, driven primarily by increases in incidence rates of ALL (APC P interaction = 0.18). When we examined trends in AAIR for the years in which histologic sub-type was available (2000-2009), the incidence of ALL NOS decreased in a similar pattern as observed for the increase in incidence of of B-cell ALL (Figure 3.2). During this same period, there was a slight increase in the AAIR of T-cell ALL. AAIRs from 2006-2010 by leukemia histologic subtype (ALL/AML/other leukemia) for Hispanic and non-Hispanic children are shown in Figure 3.3a and 3.3b. The incidence of ALL peaks at approximately 2-3 years of age for both ethnic groups and decreases until age 7-8 for Hispanics and through age 19 for non-Hispanics. In both Hispanic and non-Hispanic children, the rate of AML is highest among infants 0-1 years of age; a decrease in rates is observed from age 2-6, followed by a small, gradual increase in incidence through age 19. In figure 3.4, we present the age specific incidence rates for children diagnosed in 1992-1993 and 2009-2010 by ethnicity. In both Hispanic and non-Hispanic children, the incidence rates in children aged 3-4 were higher in more recent years (2009-2010). In Hispanic children, the incidence rates are higher in 2009-2010 in children diagnosed at 58 59 age 14 or older, relative to incidence rates from 1992-1992. In non-Hispanic children, the incidence rates were similar in all other age groups for children diagnosed in earlier and later years. Trends by age, gender and race The AAIR of ALL was greatest in children aged 1-4 years for both Hispanic and non-Hispanic ethnic groups (Figure 3.5a and 3.5b). The APC from 1992-2010 for both groups (data not shown) was similar and non-significant (APC Hispanic =0.62; 95%CI: -0.48, 1.73; APC non-Hispanic =0.61; 95%CI: -0.31, 4.54). Incidence rates were lower in children diagnosed at ages 5-9 years than 1-4 years and lower in children diagnosed at 10-19 years than both younger age groups. AAIRs of ALL for Hispanic children 10-19 years of age increased by an average of 3.14% per year (95%CI: 1.82, 4.49), but no increases were observed in non-Hispanic children (P interaction = 0.022; data not shown). The incidence of ALL was higher for males than for females in both ethnic groups (Figure 3.6). Increasing AAIRs were observed in Hispanic children from 1992-2010 for both males and females, with greater increases observed in males (APC male =1.44; 95%CI: 0.67, 2.22; APC female =1.12; 95%CI: 0.02, 2.33). Smaller, non-significant increases in incidence rates were observed from 1992-2010 for non-Hispanic children. AAIRs of ALL were highest in Hispanic White children (AAIR=5.01 per 100,000 persons) as compared to other racial/ethnic groups (Figure 3.7). A statistically significant increase in the incidence rate of ALL was observed among Hispanic White children from 1992-2010 (APC = 1.41; 95%CI: 0.80, 2.03), while no significant increase in rates was observed in any other race/ethnicity. 59 60 3.5 DISCUSSION The increase in childhood leukemia incidence rates overall from 1992-2010 appears largely to have been driven by increasing rates of ALL among Hispanic children. Although the incidence of childhood ALL was highest in younger children (1-4 years of age), the largest increase in incidence rates occurred in Hispanic children 10-19 years of age at diagnosis. The APC for 10-19 year olds from 1992-2010 was nearly five times that observed in Hispanic children diagnosed from 1-4 years of age. In contrast, the APC in non-Hispanic children was smaller and not statistically significant. The absolute change in incidence rates was greatest in Hispanic children aged 15-19 years at diagnosis (approximately 1.31 cases per 100,000 children in 1992-1993 compared to 3.15 cases per 100,000 children in 2009-2010). In the United States, with an estimated population of approximately 4.6 million Hispanic children aged 15-19 in 2012 (U.S. Census Bureau, Current Population Survey, Annual Social and Economic Supplement, 2012, (http://www.census.gov/population/hispanic/)), this amounts to more than 85 new cases in Hispanic children in this age group in 2012 than would have been expected had incidence rates remained stable from 1992-2010. Factors that may explain the increasing rates of childhood leukemia, such as changes in diagnostics or classification, or environmental or lifestyle factors, would need to differentially explain these patterns in Hispanic children compared to other ethnic groups. Exposure to factors such as parental alcohol or tobacco use, or exposure to ionizing radiation, are likely to have remained stable or decreased in prevalence over this time period. Other suspected risk factors for childhood leukemia that may explain the increase in incidence in Hispanic children, such as childhood atopic conditions, pesticide 60 61 exposure, and maternal and child weight, have increased during this time period, but may not be specific to Hispanic children and require further evaluation. A brief summary of the current literature on suspected risk factors for ALL and AML and the potential association with the observed trends in incidence rates is described below. Environmental and Lifestyle factors for ALL and AML In a recent meta-analysis, maternal consumption of alcohol during pregnancy (yes/no) was associated with risk of childhood AML (OR=1.56; 95%CI: 1.13, 2.15; 9 studies) in the offspring, but not with risk of ALL (OR=1.10; 95%CI: 0.93, 1.29; 11 studies)(25). The association was stronger among children diagnosed with AML younger than 4 years of age (OR=2.68; 95%CI: 1.85, 3.89; 5 studies), and when restricting to wine consumption during pregnancy and AML risk (OR=1.67; 95%CI: 1.21, 2.32; 4 studies)(25). No studies have evaluated whether the risk of childhood leukemia associated with alcohol consumption differs by ethnicity. However, the lack of association with childhood ALL suggests that the observed increases in ALL incidence among Hispanic children were not likely driven by increases in maternal consumption of alcohol. Exposure to paternal smoking early in life (i.e. prior to conception, during pregnancy, in early childhood, or while breastfeeding) is a suspected lifestyle risk factor for childhood ALL, though no association has been observed with maternal tobacco use. A recent meta-analysis, using data from 18 published case-control studies, observed an increased risk of childhood ALL with exposure to paternal smoking overall (OR=1.11; 95%CI:1.05, 1.16), prior to conception (OR=1.25; 95%CI:1.08, 1.46), and during 61 62 pregnancy (OR=1.24; 95%CI:1.07, 1.45)(9). A meta-analysis evaluating the effect of maternal smoking during pregnancy found no association, based on 20 published studies(26). Further, a study in California with a large proportion of Hispanic cases and controls reported similar suggestive increases in risk of childhood ALL for exposure to paternal smoking prior to conception for both Hispanic and non-Hispanic children (P heterogeneity =0.39), and no increased risk associated with maternal smoking for children of either ethnicity(27). Greater increases in risk were observed for AML, but risk estimates were unstable, and sample sizes were not sufficient to allow evaluation of risk by ethnicity(27). Adult tobacco use decreased following the publication of the 1964 S urg e on Ge n e ra l’s re po r t on S moki ng a nd He a lt h. Ac c o rding to a re po rt on tre nds in tobacco use, the percentage of current smokers in the United States decreased from 34.7% in 1970 to 20.6% in 2009 among adults aged 18 or older(28). Further, tobacco use is lower in persons of Hispanic ethnicity than in non-Hispanic White persons (22.2% v. 14.5% in 2009)(28). Tobacco use is therefore unlikely to explain an increase in incidence of ALL in Hispanic children. Day care attendance (as a surrogate for early life infections) has been suggested as a protective factor for childhood ALL, as reported in a meta-analysis of 14 case-control studies(14). One recent study observed differences in the association by Hispanic ethnicity. Protective, statistically significant effect estimates were observed for non- Hispanic White participants for day care attendance censored at different time points relative to the referent date. Among Hispanic children, when exposure was censored at 1 year prior to reference age, or at 1 year of age, elevated risk estimates were observed(29). A proposed mechanism through which early life infections may protect against childhood 62 63 leukemia incidenc e wa s de sc ribe d a s the “ d e la y e d infe c ti on” h y poth e sis (30). The “ de la y e d inf e c ti on” h y po thesis propo se s that a de la y in e x posure s to c omm on infe c ti ons during the first year of life may alter the development of the immune system, resulting in inadequate immune response. When children with an in utero preleukemic mutation are then exposed to such infections, this leads to excessive stress potentially resulting in additional chromosomal aberrations, leading to the development of leukemia. The “ d e la y e d infe c ti on” h y pothesis is sim il a r to t he “ h y gien e h y poth e sis ,” whic h was first proposed by Strachan in 1989 to explain the increase in allergy prevalence observed in western populations(31). Epidemiologic studies have noted a decreased risk of childhood allergies associated with several variables intended to measure early life infection (including higher birth order and day care attendance), supporting the hypothesis that infection early in life may protect against atopic conditions. Studies also show inverse associations between these same measures of infection (higher birth order and day care attendance) and childhood leukemia incidence. Such observations have prompted studies evaluating the association between allergies and childhood leukemia. Studies have shown an overall inverse association between childhood atopic conditions, including allergy, asthma, hay fever, eczema and hives, and ALL risk. Two meta- analyses of case-control studies observed a statistically significant protective effect of childhood ALL for allergy (8/6 studies), hay fever (5/3 studies), and eczema (5/5 studies), and a protective, non-significant effect for asthma (6/7 studies)(12, 13). In recent years, the prevalence of allergic conditions (skin allergies) in children has increased in the United States (1997-2011), with a lower prevalence of allergic conditions in Hispanic children compared to non-Hispanic children (32). The incidence of asthma in the United 63 64 States has also increased in recent years (2001-2010)(33). But, any increase in atopic conditions should result in a decrease in the incidence of childhood leukemia, and is therefore unlikely to explain the observed increases in childhood leukemia incidence, particularly in Hispanic children. One factor that may disproportionately affect Hispanic children in states covered by SEER registries is pesticides. Concern over pesticides as a potential risk factor for childhood leukemia has prompted numerous studies of the association of parental residential or occupational exposure, or ambient pesticide exposure to various classifications of pesticides (e.g. insecticides, herbicides) at different time periods early in the c hil d’s li fe (i.e . in utero, during early childhood). Meta-analyses of epidemiologic studies of childhood leukemia risk and residential exposure during pregnancy to pesticides, insecticides, and herbicides found elevated relative risk estimates of 1.54 (95%CI: 1.13, 2.11; 11 studies), 2.05 (95%CI: 1.80, 2.32; 8 studies), and 1.61 (95%CI: 1.20, 2.16; 5 studies), respectively(3). Slightly attenuated risks were observed for residential exposure to these agents in childhood(3). Meta-analyses of epidemiologic studies of maternal occupational exposure to pesticides and childhood leukemia found increased risks for maternal prenatal exposure overall (OR=2.09; 95%CI: 1.51, 2.88; 16 studies), with exposure to insecticides (OR=2.72; 95%CI: 1.47, 5.04; 6 studies) and herbicides (OR=3.62; 95%CI: 1.28, 10.3; 2 studies), and with exposure to any pesticides when restricting to ALL only (OR=2.64; 95%CI: 1.40, 5.00; 5 studies)(4). No associations were observed for paternal occupational exposure to pesticides and childhood leukemia(4). Several positive, but statistically non-significant associations were observed for ambient exposure to specific pesticides and childhood ALL based on 64 65 Pesticide Use Reporting (PUR) data and residential address(34). One study evaluated the association between childhood ALL and residential exposure to persistent organochlorine chemicals by ethnicity, measuring levels of 6 Polychlorinated Biphenyl (PCB) congeners and 6 specific chemical pesticides in carpet dust in the homes of case and control children. This study found stronger associations between PCB congeners and childhood leukemia risk in non-Hispanic White children than in Hispanic children, though levels of both PCB congeners were similar in Hispanic and non-Hispanic controls(35). To our knowledge, no other studies have reported associations between pesticide exposure and risk of childhood leukemia by ethnicity; thus, it is unclear whether Hispanic children may be at a greater risk of leukemia when exposed to pesticides, or whether the prevalence of pesticide exposure among Hispanic parents, or in residential areas with a high proportion of Hispanic families, has increased in recent decades. Either could contribute to the greater increase in incidence of childhood ALL among Hispanic children. Greater increases in obesity from 1999-2000 to 2007-2008 are observed in Hispanic women of childbearing age (aged 20-39) as compared to non-Hispanic White women(40) . A publi c a ti on of NH AN ES da ta foun d the pre va l e nc e of obe si t y ( B M I ≥ 30 ) increased from 30.6% in 1999-2000 to 39.6% in 2007-2008 in Hispanic women, and from 24.4% to 31.3% in non-Hispanic White women(40). While there is limited biological evidence for the association between maternal obesity and childhood leukemia in the offspring, it is plausible that mechanisms exist by which maternal obesity may lead to increased incidence of childhood leukemia. Future studies with data available on maternal weight characteristics should consider exploring this association. 65 66 Epidemiologic studies also suggest that birth weight is associated with increased risk of childhood leukemia. A recent meta-analysis of case-control studies, updating an earlier meta-analysis(11), found that children with high birth weight are at increased risk of both childhood ALL (OR=1.24; 95%CI: 1.16, 1.33; 23 studies) or AML (OR=1.24; 95% CI: 1.16, 1.32; 9 studies), with evidence of a dose-response trend for ALL (OR per 1Kg increase =1.18; 95%CI: 1.12, 1.23)(10). A recent pooled analysis of fetal growth and risk of childhood ALL using data from 12 studies participating in the Childhood Leukemia International Consortium (CLIC) found that children who were large-for-gestational-age were at an increased risk of childhood leukemia relative to children who were normal- for-gestational-age (OR=1.24; 95%CI: 1.13, 1.36)(41). However, to our knowledge, data on secular trends in birth weight or weight-for-gestational-age by ethnicity are not readily available, and as such, the influence of increasing birth weight on increasing incidence of ALL in Hispanic children is, at best, speculative. The observation of increasing rates among Hispanic children may reflect increasing exposure to lifestyle or environmental factors that may differ by their residential or geographic location, greater genetic susceptibility that has not yet been fully explored in epidemiologic research, or a combination of these factors. Alternatively, the increase in rates could reflect changes in the diagnosis or classification of leukemia. While improvements have been made in histologic classification of leukemia, it is not clear that there have been improvements in diagnostic practices that would identify new cases that previously would have been missed. The increase in incidence rates observed in Hispanic children may instead simply follow an earlier increase observed in White children, as Hispanic children become more acculturated and adopt a more Westernized 66 67 lifestyle. However, SEER data in regions with a historically large proportion of non- Hispanic white populations (e.g. Connecticut, Iowa, Utah), with data available as early as 1973, suggests that the increase in incidence rate among non-Hispanic White children has been small, and relatively constant from 1973-2010. Although the Hispanic classification was not available until 1992, a greater increase in incidence rates was observed from 1973-2010 when restricting to SEER registries that are likely to include Hispanic children (e.g. New Mexico, San Francisco- Oa kland ). F in a ll y , c l a ssi fic a ti on in S EE R a s “ Hispa nic ” is based on a surname algorithm. As such, the differences in the increase in incidence rates of ALL between Hispanic and non-Hispanic children may be artificially driven by changes in the classification of children as Hispanic over time. However, the observed difference in incidence rate increases by ethnicity has only been observed in childhood ALL and is not present in other childhood cancers (e.g. childhood brain tumors), suggesting that changes in ethnic classification are unlikely to be solely responsible for the results reported in this study. Conclusion Incidence rates in Hispanic children have increased significantly from 1992-2010, particularly among male children and those diagnosed between aged 10-19, while minimal increases were observed in non-Hispanic children and in younger children of either ethnicity. The increase in incidence rates among Hispanic children is unlikely to be explained by increasing exposure to known environmental risk factors (i.e. parental tobacco or alcohol use), or by changes in classification of disease, or classification of ethnicity in SEER databases. The completion of large, well-design studies of childhood 67 68 leukemia including genome wide susceptibility, environmental, and lifestyle data are needed, alongside secular trend data, to further elucidate the reasons for the observed increases in incidence rates among Hispanic children in the past two decades. 68 69 3.6 TABLES AND FIGURES Table 3.1. Age-Adjusted Incidence Rates (AAIR), Annual Percent Change (APC), And Rate Difference For All Childhood Leukemias By Ethnicity And Selected Demographic Characteristics Hispanic Non-Hispanic N AAIR a (95%CI) APC b (95%CI) RD c N AAIR a (95%CI) APC b (95%CI) RD c Rate Ratio (95%CI) Overall 2044 5.71 (5.47, 5.97) 1.16 (0.55, 1.76) 0.85 3256 3.95 (3.81, 4.09) 0.52 (0.01, 1.03) 0.33 1.45 (1.37, 1.53) Race White 1945 6.07 (5.80, 6.35) 1.24 (0.67, 1.82) 1.04 2329 4.23 (4.06, 4.41) 0.59 (-0.04, 1.22) 0.53 1.43 (1.35, 1.52) Black 59 3.89 (2.95, 5.02) NA NA 445 2.77 (2.51, 3.04) 0.48 (-0.50, 1.47) -0.05 1.41 (1.05, 1.85) Asian/Pacific Islander 14 NA NA NA 394 3.85 (3.48, 4.25) -0.39 (-1.71, 0.95) -0.45 NA American Indian 8 NA NA NA 44 3.77 (2.74, 5.06) NA NA NA Unknown 18 NA NA NA 44 NA NA NA NA Gender Male 1163 6.35 (5.99, 6.73) 1.24 (0.51, 1.99) 0.79 1803 4.27 (4.07, 4.47) 0.44 (-0.42, 1.31) 0.33 1.49 (1.38, 1.60) Female 881 5.04 (4.71, 5.38) 1.04 (-0.06, 2.15) 0.93 1453 3.61 (3.43, 3.80) 0.59 (0.01, 1.18) 0.34 1.39 (1.28, 1.52) Age <1 97 4.96 (4.03, 6.06) 0.48 (-2.20, 3.25) 1.34 155 3.99 (3.39, 4.67) 2.55 (-0.11, 5.28) 1.70 1.24 (0.95, 1.61) 1-4yrs 984 10.55 (9.83, 11.3) 0.91 (-0.11, 1.94) 1.84 1310 8.34 (7.89, 8.80) 0.54 (-0.31, 1.39) 0.86 1.27 (1.16, 1.38) 5-9yrs 445 5.24 (4.77, 5.75) -0.10 (-1.42, 1.24) -0.29 699 3.52 (3.26, 3.79) 0.10 (-1.02, 1.22) 0.21 1.49 (1.32, 1.68) 10-14yrs 364 4.21 (3.79, 4.67) 1.72 (0.17, 3.31) 0.34 553 2.60 (2.39, 2.83) 0.74 (-0.40, 1.90) 0.16 1.62 (1.42, 1.85) 15-19yrs 354 4.13 (3.71, 4.59) 2.65 (0.56, 4.79) 1.67 539 2.37 (2.18, 2.58) 0.08 (-1.45, 1.63) -0.05 1.74 (1.52, 1.99) Subtype Acute lymphocytic leukemia 1675 4.68 (4.46, 4.91) 1.31 (0.68, 1.94) 0.85 2493 3.04 (2.92, 3.16) 0.50 (-0.19, 1.19) 0.27 1.54 (1.45, 1.64) Acute myeloid leukemia 320 0.89 (0.80, 0.99) 0.63 (-1.27, 2.57) 0.14 681 0.81 (0.75, 0.87) 0.84 (0.09, 1.60) 0.12 1.10 (0.96, 1.26) Other leukemias 49 0.14 (0.10, 0.18) -0.79 (-4.72, 3.31) -0.13 82 0.10 (0.08, 0.12) -1.86 (-4.48, 0.84) -0.06 1.38 (0.95, 1.99) a AAIR per 100,000 persons for diagnoses from 2006-2010 using SEER18 data b APC for 1992-2010 using SEER13 data c Rate difference was calculated using the 2009-2010 average and 1992-1993 average rates per 100,000 persons 69 70 Table 3.2. Age-Adjusted Incidence Rates (AAIR), Annual Percent Change (APC), and Rate Difference for Childhood ALL and AML by Ethnicity and Selected Demographic Characteristics Hispanic Non-Hispanic N AAIR a (95%CI) APC b (95%CI) RD N AAIR a (95%CI) APC b (95%CI) RD Rate Ratio (95%CI) p-value d p-value e ALL 1675 4.68 (4.46, 4.91) 1.31 (0.68, 1.94) 0.85 2493 3.04 (2.92, 3.16) 0.50 (-0.19, 1.19) 0.27 1.54 (1.45, 1.64) 0.032 0.184 Race White 1605 5.01 (4.77, 5.26) 1.41 (0.80, 2.03) 1.02 1838 3.36 (3.21, 3.52) 0.64 (-0.11, 1.40) 0.49 1.49 (1.39, 1.60) 0.036 0.22 Black 43 2.85 (2.06, 3.85) NA NA 296 1.85 (1.65, 2.07) -0.25 (-2.06, 1.58) -0.25 1.54 (1.09, 2.13) Asian/PI 8 NA NA NA 300 2.94 (2.61, 3.29) 0.15 (-1.20, 1.52) -0.14 NA Amer. Indian 6 NA NA NA 28 2.40 (1.60, 3.48) NA NA NA Unknown 13 NA NA NA 31 NA NA NA NA Gender Male 962 5.26 (4.93, 5.60) 1.44 (0.67, 2.22) 0.97 1419 3.38 (3.20, 3.56) 0.45 (-0.25, 1.43) 0.31 1.56 (1.43, 1.69) 0.065 0.29 Female 713 4.07 (3.78, 4.39) 1.12 (0.02, 2.23) 0.73 1074 2.69 (2.53, 2.85) 0.53 (-0.34, 1.42) 0.23 1.52 (1.38, 1.67) 0.24 0.46 Age <1 47 2.41 (1.77, 3.20) -1.54 (-5.04, 2.08) -0.34 73 1.88 (1.47, 2.37) 1.69 (-1.76, 5.26) 0.20 1.28 (0.87, 1.87) 0.24 0.23 1-4yrs 682 9.18 (8.50, 9.90) 0.62 (-0.48, 1.73) 1.34 1142 7.27 (6.85, 7.70) 0.61 (-0.31, 4.54) 0.94 1.26 (1.15, 1.39) 0.9 0.93 5-9yrs 398 4.69 (4.24, 5.17) 0.24 (-1.11, 1.61) 0.19 600 3.02 (2.78, 3.27) 0.40 (-0.78, 1.60) 0.52 1.55 (1.37, 1.77) 0.86 0.94 10-14yrs 283 3.28 (2.90, 3.68) 2.53 (0.72, 4.38) 0.38 378 1.78 (1.60, 1.97) 0.15 (-1.19, 1.50) -0.11 1.84 (1.57, 2.16) 0.025 0.12 15-19yrs 265 3.09(2.73, 3.49) 3.48 (1.05, 5.97) 1.84 300 1.32 (1.18, 1.48) 0.75 (-1.49, 3.05) -0.09 2.34 (1.98, 2.77) 0.001 0.009 AML 320 0.89 (0.80, 0.99) 0.63 (-1.27, 2.57) 0.14 681 0.81 (0.75, 0.87) 0.84 (0.09, 1.60) 0.12 1.10 (0.96, 1.26) 0.56 0.43 Race White 294 0.91 (0.81, 1.02) 0.54 (-1.41, 2.53) 0.13 442 0.78 (0.71, 0.86) 0.83 (-0.32, 2.00) 0.12 1.16 (1.00, 1.35) 0.63 0.51 Black 15 0.98 (0.55, 1.62) NA NA 131 0.80 (0.67, 0.95) 2.91 (-0.01, 5.91) 0.28 1.22 (0.66, 2.09) Asian/PI 5 NA NA NA 85 1.01 (0.52, 1.77) -2.35 (-5.07, 