Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Childhood cancer in the offspring of male sawmill workers occupationally exposed to chlorophenate fungicides Heacock, Helen Jane 1998

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1998-345300.pdf [ 6.69MB ]
JSON: 831-1.0089357.json
JSON-LD: 831-1.0089357-ld.json
RDF/XML (Pretty): 831-1.0089357-rdf.xml
RDF/JSON: 831-1.0089357-rdf.json
Turtle: 831-1.0089357-turtle.txt
N-Triples: 831-1.0089357-rdf-ntriples.txt
Original Record: 831-1.0089357-source.json
Full Text

Full Text

C H I L D H O O D C A N C E R IN T H E OFFSPRING OF M A L E S A W M I L L W O R K E R S O C C U P A T I O N A L L Y EXPOSED TO C H L O R O P H E N A T E FUNGICIDES  By  HELEN JANE HEACOCK  B . S c , The University of Guelph, 1983 Graduate Diploma in Occupational Health, M c G i l l University, 1985 M . S c , The University of London, 1986 A THESIS SUBMITTED IN P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S FOR T H E D E G R E E OF  DOCTOR OF PHILOSOPHY In  T H E F A C U L T Y O F G R A D U A T E STUDIES Department of Health Care and Epidemiology  We accept this thesis as c ^ f o r m i n g to the required standard  T H E UNIVERSITY OF BRITISH C O L U M B I A October '1 1998 © Helen Jane Heacock, 1998  In presenting  this  thesis  in partial  fulfilment  of the requirements  for an advanced  degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. 1 further agree that permission for extensive copying of this thesis for scholarly purposes may be granted department publication  or  by his or  her representatives.  It  is  by the head of my  understood  copying  or  of this thesis for financial gain shall not be allowed without my written  permission.  Department of  ^^-QJL~fL  Qpd~SL CKAA^,  The University of British Columbia Vancouver, Canada  DE-6 (2/88)  that  t^pid g A A i ' o \ O  T A B L E OF CONTENTS Abstract  vii  List of Tables  ix  List of Figures  xii  Acknowledgements  xiii  CHAPTER ONE: INTRODUCTION 1.0  Introduction and Overview  1  1.1  Research Questions  2  1.2  Rationale  2  1.3  Designs Appropriate for Studying Reproductive Effects of Occupational Exposures  6  CHAPTER TWO: LITERATURE REVIEW 2.0  Introduction  9  2.1  Childhood Cancer  9  Descriptive Epidemiology  9  Incidence  10  Time Trends  12  International Incidence  13  Mortality and Survival  14  Etiology/Carcinogenesis  15 ii  T A B L E OF CONTENTS, cont.  2.2  2.3  Paternal Occupational Exposures  22  Paternal Non-Occupational Exposures  24  Maternal Exposures  25  Childhood Risk Factors  27  Chlorophenols  28  Description and Use  28  History of Chlorophenate Use in B C  29  Routes of Exposure  32  Occupational  32  Non-occupational  33  Distribution, Excretion and Metabolism  34  Toxicity  35  Reproductive Effects  36  Animals  36  Humans  37  Mutagenic Effects  39  Carcinogenic Effects  39  Animals  40  Humans  40  Reproductive Toxicology Male Reproduction  43 45  iii  T A B L E OF CONTENTS, cont.  2.4  Male Reproductive Toxicology  45  Human Studies  46  Animal Studies  49  Epidemiological Studies  50  Summary  54  CHAPTER THREE: METHODS 3.0  Introduction  57  3.1  Funding Process  58  3.2  Power Calculations  60  3.3  Data Sources The British Columbia Cancer Registry  60  Computerized Record Linkage: Background  61  Generating the Offspring Cohort  62  Generating the Cancer Offspring Cohort.... 64 3.4  Study Design  68  Previous Research  68  Validity and Reliability of Measures  68  Personal Identifying Information  71  Retrospective Exposure Assessment  72  T A B L E O F C O N T E N T S , cont.  Present Study  73  Cohort Analysis Person-Year Calculations  73 74  Person-Year Calculations Until Offspring Cohort all Achieve 20 Years of Age  75  Calculation of Provincial Incidence Rates..  76  Standardized Incidence Ratios  79  Nested Case-Control Analysis  CHAPTER FOUR:  80  Selection of Cases and Controls  81  Indices of Exposure  83  Determining a Baseline Model  87  RESULTS  4.0  Offspring Cohort Description  89  4.1  Characteristics of Cancer Cases  93  4.2  Associations with Adverse Reproductive Outcomes  99  4.3  Associations with Cancer in Parents  100  4.4  Cohort analysis  101  4.5  Nested case-control analysis  108  T A B L E O F C O N T E N T S , cont.  C H A P T E R F I V E : DISCUSSION A N D C O N C L U S I O N 5.0  General Comments  5.1  How the Results of this Study Relate to Previous Research 121  5.2  Limitations of the Study  123  Study Power..  123  Confounding  127  Bias  129  Chance  131  5.3  5.4  Opportunities for Future Research  118  131  Endocrine Disruptors and Male Reproduction...  131  Future Childhood Cancer Studies  133  Conclusion  134  Bibliography  136  Appendices  153  ABSTRACT  The purpose of this study was to determine whether paternal occupational exposure to chlorophenol fungicides, namely tetra- and pentachlorophenate, and their dioxin contaminants is associated with childhood cancer in the offspring of sawmill workers. The study was conducted using data from 23,829 British Columbian sawmill workers employed for at least one continuous year between 1950-1985 in eleven chlorophenate-using sawmills and to live births born at least one year after the initiation of employment in the period 1952 - 1988. Probabilistic linkage of the sawmill worker cohort to the provincial Division of Vital Statistics' marriage and birth files produced an offspring cohort of 19,674 children. The offspring cohort was then linked to the British Columbia Cancer Registry. A l l malignant cases less than twenty years of age and appearing in the cancer registry between 1969 - 1993 were included. Forty cases of cancer were identified during 259,919 person-years of followup; twenty-two cases among the girls and eighteen among the boys. Childhood cancer trends were assessed using both external and internal analyses. Standardized incidence ratios (SIRs) were calculated using the British Columbia population as a reference. Among girls, the SIR for all cancers was 1.22, (95% CI 0.76 - 1.84), for leukemia 1.00, (95% CI 0.32 - 2.33) and for brain cancer 2.02, (95% CI 0.74 - 4.40). Among boys, the SIR for all cancers was 0.89, (95% CI 0.53 - 1.41), for leukemia 0.95, (95% CI 0.35 - 2.08) and for brain cancer 0.77, (95% CI 0.16 - 2.25). A nested case-control analysis assessed the effects of paternal cumulative exposure and windows of exposure on risk of developing cancer in the offspring. Although the results produced slightly elevated dose-response relationships, most noticeably with brain neoplasms, none were statistically significant. Even though the study power was  limited, the analyses provide little evidence to support a relationship between risk of childhood cancer and paternal occupational exposure to chlorophenate fungicides in B C sawmills.  List of Tables Table 1: Sites and histologies included in incidence calculations  p  78  Table 2 : Characteristics of Offspring Cohort: Ages 0 - 2 0  p  92  Registry. A l l malignant neoplasms ("histology/3") identified among female offspring..p  94  Table 3: Results from linkage of offspring cohort to British Columbia Cancer  Table 4: Results from linkage of offspring cohort to British Columbia Cancer Registry. A l l malignant neoplasms ("histology/3") identified among male offspring  p 96  Table 5: Results from linkage of offspring cohort to British Columbia Cancer Registry. A l l in-situ ("histology/2") and uncertain behaviour ("histology/1") neoplasms identified among the offspring.  p  98  p  103  Table 6: Childhood cancer (ages 0-19) among female offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs Table 7: Childhood cancer (ages 0-19) among male offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs  p 104  Table 8: Observed and expected cases, SIRs and 95% CIs, for the most common sites, by age group (both sexes)  p  105  p  106  Table 9: Observed and expected cases, SIRs and 95% CIs, for the most common sites, by age group (girls)  Table 10: Observed and expected cases, SIRs and 95% CIs, for the most common sites, by age group (boys)  p  107  ix  List of Tables, cont. Table 11: Characteristics of cases and controls  p  108  Table 12: Paternal chlorophenate exposure characteristics of cases and controls  p  109  p  110  p  111  p  111  p  112  p  112  Table 13: Risk for developing childhood cancer (all sites) associated with cumulative hours of paternal chlorophenate exposure Table 14: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposure from one year post commencing employment until 90 days prior to conceiving (Window 1) Table 15: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposure during the 90 days prior to conception (Window 2) Table 16: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposures during pregnancy (Window 3) Table 17: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposure from birth until cancer diagnosis (Window 4) Table 18: Risk for developing childhood brain cancer associated with cumulative hours of paternal chlorophenate exposure  p  114  p  114  Table 19: Risk among girls for developing childhood brain cancer associated with cumulative hours of paternal chlorophenate exposure  x  List of Tables, cont. Table 20: Risk for developing childhood brain cancer associated with  windows of cumulative hours of paternal chlorophenate exposure: exposure from one year post commencing employment until 90 days prior to conceiving (Window 1)  p  115  p  115  Table 21: Risk for developing childhood brain cancer associated with  windows of cumulative hours of paternal chlorophenate exposure: exposure during the 90 days prior to conception (Window 2) Table 22: Risk for developing childhood brain cancer associated with  windows of cumulative hours of paternal chlorophenate exposure: exposures during pregnancy (Window 3)  p  116  p  116  Table 23: Risk for developing childhood brain cancer associated with  windows of cumulative hours of paternal chlorophenate exposure: exposure from birth until cancer diagnosis (Window 4) Table 24: Risk for developing childhood leukemia associated with  cumulative hours of paternal chlorophenate exposure  p  117  xi  List of Figures Figure 1 : Changes in fungicide formulations and application methods from 1940 to 1990 Figure 2: Time line describing cumulative exposure to chlorophenates and windows of exposure Figure 3: Diagrammatic representation of the B C sawmill worker study, and subsequent studies examining reproductive outcomes of the sawmill workers...  