0.44) -0.34 NA Amer.Indian 2 NA NA NA 12 0.83 (0.66, 1.02) NA NA NA Unknown 4 NA NA NA 11 NA NA NA NA Gender Male 168 0.91 (0.78, 1.06) 0.35 (-1.54, 2.26) 0.00 341 0.79 (0.71, 0.88) 0.51 (-0.61, 1.65) 0.08 1.15 (0.95, 1.39) 0.74 0.65 Female 152 0.87 (0.74, 1.02) 1.07 (-1.92, 4.16) 0.28 340 0.83 (0.74, 0.92) 1.22 (0.10, 2.35) 0.15 1.05 (0.86, 1.28) 0.54 0.43 Age <1 41 2.10 (1.51, 2.85) 5.40 (-0.37, 11.5) 2.65 75 1.93 (1.52, 2.42) 4.41 (1.13, 7.80) 1.42 1.09 (0.72, 1.61) 0.31 0.28 1-4yrs 92 1.24 (1.00, 1.52) 4.38 (1.39, 7.47) 0.54 142 0.90 (0.76, 1.07) -0.23 (-2.71, 2.32) -0.15 1.37 (1.04, 1.79) 0.11 0.16 5-9yrs 35 0.41 (0.29, 0.57) -3.40 (-7.14, 0.50) -0.28 84 0.42 (0.34, 0.52) -0.08 (-3.65, 3.61) -0.26 0.98 (0.64, 1.46) 0.88 0.41 10-14yrs 72 0.83 (0.65, 1.05) -1.55 (-4.91, 1.93) 0.14 155 0.73 (0.62, 0.85) 2.15 (0.33, 4.00) 0.37 1.14 (0.85, 1.52) 0.079 0.09 15-19yrs 80 0.93 (0.74, 1.16) 0.10 (-4.30, 4.72) -0.25 225 0.99 (0.87, 1.13) 0.07 (-2.09, 2.27) 0.18 0.94 (0.72, 1.22) 0.84 0.77 a AAIR per 100,000 persons, diagnoses from 2006-2010, SEER18 data b APC for 1992-2010 using SEER13 data c Rate difference calculated using the 2009-2010 average and 1992-1993 average rates d P-value for test of difference in APC, 1992-2010 e P-value for test of difference in absolute change in incidence rates from 1992-2010 70 71 Figure 3.1. Age-Adjusted Incidence Rates By Ethnicity And Subtype, SEER 13, 1992-2010 0 1 2 3 4 5 6 Age-adjusted incidence rate (per 100,000) Year of Diagnosis Hispanic ALL Non-Hispanic ALL Hispanic AML Non-Hispanic AML 71 72 Figure 3.2. Age-Adjusted Incidence Rates for ALL by Precursor Cell Subtype, SEER 18, 2000-2009 0 0.5 1 1.5 2 2.5 3 3.5 Age-Adjusted Incidence Rate (per 100,000 person years) Year of Diagnosis NOS B-cell T-cell Burkitt 72 73 Figure 3.3. Age-Specific Incidence Rates by Subtype for A) Hispanic Children, and B) Non-Hispanic Children, SEER 13, 1992-2010 a. b. 0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Age Specific incidence rates (per 100,000) Age at diagnosis Non-Hispanic ALL Non-Hispanic AML Non-Hispanic Other 0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Age Specific incidence rate (per 100,000) Age at diagnosis Hispanic ALL Hispanic AML Hispanic Other 73 74 Figure 3.4. Age-Specific Incidence Rates for Childhood ALL for A) Hispanic and B) Non-Hispanic Children Diagnosed in 1992-1993 and 2009-2010, SEER 13 a. b. 0 2 4 6 8 10 12 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Age SpecificIncidence Rate (per 100,000) Age at Diagnosis Hispanic 1992-1993 Hispanic 2009-2010 0 2 4 6 8 10 12 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Age Specific Incidence Rate (per 100,000) Age at Diagnosis Non-Hispanic 1992-1993 Non-Hispanic 2009-2010 74 75 Figure 3.5. Age-Adjusted Incidence Rates by Age at Diagnosis for A) Hispanic and B) Non-Hispanic Children, SEER 13, 1992-2010 a. b. 0 2 4 6 8 10 12 Age Adjusted Incidence Rate (per 100,000) Year of Diagnosis Hispanic 1-4yrs Hispanic 5-9yrs Hispanic 10-19yrs 0 2 4 6 8 10 12 Age Adjusted Incidence Rate (per 100,000) Year of Diagnosis Non-Hispanic 1-4yrs Non-Hispanic 5-9yrs Non-Hispanic 10-19yrs 75 76 Figure 3.6. Age-Adjusted Incidence Rates by Gender and Ethnicity, SEER 13, 1992-2010 Figure 3.7. Age-Adjusted Incidence Rates by Race/Ethnicity, SEER 13, 1992- 2010 0 1 2 3 4 5 6 Age Adjusted Incidence Rate (per 100,000) Year of Diagnosis Hispanic Male Hispanic Female Non-Hispanic Male Non-Hispanic Female 0 1 2 3 4 5 6 Age Adjusted Incidence Rate (per 100,000) Year of Diagnosis Hispanic White Non-Hispanic White Asian/Pacific Islander Non-Hispanic Black 76 77 3.7 REFERENCES 1. Howlader N, Noone AM, Krapcho M, et al. SEER Cancer Statistics Review, 1975-2010. Bethesda, MD: National Cancer Institute, 2013. 2. Wiemels J. Perspectives on the causes of childhood leukemia. Chemico- Biological Interactions 2012;196(3):59-67. 3. Turner MC, Wigle DT, Krewski D. Residential Pesticides and Childhood Leukemia: A Systematic Review   and Meta-Analysis. Environmental Health Perspectives 2009. 4. Wigle DT, Turner MC, Krewski D. A Systematic Review and Meta-analysis of Childhood Leukemia and Parental Occupational Pesticide Exposure. Environmental Health Perspectives 2009;117(10):1505-13. 5. Kheifets L, Ahlbom A, Crespi CM, et al. A pooled analysis of extremely low- frequency magnetic fields and childhood brain tumors. American Journal of Epidemiology 2010;172(7):752-61. 6. Ahlbom A, Day N, Feychting M, et al. A pooled analysis of magnetic fields and childhood leukaemia. Br J Cancer 2000;83(5):692-8. 7. Greenland S, Sheppard AR, Kaune WT, et al. A pooled analysis of magnetic fields, wire codes, and childhood leukemia. Childhood Leukemia-EMF Study Group. Epidemiology 2000;11(6):624-34. 8. Milne E, Greenop KR, Scott RJ, et al. Parental Prenatal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia. American journal of epidemiology 2011;175(1):43-53. 9. Liu R, Zhang L, McHale CM, et al. Paternal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia: Systematic Review and Meta-Analysis. Journal of Oncology 2011;2011:1-16. 10. Caughey RW, Michels KB. Birth weight and childhood leukemia: A meta- analysis and review of the current evidence. International Journal of Cancer 2008;124(11):2658-70. 11. Hjalgrim LL. Birth Weight as a Risk Factor for Childhood Leukemia: A Meta- Analysis of 18 Epidemiologic Studies. American Journal of Epidemiology 2003;158(8):724-35. 12. Linabery AM, Jurek AM, Duval S, et al. The Association Between Atopy and Childhood/Adolescent Leukemia: A Meta-Analysis. American Journal of Epidemiology 2010;171(7):749-64. 13. Dahl S, Schmidt LS, Vestergaard T, et al. Allergy and the risk of childhood leukemia: a meta-analysis. Leukemia 2009;23(12):2300-4. 14. Urayama KY, Buffler PA, Gallagher ER, et al. A meta-analysis of the association between day-care attendance and childhood acute lymphoblastic leukaemia. International Journal of Epidemiology 2010;39(3):718-32. 15. Chokkalingam AP, Buffler PA. Genetic susceptibility to childhood leukaemia. Radiation Protection Dosimetry 2008;132(2):119-29. 16. Yan J, Yin M, Dreyer ZE, et al. A meta-analysis of MTHFR C677T and A1298C polymorphisms and risk of acute lymphoblastic leukemia in children. Pediatr Blood Cancer 2012;58(4):513-8. 77 78 17. Healy J, Richer C, Bourgey M, et al. Replication analysis confirms the association of ARID5B with childhood B-cell acute lymphoblastic leukemia. Haematologica 2010;95(9):1608-11. 18. Papaemmanuil E, Hosking FJ, Vijayakrishnan J, et al. Loci on 7p12.2, 10q21.2 and 14q11.2 are associated with risk of childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1006-10. 19. Prasad RB, Hosking FJ, Vijayakrishnan J, et al. Verification of the susceptibility loci on 7p12.2, 10q21.2, and 14q11.2 in precursor B-cell acute lymphoblastic leukemia of childhood. Blood 2010;115(9):1765-7. 20. Treviño LR, Yang W, French D, et al. Germline genomic variants associated with childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1001-5. 21. Sherborne AL, Hosking FJ, Prasad RB, et al. Variation in CDKN2A at 9p21.3 influences childhood acute lymphoblastic leukemia risk. Nature Genetics 2010;42(6):492-4. 22. Chokkalingam AP, Hsu LI, Metayer C, et al. Genetic variants in ARID5B and CEBPE are childhood ALL susceptibility loci in Hispanics. Cancer Causes Control 2013;24(10):1789-95. 23. Han S, Lee K-M, Park SK, et al. Genome-wide association study of childhood acute lymphoblastic leukemia in Korea. Leukemia Research 2010;34(10):1271-4. 24. Bonaventure A, Goujon-Bellec S, Rudant J, et al. Maternal smoking during pregnancy, genetic polymorphisms of metabolic enzymes, and childhood acute leukemia: the ESCALE Study (SFCE). Cancer causes & control : CCC 2011;23(2):329-45. 25. Latino-Martel P, Chan DSM, Druesne-Pecollo N, et al. Maternal Alcohol Consumption during Pregnancy and Risk of Childhood Leukemia: Systematic Review and Meta-analysis. Cancer Epidemiology Biomarkers & Prevention 2010;19(5):1238-60. 26. Klimentopoulou A, Antonopoulos CN, Papadopoulou C, et al. Maternal smoking during pregnancy and risk for childhood leukemia: A nationwide case-control study in Greece and meta-analysis. Pediatric Blood & Cancer 2011;58(3):344-51. 27. Chang JS. Parental Smoking and the Risk of Childhood Leukemia. American journal of epidemiology 2006;163(12):1091-100. 28. Trends in Tobacco Use. American Lung AssociationResearch and Program Services Epidemiology and Statistics Unit, 2011. 29. Ma X, Buffler PA, Wiemels JL, et al. Ethnic difference in daycare attendance, early infections, and risk of childhood acute lymphoblastic leukemia. Cancer Epidemiol Biomarkers Prev 2005;14(8):1928-34. 30. Inaba H, Greaves M, Mullighan CG. Acute lymphoblastic leukaemia. Lancet 2013;381(9881):1943-55. 31. Chang JS, Wiemels JL, Buffler PA. Allergies and childhood leukemia. Blood Cells, Molecules, and Diseases 2009;42(2):99-104. 32. Jackson KD HL, Akinbami LJ. Trends in allergic conditions among children: United States, 1997-2011. Hyattsville, MD: National Center for Health Statistics, 2013, 78 79 33. Akinbami LJ MJ, Baily C, Zahran HS, King M, Johnson CA, Liu X. Trends in Asthma Prevalence, Health Care Use, and Mortality in the United States, 2001- 2010. Hyattsville, MD: National Center for Health Statistics, 2012, 34. Rull RP, Gunier R, Von Behren J, et al. Residential proximity to agricultural pesticide applications and childhood acute lymphoblastic leukemia. Environ Res 2009;109(7):891-9. 35. Ward MH, Colt JS, Metayer C, et al. Residential exposure to polychlorinated biphenyls and organochlorine pesticides and risk of childhood leukemia. Environ Health Perspect 2009;117(6):1007-13. 36. Cynthia O, Margaret C. Prevalence of Obesity Among Children and Adolescents: United States, Trends 1963 –1965 Through 2007 –2008. Center for Disease Control, National Center for Health Statistics, 2010. 37. Freedman DS, Khan LK, Serdula MK, et al. Racial and ethnic differences in secular trends for childhood BMI, weight, and height. Obesity (Silver Spring) 2006;14(2):301-8. 38. Aldhafiri FK, McColl JH, Reilly JJ. Prevalence of being underweight and overweight and obesity at diagnosis in UK patients with childhood acute lymphoblastic leukaemia 1985-2002. J Hum Nutr Diet 2013. 39. Stamatakis E, Wardle J, Cole TJ. Childhood obesity and overweight prevalence trends in England: evidence for growing socioeconomic disparities. Int J Obes (Lond) 2010;34(1):41-7. 40. Flegal KM, Carroll MD, Ogden CL, et al. Prevalence and trends in obesity among US adults, 1999-2008. JAMA 2010;303(3):235-41. 41. Milne E, Greenop KR, Metayer C, et al. Fetal growth and childhood acute lymphoblastic leukemia: Findings from the childhood leukemia international consortium. Int J Cancer 2013;133(12):2968-79. 79 80 CHAPTER 4: Gene-Smoking Interactions and Risk of Childhood Acute Lymphocytic Leukemia among Hispanic children in a Genome-Wide Association Study 4.1 ABSTRACT Known genetic and environmental risk factors explain only a small proportion of the incidence of childhood acute lymphocytic leukemia (ALL), however, genetic variants may increase a chi ld’s susce pti bil it y to the e f fe c ts of pa re ntal smok ing e a rl y in li fe . In this analysis, we used novel statistical methods to scan the genome for gene-parental smoking interactions to identify potential susceptibility loci. Participants were self- identified Hispanic cases and controls from phases I-III of the California Childhood Leukemia Study (CCLS). Cases (N=380) were children <15 years of age at diagnosis, identified from participating hospitals in California. Controls (N=454) were matched to cases on date of birth, gender, and maternal race. Detailed maternal and paternal smoking history was collected by self-administered questionnaire (paternal data, phase I) or in- person interview (maternal data and paternal data from phase II/III). Genome-wide sequencing was conducted using DNA from archival dried blood spot samples using the Illumina Human OmniExpress v.1 platform. Data was evaluated for the presence of gene- parental smoking interaction using a series of traditional analytic methods and novel, statist ica ll y e f fic ient t wo step sc a nnin g met hods, im pleme nted using “ Gx Esc a n” (http://biostats.usc.edu/software). We sought to replicate our most significant SNPs in two independent case-only studies of childhood leukemia in France (Etude Sur les Cancers et let L e u c e ’mie s de l’Enf a nt (E S C A LE) stud y , n= 441 ), a nd Austra li a (A US - 80 81 ALL study, n=285). We identified a list of 8 SNPs for replication based on two-step scanning procedures. Of these, one SNP for maternal smoking prior to pregnancy and during pregnancy was statistically significant in the AUS-ALL study both for all ALL diagnoses and when restricting to B-Cell ALL diagnoses. Genotyping data for this SNP was not available in the ESCALE study. The strongest association in both the CCLS and AUS-ALL studies was observed for maternal smoking during pregnancy when restricting to B-Cell ALL (summary OR = 4.40; 95% CI: 2.53, 7.64). Novel two-step scanning methods for gene-environment interactions (genome-wide interaction scan (GWIS) analysis) can be used to evaluate GxE interactions in studies with relatively small sample sizes while maintaining sufficient power to detect potential susceptibility loci. Results from the CCLS analysis and replication in the ESCALE and AUS-ALL studies indicate potential novel susceptibility loci for maternal smoking during pregnancy and risk of B- cell ALL. However, additional studies should be conducted to confirm these results in a larger study population. 4.2 INTRODUCTION Genome-wide association studies (GWAS) of childhood acute lymphocytic leukemia (ALL) have identified several promising associations between polymorphisms in ARID5B, IKZF1, CEBPE, and CDKN2A/B and risk of ALL (1-5). Similarly, a study conducted in Hispanic and non-Hispanic children in the California Childhood Leukemia Study (CCLS) found the same associations for ARID5B (all subtypes), and CEBPE (when restricting to B-cell ALL) for both Hispanic and non-Hispanic subjects. The associations for two IKZF1 variants and risk of ALL were present among non-Hispanic white children, 81 82 but not among Hispanic children; mixed results were observed for variants in CDKN2A (6). Other GWAS and candidate gene studies have found additional variants that may be associated with increased risk of childhood ALL, but results have not yet been validated (7-9). Numerous epidemiologic studies have evaluated the association between parental smoking at different time periods and risk of childhood ALL, as characterized in 5 meta- analyses (10-14). Overall, summary results indicate that paternal smoking is associated with an approximate 10-20% increase in risk, while no apparent association was found between maternal smoking and risk of ALL. The most recent meta-analysis, summarizing data from 18 published case-control studies, found that children exposed to paternal smoking (overall, prior to conception, and during pregnancy) were statistically significantly more likely to be diagnosed with childhood ALL than children not exposed to tobacco smoke (Overall: OR=1.11, 95%CI: 1.05, 1.18; Prior to conception: OR=1.24, 95%CI: 1.07, 1.43; During pregnancy: OR=1.25, 95%CI: 1.08, 1.46)(10). A second meta-analysis using 10 of these studies found similar results for smoking (never/ever) around the time of conception, and greater risk observed for higher levels of smoking (>20 cigarettes/day) (around conception: OR=1.15, 95%CI: 1.06, 1.24; OR >20cig/day =1.44, 95%CI: 1.24, 1.68)(11). For maternal smoking during pregnancy, a meta-analysis based on 20 published studies found no evidence of an increase in ALL risk with for exposed children (OR = 1.03; 95%CI: 0.96, 1.12)(13). It is likely that both genetic and environmental risk factors contribute to childhood ALL. The strength of association between paternal or maternal smoking and risk of childhood leukemia may be higher for children with greater genetic susceptibility to 82 83 carcinogenic chemicals in tobacco smoke. If only a subset of all exposed children are susceptible, the average risk of childhood leukemia across all children in a population may be modest compared to those who are genetically susceptible. To explore this hypothesis, four published studies have evaluated gene-environment interactions by a ssessing a c hil d’s risk of A LL a ssocia t e d with pa re ntal smok in g b y v a r iants in g e ne s involved in the activation or detoxification of chemicals in tobacco smoke (9, 12, 15, 16). A case-control study conducted to evaluate the interaction between maternal smoking during pregnancy and six different SNPs found that maternal smoking was associated with ALL among children with genetic variants in the CYP1A1( *2A) or GSTM1 (null) genes (15). Another case-control study found that among children with a high-risk haplotype for 5 CYP1A1 polymorphisms (including CYP1A1 *2A), risk of ALL was significantly increased among children whose fathers smoked during pregnancy (OR=2.8; 95%CI: 1.5, 5.3) (12). The remaining studies found statistically non-significant interactions for maternal smoking around the time of pregnancy (by trimester, and in early childhood) and 3 CYP1A1 polymorphisms (16), or for maternal smoking during pregnancy and 7 SNPs in 5 genes, including CYP1A1, NQO1, CYP2E1*5, EPHX1, and NAT2*5 (9). These candidate gene studies have focused on SNPs for interaction analyses that are likely involved in tobacco smoke metabolism, but have not looked at any interactions with variants, or genes, that have been discovered in GWAS. In the current study, we propose to use an agnostic approach to scan the genome for gene-parental smoking interactions for childhood ALL using novel two-step statistical methods. No study to date has reported on potential gene-environment interactions for parental smoking and ALL using genome-wide data, possibly because only recently have 83 84 statistically efficient methods for genome-wide interaction analysis become available. We conducted the current analysis to identify additional variants that may be associated with susceptibility to tobacco smoke and risk of childhood ALL using existing genome- wide data from the California Childhood Leukemia Study. 4.3 MATERIALS AND METHODS Subjects The current analysis includes participants from phases I, II, and III of the California Childhood Leukemia Study (CCLS), an ongoing population based case-control study. Phase I of the CCLS (1995-1999) included participants from 17 counties in the San Francisco Bay area. Phases II (2000-2002) and III (2003-2008) include cases and controls from a total of 35 counties in northern and central California. Cases (n=380) were identified from participating hospitals in Northern and Central California at the time of diagnosis and include children of Hispanic ethnicity aged 0-14 living in counties participating in CCLS. Cases were diagnosed with childhood ALL between 1995 and 2008. Eighty-five percent of eligible cases participated in the study. Controls (n=454) were identified from the California Birth Registry and were matched to cases on date of birth, gender, and maternal race. The recruitment process for controls is detailed elsewhere(17). Briefly, up to 6 potential controls were identified from birth certificate files based on the matching criteria. If the first randomly selected subject declined participation, the next eligible control was identified for recruitment by study interviewers. The participation rate among controls was 59%. 84 85 Tw o indepe nde nt c a s e - onl y studi e s of c hil dhood leuke mi a fr om F ra n c e (E tude S ur les C a nc e rs e t les L e uc e mi e s de l’Enf a nt (E S C A L E) stud y n= 441) (18) and Australia (AUS-ALL study; n=285)(11, 18) served as replication data sets. In the ESCALE study, cases were children diagnosed with ALL in 2003-2004 and identified through the French National Registry of Childhood Hematopoietic Malignancies (NRCH). In the AUS-ALL study, cases were identified from one of 10 pediatric oncology centers in Australia between 2003-2007. In Australia, nearly all children are diagnosed and treated at these 10 oncology centers. Exposure to parental smoking Maternal smoking history was collected by in-person interview with the mothers of cases and controls for all participants in the study. Paternal smoking history was collected using either self-administered questionnaires (phase I) or in-person interview (phase II/III) with the mothers of cases and controls. I n thi s stud y , ‘ e ve r ’ ha ving sm oked wa s de fine d a s ha vin g e ve r smok e d 100 c igar e tt e s in the pa r e nts’ li fe ti m e . Information was collected on whether fathers smoked during the three months prior to conception, and whether mothers ever smoked during the three months prior to conception, during pregnancy, while breast- fe e din g , or in e a rl y c h il dhood (f rom birth to t he c hil d’s 3 rd birthday or date of diagnosis). For any time periods during which the mother smoked, average dose information (cigarettes/day) was collected. The ESCALE study included information on maternal smoking during pregnancy only, collected by phone interview with the mothers of cases. The AUS-ALL study included information on annual lifetime maternal smoking, collected by self-administered 85 86 questionnaire, or by follow-up phone interview when necessary. Children in the AUS- ALL study were considered exposed to maternal smoking prior to conception if the mother smoked a) 2 calendar years prior to the year of birth of the child if the child was born in January-May, or b) 1 calendar year prior if the child was born June-December. The child was considered exposed to maternal smoking during pregnancy if the mother smoked a) in the calendar year prior to the year of birth of the child if the child was born in January-May, or b) in the same calendar year of the birth if the child was born in June- December. Genotyping Genome-wide sequencing was conducted using DNA from archival dried blood spot samples using the Illumina Human OmniExpress v.1 platform (Illumina Inc, San Diego, CA, http://www.illumina.com) at the University of California Berkeley School of Public Health Genetic Epidemiology and Genomics Laboratory. DNA was extracted from blood spots using the QIAamp DNA Mini Kit (QIAGEN, USA, Valencia, CA); blood spots were available for 87% of interviewed participants. Of the 730,525 SNPs genotyped, 96,488 SNPs were excluded from final analyses for the following reasons (figure 4.1a): a) non-autosomal (21,167 SNPs), b) Hardy Weinberg equilibrium p-value < 10 -5 (1,034 SNPs), c) SNP call rate <98% (5,351 SNPs), d) minor allele frequency (MAF) <2% (68,936 SNPs), resulting in a total of 634,037 SNPs remaining for inclusion in all analyses. A total of 848 samples were genotyped, including 5 duplicate case specimens and 5 duplicate control specimens. After removing duplicates, 380 cases, and 454 controls 86 87 had genotyping data that met quality control criteria for inclusion in the final analysis. Individual samples were excluded from final analyses for the following reasons (figure 4.1b): a) sex check failure (2 samples), b) call rate <98% (1 sample). For approximately 1% of cases and controls (5 cases and 5 controls), duplicate specimens were analyzed to assess concordance. Concordance was greater than 99.9% for all duplicates. For each pair of duplicate specimens, the sample with the higher call rate was used in analysis. Replication dataset. Genotyping for cases in the ESCALE study was completed using blood samples obtained at study enrollment using the Human 370-Quad Illumina beadchip. In the AUS-ALL study, genotyping was completed using one of 3 platforms: Illumina 610K-Quad beadchip, Illumina 370K-Duo beadchip, or Illumina 370K-Quad beadchip. Statistical Analysis In the CCLS, data was evaluated for the presence of any gene-environment interactions using a series of traditional methods, and newer, statistically more efficient two-step procedures. Analyses were completed for 1) All ALL cases and controls, 2) B- cell ALL cases only and all controls, 3) ALL cases diagnosed between 1-10 years of age (low-risk ALL), and controls 1-10 years of age at reference date. Two traditional approaches included: 1) Case-control analysis with an interaction term to test the null hypothesis β GxE =0, using the model: Logit( P r( D = 1 | G) = β 0 + β G G + β E E + β GxE GxE. (1) 87 88 2) Case-only analysis to test the null hypothesis γ GxE = 0, using the model: Logit(Pr(E=1|G, D=1) = γ 0 + γ GxE G (2) Three two-step procedures were also used to evaluate GxE interactions in the CCLS data. Two-step approaches may be more statistically efficient than traditional approaches to GxE interaction analysis. In this analysis, we used an initial screen in the first step to select SNPs to be tested in the second step, with a predetermined threshold level of sig nific a n c e , α 1 . Then, in the second step, we used a formal test for interaction using model (1) above, testing only SNPs that were selected from the first step, with a Bonferroni adjustment to correct for multiple comparisons. We used three different two- step scanning procedures to evaluate the presence of gene-environment interactions: 1) 2-step, DG | GxE, in which the marginal disease-gene (DG) association was used to screen SNPs at a step- 1 si g nific a nc e leve l α 1 (19). 