Acknowledgements I wish primarily to thank my husband, Jason, who never lost faith in my ability to successfully complete this thesis. For his love, encouragement, wonderful sense of humour, patience and support, I am eternally grateful. My children Alexander and Catherine, who have only ever known me as a student, always kept me in touch with the joys of life outside the academic world. To these three individuals, I dedicate this thesis. I would also like to acknowledge friends and family, who knew when and when not to ask about my PhD work; who celebrated my accomplishments, encouraged my dreams and were there to support me when the road ahead seemed bleak. Specifically, I would like to thank my parents, Ron and Sheila, my parents in-law, Goldie and Albert, and my good friends, Lori, Anne, Tina, Cathy and Ruth. The sawmill worker cohort group, without which this thesis would have never occurred provided not only rich data sets instrumental to this thesis, but a team of researchers with incredible depth and breadth of knowledge in epidemiology, biostatistics, computing, as well as occupational, environmental and community health. Thanks are also due to my PhD thesis committee, Clyde, Paul, Richard and Robert, all of whom have been or continue to be involved with the sawmill worker study. Their enthusiasm for the project, guidance and thoughtful comments were greatly appreciated. The administrative staff in the Department of Health Care and Epidemiology, namely, Sherley, Virginia, Laurel, Stephanie and Zeba have been instrumental in ensuring that all necessary requirements were met. Their cheerfulness, knowledge and assistance is appreciated. Finally, I would like to thank two other individuals, Chris Bajdik and Joanne Moody, for freely contributing of their time to help with some of the statistical and linkage aspects of this thesis.  CHILDHOOD C A N C E R IN THE OFFSPRING OF M A L E S A W M I L L W O R K E R S O C C U P A T I O N A L L Y EXPOSED TO C H L O R O P H E N A T E FUNGICIDES  CHAPTER ONE: INTRODUCTION 1.0 Introduction and Overview Fabia and Thuy first examined the relationship between paternal occupational exposures and childhood cancer in 1974 (Fabia and Thuy 1974). Since then, a number of epidemiological investigations have explored this potential association, although none have established causal relationships.  Due to the rare occurrence of childhood cancer, most of the published studies have used the case-control design, and hence are subject to bias; especially selection and recall bias as well as exposure misclassification. A retrospective cohort study with accurate historical exposure data is a research design that is less prone to these biases and therefore, more likely to correctly identify cause and effect relationships.  Data gathered for a large retrospective cohort investigation of cancer incidence and mortality among 26,500 British Columbian sawmill workers occupationally exposed to chlorophenate fungicides between 1950 - 1985 (Hertzman et al. 1997), and a nested casecontrol study of birth defects among 19,700 offspring of these sawmill workers (DimichWard et al. 1996) presented a unique opportunity conduct a cohort study of paternal occupational exposures to chlorophenate fungicides and childhood cancer. Access to data sets gathered for the two above-mentioned studies, along with computerized records 1  from the British Columbia Cancer Registry, made it possible to examine childhood cancer in a cost-effective and time-efficient manner.  1.1 Research Questions 1. Did occupational exposure to chlorophenate fungicides increase the risk for developing cancer in the offspring of a cohort of 23,829 male sawmill workers employed in eleven British Columbian sawmills for at least one continuous year between January 1, 1950 and December 31, 1985?  2.  Is there a dose-response relationship between father's exposure and risk of childhood cancer in offspring?  3.  Did timing of exposure play a role in the development of childhood cancer? Four windows of exposure were investigated: (i) any exposure from one year post commencing employment to three months before conception, (ii) any exposure in the three months before conception, (iii) any exposure during pregnancy, and (iv) any exposure from birth until diagnosis of cancer.  1.2 Rationale The rationale for conducting this study relates not only to its relevance to current scientific interest in male reproductive toxicology (e.g. are environmental endocrine disruptors, such as pesticides and dioxins, associated with adverse reproductive outcomes?), but also to the unique opportunity that availed itself. Complete data sets  containing both exposure and birth data gathered for a large retrospective cohort study of male sawmill workers exposed to chlorophenates and access to the British Columbia Cancer Agency's cancer registry permitted an efficient method to examine childhood cancer. Teschke et al. (1998) have recently described the multiple opportunities that exist for epidemiological research emanating from data gathered for a traditional occupational cohort study. The authors point out that efforts to create a database for a large retrospective cohort study are immense and that using such a database to investigate other outcomes and/or exposures will result not only in considerable savings in time and money, but will also provide a more comprehensive understanding of the determinants of health for the population under study. As an example, the paper describes the many "offshoot" studies that have resulted from the original retrospective cohort study of sawmill workers exposed to chlorophenates. In a time when government and other funding agencies are cutting back the money available for scientific research, it makes economic sense to fully utilize data gathered for one study to answer other related research questions.  