2) 2-step, EG | GxE, in which the marginal environment-gene (EG) association was used to screen SNPs, again at a step- 1 si g nifi c a nc e lev e l α 1 (20). 3) 2-step, EDG | GxE, in which a test statistic (2 degrees of freedom) combining the results of DG and EG associations was used to screen SNPs at a step-1 significance leve l α 1 (21) We used unconditional logistic regression models in step 1 to evaluate marginal associations (DG, EG, or EDG) with childhood ALL where genotype was modeled using a log additive approach. Models were adjusted for gender, age at diagnosis or reference age, and four principal components (PC). In all step 2 models, formal tests of interaction were conducted using model (1), adjusting for the same covariates. 88 89 Analyses using both traditional and two-step methods were completed using PLINK software for whole genome association analysis v.1.07, available at http://pngu.mgh.harvard.edu/purcell/plink/, and R v2.15.168, available at http://www.r- project.org/. Principal components analysis to model ancestry differences was completed using EIGENSTRAT within the EIGENSOFT statistical package, v 3.0, available at http://genepath.med.harvard.edu/~reich/Software.htm. Replication Replication of top SNPs was conducted in two independent case-only studies of childhood ALL: the ESCALE study in France (N=441), and in the AUS-ALL study in Australia (N=285). For the ESCALE study, we analyzed the association between maternal smoking during pregnancy and SNPs selected for replication from the main GxE interaction analysis in a case-only analysis. For the AUS-ALL study, we also used a case-only approach to analyze both maternal smoking prior to pregnancy and maternal smoking during pregnancy with each of the identified SNPs. A final list for replication was obtained from initial GWIS analyses of the CCLS data, which were completed using GxEscan. A list of the most highly significant SNPs was generated based on all two-step approaches for each subgroup (ALL, B-cell ALL only, ALL diagnosed from 1-10 years) (see Supplemental Tables 4.1-4.3). SNPs were first sorted by p-value to identify a list of most promising candidates for replication. When the original SNP was available on the replication platforms in the ESCALE or AUS-ALL studies, that SNP was sent for replication. For the remaining SNPs, SNAP (www.broadinstitute.org/mpg/snap/) was used to find all other SNPs that were available 89 90 on the replication platforms and that were in linkage disequilibrium (LD) with our top candidates for replication in both non-Hispanic and Hispanic populations, using the 1000 genomes project and a minimum R 2 of 0.8. From the list of possible SNPs, the SNP with the highest LD (in both populations) with the SNP from our list of candidates was sent for replication. Replication analyses were completed using model (2) above in a case-only analysis to evaluate interaction. In the ESCALE study, analyses were completed for maternal smoking during pregnancy. In the AUS-ALL study, analyses were completed both for maternal smoking prior to pregnancy and maternal smoking during pregnancy. No replication was completed for maternal smoking during early childhood or paternal smoking during pregnancy, as exposure data was not available in either of the replication studies. All models were adjusted for age, gender, and study-specific principal components to account for population stratification. We conducted a meta-analysis to compute summary effect estimates for SNPs that replicated in both populations, or in one population when the SNP was not available on the platform used in the other population. Variance-weighted summary effect estimates and corresponding 95% confidence intervals were computed using results from both the CCLS and replication studies based on case-only analysis results from each study, using STATA v13.1. 4.4 RESULTS 90 91 The age and gender distributions were similar for cases and controls, and were similar across study site (Table 4.1). The majority of children (80.7%-83.5%) were between 1-10 years of age at the index date (diagnosis date (cases) or reference date (controls)). In the CCLS, 85.0% of all cases were B-cell ALL, as were 90.0% of cases in the ESCALE study, and 88.8% of cases in the AUS-ALL study. Paternal smoking prior to conception was more prevalent among cases than controls (29.4% v. 25.6%) in the CCLS. Maternal smoking prior to conception was less common in this sample, but was approximately the same in both cases and controls (11.3% v. 11.2%). In the AUS-ALL sample, maternal smoking prior to conception was far more prevalent (26.3%). Maternal smoking during pregnancy was greater in cases than in controls (8.7% v. 6.8%) in the CCLS, and was more common among case mothers in both the ESCALE (24.7%) and AUS-ALL (20.7%) studies. Maternal smoking while breastfeeding was uncommon among both cases and controls, while maternal smoking in early childhood was slightly more prevalent in cases than in controls (12.9% v. 11.9%) in the CCLS. The 20 most promising SNPs from our primary analysis in the CCLS are presented in table 4.2. This list was created after ranking all SNPs identified by the 2-step GxE testing procedures in the CCLS dataset by p-value. Of these, 5 SNPs were sent for replication in the ESCALE study (rs12562620, rs2001350, rs6930840, rs10821938, rs11876226), and 6 SNPs were sent for replication in the AUS-ALL study (rs7421154, rs2001350, rs6930840, rs10821938, rs3112511, rs11876226) because the SNPs were previously genotyped in the ESCALE or AUS-ALL studies, and the observed association was found for maternal smoking during pregnancy (exposure data available for ESCALE and AUS-ALL). Two additional SNPs are shown in the replication table for maternal 91 92 smoking prior to pregnancy (Table 4.6) for the AUS-ALL study. These SNPs were in LD with two of the highest-ranking SNPs in the CCLS. Of the top 20 SNPs, statistically significant SNPs were identified on 9 different chromosomes. Of these, strong LD was identified between SNPs in chromosomes 1, 6, 17, and 21. The minor allele frequencies for each SNP are shown in Table 4.2. Minor allele frequencies (MAF) for CCLS controls were similar but not identical to the MAF reported in HapMap3 for the MEX ethnic group. MAF for replication SNPs evaluated in cases in the ESCALE and AUS-ALL studies were generally similar to one another, and similar to the MAF reported in HapMap3 for the CEU population, with the exception of rs10821938 (ARID5B; MAF HapMap-CEU =0.40; MAF ESCALE-ALL =0.55; MAF ESCALE- bcell =0.57; MAF AUS-ALL-ALL =0.63; MAF AUS-ALL-bcell =0.53). Interaction effects for the top SNPs for exposure to maternal smoking during pregnancy (8 of the top 20 SNPs), by childhood ALL subtype (ALL, B-cell ALL, ALL diagnosed between 1-10 years of age) in the CCLS data are presented in Table 4.3. Risk estimates were calculated using log-additive models with an interaction term and treating g e not y p e a s a c onti nuous va ria ble. The ‘te st’ c olu mn indi c a tes the sc r e e ni ng t e st (ste p 1 ) that resulted in the evaluation of that SNP in the step 2 GxE analyses. The odds ratios for each genotype-smoking level combination are reported using the unexposed homogenous major allele group a s the re fe re nt (“ A A” indi c a te s a homoz y g ous major allele genotype, “ AB” indi c a ted a h e te roge nous g e not y pe s, “ B B ” indi c a t e s a homozygous variant g e not y p e ). Ma te rna l sm o king durin g pr e g na nc y is c ode d a s ‘ y e s’ or ‘no’ . No SNPs in the ALL, B-cell ALL, or age-restricted strata reached statistical significance based on the pre- se t α thr e shol d a dj usted for mul ti ple testing , usin g the 2 - 92 93 step subset methods. When statistical testing was completed in step-2 using a weighted approach (wei g hted α = 0.005 for the fir st bin), r s10821938 wa s statis ti c a ll y sig nifi c a nt when restricting to children between 1-10 years of age at diagnosis/referent date (P=0.003). In analyses of all childhood ALL cases, 2 SNPs (rs11144218, rs11876226) were close to statistical significance after correction for multiple comparisons (OR interaction =0.22; p=0.0039; OR interaction =0.23; p=0.0044, respectively, table 4.3). Rs11144218 was not in LD with any SNPs on the platforms used in the ESCALE or AUS-ALL studies (R 2 <0.8), so we were unable to evaluate this SNP in our replication samples. When restricting to B-cell ALL, 2 SNPs (rs7421154 and rs2001350) had suggestive, but not statistically significant effects after correction for multiple comparisons (OR interaction =4.99; P=0.0062; OR interaction =6.93; P=0.0026, respectively, table 4.3). Both SNPs suggest a large excess risk of childhood leukemia among children carrying the homozygous variant genotype and having exposure to maternal smoking during pregnancy (OR BB,smk =6.63; OR BB,smk =24.6, respectively). When restricting to ALL diagnosed between 1-10 years of age, suggestive associations for the 2 SNPs associated with B-cell ALL (i.e. rs7421154 and rs2001350) were also found for this age group (OR interaction =5.61; P=0.0060; OR interaction =7.18; P=0.0040, respectively, table 4.3), with large increases in risk observed in the homozygous variant, exposed group (OR BB,smk =8.90; OR BB,smk =24.9, respectively). Two additional SNPs with suggestive but statistically non-significant associations after 93 94 adjustment for multiple comparisons include rs10829138 and rs7260166 (OR interaction =0.24; P=0.0030; OR interaction =0.22; P=0.0036, respectively). The results of the replication analysis for the 6 SNPs evaluated in the French- ESCALE or Australian-ALL case-only studies for maternal smoking during pregnancy are shown in Table 4.4. The minor allele (A) for SNP rs7421154 was associated with elevated risk among children whose mothers smoked during pregnancy in the CCLS study (OR interaction =3.41, P=0.028) and among children in the Australia-ALL study (OR interaction =2.28, P=0.018). This SNP was not genotyped in the ESCALE study, nor was a SNP available for testing that was in close linkage disequilibrium with rs7421154. The minor allele (G) for SNP rs2001350 also was associated with elevated risk among children whose mothers smoked during pregnancy in the CCLS (OR interaction =3.41, P=0.028), and marginally associated with risk in the Australia-ALL study (OR interaction =1.75, P=0.10), and statistically significantly associated in the French- ESCALE study, but in the reverse direction (OR interaction =0.45, P=0.016). None of the other 4 SNPs investigated in the replication analysis reached statistical significance. Similar, but slightly stronger effects were observed when restricting to B-cell ALL. Results from the CCLS data evaluating interactions between maternal smoking prior to pregnancy and each of the top SNPs by ALL subtype (ALL, B-cell ALL, ALL diagnosed between 1-10 years of age) are presented in table 4.5. SNPs associated with the highest ranked interaction p-values for childhood ALL include: rs10864233, rs7771614, rs3112511, rs2833930, rs2254562 (P<0.005, table 4.5). When restricting to B-cell ALL, the strongest interaction odds ratio was observed for rs7771614 (OR interaction =2.50; P=0.00075, table 4.5), with an excess risk observed in 94 95 the homozygous variant, exposed group (OR BB,smk =2.72). Strong, but statistically non- significant interactions (after multiple comparisons adjustment using two step procedures) were observed for rs10864233, rs12618745, rs1338472, rs3112511, rs2833930, and rs2254562 with maternal smoking prior to pregnancy. SNP rs7421154 had a positive interaction odds ratio in the CCLS (ALL: OR=2.66; P=0.04; B-cell: OR=3.62; P=0.0098, table 4.5) and was statistically significant in the AUS-ALL replication, with slightly stronger results observed in the B-cell ALL subgroup (ALL: OR=1.95, P=0.03; B-cell: OR=2.27, P=0.014, table 4.6). In a subgroup analysis in the CCLS data, of all cases diagnosed between 1-10 years of age, rs12562620 had a positive interaction odds ratio (OR interaction =2.84; P = 0.0067 we i g hted α = 0.005 table 4.5). An excess risk was observed in the exposed, homozygous variant group (OR BB,smk =2.73). Other promising results were observed for rs10864233 and rs7421154 (OR interaction =3.13; P=0.0019; OR interaction =4.72; P=0.0035, respectively). The results of the replication analysis for the 6 SNPs when restricted to B-cell ALL are shown in Table 4.6. SNP rs6930840 (OR interaction =2.92, p=0.00083) was tested in both the French (ESCALE) and Australia (AUS-ALL) studies as a substitute for rs1338472, due to the availability of the genotype data and its strong LD with the original CCLS SNP (R>0.8, CEU and MEX, using HapMap3). However, results for rs6930840 were attenuated in the CCLS sample (P=0.006), and statistically significant results were not observed in the replication set (P>0.05, table 6). No statistically significant associations were observed for SNPs rs3112511 (tested in the AUS-ALL group P>0.05, table 6) or rs2833946 (tested as a substitute for rs2833930, rs2254562 after selection by 95 96 LD) (P>0.05). Results were unable to be replicated for rs10864233 and rs7771614 because no SNPs in LD (R 2 >0.8) with genotypes from the AUS-ALL platform were available. Results for the interaction between maternal smoking in early childhood and each SNP with risk of childhood leukemia by subgroup (ALL, B-cell ALL, ALL diagnosed between 1-10) are shown in table 4.7. Neither the ESCALE nor AUS-ALL studies collected data on maternal smoking in early childhood (defined in the CCLS as smoking between b irth a nd the c h il d’s 3 rd birthday or date of diagnosis, whichever came first), so results from this analysis were unable to be replicated. Although no SNPs reached genome-wide significance using two-step approaches in this analysis, the SNPs with the highest ranked p-values are presented in table 4.7. Rs2014604, rs4791665, rs3785623, and rs3112511, are all located in the intron region of the PIGL gene (phosphatidylinositol glycan anchor biosynthesis, class L) and are in LD with one another (R 2 >0.8, CEU, MEX (HapMap3)). These SNPs approached statistical significance in the analysis of all ALL cases (P<0.003 for all), and when restricting to B-cell ALL (P<0.003 for all). Slightly attenuated interaction odds ratios were observed when restricting to cases diagnosed between 1-10 years of age. Summary effect estimates for rs7421154 are presented in figure 4.2. Each summary effect estimate was computed using inverse variance weighting, with greater weights given to the effect estimate with the smaller variance. For each exposure (maternal smoking during pregnancy, maternal smoking prior to pregnancy), and for each subtype (any ALL diagnosis, B-cell ALL only), case-only effect estimates were greater for the CCLS study than for the AUS-ALL study. AUS-ALL study results were given 96 97 greater weight due to a smaller variance resulting from a greater number of cases included in analysis (n=441 v. n=380). The interaction OR for rs7421154 was greater for maternal smoking during pregnancy, relative to maternal smoking prior to pregnancy, and was greater for both exposures when restricting to B-cell ALL diagnoses. In the B- cell ALL subgroup, the interaction OR for maternal smoking during pregnancy was 4.40 (2.53, 7.64). The Locus Zoom plot for rs7421154 is shown in Figure 4.3. rs7421154 is not located in a gene, but is flanked by PLEK, APLF, FBX048, and LOC391383 (not shown). Two other SNPs in the same region had moderate recombination with this SNP, but had higher p-values than rs7421154. 4.5 DISCUSSION In this analysis, we evaluated gene-parental smoking interactions for risk of childhood ALL using novel two-step statistical methods for GxE interaction analysis. We identified one SNP (rs7421154) for maternal smoking prior to pregnancy and during pregnancy that replicated in an independent study of childhood leukemia (AUS-ALL) both for any type of ALL and when restricting to B-cell ALL in children. Genotyping data for SNP rs7421154 was not available in our second replication study (ESCALE), nor was a suitable substitute SNP in LD with rs7421154, therefore we were not able to attempt replication in ESCALE. The strongest association for rs7421154 in both the CCLS and AUS-ALL studies was observed for maternal smoking during pregnancy when restricting to B-cell ALL (summary OR = 4.40; 95% CI: 2.53, 7.64). A positive interaction OR also was found for SNP rs2001350 in the CCLS and in the AUS-ALL study when considering maternal smoking during pregnancy. The 97 98 interaction did not reach statistical significance in the AUS-ALL study (P=0.10 for ALL and P=0.078 for B-cell ALL). SNP rs2001350 did reach statistical significance in the ESCALE study, however the direction of the interaction OR was in the opposite direction as observed in the CCLS and AUS-ALL. We were unable to replicate several of the most highly significant SNPs from the CCLS analysis because no genotyping data was available on the platforms used in the ESCALE or AUS-ALL studies on that SNP or another SNP in LD with the SNP identified from CCLS. The biological role of the 2 SNPs with respect to tobacco susceptibility is unclear. The SNP rs7421154 is located in a non-coding region on chromosome 2, flanked by several genes including PLEK and APLF, FBX048, and LOC391383. Very little is known about the function of any these genes. The potential biological role of the variant with respect to susceptibility to early life tobacco smoke is unclear. SNP rs2001350 is located in the intron region of NFE2L2 (alternatively, NRF2), which also is located on chromosome 2. NFE2L2 is part of a family of three genes that encode basic leucine zipper (bZIP) proteins. NFE2L2 has been shown to interact with PMF1 to mediate SSAT transcriptional induction; mutations in either NFE2L2 or PMF1 disrupt the ability to regulate SSAT (22, 23). Superinduction of SSAT has been shown to have antineoplastic activity(22, 23). A polymorphism at rs2001350 could induce susceptibility to childhood leukemia, a susceptibility that may be exacerbated by exposure to carcinogens in tobacco smoke early in life. If functional, SNPs that alter activity of NFE2L2 could influence susceptibility to leukemia among children exposed to carcinogens in tobacco smoke early in life. Cigarette smoke has been shown to contain high levels of oxidants, leading to increased 98 99 oxidative stress and accumulation of reactive oxygen species (ROS) (24). NFE2L2 has been found to play a role in antioxidant activities, the detoxification of ROS, and is involved in the regulation of additional antioxidant and detoxification genes(25). For example, NFE2L2 regulates NQO1, an antioxidant enzyme that is involved in the detoxification of ROS and other chemicals found in tobacco smoke (26). A variant in NFE2L2 may result in leukomogenesis through an accumulation of ROS, by increasing oxidative stress, or through downregulation of NQO1, which may then lead to excess ROS in the body. A study of NQO1 found an increased risk of childhood ALL associated with variants that result in decreased activity of the NQO1 enzyme, supporting this hypothesis (27). NFE2L2 also has been shown to interact with tobacco smoke to produce FEV 1 decline (28) and to increase risk of colorectal adenomas (29). A NFE2L2 SNP that diminishes gene activity may lead to increased susceptibility to ROS or oxidative stress from tobacco smoke and an increased risk of ALL in children carrying this variant, when exposed to tobacco smoke early in life. Alternatively, cigarette smoke may initiate NFE2L2 activity and influence enzyme expression. Studies have shown that cigarette smoke exposure activates NFE2L2, which orchestrates the first line of defense, via antioxidant response, against tobacco smoke (30). Another study in human alveolar epithelial cells concluded that Nrf2 (the enzyme regulated by NFE2L2) was involved in cellular defense against oxidative stress and toxicity from tobacco smoke via increased expression of Nrf2 in the presence of cigarette smoke extract (31). Although there is evidence to suggest a possible mechanism through which NFE2L2 and maternal smoking may interact to change risk of childhood ALL, the findings from this study are unclear. The results from the CCLS and the Australia-ALL 99 100 data suggest a positive interaction between maternal smoking and the NFE2L2 variant; however, the data from ESCALE suggest a statistically significant association in the reverse direction. There may be differences in population structure between our sample of Hispanic children and the non-Hispanic White children in the replication set in the ESCALE study, however the Australia study is also non-Hispanic White. This is the first study of childhood leukemia to evaluate gene-parental smoking interactions using genome-wide association data (GWAS). We were able replicate two of our most promising SNPs in independent replication sets in France and Australia. Although this study is the largest population-based study of childhood leukemia in the United States, our sample size is small for detection of gene-environment interactions. Our analysis is also unique because it examines GxE interactions in Hispanic children; a limitation of our findings, because there are no other large population-based studies of childhood leukemia with GWAS data in Hispanic children, is that both our replication samples consisted of White children. We were able to send 8 of our 20 most promising SNPs for replication in either the ESCALE or AUS-ALL studies, or both. We were unable to test the remaining 12 SNPs because the ESCALE or Australian study did not have the maternal smoking variable for the time period tested or because the exact SNP or suitable marker SNP was not available. In this study, we applied novel, statistically efficient methods to evaluate gene- environment interactions and risk of childhood ALL that provide substantial increases in power over traditional methods of interaction analysis in case-control studies. We were therefore able to prioritize SNPs for replication using methods that are highly efficient compared to traditional methods of GxE analysis. Development of statistically efficient 100 101 procedures is especially important when completing studies of genetic-environmental interaction of rare diseases, to allow us to better understand the interplay of genetic susceptibility and environment in diseases such as childhood leukemia. The use of several screening methods allowed us to identify SNPs for testing that would not have been identified using only one of the procedures. While the disease-gene (DG) screening methods identified the most high priority SNPs, the two SNPs with statistically significant interaction odds ratios in the replication analysis were identified through the environment-gene (EG) and environment-disease-gene (EDG) screening procedures. Our findings suggest that genome-wide interaction analyses are feasible in studies of childhood cancer when applying efficient statistical methods, such as the EG, DG, or EDG 2-step procedures. We also found evidence supporting the role of two SNPs in genes that may modify the effect of early life exposure to tobacco smoke in risk of childhood ALL. Application of these procedures to a larger sample, including Non- Hispanic whites and participants from additional international studies may be feasible in the future and should be used to obtain more precise estimates of potential gene-parental smoking interactions. 101 102 4.6 TABLES AND FIGURES Table 4.1. Demographic Characteristics Of Cases And Controls In The Primary Analysis (CCLS), And Of Cases In The Replication Analysis (ESCALE and AUS- ALL) CCLS ESCALE AUS-ALL (1995-2008) (2003-2004) (2003-2007) Cases n(%) Controls n(%) Cases n(%) Cases n(%) N=380 N=454 N=441 N=285 Male 201 (52.9) 240 (52.9) 234 (53.0) 158 (55.4) Age at diagnosis (years) a Mean (SD) 5.3 (3.6) 5.3 (3.4) 5.6 (3.6) 6.0 (3.7) Median (IQR) 4.1 (2.6, 7.4) 4.4 (2.7, 7.5) 4.4 (2.7, 7.6) 4.9 (3.2, 8.5) <1yr 16 (4.2) 14 (3.1) 8 (1.8) 7 (2.5) 1-10yrs 310 (81.6) 379 (83.5) 380 (86.2) 230 (80.7) >10yrs 54 (14.2) 61 (13.4) 53 (12.0) 48 (16.7) Histologic tumor type B-cell 323 (85.0) -- 397 (90.0) 253 (88.8) Other 57 (15.0) -- 44 (10.0) 32 (11.2) Paternal smoking prior to conception No 250 (70.6) 328 (74.4) -- -- Yes 104 (29.4) 113 (25.6) -- -- Unknown 26 13 -- -- Maternal smoking prior to conception n=832 No 336 (88.7) 402 (88.7) -- 210 (73.7) Yes 43 (11.3) 52 (11.2) -- 75 (26.3) 1-5 cigarettes/day 22 (5.8) 36 (8.0) -- -- 6+ cigarettes/day 21 (5.5) 15 (2.2) -- -- Unknown 1 0 -- -- Missing dose information 0 1 -- -- Maternal smoking during pregnancy n=831 No 346 (91.3) 423 (93.2) 332 (75.3) 226 (79.3) Yes 33 (8.7) 31 (6.8) 109 (24.7) 59 (20.7) 1-5 cigarettes/day 21 (5.6) 23 (5.1) -- -- 6+ cigarettes/day 10 (2.6) 8 (1.7) -- -- Unknown 1 0 -- -- Missing dose information 2 0 -- -- Maternal smoking while breastfeeding n=799 No 355 (97.0) 421 (97.2) -- -- Yes 11 (3.0) 12 (2.8) -- -- 1-5 cigarettes/day 6 (1.6) 9 (2.1) -- -- 6+ cigarettes/day 4 (1.1) 2 (0.5) -- -- Unknown 13 20 -- -- Missing dose information 1 1 -- -- 102 103 Table 4.1 continued CCLS ESCALE AUS-ALL (1995-2008) (2003-2004) (2003-2007) Cases n(%) Controls n(%) Cases n(%) Cases n(%) N=380 N=454 N=441 N=285 Maternal smoking in early childhood n=811 No 330 (87.1) 400 (88.1) -- -- Yes 49 (12.9) 54 (11.9) -- -- 1-5 cigarettes/day 26 (7.0) 24 (5.5) -- -- 6+ cigarettes/day 16 (4.3) 15 (3.4) -- -- Unknown 1 0 -- -- Missing dose information 7 15 -- -- a Reference age for controls 103 104 Table 4.2. Characteristics Of The 20 Most Significant Snps Identified In The CCLS Analysis, Using Two-Step Methods For Any Exposure (Maternal Smoking Prior To Pregnancy, During Pregnancy, Or In Early Childhood), And For Any Leukemia, When Restricting To B-Cell ALL, Or When Restricting To ALL Diagnosed Between 1-10 Years Of Age. CCLS ESCALE AUS-ALL SNP Repl. Chr Loc. Gene LD Minor Allele Controls MAF ALL cases MAF Bcell cases MAF HAPMAP MEX MAF ALL cases MAF Bcell cases MAF ALL cases MAF Bcell cases MAF HAPMAP CEU MAF rs12562620 E 1 216286427 USH2A a G 0.47 0.48 0.46 0.56 0.67 0.68 -- -- 0.65 rs10864233 * 1 216295188 USH2A a A 0.49 0.49 0.48 -- -- -- -- -- 0.68 rs7421154 A 2 68665945 PLEK||LOC391383 A 0.13 0.10 0.09 0.08 -- -- 0.12 0.12 0.09 rs2001350 E,A 2 178100425 NFE2L2 G 0.09 0.10 0.10 0.15 0.10 0.10 0.11 0.10 0.09 rs12618745 ** 2 195651706 LOC343981||LOC391470 G 0.04 0.07 0.07 -- -- -- -- -- 0.05 rs6930840 A 6 45951830 CLIC5 b,d A 0.39 0.30 0.31 0.39 0.36 0.36 0.34 0.33 0.32 rs1338472 * 6 45968780 CLIC5 b,c A 0.40 0.31 0.32 0.41 -- -- 0.31 rs7771614 * 6 45975798 CLIC5 c,d A 0.44 0.34 0.34 -- -- -- -- -- 0.41 rs11144218 ** 9 77696918 NMRK1 A 0.46 0.36 0.36 0.35 -- -- -- -- 0.58 rs10821938 E,A 10 63724773 ARID5B A 0.49 0.62 0.63 0.45 0.55 0.57 0.53 0.53 0.40 rs2014604 **** 17 16176883 PIGL e G 0.47 0.55 0.56 0.59 -- -- -- -- 0.48 rs4791665 **** 17 16196056 PIGL e G 0.47 0.54 0.56 0.60 -- -- -- -- 0.47 rs3785623 **** 17 16201160 PIGL e G 0.46 0.55 0.56 0.57 -- -- -- -- 0.48 rs3112511 A 17 16207526 PIGL e A 0.50 0.40 0.39 0.40 -- -- 0.47 0.47 0.52 rs11876226 E,A 18 22653562 ZNF521 C 0.14 0.22 0.22 0.22 0.18 0.18 0.17 0.17 0.14 rs12459746 ** 19 38857085 CATSPERG G 0.40 0.32 0.32 -- -- -- -- -- 0.24 rs11880103 ** 19 50827700 KCNC3 A 0.45 0.51 0.53 0.53 -- -- -- -- 0.42 rs7260166 *** 19 9339399 OR7D11P||OR7D1P G 0.25 0.33 0.33 0.24 -- -- -- -- 0.25 rs2833930 * 21 34014768 SYNJ1 f A 0.23 0.31 0.30 0.28 -- -- -- -- 0.29 rs2254562 * 21 34059352 SYNJ1 f G 0.23 0.30 0.30 0.28 -- -- -- -- 0.27 *not available on Illumina 317K, 370D, 370Q, or 610Q platforms; different SNP in LD was replicated **Not sent for replication because SNP was not on platform, and no SNPs in LD on E or A platforms ***Not sent for replication because only replicated in ALL and B-cell ALL, and significant results were observed in DX 1-10 group ****Not sent for replication because results for maternal smoking in early childhood a: LD>0.8 CEU, no data MEX b,e,f,g: LD>0.8 CEU, MEX c,d: LD>0.6 CEU, no data MEX 104 105 Table 4.3. Interaction Effects For Maternal Smoking During Pregnancy By Leukemia Subgroup, CCLS Any ALL B-cell ALL ALL diagnosed 1-10 years of age SNP Chr Gene^ Rep † Test N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR p- value rs7421154 2 PLEK|| LOC391383 A EDG EG 833 No 1.00 0.64 0.41 3.41 0.028 777 No 1.00 0.58 0.33 4.99 0.0062 689 No 1.00 0.68 0.46 5.61 0.0060 Yes 0.96 2.08 4.54 Yes 0.80 2.30 6.63 Yes 0.62 2.35 8.90 rs2001350 2 NFE2L2 E, A EG 833 No 1.00 0.94 0.88 4.14 0.015 777 No 1.00 0.86 0.75 6.83 0.0026 689 No 1.00 0.88 0.77 7.18 0.0040 Yes 0.90 3.64 14.1 Yes 0.71 4.17 24.6 Yes 0.63 3.96 24.9 rs6930840 a 6 CLIC5 A NA 833 No 1.00 0.64 0.41 2.26 0.027 777 No 1.00 0.66 0.44 2.04 0.078 689 No 1.00 0.64 0.41 1.85 0.14 Yes 0.83 1.21 1.76 Yes 0.87 1.18 1.59 Yes 0.85 1.01 1.20 rs1338472 a 6 CLIC5 * DG 833 No 1.00 0.64 0.42 2.77 0.007 777 No 1.00 0.65 0.42 2.68 0.018 689 No 1.00 0.63 0.40 2.46 0.035 Yes 0.72 1.29 2.57 Yes 0.73 1.28 2.22 Yes 0.70 1.09 1.68 rs11144218 9 NMRK1 ** DG 832 No 1.00 0.74 0.55 0.22 0.0039 776 No 1.00 0.74 0.55 0.25 0.010 688 No 1.00 0.73 0.54 0.25 0.017 Yes 4.66 0.76 0.12 Yes 4.19 0.78 0.15 Yes 4.14 0.76 0.14 rs10821938 10 ARID5B E, A EDG DG 833 No 1.00 1.82 3.31 0.49 0.062 777 No 1.00 1.98 3.92 0.35 0.013 689 No 1.00 1.89 3.59 0.24 0.0030 § Yes 3.09 2.76 2.47 Yes 4.34 3.01 2.09 Yes 5.80 2.66 1.22 rs11876226 18 ZNF521 E, A EDG DG 833 No 1.00 1.88 3.53 0.23 0.0044 777 No 1.00 1.88 3.53 0.24 0.008 689 No 1.00 1.75 3.05 0.25 0.030 Yes 2.31 1.01 0.44 Yes 2.26 1.03 0.47 Yes 1.95 0.86 0.38 rs7260166 19 OR7D11P|| OR7D1P *** EDG DG 833 No 1.00 1.54 2.38 0.36 0.010 777 No 1.00 1.54 2.38 0.37 0.015 689 No 1.00 1.75 3.06 0.22 0.0036 Yes 2.41 1.34 0.74 Yes 2.35 1.33 0.75 Yes 2.56 0.98 0.37 ^SNPs with one gene listed are located in the intron region of that gene. SNPs with 2 genes listed (separated by ||) are not located in a gene; genes listed are flanking genes † S e n t f o r r e p l i c a t i o n ( A L L a n d B -cell ALL only); E = ESCALE study in France, A = AUS-ALL study in Australia. §Significant in weighted approaches for ALL diagnosed from 1-10 years of age *Not sent for replication, but different SNP in LD sent **Not sent for replication because no SNPs in LD on E or A platforms ***Not sent for replication because only replicated in ALL and B-cell ALL (significant for DX1-10) a: LD>0.8, CEU, MEX (HapMap3) 105 106 Table 4.4. Replication Results For Maternal Smoking During Pregnancy By Leukemia Subgroup, CCLS, ESCALE, AUS- ALL Any Acute Lymphocytic Leukemia Bcell Acute Lymphocytic Leukemia ALL CCLS ESCALE AUS-ALL CCLS ESCALE AUS-ALL SNP Chr Gene^ N OR* p-value N OR* p-value N OR* p-value N OR* p-value N OR* p-value N OR* p-value rs7421154 2 PLEK|| LOC391383 833 3.41 0.028 -- -- -- 285 2.28 0.018 777 4.99 0.0062 -- -- -- 253 2.98 0.003 rs2001350 2 NFE2L2 833 4.14 0.015 441 0.45 0.016 285 1.75** 0.10 777 6.83 0.0026 397 0.39 0.012 253 1.92** 0.078 rs6930840 a 6 CLIC5 833 2.26 0.027 441 1.11 0.52 285 0.72 0.16 777 2.04 0.078 397 1.14 0.48 253 0.73 0.22 rs1338472 a 6 CLIC5 833 2.77 0.007 -- -- -- -- -- -- 777 2.68 0.018 -- -- -- -- -- -- rs10821938 10 ARID5B 833 0.49 0.062 422 1.11 0.55 285 1.20 0.99 777 0.35 0.013 381 1.21 0.30 253 1.09 0.70 § rs11876226 18 ZNF521 833 0.23 0.0044 441 1.03 0.88 285 1.38 0.32 777 0.24 0.008 397 0.97 0.88 253 1.29 0.42 ^SNPs with one gene listed are located in the intron region of that gene. SNPs with 2 genes listed (separated by ||) are not located in a gene; genes listed are flanking genes CCLS: models adjusted for gender, age at diagnosis (continuous), and the first 4 principal components ESCALE: models adjusted for gender, age at diagnosis (8-class variable) and the first 7 principal components. § SNP imputed using impute V2 for French replication *OR (CCLS) = Interaction OR for product term *OR (ESCALE and AUS-ALL) = Assoc between SNP and maternal smoking during pregnancy among cases **Originally computed using minor allele homozygous as referent. ORs here have been inverted. a: LD>0.8 CEU, MEX 106 107 Table 4.5. Interaction Effects For Maternal Smoking Prior To Pregnancy By Leukemia Subgroup, CCLS Any ALL B-cell ALL ALL diagnosed 1-10 years of age SNP Chr Gene^ Rep † Test N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P rs12562620 a 1 USH2A * EDG EG 833 No 1.00 0.96 0.92 2.32 0.011 777 No 1.00 0.92 0.85 2.50 0.0075 689 No 1.00 0.93 0.87 2.84 0.0067 Yes 0.55 1.23 2.75 Yes 0.51 1.18 2.72 Yes 0.39 1.03 2.73 rs10864233 a 1 USH2A * EG 833 No 1.00 0.89 0.79 2.63 0.0021 777 No 1.00 0.85 0.72 2.85 0.0016 689 No 1.00 0.86 0.74 3.13 0.0019 Yes 0.45 1.05 2.46 Yes 0.42 1.01 2.43 Yes 0.32 0.85 2.30 rs7421154 2 PLEK|| LOC391383 A EG 833 No 1.00 0.65 0.42 2.66 0.040 777 No 1.00 0.58 0.34 3.62 0.0098 689 No 1.00 0.68 0.46 4.72 0.0035 Yes 0.81 1.39 2.40 Yes 0.73 1.54 3.27 Yes 0.46 1.47 4.70 rs2001350 2 NFE2L2 A EG 833 No 1.00 0.97 0.94 2.52 0.050 777 No 1.00 0.88 0.78 3.66 0.0088 689 No 1.00 0.91 0.82 3.92 0.0089 Yes 0.79 1.92 4.70 Yes 0.68 2.19 7.10 Yes 0.49 1.74 6.19 rs12618745 2 LOC343981|| LOC391470 * EDG 833 No 1.00 2.55 6.50 0.21 0.0071 777 No 1.00 2.78 7.75 0.16 0.0038 689 No 1.00 2.45 6.01 0.12 0.0065 Yes 1.32 0.71 0.39 Yes 1.38 0.61 0.27 Yes 1.08 0.32 0.09 rs1338472 b 6 CLIC5 ** DG 833 No 1.00 0.62 0.38 2.92 0.00083 777 No 1.00 0.62 0.39 2.82 0.0026 689 No 1.00 0.61 0.37 2.43 0.016 Yes 0.50 0.91 1.63 Yes 0.52 0.91 1.60 Yes 0.45 0.66 0.98 rs7771614 b 6 CLIC5 ** DG 832 No 1.00 0.61 0.37 2.53 0.0032 776 No 1.00 0.92 0.85 2.50 0.00075 688 No 1.00 0.59 0.35 2.13 0.041 Yes 0.51 0.79 1.22 Yes 0.51 1.18 2.72 Yes 0.45 0.57 0.72 rs3112511 17 PIGL A EDG DG 832 No 1.00 0.62 0.38 2.60 0.0023 776 No 1.00 0.59 0.35 2.49 0.0043 688 No 1.00 0.59 0.35 2.47 0.0080 Yes 0.45 0.72 1.16 Yes 0.47 0.70 1.03 Yes 0.36 0.53 0.77 ^SNPs with one gene listed are located in the intron region of that gene. SNPs with 2 genes listed (separated by ||) are not located in a gene; genes listed are flanking genes † S e n t f o r r e p l i c a t i o n ( A L L a n d B -cell ALL only); E = ESCALE study in France, A = AUS-ALL study in Australia. *SNP not available on AUS-ALL platform, no SNP in LD available **SNP not available on AUS-ALL platform, other SNP in LD sent for replication a: LD>0.8 CEU (1000 Genomes), no data MEX b: LD>0.6 CEU (1000 Genomes), no data MEX c: LD>0.8 CEU, MEX (HapMap3) 107 108 Table 4.5 Continued Any ALL B-cell ALL ALL diagnosed 1-10 years of age N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P rs2833930 c 21 SYNJ1 ** DG 831 No 1.00 1.80 3.23 0.38 0.0035 776 No 1.00 1.72 2.96 0.34 0.0035 688 No 1.00 1.79 3.19 0.36 0.0095 Yes 1.82 1.23 0.83 Yes 1.89 1.11 0.65 Yes 1.39 0.90 0.59 rs2254562 c 21 SYNJ1 ** DG 819 No 1.00 1.75 3.05 0.38 0.0043 771 No 1.00 1.70 2.88 0.34 0.0037 678 No 1.00 1.72 2.97 0.37 0.012 Yes 1.86 1.25 0.84 Yes 1.92 1.11 0.65 Yes 1.42 0.92 0.59 ^SNPs with one gene listed are located in the intron region of that gene. SNPs with 2 genes listed (separated by ||) are not located in a gene; genes listed are flanking genes † S e n t f o r r e p l i c a t i o n ( A L L a n d B -cell ALL only); E = ESCALE study in France, A = AUS-ALL study in Australia. *SNP not available on AUS-ALL platform, no SNP in LD available **SNP not available on AUS-ALL platform, other SNP in LD sent for replication a: LD>0.8 CEU (1000 Genomes), no data MEX b: LD>0.6 CEU (1000 Genomes), no data MEX c: LD>0.8 CEU, MEX (HapMap3) 108 109 Table 4.6. Replication Results For Maternal Smoking Prior To Pregnancy By Leukemia Subgroup, CCLS, ESCALE, AUS- ALL Acute Lymphocytic Leukemia Bcell Acute Lymphocytic Leukemia CCLS AUS-ALL CCLS AUS-ALL SNP Chr Gene^ LD Minor Allele CCLS controls MAF N OR* P N OR* p-value N OR* P N OR* P rs7421154§ 2 PLEK|| LOC391383 A 0.13 833 2.66 0.040 285 1.95 0.03 777 3.62 0.0098 253 2.27 0.014 rs2001350 2 NFE2L2 G 0.09 833 2.52 0.050 285 1.56** 0.15 777 3.66 0.0088 253 1.47** 0.27 rs6930840 6 CLIC5 a A 0.39 833 2.48 0.0060 285 0.80 0.30 777 2.43 0.015 253 0.80 0.34 rs1338472 6 CLIC5 a A 0.40 833 2.92 0.00083 -- -- -- 777 2.82 0.0026 -- -- -- rs3112511§ 17 PIGL b A 0.50 832 2.60 0.0023 285 0.97 0.86 776 2.49 0.0043 253 0.91 0.65 rs9944524 17 PIGL b G NA -- -- -- 285 1.00 0.98 -- -- -- 253 1.05 0.85 rs2833930 21 SYNJ1 c A 0.23 831 0.38 0.0035 -- -- -- 776 0.34 0.0035 -- -- -- rs2833946 21 PAXBP1 c A NA -- -- -- 285 1.08 0.73 -- -- -- 253 1.07 0.75 ^SNPs with one gene listed are located in the intron region of that gene. SNPs with 2 genes listed (separated by ||) are not located in a gene; genes listed are flanking genes CCLS: models adjusted for gender, age at diagnosis (continuous), and the first 4 principal components *OR (CCLS) = Interaction OR for product term *OR (AUS-ALL) = Assoc between SNP and maternal smoking during pregnancy **Originally computed using minor allele homozygous as referent. ORs here have been inverted. § SNP imputed for AUS-ALL group a,c: LD>0.8 CEU, MEX b: LD>0.8 CEU, LD>0.5, MEX 109 110 Table 4.7. Interaction Effects For Maternal Smoking In Early Childhood By Leukemia Subgroup, CCLS Any Acute Lymphocytic Leukemia B-cell Acute Lymphocytic Leukemia ALL diagnosed 1-10 years of age SNP Chr Gene Test N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P N Smk AA OR AB OR BB OR Int. OR P rs2014604 a 17 PIGL DG 833 No 1.00 1.54 2.37 0.39 0.0027 777 No 1.00 1.62 2.64 0.38 0.0026 689 No 1.00 1.57 2.48 0.41 0.0085 Yes 3.18 1.92 1.16 Yes 3.23 2.00 1.24 Yes 2.58 1.68 1.09 rs4791665 a 17 PIGL DG 833 No 1.00 1.53 2.34 0.39 0.0029 777 No 1.00 1.62 2.61 0.38 0.0027 689 No 1.00 1.56 2.44 0.42 0.009 Yes 3.16 1.91 1.16 Yes 3.23 1.99 1.23 Yes 2.57 1.67 1.08 rs3785623 a 17 PIGL DG 833 No 1.00 1.56 2.44 0.37 0.0015 777 No 1.00 1.64 2.71 0.35 0.0014 689 No 1.00 1.58 2.51 0.38 0.0051 Yes 3.39 1.94 1.11 Yes 3.46 2.01 1.17 Yes 2.76 1.68 1.02 rs3112511 a 17 PIGL EDG DG 832 No 1.00 0.61 0.37 2.75 0.000996 776 No 1.00 0.58 0.34 2.76 0.0013 688 No 1.00 0.59 0.34 2.63 0.0035 Yes 0.49 0.81 1.36 Yes 0.48 0.77 1.24 Yes 0.43 0.67 1.04 rs12459746 19 CATSPERG DG 831 No 1.00 0.76 0.57 0.33 0.0051 775 No 1.00 0.73 0.53 0.36 0.013 687 No 1.00 0.72 0.52 0.36 0.02 Yes 2.49 0.63 0.16 Yes 2.34 0.61 0.16 Yes 2.16 0.56 0.14 a: LD>0.8 CEU, MEX (HapMap3) 110 111 Figure 4.1a. SNP Exclusion Criteria Remaining SNPs = 634,037 Filtered based on minor allele frequency 68,936 SNPs removed (MAF <2%) Filtered based on SNP call rate 5,351 SNPs removed (call rate <98%) Filtered based on HWE in controls 1,034 SNPs removed (p < 1 x 10 -5 ) Removal of non-autosomal SNPs 21,167 SNPs removed 730,525 SNPs 111 112 Figure 4.1b. Sample Exclusion Criteria Final sample size = 834 subjects (380 cases, 454 controls) Cryptic relatedness assessed No samples removed Duplicates removed 10 duplicates removed Genotyping call rate calculated 1 sample removed (call rate <98%) Sex check performed 2 samples removed (failed sex check) 848 samples genotyped 112 113 Figure 4.2. Summary Effect Estimates From Case-Only Analyses For Rs7421154, By Smoking Exposure And Subtype, For Results From The CCLS And AUS-ALL Studies 113 114 Figure 4.3. Locus Zoom Plot for rs7421154 0 1 2 3 4 5 - log 10 (p −value) 0 20 40 60 80 100 Recombination rate (cM/Mb) rs7421154 0.2 0.4 0.6 0.8 r 2 PPP3R1 CNRIP1 PLEK FBXO48 APLF 68.5 68.6 68.7 68.8 Position on chr2 (Mb) Plotted SNPs 114 115 4.7 SUPPLEMENTARY MATERIAL Supplemental table 4.1 Characteristics of Top SNPs Identified from Two-Step Scans by Tobacco Smoke Exposure for Diagnoses with ALL 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Maternal smoking in early childhood 277-340 SNPs tested 2=0.000135- 0.000181 139-189 SNPs tested 2=0.000265- 0.00036 rs3112511 17 16207526 A 832 13.41 0.00025 1.01 3.29 0.001 DG N Y N N 3 rs12459746 19 38857085 G 831 12.88 0.00033 -1.1 -2.8 0.0051 DG N N - N 1 rs931370 4 188105792 A 833 17.49 0.00016 -1.29 -2.66 0.0077 DGEG N Y N N 1 rs16997129 4 110989719 A 831 13.52 0.00024 -0.92 -2.64 0.0082 EG N Y N N 2 rs3799732 6 125574509 A 833 12.2 0.00048 0.86 2.6 0.0092 EG N N - N 1 rs12945877 17 16238203 A 829 12.39 0.00043 -0.79 -2.45 0.014 DG N N - N 2 rs15739 17 16229232 G 833 12.39 0.00043 -0.79 -2.45 0.014 DG N N - N 2 rs8042848 15 34008989 G 833 14.48 0.00014 0.73 2.37 0.018 DG N Y N N 1 rs1375681 4 62037705 A 833 12.42 0.00043 -1.09 -2.32 0.02 EG N N - N 1 rs17086349 4 66627407 G 833 13.62 0.00022 0.88 2.31 0.021 EG N Y N N 1 Maternal smoking prior to pregnancy 270-370 SNPs tested 2=0.000135- 0.000185 124-189 SNPs tested 2=0.000265- 0.000403 rs1338472 6 45968780 A 833 12.32 0.00045 1.07 3.34 0.00082 DG N N - N 3 rs3112511 17 16207526 A 832 13.41 0.00025 0.96 3.05 0.0023 DG N Y N N 3 rs7771614 6 45975798 A 832 14.78 0.00012 0.93 2.95 0.0032 DG N Y N N 2 rs2833930 21 34014768 A 831 14.02 0.00018 -0.98 -2.92 0.0035 DG N Y N N 1 rs2254562 21 34059352 G 819 12.35 0.00044 -0.96 -2.85 0.0043 DG N N - N 1 rs7951773 11 11868549 A 833 12.42 0.00043 -1.15 -2.76 0.0058 DG N N - N 1 rs6930840 6 45951830 A 833 13.14 0.00029 0.91 2.75 0.006 DG N N - N 2 rs12618745 2 195651706 G 833 16.67 0.00024 -1.55 -2.69 0.0071 DGEG N Y N N 1 rs2792833 6 49871812 A 833 17.29 0.00018 -0.92 -2.62 0.0088 DGEG DG N Y N N 1 rs16997129 4 110989719 A 831 18.69 0.000088 -0.91 -2.59 0.0097 DGEG EG N Y N N 2 115 116 Supplemental table 4.1 Continued 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Maternal smoking during pregnancy 284-370 SNPs tested 2=0.000135- 0.000176 138-189 SNPs tested 2=0.000265- 0.000362 rs11144218 9 77696918 A 832 12.69 0.00037 -1.51 -2.88 0.0039 DG N N - N 1 rs11876226 18 22653562 C 833 15.66 0.0004 -1.46 -2.85 0.0044 DGEG DG N N - N 2 rs1338472 6 45968780 A 833 12.32 0.00045 1.02 2.69 0.0071 DG N N - N 3 rs10807935 7 46611204 A 833 12.62 0.00038 -1.14 -2.68 0.0074 EG N N - N 1 rs12156358 8 74283845 A 833 14.15 0.00017 -1.04 -2.52 0.012 EG N Y N N 1 rs13111781 4 136653594 A 833 15.29 0.00048 1.03 2.45 0.014 DGEG N N - N 1 rs847005 6 108002078 A 833 12.87 0.00033 -1.11 -2.45 0.014 DG N N - N 1 rs12639259 3 7063064 A 833 15.49 0.00043 -2 -2.41 0.016 DGEG N N - N 1 rs6890669 5 167863193 A 833 12.94 0.00032 -0.97 -2.22 0.027 EG N N - N 1 rs6930840 6 45951830 A 833 13.14 0.00029 0.82 2.22 0.027 DG N N - N 2 Maternal smoking in early childhood + paternal smoking prior to pregnancy 308-369 SNPs tested 2=0.000136- 0.000162 158-188 SNPs tested 2=0.000266- 0.000316 rs17068331 6 139416226 A 794 13.4 0.00025 1.53 3.09 0.002 EG N Y N N 1 rs3112511 17 16207526 A 793 15.68 0.00039 1.12 2.79 0.0052 DGEG DG N N - N 3 rs11876226 18 22653562 C 794 15.71 0.00039 -1.37 -2.62 0.0088 DGEG DG N N - N 2 rs12945877 17 16238203 A 790 12.39 0.00043 -1.05 -2.59 0.0096 DG N N - N 2 rs15739 17 16229232 G 794 12.39 0.00043 -1.05 -2.59 0.0096 DG N N - N 2 rs1338472 6 45968780 A 794 12.32 0.00045 