Specifically, the following reasons provide a basis for embarking on this study; the first two relate to relevance and the second two to opportunity:  1. Chlorophenols are a class of pesticides that command considerable attention. In the original review of chlorophenols, the International Agency for Research in Cancer (IARC) classified chlorophenoxy herbicides and chlorophenols as "possibly carcinogenic to humans" (IARC 1986). This classification remained the same in the  most recent review of chlorophenols (IARC 1998). IARC has established an "International Register of Workers Exposed to Chlorophenoxy Herbicides and Contaminants", which is a multi-country study attempting to quantify past exposures to chlorophenoxy herbicides, chlorophenols and their contaminants, and link these findings with health outcomes in workers exposed to such agents (IARC working Group 1990, Saracci et al. 1991, Bas Bueno de Mesquita et al. 1993, Kogevinas et al. 1995). The IARC monographs evaluating the carcinogenic risk of chlorophenols suggest gaps in knowledge exist and the monograph pertaining specifically to pentachlorophenol reported no data on humans was available to the working group (IARC 1986, IARC 1991). Two studies utilizing data originally collected for the sawmill worker cohort study (Hertzman et al. 1997) have explored reproductive end points associated with paternal occupational exposure to chlorophenates. DimichWard et al. (1996) found no effect of exposure on gestational age, prematurity, stillbirths and neonatal death among the offspring. However, the same study found increased risks for the following congenital anomalies: eye, genitourinary tract, spina bifida and anencephaly. Heacock et al. (1998) found no increased risk for infertility with exposure to chlorophenates among the sawmill worker cohort. Hence, this study will provide information to complete the spectrum of reproductive effects associated with paternal occupational chlorophenate exposure.  2. Health Canada's Bureau of Chronic Disease Epidemiology initiated the Canadian Childhood Cancer Control Program in 1992. The program will be conducted in partnership with health care providers, researchers, provincial, territorial and federal 4  governments, voluntary agencies and other parties concerned with childhood cancer. It is a nationwide data collection system, the objective of which is to gather information on childhood cancer cases in order to reduce incidence and improve prognosis (Gibbons et al. 1994). The Program consists of four principle areas: (i) case-control studies of childhood cancer examining risk factors associated with childhood cancer, (ii) an assessment of the late effects of childhood cancer among survivors, (iii) a pediatric tissue bank, and (iv) an evaluation of childhood cancer treatment and survival trends across Canada. The thesis research will add to the database being generated by this initiative.  3. Interest in male reproductive health, especially as it relates to occupational exposures, has increased in the past few years. In November 1993, the Report of the Royal Commission on New Reproductive Technologies (Proceed With Care 1993) reported, "the best way to reduce reproductive harm in the workplace is by developing a knowledge base that provides a solid foundation for definitive preventative action. The fundamental problem is the lack of information on reproductive health in general and on the specific occupational factors that might cause reproductive harm. This information can be generated through research that seeks to produce new data, through analysis and linkage of existing data bases, and through expansion of the scope of existing occupational health and safety research to include reproductive health." This research linked existing data sets to explore reproductive effects of occupational exposure to chlorophenols, thereby producing new information in a cost-effective manner. This study will also provide a database containing the cases of  5  childhood cancer among the offspring of the sawmill worker cohort. This database will enable future researchers to further follow the offspring cohort into adult life.  4. Few cohort studies of paternal occupational exposures and childhood cancer exist. To my knowledge, none have examined paternal exposure to chlorophenols. The British Columbia cohort of sawmill workers is the largest and best-defined chlorophenate exposed population in Canada. The sawmill worker offspring cohort has been established. The British Columbia Cancer Registry has complete incidence files for all childhood cancers from January 1, 1969 onward. These data sets provided an unprecedented opportunity, in a cost and time efficient manner, to investigate the reproductive carcinogenicity of chlorophenates.  