0.89 2.44 0.015 DG N N - N 3 rs4973760 3 27349047 A 794 12.85 0.00034 -1 -2.33 0.02 EG N N - N 1 rs7771614 6 45975798 A 793 14.78 0.00012 0.85 2.27 0.023 DG N Y N N 2 rs7514201 1 52412005 C 794 13.68 0.00022 1.76 2.26 0.024 EG N Y N N 1 rs879717 5 143587990 A 794 17.51 0.00016 -1.07 -2.26 0.024 DGEG DG N Y N N 1 116 117 Supplemental table 4.1 Continued 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Paternal smoking prior to pregnancy 285-370 SNPs tested 2=0.000135- 0.000175 138-189 SNPs tested 2=0.000265- 0.000362 rs13022554 2 166752432 G 792 16.97 0.00021 -0.7 -2.48 0.013 DGEG N Y N N 1 rs9947632 18 36062561 A 795 12.8 0.00035 -0.63 -2.5 0.013 EG N N - N 1 rs9838484 3 1673173 A 795 15.46 0.00044 0.64 2.47 0.014 DGEG DG N N - N 1 rs1517516 3 16475215 A 794 12.34 0.00044 0.73 2.39 0.017 EG N N - N 1 rs913905 13 30720425 G 795 17.34 0.00017 -0.57 -2.38 0.017 DGEG DG N Y N N 1 rs2568222 2 85377979 A 795 15.77 0.00038 0.57 2.36 0.018 DGEG DG N N - N 1 rs4877062 9 91820998 G 795 16.03 0.00033 -0.55 -2.36 0.018 DGEG EG N N - N 1 rs11690345 2 5196902 G 795 16.72 0.00023 0.52 2.27 0.023 DGEG EG N Y N N 1 rs6965881 7 52470990 A 795 13.12 0.00029 0.52 2.28 0.023 DG N N - N 1 rs17005406 1 219579930 G 795 15.52 0.00043 1.02 2.23 0.026 DGEG DG N N - N 1 117 118 Supplemental table 4.2 Characteristics of Top SNPs Identified from Two-Step Scans by Tobacco Smoke Exposure for Diagnoses with B-Cell ALL 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Maternal smoking in early childhood 274-340 SNPs tested 2=0.000140- 0.000182 141-180 SNPs tested 2=0.000278- 0.000355 rs3112511 17 16207526 A 776 15.95 0.00034 1.02 3.21 0.0013 DGEG DG N N - N 3 rs3785623 17 16201160 A 777 13.28 0.00027 1.04 3.19 0.0014 DG N N - N 3 rs2014604 17 16176883 A 777 12.9 0.00033 0.97 3.01 0.0026 DG N N - N 2 rs4791665 17 16196056 A 777 12.66 0.00037 0.96 3 0.0027 DG N N - N 2 rs4775362 15 61433932 G 777 12.94 0.00032 -0.97 -2.8 0.0051 DG N N - N 1 rs12459746 19 38857085 G 775 13.7 0.00021 -1.02 -2.5 0.013 DG N Y N N 1 rs11880103 19 50827700 A 777 22.61 0.000012 -0.85 -2.45 0.014 DGEG EG N Y N N 1 rs15739 17 16229232 G 777 13.4 0.00025 -0.8 -2.42 0.015 DG N Y N N 3 rs12945877 17 16238203 A 773 13.39 0.00025 -0.8 -2.42 0.016 DG N Y N N 3 rs17132136 11 98406723 A 777 18 0.00012 2.4 2.4 0.017 DGEG DG N Y N N 1 Maternal smoking prior to pregnancy 270-340 SNPs tested 2=0.000147- 0.000185 128-180 SNPs tested 2=0.000278- 0.000391 rs12618745 2 195651706 G 777 15.62 0.0004 -1.84 -2.89 0.0038 DGEG N N - N 1 rs3112511 17 16207526 A 776 16.35 0.00028 0.91 2.85 0.0043 DGEG DG N N - N 3 rs3785623 17 16201160 A 777 13.28 0.00027 0.86 2.7 0.007 DG N N - N 3 rs6565029 16 82524375 A 777 12.5 0.00041 -0.95 -2.7 0.007 DG N N - N 1 rs7771614 6 45975798 A 776 13.78 0.00021 0.92 2.69 0.0071 DG N Y N N 1 rs12562620 1 216286427 G 777 15.44 0.00044 0.91 2.68 0.0075 DGEG EG N N - N 1 rs12945877 17 16238203 A 773 13.39 0.00025 -0.88 -2.65 0.0081 DG N Y N N 3 rs15739 17 16229232 G 777 13.4 0.00025 -0.88 -2.65 0.0081 DG N Y N N 3 rs2001350 2 178100425 G 777 13.08 0.0003 1.3 2.62 0.0088 EG N N - N 2 rs4783112 16 82526157 G 776 12.95 0.00032 -0.9 -2.55 0.011 DG N N - N 1 118 119 Supplemental table 4.2 Continued 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Maternal smoking during pregnancy 271-342 SNPs tested 2=0.000146- 0.000185 144-180 SNPs tested 2=0.000278- 0.000347 rs2001350 2 178100425 G 777 12.84 0.00034 1.92 3.02 0.0026 EG N N - N 2 rs7421154 2 68665945 A 777 17.1 0.00019 1.61 2.74 0.0062 DGEG EG N Y N N 1 rs11876226 18 22653562 C 777 15.32 0.00047 -1.42 -2.65 0.0081 DGEG DG N N - N 2 rs10807935 7 46611204 A 777 13 0.00031 -1.19 -2.63 0.0086 EG N N - N 1 rs12319134 12 32700402 A 777 18.2 0.00011 2.15 2.59 0.0096 DGEG N Y N N 1 rs10821938 10 63724773 C 777 29.48 0.0000004 1.05 2.5 0.013 DGEG DG N Y N N 1 rs12156358 8 74283845 A 777 15.16 0.000099 -0.97 -2.31 0.021 EG N Y N N 1 rs9944770 18 68140829 G 777 15.67 0.0004 -2.57 -2.27 0.023 DGEG N N - N 1 rs9354812 6 69706344 A 777 16.27 0.00029 1.1 2.19 0.029 DGEG DG N N - N 1 rs1957240 14 89438126 A 777 16.89 0.00021 -1.06 -2.17 0.03 DGEG EG N Y N N 1 Maternal smoking in early childhood + paternal smoking prior to pregnancy 282-356 SNPs tested 2=0.00014- 0.000177 153-180 SNPs tested 2=0.000278- 0.000327 rs17068331 6 139416226 A 740 12.93 0.00032 1.66 3.2 0.0014 EG N N - N 1 rs2014604 17 16176883 A 740 16.3 0.00029 1.23 2.97 0.003 DGEG DG N N - N 2 rs3785623 17 16201160 A 740 17.06 0.0002 1.23 2.96 0.003 DGEG DG N Y N N 3 rs4791665 17 16196056 A 740 15.99 0.00034 1.22 2.94 0.0033 DGEG DG N N - N 2 rs3112511 17 16207526 A 739 17.76 0.00014 1.08 2.71 0.0067 DGEG DG N Y N N 3 rs15739 17 16229232 G 740 15.32 0.00047 -1.07 -2.62 0.0087 DGEG DG N N - N 3 rs12945877 17 16238203 A 736 15.33 0.00047 -1.07 -2.62 0.0088 DGEG DG N N - N 3 rs11876226 18 22653562 C 740 15.51 0.00043 -1.36 -2.49 0.013 DGEG DG N N - N 2 rs7372490 3 59367296 A 740 12.92 0.00032 2.82 2.42 0.015 DG N N - N 1 rs17056509 4 172041974 G 739 12.28 0.00046 1.97 2.27 0.023 DG N N - N 1 rs2256531 6 80807454 A 740 14 0.00018 1.12 2.27 0.023 DG N Y N N 1 119 120 Supplemental table 4.2 Continued 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Paternal smoking prior to pregnancy 299-340 SNPs tested 2=0.000147- 0.000167 155-180 SNPs 2=0.000278- 0.000323 rs1486230 3 131195753 C 740 12.81 0.00034 0.76 2.93 0.0034 DG N N - N 1 rs7622001 3 131183784 G 740 12.81 0.00034 0.76 2.93 0.0034 DG N N - N 1 rs7909593 10 12402466 C 740 12.71 0.00036 -1.08 -2.72 0.0066 EG N N - N 1 rs9947632 18 36062561 A 740 12.73 0.00036 -0.72 -2.71 0.0067 EG N N - N 1 rs13022554 2 166752432 G 737 15.55 0.00042 -0.76 -2.58 0.0099 DGEG N N - N 1 rs1571046 1 101299216 A 740 16.93 0.00021 -0.64 -2.58 0.0099 DGEG EG N Y N N 1 rs2568222 2 85377979 A 740 17.07 0.0002 0.65 2.52 0.012 DGEG DG N Y N N 1 rs1517516 3 16475215 A 739 12.26 0.00046 0.79 2.47 0.014 EG N N - N 1 rs913905 13 30720425 G 740 15.47 0.00044 -0.61 -2.44 0.015 DGEG N N - N 1 rs7689213 4 187730631 G 740 12.72 0.00036 -0.67 -2.4 0.016 DG N N - N 1 120 121 Supplemental table 4.3 Characteristics of Top SNPs Identified from Two-Step Scans by Tobacco Smoke Exposure for Diagnoses with B-Cell ALL 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Maternal smoking in early childhood 264-304 SNPs tested 2=0.000164- 0.000189 133-152 SNPs tested 2=0.000329- 0.000376 rs3112511 17 16207526 A 688 12.73 0.00036 0.97 2.92 0.0035 DG N N - N 2 rs17628268 4 113330772 G 689 13.79 0.0002 1.16 2.44 0.015 EG N Y N N 2 rs6986417 8 55355086 G 689 15.49 0.00043 -0.75 -2.35 0.019 DGEG N N - N 1 rs10757058 9 19419528 G 688 18.99 0.000075 -1.31 -2.33 0.02 DGEG DG N Y N N 1 rs11204447 20 59984927 G 689 12.4 0.00043 -0.88 -2.32 0.02 EG N N - N 1 rs12459746 19 38857085 G 687 12.52 0.0004 -1.03 -2.33 0.02 DG N N - N 1 rs10195963 2 217862696 A 689 13.03 0.00031 -0.87 -2.3 0.021 EG N N - N 1 rs6016373 20 39154095 G 684 12.24 0.00047 0.96 2.26 0.024 DG N N - N 1 rs999562 19 5091332 A 689 13.62 0.00022 -0.91 -2.24 0.025 EG N Y N N 1 rs9899630 17 9068448 A 689 19.36 0.000062 -1.03 -2.19 0.029 DGEG DG N Y N N 1 Maternal smoking prior to pregnancy 247-303 SNPs tested 2=0.000165- 0.000202 115-133 SNPs tested 2=0.000376- 0.000435 rs10864233 1 216295188 A 689 13.87 0.0002 1.14 3.11 0.0019 EG N Y N N rs7421154 2 68665945 A 689 13.68 0.00022 1.55 2.92 0.0035 EG N Y N N 2 rs7260166 19 9339399 G 689 16.15 0.00031 -1.27 -2.78 0.0054 DGEG DG N N - N 2 rs12562620 1 216286427 G 689 23.15 0.0000094 1.04 2.71 0.0067 DGEG EG N Y N N 1 rs3112511 17 16207526 A 688 12.73 0.00036 0.9 2.65 0.008 DG N N - N 2 rs2792833 6 49871812 A 689 16.71 0.00024 -1.03 -2.62 0.0089 DGEG DG N Y N N 1 rs2068721 1 216252742 C 689 15.34 0.00047 -0.97 -2.61 0.0092 DGEG EG N N - N 1 rs1544299 1 216264388 C 689 15.49 0.00043 -0.95 -2.55 0.011 DGEG EG N N - N 1 rs2452153 11 88135392 C 689 13.47 0.00024 -1.84 -2.54 0.011 DG N Y N N 1 rs874690 17 68102616 A 689 13.02 0.00031 -1.92 -2.43 0.015 EG N N - N 2 121 122 Supplemental table 4.3 Continued 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Maternal smoking during pregnancy 276-303 SNPs tested 2=0.000165- 0.000181 133-148 SNPs tested 2=0.000338- 0.000376 rs10821938 10 63724773 C 689 23.38 0.0000084 1.42 2.96 0.0031 DGEG DG N Y N Y 1 rs7260166 19 9339399 G 689 16.41 0.00027 -1.52 -2.91 0.0036 DGEG DG N N - N 2 rs7421154 2 68665945 A 689 19.54 0.000057 1.72 2.75 0.006 DGEG EG N Y N N 2 rs4506592 10 63727187 G 689 34.9 0.000000026 1.09 2.34 0.019 DGEG DG N Y N N 1 rs1820325 2 166534206 G 689 13.66 0.00022 -1.32 -2.33 0.02 EG N Y N N 1 rs7090445 10 63721176 G 689 37.2 8.4E-09 -1 -2.25 0.024 DGEG DG N Y N N 1 rs4975714 5 1934666 G 689 12.93 0.00032 2.2 2.24 0.025 DG N N - N 1 rs17628268 4 113330772 G 689 12.36 0.00044 1.27 2.21 0.027 EG N N - N 2 rs13147855 4 10691942 A 689 18.66 0.000089 1.05 2.18 0.029 DGEG EG N Y N N 1 rs874690 17 68102616 A 689 12.5 0.00041 -2.38 -2.16 0.03 EG N N - N 2 Maternal smoking in early childhood + paternal smoking prior to pregnancy 288-303 SNPs tested 2=0.000165- 0.000174 133-155 SNPs tested 2=0.000323- 0.000376 rs1338472 6 45968780 A 658 12.26 0.00046 1.06 2.53 0.011 DG N N - N 1 rs1945996 11 56742853 A 658 12.69 0.00037 1.56 2.51 0.012 EG N N - N 1 rs4973760 3 27349047 A 658 13.2 0.00028 -1.19 -2.5 0.013 EG N N - N 1 rs7771614 6 45975798 A 657 15.38 0.00046 1.04 2.42 0.015 DGEG DG N N - N 1 rs7971297 12 131872850 A 658 18.62 0.000091 -1.52 -2.41 0.016 DGEG DG N Y N N 1 rs12518320 5 172794876 A 658 12.12 0.0005 -1.07 -2.38 0.017 DG N N - N 1 rs2095074 9 16251695 G 658 12.14 0.00049 1.79 2.37 0.018 EG N N - N 1 rs1251596 12 63703062 G 658 21.71 0.000019 2.09 2.26 0.024 DGEG EG N Y N N 1 rs10137048 14 26536968 A 657 14.18 0.00017 1.36 2.18 0.029 DG N Y N N 1 rs12488675 3 1492235 A 658 14.5 0.00014 -1.4 -2.18 0.03 EG N Y N N 1 122 123 Supplemental table 4.3 Continued 1=0.0005 1=0.00025 SNP CHR Location Min. All. N Step 1 χ 2 Value Step 1 P Beta GxE Step 2 T- statistic Step 2 P 2 Step Method Statistically significant Tested in Step 2 Statistically significant Sign. weighted? # of exposures with top hit Paternal smoking prior to pregnancy 241-304 SNPs tested 2=0.000164- 0.000207 rs6530778 8 14313293 G 658 15.72 0.00039 -0.82 -3.15 0.0016 DGEG N N - N 1 rs11132376 4 186928529 A 657 12.83 0.00034 0.99 3.04 0.0023 EG N N - N 1 rs7567289 2 6389676 A 657 13.88 0.0002 1.12 2.97 0.003 EG N Y N N 1 rs12616042 2 182062879 G 658 13.45 0.00025 0.74 2.56 0.011 EG N Y N N 1 rs13022554 2 166752432 G 655 19.74 0.000052 -0.83 -2.55 0.011 DGEG EG N Y N N 1 rs6668037 1 239332114 A 658 15.36 0.00046 -0.82 -2.52 0.012 DGEG EG N N - N 1 rs499966 13 32998610 A 658 13.24 0.00027 -0.8 -2.45 0.014 EG N N - N 1 rs9947632 18 36062561 A 658 12.15 0.00049 -0.66 -2.4 0.017 EG N N - N 1 rs9610489 22 36713531 G 658 12.85 0.00034 -0.67 -2.37 0.018 DG N N - N 1 rs2568222 2 85377979 A 658 16.58 0.00025 0.65 2.33 0.02 DGEG DG N Y N N 1 123 124 4.8 REFERENCES 1. Papaemmanuil E, Hosking FJ, Vijayakrishnan J, et al. Loci on 7p12.2, 10q21.2 and 14q11.2 are associated with risk of childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1006-10. 2. Prasad RB, Hosking FJ, Vijayakrishnan J, et al. Verification of the susceptibility loci on 7p12.2, 10q21.2, and 14q11.2 in precursor B-cell acute lymphoblastic leukemia of childhood. Blood 2010;115(9):1765-7. 3. Treviño LR, Yang W, French D, et al. Germline genomic variants associated with childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1001-5. 4. Healy J, Richer C, Bourgey M, et al. Replication analysis confirms the association of ARID5B with childhood B-cell acute lymphoblastic leukemia. Haematologica 2010;95(9):1608-11. 5. Sherborne AL, Hosking FJ, Prasad RB, et al. Variation in CDKN2A at 9p21.3 influences childhood acute lymphoblastic leukemia risk. Nature Genetics 2010;42(6):492-4. 6. Chokkalingam AP, Hsu LI, Metayer C, et al. Genetic variants in ARID5B and CEBPE are childhood ALL susceptibility loci in Hispanics. Cancer Causes Control 2013;24(10):1789-95. 7. Han S, Lee K-M, Park SK, et al. Genome-wide association study of childhood acute lymphoblastic leukemia in Korea. Leukemia Research 2010;34(10):1271-4. 8. Almalte Z, Samarani S, Iannello A, et al. Novel associations between activating killer-cell immunoglobulin-like receptor genes and childhood leukemia. Blood 2011;118(5):1323-8. 9. Bonaventure A, Goujon-Bellec S, Rudant J, et al. Maternal smoking during pregnancy, genetic polymorphisms of metabolic enzymes, and childhood acute leukemia: the ESCALE Study (SFCE). Cancer causes & control : CCC 2011;23(2):329-45. 10. Liu R, Zhang L, McHale CM, et al. Paternal smoking and risk of childhood acute lymphoblastic leukemia: systematic review and meta-analysis. J Oncol 2011;2011:854584. 11. Milne E, Greenop KR, Scott RJ, et al. Parental Prenatal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia. American journal of epidemiology 2011;175(1):43-53. 12. Lee K-M, Ward MH, Han S, et al. Paternal smoking, genetic polymorphisms in CYP1A1 and childhood leukemia risk. Leukemia Research 2009;33(2):250-8. 13. Klimentopoulou A, Antonopoulos CN, Papadopoulou C, et al. Maternal smoking during pregnancy and risk for childhood leukemia: A nationwide case-control study in Greece and meta-analysis. Pediatric Blood & Cancer 2011;58(3):344-51. 14. Boffetta P, Trédaniel J, Greco A. Risk of childhood cancer and adult lung cancer after childhood exposure to passive smoke: A meta-analysis. Environmental Health Perspectives 2000;108(1):73-82. 15. Clavel J, Bellec S, Rebouissou S, et al. Childhood leukaemia, polymorphisms of metabolism enzyme genes, and interactions with maternal tobacco, coffee and alcohol consumption during pregnancy. Eur J Cancer Prev 2005;14(6):531-40. 124 125 16. Infante-Rivard C, Krajinovic M, Labuda D, et al. Parental smoking, CYP1A1 genetic polymorphisms and childhood leukemia (Québec, Canada). Cancer causes & control : CCC 2000;11(6):547-53. 17. Chang JS. Parental Smoking and the Risk of Childhood Leukemia. American journal of epidemiology 2006;163(12):1091-100. 18. Orsi L, Rudant J, Bonaventure A, et al. Genetic polymorphisms and childhood acute lymphoblastic leukemia: GWAS of the ESCALE study (SFCE). Leukemia 2012:1-4. 19. Kooperberg C, Leblanc M. Increasing the power of identifying gene x gene interactions in genome-wide association studies. Genet Epidemiol 2008;32(3):255-63. 20. Murcray CE, Lewinger JP, Gauderman WJ. Gene-environment interaction in genome-wide association studies. Am J Epidemiol 2009;169(2):219-26. 21. Gauderman WJ, Zhang P, Morrison J, et al. Finding GWAS signals in the lower Manhattan by testing GxE interactions. International Genetic Epidemiology Society Annual Conference; Stevenson, WA, 2012. 22. Wang Y, Devereux W, Stewart TM, et al. Characterization of the interaction between the transcription factors human polyamine modulated factor (PMF-1) and NF-E2-related factor 2 (Nrf-2) in the transcriptional regulation of the spermidine/spermine N1-acetyltransferase (SSAT) gene. The Biochemical journal 2001;355(Pt 1):45-9. 23. Wang Y, Devereux W, Stewart TM, et al. Cloning and characterization of human polyamine-modulated factor-1, a transcriptional cofactor that regulates the transcription of the spermidine/spermine N(1)-acetyltransferase gene. The Journal of biological chemistry 1999;274(31):22095-101. 24. Pryor WA. Biological effects of cigarette smoke, wood smoke, and the smoke from plastics: the use of electron spin resonance. Free radical biology & medicine 1992;13(6):659-76. 25. Rangasamy T, Cho CY, Thimmulappa RK, et al. Genetic ablation of Nrf2 enhances susceptibility to cigarette smoke –induced emphysema in mice. Journal of Clinical Investigation 2004;114(9):1248-59. 26. Kobayashi M, Yamamoto M. Nrf2-Keap1 regulation of cellular defense mechanisms against electrophiles and reactive oxygen species. Advances in enzyme regulation 2006;46:113-40. 27. Krajinovic M, Sinnett H, Richer C, et al. Role of NQO1, MPO and CYP2E1 genetic polymorphisms in the susceptibility to childhood acute lymphoblastic leukemia. Int J Cancer 2002;97(2):230-6. 28. Masuko H, Sakamoto T, Kaneko Y, et al. An interaction between Nrf2 polymorphisms and smoking status affects annual decline in FEV1: a longitudinal retrospective cohort study. BMC medical genetics 2011;12:97. 29. Tijhuis MJ, Visker MH, Aarts JM, et al. NQO1 and NFE2L2 polymorphisms, fruit and vegetable intake and smoking and the risk of colorectal adenomas in an endoscopy-based population. Int J Cancer 2008;122(8):1842-8. 30. Muller T, Hengstermann A. Nrf2: friend and foe in preventing cigarette smoking- dependent lung disease. Chemical research in toxicology 2012;25(9):1805-24. 125 126 31. Kosmider B, Messier EM, Chu HW, et al. Human alveolar epithelial cell injury induced by cigarette smoke. PLoS One 2011;6(12):e26059. 126 127 Chapter 5: National Institute of Health R21 Grant “Genome- Wide Interaction Scan (GWIS) Analysis of Pesticide Exposure and Risk of Childhood Acute Lymphocytic Leukemia (ALL)” National Institute of Health, R21 Grant, submitted October 2013 Resubmission July 2014 5.1 NARRATIVE Relevance to Public Health We propose to conduct a gene x environment interaction analysis to evaluate genetic susceptibility to ambient pesticide exposure for acute lymphocytic leukemia (ALL). The study will include children enrolled in a large, population-based case-control study of childhood leukemia. Unique features of the study include state-of-the-art exposure models to obtain precise estimates of ambient pesticide exposure, and novel statistical methods designed to maximize statistical power for gene-environment interaction analysis that incorporate genome wide data. The results of this study may be used to impact regulatory policy on allowable levels of early life exposures to ambient pesticides, and to inform directed analyses of gene-pesticide interactions using biomarkers of pesticide exposure for the most likely causal agents and targeted sequencing in genetic regions of interest. 5.2 ABSTRACT Childhood leukemia is the most common type of cancer in children under 15 years of age, representing more than 30% of all childhood cancers. The incidence rate of 127 128 childhood acute lymphocytic leukemia (ALL), the most common leukemia in children, has been increasing 0.8% per year since 1973, with the greatest increase in recent years among Hispanic children (1.3% per year from 1992-2010). Although survival rates for childhood leukemia are high, treatment often results in devastating long-term physiologic and psychological effects. More than 95% of survivors have a chronic health condition as an adult, of which more than 80% are disabling or life-threatening. Childhood ALL is a complex disease in which both genetic and environmental factors appear to contribute. Molecular epidemiologic studies have found that a prenatal genetic mutation alone is not sufficient for leukemia development, but that a second c ritica l ‘hit’ e a rl y in li f e , suc h a s a n e nvironm e ntal c ha ll e n ge , is ne c e ssar y for A LL initiation. Meta-analyses have implicated parental occupational (exposure assigned based on job title) and residential exposure to pesticides (self-report of regularly used pesticides) as risk factors for childhood ALL. Studies on the role of genetic susceptibility to pesticides (i.e., gene-pesticide interactions) on childhood ALL are promising but limited to date. To fully evaluate the association between ambient pesticides and childhood ALL risk, more accurate and specific methods of pesticide exposure assessment are needed to minimize misclassification, and efficient statistical analytic approaches are needed to maximize power to evaluate gene-pesticide interactions. We propose to conduct a Genome-Wide Interaction Scan (GWIS) to evaluate the association between childhood ALL and genetic susceptibility to ambient pesticide exposure, among children enrolled in the California Childhood Leukemia Study (CCLS), a large, population-based case-control study (997 cases/1230 controls). We will use state-of-the-art pesticide exposure models, incorporating Pesticide Use Reporting 128 129 (PUR) data, Land-Use data, and complete residential histories of children from 1 year prior to birth until diagnosis to obtain precise estimates of ambient pesticide exposure to minimize exposure misclassification. To evaluate gene-pesticide interactions and childhood ALL risk, we will use novel, statistically efficient methods for GxE interaction analysis that utilize available genome-wide data. This analysis is only feasible given our unique clinical and genetic resource (CCLS, designed and assembled by research team at UC Berkeley) and recent development of highly statistically efficient GxE interaction scan procedures and sophisticated ambient pesticide exposure models. The results of this study may be used to inform regulatory policy on allowable levels of early life exposures to ambient pesticides, and to direct subsequent analyses of gene- pesticide interactions using biomarkers of pesticide exposure for the most likely causal agents and targeted sequencing in genetic regions of interest. 5.3 SPECIFIC AIMS Childhood leukemia is the most frequent cancer in young children (0-14 years of age), representing more than 30% of all childhood cancers. Acute lymphocytic leukemia (ALL) is the most common subtype, constituting approximately 80% of all childhood leukemias. The incidence of childhood ALL has been growing by 0.8% per year since 1973, with the greatest increase in recent years among Hispanic children(1). The reasons for the increase or why incidence has been growing more rapidly in this ethnic group is unknown. Although survival rates for childhood ALL are high, treatment is extremely taxing to the child and often results in devastating long-term physiologic and 129 130 psychological effects. More than 95% of survivors have a chronic health condition as an adult, and more than 80% are disabling or life-threatening disorders(2). While both genetic and environmental factors appear to contribute to childhood ALL, it is thought that less than 10% of all cases may be explained by recognized genetic or environmental risk factors(3). Prenatal mutations or rearrangements are common in childhood ALL, but molecular work has shown that these changes are not sufficient to initiate disease. Prenatal mutations may combine with other pre or postnatal environmental exposures or genetic factors to initiate disease. A growing number of studies have evaluated the role of genetic susceptibility to childhood ALL using genome wide association (GWAS) data(4-8), however, few studies have evaluated gene-pesticide interactions(9, 10). Meta-analyses have implicated parental occupational (exposure assigned based on job title)(11) and residential exposure to pesticides (self-report of regularly used pesticides)(12) as risk factors for childhood ALL. To elucidate the association between ambient pesticides and childhood ALL risk, a more accurate and comprehensive exposure modeling approach is needed. Further, efficient statistical analytic approaches that allow incorporation of both environmental (e.g., pesticide) and genome-wide data are necessary for accurate estimation of the contribution of pesticide exposure to ALL risk. The proposed study will evaluate the association between childhood ALL and ambient pesticide exposure at critical windows early in life in the California Childhood Leukemia Study (CCLS), a large, population-based case-control study (998 cases/1230 controls). We will use established, state-of-the-art exposure models, which utilize Pesticide Use Reporting (PUR) data, Land-Use Data, and complete residential histories 130 131 of all children from 1 year prior to birth until diagnosis to obtain precise estimates of exposure to ambient pesticides. These models decrease non-differential exposure misclassification compared to exposure assessment methods using PUR data alone. We will explore the association between ALL risk, early life pesticide exposure, and genetic susceptibility using GWAS data and efficient statistical methods that maintain adequate power to detect gene-environment interactions. Our hypotheses and specific aims are: Aim 1: Evaluate the risk of childhood ALL associated with exposure to ambient pesticides, using state-of-the-art exposure models that incorporate PUR data, Land- Use data, and complete residential history for children in the California Childhood Leukemia Study (CCLS, 1995-2008). Hypothesis 1: We hypothesize that children are at an increased risk of ALL with early life exposure to high levels of ambient pesticides, including herbicides, insecticides, and fungicides of specific toxicological (e.g., endocrine disruptors) or physiochemical classes (e.g., organophosphates). Aim 2: Evaluate gene-environment (GxE) interactions and risk of childhood ALL associated with exposure to ambient pesticides for Hispanic and non-Hispanic children enrolled in CCLS using precise exposure estimates from Aim 1 and novel statistical methods for GxE analysis.  Aim 2.1 Use ambient pesticide exposure estimates and available GWAS data for Hispanic children (CCLS, 1995-2008) to scan the genome for susceptibility loci using two-step approaches for GxE interaction analysis. 131 132  Aim 2.2 Replicate results in non-Hispanic (CCLS, 1995-2008), and Hispanic children (CCLS, 2009-2014). Hypothesis 2: We hypothesize that a gene-environment interaction scan (GWIS) analysis will show that children with genetic susceptibility to pesticides are at an increased risk of A L L wh e n e x pose d e a rl y in li fe , or whe n the c hi ld’s pa re nts we r e e x pose d to hig h leve ls of pesticides prior to conception or during pregnancy. We will apply new analytic approaches to detect gene-environment interactions, using a Genome-Wide Interaction Scan (GWIS) that provides substantially higher statistical power than traditional analytic approaches. The analysis will be completed in a large case-control study with a high proportion of Hispanic children from central California, an agricultural region with high levels of ambient pesticides. The focus on Hispanic children is especially important given the high burden of childhood ALL in this population. Results of this project may inform the development of focused gene- environment interaction studies through the evaluation of precise pesticide exposure assessment models and efficient statistical methods for genome-wide interaction analysis. Results may be used to identify the most likely causal agents for directed analyses of gene-pesticide interactions that incorporate biomarkers of pesticide exposure and targeted sequencing in genetic regions of interest. 5.4 RESEARCH STRATEGY Significance 132 133 Leukemia is the most common form of cancer among children 0-14 years of age, yet the causes of childhood leukemia remain elusive. More than 3,600 cases of childhood leukemia are expected to be diagnosed in the US in 2013, representing almost a third of all childhood cancers(13). Acute lymphocytic leukemia (ALL) is the most common subtype of leukemia in children, with approximately 80% of all cases belonging to this group. Demographic characteristics of children at high risk include: male sex, age (between 2-6 years for ALL), and Hispanic ethnicity. Ionizing radiation remains the only established causal environmental risk factor for childhood leukemia. Other lifestyle or environmental factors have been suggested as causal agents, with the strongest evidence for residential or occupational exposure to pesticides(11, 12). Meta-analyses also suggest elevated ALL risk for non-ionizing radiation(14-16), paternal tobacco use(17, 18), high birth weight(19, 20), with protective effects observed for atopic conditions (e.g. allergy, asthma, hay fever)(21, 22) and day care attendance (as a surrogate for early life immune challenge)(23). Since 1973, incidence rates of ALL have increased by approximately 0.8% per year in children of all races/ethnicities (Figure 1), with the greatest increase in more recent years among Hispanic children (1.3% per year from 1992-2010)(1). Reasons for this increase have not been established. Despite high survival rates, treatment for childhood leukemia is extremely difficult on a child, and frequently results in long-term medical problems. More than 95% of childhood cancer survivors have a chronic health condition as an adult, and more Figure 5.1. Age-Adjusted Incidence Rates for Childhood ALL from 1973- 2010, SEER 9 2 2.5 3 3.5 4 4.5 Age-Adjusted Incidence Rate (per 100,000) Year of Diagnosis 133 134 than 80% of survivors have a disabling or life-threatening chronic condition. Survivors have high prevalence of pulmonary (65%), auditory (62%), endocrine/reproductive (62%), cardiac (56%) or neurocognitive (48%) disorders(2). Environmental exposures in susceptible children may explain some of the increased incidence of ALL. Childhood ALL is a complex disease in which both genetic and environmental factors appear to contribute. While chromosomal alterations or mutations are recognized initiating events in many cases of ALL, they are not considered sufficient causes of disease. For example, the TEL-AML1 translocation, a common ALL mutation, is found in 1 in 100 children at birth, yet only 1 in 25,000 children develop ALL(3). It is hypothesized that a second critical ‘hit’ early in life(24), such as an environmental challenge, is necessary for the initiation of ALL in susceptible children. These mutations, many of which have a prenatal origin(25-27), may serve as the backdrop against which other factors, including in utero exposures, genetic factors, and postnatal exposures and mutations, may trigger disease onset(28). There may be a strong role for genetic susceptibility, via main effects of genes or interaction with other factors, in a t lea st one of the two or more ‘hits’ re quire d fo r the de ve lopm e nt of A LL . The role of other genetic susceptibility factors, such as single nucleotide polymorphisms (SNPs), has been investigated in several studies of non-Hispanic white children. Results from genome-wide association studies (GWAS) suggest an association between SNPs in ARID5B, IKZF1, CEBPE, and CDKN2A/B and childhood ALL risk(4-8), but the potential interaction with environmental exposures has not been investigated. A recent review of childhood leukemia suggests that less than 10% of all cases can be explained by recognized genetic or environmental risk factors(3). Additional genetic variants, alone or 134 135 in combination with environmental carcinogens (e.g., pesticides), may explain remaining cases of ALL. Toxicological and experimental studies suggest biologically plausible mechanisms for a causal association between early life exposure to pesticides (e.g., in utero or early childhood exposure) and risk of childhood ALL. Epidemiologic data is needed to elucidate the risk associated with specific pesticides, given the greater susceptibility of children to harmful agents. Children may be at a higher risk of toxicity than adults due to increased metabolic rates, differential detoxification of chemicals, and exposure during critical developmental stages(29). Oxidative stress or endocrine disruptor mediated mechanisms are the most probable link between pesticides and carcinogenesis. An overproduction of Reactive Oxygen Species (ROS) resulting from parental or childhood exposure to pesticides may cause extensive DNA and protein damage in cells. A specific target of ROS, such as tumor suppressor genes or oncogenes, has not been established with respect to childhood leukemia, but more research is needed. Alternatively, endocrine disruptors found in pesticides may lead to changes in gene network expression by binding and displacing endogenous ligands. Inflammatory or epigenetic mechanisms may also be involved, but research is in a nascent stage of development(30). Both parental occupational and residential exposure to pesticides have been implicated as risk factors for childhood leukemia, using self-reported data. Meta- analyses of epidemiologic studies of residential exposure during pregnancy to unspecified pesticides, insecticides, and herbicides found relative risk estimates of 1.54 (95%CI: 1.13, 2.11; 11 studies), 2.05 (95%CI: 1.80, 2.32; 8 studies), and 1.61 (95%CI: 1.20, 2.16; 5 135 136 studies), respectively(12). For residential exposure in childhood, increased risk estimates were observed both for exposure to unspecified pesticides (OR=1.38; 95%CI: 1.12, 1.70; 9 studies), and insecticides (OR=1.61; 95%CI: 1.33, 1.95; 7 studies)(12). Meta-analyses of epidemiologic studies of maternal occupational exposure to pesticides and childhood leukemia observed increased risks for maternal prenatal exposure overall (OR=2.09; 95%CI: 1.51, 2.88; 16 studies), with exposure to insecticides (OR=2.72; 95%CI: 1.47, 5.04; 6 studies) and herbicides (OR=3.62; 95%CI: 1.28, 10.3; 2 studies), and with exposure to any pesticides when restricting to ALL only (OR=2.64; 95%CI: 1.40, 5.00; 5 studies)(11). No association was observed for paternal occupational exposure to pesticides and childhood leukemia (OR=1.09; 95%CI: 0.88, 1.34; 30 studies)(11). Several positive, but statistically non-significant associations were observed for ambient exposure to specific pesticides and childhood ALL based on Pesticide Use Reporting (PUR) data only and residential address, using a subset of participants from phases I and II (1995-2002) of the California Childhood Leukemia Study (CCLS)(31). The current proposal will use a larger, updated sample of CCLS children with more detailed ambient exposure measures and novel statistical models to evaluate gene x pesticide interactions using genome-wide data. The addition of data on genetic susceptibility and more comprehensive and accurate estimates of ambient pesticide exposure may lead to better estimates of childhood ALL risk. Genetically susceptible children may be at a higher risk of ALL when exposed to ambient pesticides early in life. Risk may have been underestimated in previous studies that focused on reporting marginal effects (i.e., environment only), without evaluating effects that occur in children of differing genetic susceptibility (gene x 136 137 environment [GxE] interactions). A recent candidate gene study of MDR1 variants, indoor pesticide exposure and childhood ALL found that children with a CGC haplotype for 3 MDR1 SNPs were less susceptible to the effects of pesticides that children with a variant haplotype(9). One additional study evaluated gene-pesticide interactions for 4 genes and residential pesticide exposure, and found increased interaction ORs for two variants in CYP1A1(10). However, no studies have used efficient large-scale approaches to evaluate genome-wide GxE interactions, or have used such approaches with advanced models of ambient pesticide exposure. Overall impact of our approach. This GxE interaction analysis will identify potential susceptibility loci relevant to ambient pesticide exposure and childhood ALL. Results may contribute to a growing body of scientific literature needed to accurately inform the public about potential risks of pesticide exposure among children, and men and women of childbearing age, and to impact regulatory policy on allowable levels of pesticide exposure. The US Environmental Protective Agency (EPA) has prioritized the support of re se a rc h to “ e stablis h a sound sc ientific found a ti on” in orde r to be tt e r unde rsta nd the risks to children associated with pesticide exposure, aligned with the organization mi ssi on to prote c t c hi ldre n’s e nvironm e ntal he a lt h (29). Identification of gene x environment interactions may help to classify children at risk of greatest harm from ambient pesticide exposure. This project will test a new statistical method for large-scale analysis of gene-environment interactions, using genome-wide data, and risk of childhood ALL. The results of this project may help to refine etiological hypotheses for future grants. Results may be used to inform directed analyses of gene-pesticide 137 138 interactions for the most likely causal agents incorporating biomarkers of pesticide exposure and targeted sequencing in genetic regions of interest. Our proposed study is aligned with the 2012-2017 NIEHS strategic plan, goals 1 (understand the role of biological mechanisms in determining susceptibility to environmental stressors), 3 (advance characterization of environmental exposures through improved exposure assessment), 6 (understand the disproportionate risks of disease), and 9 (inspire scientists to move environmental health science forward, and train the next generation of leaders). Innovation The results of the proposed analysis will contribute to the understanding of the role of ambient pesticide exposure and susceptibility to childhood ALL risk. The study will be used to advance our understanding of the natural history of childhood ALL, and examine the contribution of genetic susceptibility to ALL following exposure to ambient pesticides during critical windows of fetal and child development. The project will incorporate both advanced exposure assessment modeling techniques to estimate exposure to ambient pesticides and novel GxE statistical scanning methods to identify gene-pesticide interactions, using existing genome-wide genotyping data in a large, population-based study of childhood ALL. The specific innovations include: 1. This study will be the first published study of childhood ALL to evaluate the risk associated with pesticides using precise estimates of ambient exposure that 138 139 utilize both PUR and Land-Use data with complete residential history from 1 year prior to birth until diagnosis (or reference date). 2. Established, validated, state-of-the-art exposure models developed under the direction of Dr. Myles Cockburn (Co-Investigator to this project), will be used to calculate precise estimates of exposure to ambient pesticides (by specific chemical and by toxicological or physiochemical class). These exposure models have been shown to provide more precise measures of exposure, in comparison to models utilizing PUR data alone, thereby reducing non-differential misclassification(32). 3. Novel, statistically efficient analytical methods, developed by Dr. James Gauderman (Co-Investigator of this project) will be used to evaluate GxE interactions and risk of childhood ALL, using genome-wide genetic data. Studies of childhood ALL have historically been underpowered to detect susceptibility loci reaching genome-wide significance, due to small sample sizes resulting from the rarity of the disease, and potential overcorrection due to necessary adjustment for testing of high numbers of single nucleotide polymorphisms. Several statistical methods have recently been proposed that utilize two-step scanning or ranking methods for GxE interaction analysis using GWAS data. Two-step methods provide substantial improvements in power over traditional methods for interaction analysis (i.e. standard case-case control models with an interaction term)(33). In Aim 2, we will use these new statistical methods to conduct a genome-wide interaction scan (GWIS) of ambient pesticide exposure among Hispanic children to identify susceptibility loci using available genome-wide genetic data. 139 140 4. This study will focus on Hispanic children, a traditionally underrepresented ethnic group experiencing high and growing rates of childhood ALL. Our sample includes children living in California, which has the largest agricultural pesticide use in the United States(34). A considerable proportion of children are from central California, an agricultural region with high rates of pesticide application. 5. Methods developed and tested in this study of childhood ALL can be applied to other phenotypes that may be impacted by both genetic and environmental factors, and provide a standard procedure for evaluating susceptibility to other risk factors for childhood ALL, such as air pollution or early life immune challenge. 6. The methods and results of this study may contribute to the design of refined pesticide and genetic studies used to inform regulatory policy on use of pesticides. To realize their mission to protect the health of children, the US EPA Office of P e sti c ide P rogr a ms ( O P P ) ha s prior it iz e d 1) “ e stablis hing a sound sc ientific found a ti on” to “ im prove our unde rsta ndin g of t he potential risks to c hil dre n fr om pe sti c ide use ” a nd 2) ris k a ssessment/m a na ge me nt that “ c onsi de rs the un iquene ss of c hil dre n” to “inf o rm de c i sions that a re prote c ti ve of c hil dre n” (29). Approach Overview. This study will evaluate the association between ambient pesticide exposure and childhood ALL risk using developed exposure models that incorporate PUR data, land-use data, and complete residential history. We will explore this association using existing GWAS data to evaluate gene-environment interactions using novel statistical methods for GxE interaction analysis, recently developed at USC. 140 141 The Research Team. The CCLS is one of the largest and most comprehensive ongoing studies of childhood leukemia in the US. The project PI, Dr. Roberta McKean- Cowdin (USC), Dr. Catherine Metayer (UC Berkeley), Dr. James Gauderman (USC), Dr. Myles Cockburn (USC), and Ms. Jessica Barrington-Trimis (USC) have worked closely together on components of the CCLS and all contribute expertise to completing large- scale genomics and environmental studies. Dr. Roberta McKean-Cowdin is a Co-Investigator of the CCLS, and Local PI for the southern California study center at USC. She has worked closely with investigators at UC Berkeley for many years on childhood cancer and has been responsible for assisting Dr. Metayer and colleagues on the expansion of this large statewide study of childhood leukemia into southern California. Dr. McKean-Cowdin is a cancer epidemiologist with experience in environmental and genetic epidemiology in the West Coast Childhood Brain Tumor Study (NCI R01 CA116724-01A1), Gene-Environment Factors in Childhood Brain Tumors (R01 CA 116724), Air Pollution and Childhood Brain Tumors (AQMD BTAP006), and the California Childhood Leukemia Study. Dr. Catherine Metayer, Assistant Professor, UC Berkeley School of Public Health is the Associate Director for Research at the CCLS. Dr. Metayer was instrumental in the design and collection of data for the CCLS and is actively managing recruitment procedures and data collection. Dr. Metayer has led the analysis and drafting of numerous CCLS manuscripts and is a lead investigator of the Childhood Leukemia International Consortium(35). 141 142 Dr. Jim Gauderman, Professor, Department of Preventive Medicine, USC Keck School of Medicine is Chair of the Biostatistics Division. He has expertise in the development of statistical methods for gene-environment interaction analysis, and has developed the specific statistical methods that will be used in this analysis. He will oversee the analysis of gene x pesticide interactions. Dr. Myles Cockburn, Professor, Department of Preventive Medicine, USC Keck School of Medicine has extensive experience in creating geospatial land-use regression models to for precise estimation of exposure to ambient pesticides in California. He will be primarily responsible for directing the ambient pesticide exposure assessment and analysis for children in this study using established regression models. Jessica Barrington-Trimis is a PhD student in Epidemiology at USC under the mentorship of Dr. McKean-Cowdin and Dr. Gauderman, and will begin a post-doctoral position in July 2014. Her dissertation is on analysis of genetic and environmental risk factors for childhood cancer. Ms. Barrington-Trimis has completed analyses of GxE interactions and risk of childhood brain tumors(36) and childhood ALL. Her analysis of childhood ALL uses data from the CCLS and GxEscan, a program implementing two- step scanning methods to evaluate GxE interactions using GWAS data. Participants. Analysis of exposure to ambient pesticides and childhood ALL risk will be conducted in Hispanic and non-Hispanic cases and controls enrolled in phases I- III (1995-2008) of the CCLS (n=998 cases/1230 controls). GWIS analyses to evaluate gene-pesticide interactions will use genome-wide genetic data, available for Hispanic participants in phases I-III (380 cases/454 controls). Replication of the most promising 142 143 SNPs (see Statistical Analysis, Aim 2.2) will be completed in non-Hispanic white children in phases I-III (1995-2008) for approximately 550 cases and 650 controls. Additional replication will be completed in Hispanic children enrolled in phases IV-V (2009-2014 [ongoing]; n~250 cases). Completion of GWAS for these groups is planned in the current phase of the CCLS. Cases are children diagnosed with childhood acute lymphocytic leukemia (ALL) at 0-14 years of age, and identified through rapid case ascertainment (within 72 hours) from participating hospitals. Controls were frequency matched to cases on date of birth, gender, and maternal race using birth certificate files. Both cases and controls were living in California at the time of diagnosis/reference date, in one of 38 participating counties in northern, central, and southern California. Outcomes. Childhood leukemia was defined according to the International Classification of Childhood Cancer (ICCC) which uses the World Health Organization International Classification of Diseases for Oncology (ICDO-3) histology codes to define diagnosis with childhood leukemia: 9811-9818, 9835-9837 (precursor cell acute lymphoid leukemias); 9823, 9826, 9832-9833, 9940 (mature B-cell lymphoid leukemias); 9827, 9831, 9834, 9948 (Mature T-cell and NK cell lymphoid leukemias); 9820 (lymphoid leukemia, NOS). Precursor cell acute lymphocytic leukemias are further classified by molecular subtype: B-cell (9836); T-cell (9837); NOS (9835); Burkitt cell leukemia (9826). 