1.3 Designs Appropriate for Studying Reproductive Effects of Occupational Exposures Few studies have been published where offspring have been identified from an occupational cohort and analyses on reproductive outcome attempted. Two studies have been published on the reproductive performance of the B C sawmill worker cohort: (1) a study of birth defects among the offspring of sawmill workers (Dimich-Ward et al. 1996) and (2) a study of fertility of sawmill workers (Heacock et al. 1998). Another three have been published examining cancer incidence among offspring of (1) farmers (Kristensen et al. 1996), (2) male metal workers (Bonde et al. 1992) and (3) male printing workers (Kristensen and Andersen 1992). Defining appropriate methods by which to study reproductive effects of paternal occupational exposures is therefore a bit like travelling in  uncharted territory. Different methods may be more suited to different reproductive endpoints. For instance, when the endpoint occurs at birth, such as birth weight or birth defects, the fathers' occupational and genetic contributions may have a greater impact on the outcome than a study of postnatal disease, such as learning difficulties or childhood cancer, which would include environmental contribution from the child as well as prenatal occupational and genetic contributions from the father. The study examining fertility of sawmill workers took a cohort approach; initially an external cohort analysis was used to calculate the Standardized Fertility Ratio (SFR). This analysis was accompanied by an internal cohort analysis using the Mantel- Haenszel rate ratio  :  estimator ( R R M H ) . The study of birth defects utilized a nested case-control approach to examine windows of exposure. One of the authors subsequently re-ran the analysis using a cohort approach with an internal reference population and obtained similar results (personal communication, Dr. S. Marion). A l l three of the (Scandinavian) studies investigating childhood cancer were able to use direct record linkage because of each country's system of providing the population with unique identifying numbers. To define the offspring cohorts, occupational and/or union registers were linked to Central Population Registers, and then these cohorts were linked to national population-based cancer registries to ascertain cancer outcome. A cohort design enabled the researchers to examine the relationship between paternal occupational exposure and childhood malignancy. A l l three studies used indirect standardization with an external reference population; the one examining offspring of farmers also used Poisson regression modeling with internal reference groups.  Annegers and Johnson, (1992) suggested that studying the relationship between very rare diseases, such as childhood cancer, and rare exposures, such as paternal occupational exposure to chlorophenates, is problematic with respect to finding an appropriate design. The problem with cohort designs is the small number of cases that develop over long periods of time. For instance, the study of childhood malignancy among the offspring of male printing workers produced eight cases among 12,440 children (112,668 personyears of follow-up) between 1965 - 78. Similarly, the study of malignancy among the offspring of male metalworkers produced 26 cases between 1968 - 1986 (16,489 children, 183,866 person-years of follow-up): Low case numbers make it difficult to assess individual sites and histologies. This may be important i f a putative exposure adversely affects only certain sites or histologies. The limited case numbers in cohort studies also makes stratification by exposure level difficult thereby potentially negating the assessment of dose-response relationships. Case-control studies usually include more cases, but often have less reliable exposure information (based on-birth and death certificate information and/or people's memory), any may suffer from recall bias, interviewer bias and multiple exploratory analyses. For studies nested in occupational cohorts, however, the exposure data should be of better quality, socio-economic differences will be minimized and recall bias would not be a problem. The data available for this thesis permitted cohort (using both external and internal reference populations) and nested case-control analyses.  CHAPTER TWO: LITERATURE  REVIEW  2.0 Introduction During the last few years there has been considerable scientific as well as public interest in the reproductive effects of "environmental endocrine disruptors", including pesticides and dioxins. At the same time, international collaborations have been established to investigate the etiology and epidemiology of childhood cancer. This chapter will provide a literature review on (i) the epidemiology of childhood cancer, including descriptive epidemiology, what is known about the etiology of the disease, and parental occupational risk factors, (ii) chlorophenols, focussing on the known health effects, and (iii) an overview of male mediated reproductive toxicology.  2.1 Childhood Cancer 2.1 (i) Descriptive Epidemiology Although childhood cancer is a rare event, it is the leading medical cause of death in children under 15 years of age. It is surpassed only by accidents in