143 144 Pesticide exposure modeling. We will use established ambient pesticide exposure models developed to increase precision in exposure measurements by incorporating both Pesticide Use Reporting (PUR) and Land-Use data, in combination with complete residential history(37). Models utilizing both PUR and Land-Use data reduce non-differential misclassification of exposure, thereby reducing attenuation of true ORs(32). PUR data: The California Department of Pesticide Regulation (CA DPR) began collecting information on restricted-use pesticide application in 1974, and began collecting complete reports of all agricultural pesticide application in 1990. PUR data includes monthly reports of all ingredients (active and inactive), the amount of pesticide applied (pounds), the location, acreage, and type of crop to which the pesticide is applied, and information on the date, frequency and method of application (e.g. air, ground). PUR reports are then linked to Public Land Survey System (PLSS) data, which divides land into 1-square mile sections. Land-Use data: The California Department of Water Resources (CDWR) conducts surveys of land use (e.g. location and type of crop) every 7-10 years. The proposed pesticide exposure model uses detailed information from historical CDWR surveys in combination with PUR data to compute precise monthly estimates of each pesticide application. Exposure models utilizing both PUR and Land-Use data have shown to provide exposure estimates with high specificity, thereby reducing the potential for non-differential misclassification bias, compared to models with PUR data alone(32). Geocoding: We will use complete residential history for all participants from 1 year prior to birth until date of diagnosis/reference date to assign an overall exposure to 144 145 each pesticide of interest (Table 1). Geocoded residential address information is available for participants in the CCLS. Individual pesticide exposure: Total pesticide exposure for specific time periods will be calculated using monthly estimates of pesticide exposure and complete residential histories during the relevant time period. We will evaluate the risk associated with exposure to specific pesticides (Table 1) during the 3 months prior to conception, during pregnancy (by trimester), during the first year of life, and in early childhood (from birth until date of diagnosis/reference date). All addresses recorded from 1 year prior to birth until date of diagnosis/reference date have been collected and geocoded to obtain spatial coordinates (latitude and longitude). A 500-meter buffer will be drawn around each geocoded address and monthly ambient exposure will be calculated by summing the total poundage of pesticides applied by the proportion of land area treated in that parcel. We will use these monthly estimates for each pesticide to assign overall exposure (from 1 year prior to birth until date of diagnosis/reference date), and exposure during each relevant time period. Using obtained estimates, we will explore associations using continuous exposure variables, overall and in each relevant time period for each pesticide (Table 1). For those situations in which a given exposure variable is highly skewed, we will categorize exposure variables in our models to minimize the influence of extreme values. Pesticides of interest: We have selected the following pesticides for inclusion in this analysis: 145 146 Table 5.1. Carcinogenic Potential Of Selected Pesticides And Childhood Leukemia Risk Pesticide EPA Carcinogenic Potential Pounds/# of appl.* Effect estimates for childhood leukemia Benzimidazoles Benomyl (F) ^ #† Possible 118,601/7,945 3.3 (1.2-8.8) a,h,k(38) ; 2.8 (1.1-6.9) a,d,k(38) Chlorinated phenols 2,4-D (H) ^ #† Not classifiable 2,065/112 1.61 (1.20-2.16) c,l(12) ; 3.62 (1.28-10.3) b,g(11) Organophosphates 3.7 (1.0-13.1) b,e(38) ; 3.3 (1.1-9.6) b,f(38) ; 2.47 (1.43-4.28) i,g(10) ; 2.13 (1.30-3.47) h,i(10) Azinphos-methyl (I) ‡ Not likely 185,055/2,808 2.47 (1.43-4.28) i,g(10) ; 2.13 (1.30, 3.47) h,i(10) Chlorpyrifos (I) #† ‡ Evidence of non-carcinogenicity 2,041,814/42,835 2.05 (1.80-2.32) c,g,m(12) ; 1.61 (1.33, 1.95) c,h,m(12) ; 2.72 (1.47-5.04) b,g(11) ; 1.79 (1.34-2.40) g,j(10) Diazinon (I) ^ #‡ Not likely 1,053,407/30,196 2.05 (1.80-2.32) c,g,m(12) ; 1.61 (1.33, 1.95) c,h,m(12) ; 2.72 (1.47-5.04) b,g(11) ; 1.79 (1.34-2.40) g,j(10) Pyrethroids 2.47 (1.43-4.28) g,i(10) ; 2.13 (1.30, 3.47) h,i(10) ; 1.8 (1.1-2.9) b,c,h(39) Permethrin (I) † Possible 385,581/48,305 2.47 (1.17-5.25) b,c,h(39) Thiocarbamates Mancozeb (F) ^† Probable 611,197/13,539 2.3 (1.0-5.2) a,h(38) ; 3.1 (1.1-9.0) a,e,k (38) Other pesticides Glyphosate (H) Evidence of non-carcinogenicity 4,641,560/117,647 1.61 (1.20-2.16) c,l(12) ; 3.62 (1.28-10.3) b,g(11) Paraquat dichloride (H) # Evidence of non-carcinogenicity 976,158/33,828 3.5 (1.0-12.7) b,d(38) *PUR data, 2000 (F) = fungicide; (H) = herbicide; (I) = insecticide ^developmental/reproductive toxin #suspected genotoxin †s us pe c t e d e n do c r i n e d i s r upto r ‡c h o l i ne s t e r a s e i nhi bi t o r a: paternal occupational exposure b: maternal occupational exposure c: residential exposure d: exposure during the year prior to conception e: exposure during the 1st trimester f: exposure during the 3rd trimester g: exposure during pregnancy h: exposure during first year of life/early childhood i: residential use of moth killers j: residential use of cockroach, ant, fly, bee, or wasp killers k: association in males only l: exposure to herbicides m: exposure to insecticides Genotyping. Genome-wide genotyping for Hispanic cases for phases I-III of the CCLS was conducted on archived dried blood spot samples using the Illumina Human OmniExpress v.1 platform (Illumina Inc, San Diego, CA, http://www.illumina.com) at the University of California Berkeley School of Public Health Genetic Epidemiology and Genomics Laboratory, which included a total of 730,525 SNP markers. DNA was extracted from blood spots using the QIAamp DNA Mini Kit (QIAGEN, USA, Valencia, CA); blood spots were available for 87% of interviewed participants. SNPs were excluded from final analyses for the following reasons: a) non-autosomal (21,167 SNPs), b) Hardy Weinberg equilibrium p-value<10 -5 , c) SNP call rate <98%, d) minor allele frequency (MAF) <2%. A total of 634,037 SNPs remained for inclusion in analyses. Individual samples will be excluded from final analyses for any discrepancies between sex chromosome and reported sex, or call rates <98%. For approximately 1% of cases and controls, duplicate specimens were analyzed to assess concordance. Concordance 146 147 was greater than 99.9% for all duplicates. For each pair of duplicate specimens, the sample with the higher call rate was used in analysis. Genotyping has been completed and is currently available for all Hispanic subjects in the CCLS. Genome-wide genotyping data is planned for completion through the CCLS primary grant. Candidate genotyping data is available for all additional CCLS cases and controls. Statistical Analysis. Demographic information (including complete residential history from birth until date of diagnosis/reference date) and genotyping data will be obtained from existing CCLS study databases at the UC Berkeley. Data will be transferred via secure, password protected, web-based data portal, in accordance with current IRB security measures and mandates for the protection of confidential data. Aim 1: Evaluate the risk of childhood ALL associated with exposure to ambient pesticides. Pesticide exposure (overall and by specific time window) will be modeled as a continuous variable, and in situations where the exposure is highly skewed, we will categorize exposure to minimize the influence of extreme observations, using unexposed children as the referent group for each pesticide (Table 1) at each relevant time period: 3 months prior to pregnancy, during pregnancy, during the first year of life, or in early childhood (from birth until date of diagnosis/reference date). We will also compute an overall exposure (for all time periods combined from 1 year prior to birth until date of diagnosis/reference date), and will explore weighted models (e.g., with greater weight assigned to in utero exposure) for overall exposure in additional analyses. Unconditional logistic regression models will be used to evaluate the association between each pesticide 147 148 or pesticide class and risk of childhood ALL overall, and for each specific time period, adjusting for matching factors (date of birth, gender, and maternal race), and factors hypothesized a priori to confound the association (household income, socioeconomic status (SES), and rural/urban residence location). Correlations of ambient pesticides with residential or occupational sources of exposure will be evaluated. We will additionally evaluate whether results differ by race/ethnicity using formal tests for statistical interaction. Aim 2.1 Use ambient pesticide exposure estimates and GWAS data to evaluate GxE interactions. As background, traditional approaches for analyzing GxE interaction include 1) Case- c ontrol a na l y sis with a n int e ra c ti on ter m to test the null h y pothesis β GxE = 0, using the model: L o g it (Pr( D = 1 | G) = β 0 + β G G + β E E + β GxE GxE [model (1)]; 2) Joint G, GxE test (2 de gre e of f re e dom ), using model (1) to test th e null h y pothesis β GxE =β G =0. This 2df test may provide greater power to detect a statistically significant interaction that the 1df test of the interaction term alone(40); 3) Case-only analysis to test the null hypothesis γ GxE = 0, using the model: Logit(Pr(E=1|G,D=1)= γ 0 + γ GxE G (41). Recently, several more efficient two-step procedures for GxE scans have been developed. For example, the EDGxE method(33) screens all available SNPs using a 2- degree-of-freedom joint test of disease vs. gene and environment vs. gene association. The p-values from Step 1 are used to prioritize SNPs for formal GxE testing (based on model (1) ) in S tep 2. Th e ED Gx E a nd a r e late d ‘ c oc ktail’ method provide g r e a te r powe r than traditional approaches to detect GxE interaction in a GWAS. GxE interaction analyses (both traditional and two-step approaches) will be implemented using GxEscan 148 149 (http://www.biostats.usc.edu/software), a program executed in R, which was recently developed by co-investigator Dr. James Gauderman at USC(33). Analyses will be completed using continuous exposure variables, with additional analyses utilizing categorized variables for exposures with a skewed distribution. All models will include adjustment for gender, age at diagnosis or reference age, and principal components to account for ancestry. Although there will be overlap in the lists of top SNPs identified from these various GxE scanning approaches, we also anticipate some unique ones given that each approach utilizes information in a different way and has optimal power for different underlying models of interaction(33). We will merge the top lists from various scans to form a single list that will be subjected to replication analysis as described below. Aim 2.2 Replicate results in non-Hispanic and Hispanic children. Analyses will be conducted to replicate results from our preliminary GWIS analysis in a separate sample of non-Hispanic children enrolled in phases I-III of the CCLS (n~550 cases/650 controls), and, independently, in an analysis of Hispanic children enrolled in phases IV-V of the CCLS (250 cases). To obtain a final list for replication, 1) GWIS analyses will be completed using GxEscan for each subgroup (ALL cases, B-cell ALL only); 2) A list of top SNPs will be generated based on all approaches. SNPs will be sorted by p-value to identify a list of most promising candidates for replication; 3) SNAP (www.broadinstitute.org/mpg/snap/) will be used to find all SNPs in linkage disequilibrium with our most promising candidates for replication in both non-Hispanic and Hispanic populations, using the 1000 genomes project and a minimum R 2 of 0.8. An 149 150 estimated 50-300 SNPs will then be sent for replication, which will be completed using model (1) above (see Aim 2.1). Power calculations. Aim 1: With a sample size of approximately 2200, we are adequately powered to detect an odds ratio of at least 1.35 for our main effect analysis of ambient pesticide exposure for all ALL cases ( β=0.83) or when restricting to B-cell ALL cases only ( β=0.8), with a pesticide exposure prevalence of 20% or greater. Aim 2: Our novel two-step methods are more statistically efficient than a traditional case-control interaction analysis with a product term (model (1) above)(33). Two-step methods provide power that is more comparable, and somewhat higher for most models, to that of a case-only analysis. Using the program Quanto (http://biostats.usc.edu/software), we computed power to detect GxE interaction for a case-only analysis. Table 2 shows the power of a to detect an interaction OR of 2.6 for a range of minor allele and pesticide exposure frequencies, using a genome-wide significance of 5.0x10 -8 , and assuming a main environment OR of 1.5 and no main gene effect. With 380 Hispanic cases, we anticipate sufficient power to detect an interaction OR of 2.6 using a case-only approach with a minor allele frequency of at least 25%, and a pesticide exposure prevalence of at least 20%. Two-step methods will use the complete sample (n=834) and, again, should provide similar, or greater power than case-only analyses (as shown in table 2) to detect GxE interactions. Minor Allele Frequency Exposure Frequency 15% 20% 25% 15% 0.36 0.47 0.51 20% 0.57 0.67 0.70 25% 0.71 0.78 0.80 30% 0.79 0.84 0.84 35% 0.83 0.86 0.86 Table 5.2. Power for varying exposure and minor allele frequencies, case-only analysis 150 151 Potential problems and their solutions. Sample size. The CCLS is the largest study of childhood leukemia in the United States, but the sample size required for traditional methods of GxE interaction analysis using genome-wide data are prohibitively large. We have therefore elected to use novel methods that are statistically more efficient than traditional methods. Our power calculations suggest that we are adequately powered to detect modest effects in both our primary analysis of the risk associated with ambient pesticide exposure, and in our GWIS analysis of GxE interactions using novel 2-step methods. Heterogeneous ethnicity of participants. We are including the 2 major racial/ethnic groups impacted by childhood ALL (i.e., Hispanic and non-Hispanic white children) in our primary analysis of the association between ambient pesticide exposure and childhood ALL risk. Inclusion of children of differing ethnicities will allow us to better understand the ethnic-specific risks associated with ambient pesticide exposure. Such inclusion also permits us to explore risk in Hispanic children, a group experiencing high and rising rates of childhood ALL and with high exposure to pesticides. Timeline. The proposed analysis can be completed in the following two-year time frame: Tasks Jul 2014-Dec 2014 Jan 2015-Jun 2015 Jul 2015-Dec 2015 Jan 2016-Jun 2016 Receive data from CCLS; prepare data for analysis Calculate pesticide exposure using exposure models Evaluate association between pesticides and ALL Evaluate GxE interactions using GWAS data Prepare findings for presentation/publication 151 152 5.5 REFERENCES: 1. N H, AM N, M K, et al. SEER Cancer Statistics Review, 1975-2009 (Vintage 2009 populations). In: N H, AM N, M K, et al., eds. Bethesda, MD: National Cancer Institute, 2012. 2. Hudson MM, Ness KK, Gurney JG, et al. Clinical ascertainment of health outcomes among adults treated for childhood cancer. JAMA 2013;309(22):2371- 81. 3. Wiemels J. Perspectives on the causes of childhood leukemia. Chemico- Biological Interactions 2012;196(3):59-67. 4. Healy J, Richer C, Bourgey M, et al. Replication analysis confirms the association of ARID5B with childhood B-cell acute lymphoblastic leukemia. Haematologica 2010;95(9):1608-11. 5. Papaemmanuil E, Hosking FJ, Vijayakrishnan J, et al. Loci on 7p12.2, 10q21.2 and 14q11.2 are associated with risk of childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1006-10. 6. Prasad RB, Hosking FJ, Vijayakrishnan J, et al. Verification of the susceptibility loci on 7p12.2, 10q21.2, and 14q11.2 in precursor B-cell acute lymphoblastic leukemia of childhood. Blood 2010;115(9):1765-7. 7. Sherborne AL, Hosking FJ, Prasad RB, et al. Variation in CDKN2A at 9p21.3 influences childhood acute lymphoblastic leukemia risk. Nature Genetics 2010;42(6):492-4. 8. Treviño LR, Yang W, French D, et al. Germline genomic variants associated with childhood acute lymphoblastic leukemia. Nature Genetics 2009;41(9):1001-5. 9. Urayama KY, Wiencke JK, Buffler PA, et al. MDR1 gene variants, indoor insecticide exposure, and the risk of childhood acute lymphoblastic leukemia. Cancer Epidemiol Biomarkers Prev 2007;16(6):1172-7. 10. Infante-Rivard C, Labuda D, Krajinovic M, et al. Risk of childhood leukemia associated with exposure to pesticides and with gene polymorphisms. Epidemiology 1999;10(5):481-7. 11. Wigle DT, Turner MC, Krewski D. A Systematic Review and Meta-analysis of Childhood Leukemia and Parental Occupational Pesticide Exposure. Environmental Health Perspectives 2009;117(10):1505-13. 12. Turner MC, Wigle DT, Krewski D. Residential Pesticides and Childhood Leukemia: A Systematic Review and Meta-Analysis. Environmental Health Perspectives 2009. 13. Cancer Facts & Figures 2013. Atlanta, GA: American Cancer Society, 2013, 14. Kheifets L, Ahlbom A, Crespi CM, et al. A pooled analysis of extremely low- frequency magnetic fields and childhood brain tumors. American Journal of Epidemiology 2010;172(7):752-61. 15. Ahlbom A, Day N, Feychting M, et al. A pooled analysis of magnetic fields and childhood leukaemia. Br J Cancer 2000;83(5):692-8. 16. Greenland S, Sheppard AR, Kaune WT, et al. A pooled analysis of magnetic fields, wire codes, and childhood leukemia. Childhood Leukemia-EMF Study Group. Epidemiology 2000;11(6):624-34. 152 153 17. Milne E, Greenop KR, Scott RJ, et al. Parental Prenatal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia. American journal of epidemiology 2011;175(1):43-53. 18. Liu R, Zhang L, McHale CM, et al. Paternal Smoking and Risk of Childhood Acute Lymphoblastic Leukemia: Systematic Review and Meta-Analysis. Journal of Oncology 2011;2011:1-16. 19. Caughey RW, Michels KB. Birth weight and childhood leukemia: A meta- analysis and review of the current evidence. International Journal of Cancer 2008;124(11):2658-70. 20. Hjalgrim LL. Birth Weight as a Risk Factor for Childhood Leukemia: A Meta- Analysis of 18 Epidemiologic Studies. American Journal of Epidemiology 2003;158(8):724-35. 21. Linabery AM, Jurek AM, Duval S, et al. The Association Between Atopy and Childhood/Adolescent Leukemia: A Meta-Analysis. American Journal of Epidemiology 2010;171(7):749-64. 22. Dahl S, Schmidt LS, Vestergaard T, et al. Allergy and the risk of childhood leukemia: a meta-analysis. Leukemia 2009;23(12):2300-4. 23. Urayama KY, Buffler PA, Gallagher ER, et al. A meta-analysis of the association between day-care attendance and childhood acute lymphoblastic leukaemia. International Journal of Epidemiology 2010;39(3):718-32. 24. Greaves MF. Aetiology of acute leukaemia. Lancet 1997;349(9048):344-9. 25. Wiemels JL, Cazzaniga G, Daniotti M, et al. Prenatal origin of acute lymphoblastic leukaemia in children. Lancet 1999;354(9189):1499-503. 26. Wiemels JL, Xiao Z, Buffler PA, et al. In utero origin of t(8;21) AML1-ETO translocations in childhood acute myeloid leukemia. Blood 2002;99(10):3801-5. 27. McHale CM, Wiemels JL, Zhang L, et al. Prenatal origin of TEL-AML1-positive acute lymphoblastic leukemia in children born in California. Genes Chromosomes Cancer 2003;37(1):36-43. 28. Buffler PA, Kwan ML, Reynolds P, et al. Environmental and genetic risk factors for childhood leukemia: appraising the evidence. Cancer Invest 2005;23(1):60-75. 29. Protecting Children's Health. In: Programs USEPAOoP, ed. Washington, DC, 2010. 30. Alavanja MC, Ross MK, Bonner MR. Increased cancer burden among pesticide applicators and others due to pesticide exposure. CA Cancer J Clin 2013;63(2):120-42. 31. Rull RP, Gunier R, Von Behren J, et al. Residential proximity to agricultural pesticide applications and childhood acute lymphoblastic leukemia. Environ Res 2009;109(7):891-9. 32. Rull RP, Ritz B. Historical pesticide exposure in California using pesticide use reports and land-use surveys: an assessment of misclassification error and bias. Environ Health Perspect 2003;111(13):1582-9. 33. Gauderman WJ, Zhang P, Morrison JL, et al. Finding novel genes by testing g × e interactions in a genome-wide association study. Genet Epidemiol 2013;37(6):603-13. 153 154 34. United States Department of Agriculture, N.A.S.S. 2002 Census of Agriculture: United States, Summary and State Data. Washington, DC, USA: United States Department of Agriculture, 2004. 35. Metayer C, Milne E, Clavel J, et al. The Childhood Leukemia International Consortium. Cancer Epidemiol 2013;37(3):336-47. 36. Barrington-Trimis JL, Searles Nielsen S, Preston-Martin S, et al. Parental Smoking and Risk of Childhood Brain Tumors by Functional Polymorphisms in Polycyclic Aromatic Hydrocarbon Metabolism Genes. PLoS One 2013 in press. 37. Cockburn M, Mills P, Zhang X, et al. Prostate cancer and ambient pesticide exposure in agriculturally intensive areas in California. Am J Epidemiol 2011;173(11):1280-8. 38. Monge P, Wesseling C, Guardado J, et al. Parental occupational exposure to pesticides and the risk of childhood leukemia in Costa Rica. Scand J Work Environ Health 2007;33(4):293-303. 39. Ferreira JD, Couto AC, Pombo-de-Oliveira MS, et al. In utero pesticide exposure and leukemia in Brazilian children < 2 years of age. Environ Health Perspect 2013;121(2):269-75. 40. Kraft P, Yen YC, Stram DO, et al. Exploiting gene-environment interaction to detect genetic associations. Hum Hered 2007;63(2):111-9. 41. Piegorsch WW, Weinberg CR, Taylor JA. Non-hierarchical logistic models and case-only designs for assessing susceptibility in population-based case-control studies. Stat Med 1994;13(2):153-62. 154

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Genomic risk factors associated with Ewing Sarcoma susceptibility

PDF

Age related macular degeneration in Latinos: risk factors and impact on quality of life

PDF

The environmental and genetic determinants of cleft lip and palate in the global setting

PDF

The role of heritability and genetic variation in cancer and cancer survival

PDF

Genetic epidemiological approaches in the study of risk factors for hematologic malignancies

PDF

Identifying genetic, environmental, and lifestyle determinants of ethnic variation in risk of pancreatic cancer

PDF

Genetic variation in the base excision repair pathway, environmental risk factors and colorectal adenoma risk

PDF

Genes and environment in prostate cancer risk and prognosis

PDF

Genetic risk factors in multiple myeloma

PDF

Hormonal and genetic risk factors of endometrial cancer and trends in incidence and survival of adult acute lymphoblastic leukemia

PDF

Common immune-related factors and risk of non-Hodgkin lymphomy

PDF

The effect of cytomegalovirus on gene expression of pediatric acute lymphoblastic leukemia

PDF

The interplay between tobacco exposure and polygenic risk score for growth on birthweight and childhood acute lymphoblastic leukemia

PDF

Factors that influence mammographic density: role of estrogen metabolism genes, biomarkers of inflammation, and lifestyle

PDF

Dietary carcinogens and genetic variation in their metabolism: epidemiological studies on the risk of selected cancers

PDF

Characterizing the genetic and environmental contributions to ocular and central nervous system health

PDF

Environmental risk factors of Multiple Sclerosis: a twin study

PDF

Body size and the risk of prostate cancer in the multiethnic cohort

PDF

Native American ancestry among Hispanic Whites is associated with higher risk of childhood obesity: a longitudinal analysis of Children’s Health Study data

PDF

Early childhood health experience & adult phenotype in twins

Asset Metadata

Creator Barrington-Trimis, Jessica L. (author)

Core Title Genetic and environmental risk factors for childhood cancer

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Publication Date 06/07/2015

Defense Date 05/01/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag childhood brain tumors,childhood cancer,childhood leukemia,GxE interactions,OAI-PMH Harvest,parental tobacco use,secular trends

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Gauderman, William James (committee chair), McKean-Cowdin, Roberta (committee chair), Asgharzadeh, Shahab (committee member), Stern, Mariana C. (committee member), Thomas, Duncan C. (committee member)

Creator Email jtrimis@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-416748

Unique identifier UC11295849

Identifier etd-Barrington-2537.pdf (filename),usctheses-c3-416748 (legacy record id)

Legacy Identifier etd-Barrington-2537.pdf

Dmrecord 416748

Document Type Dissertation

Format application/pdf (imt)

Rights Barrington-Trimis, Jessica L.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA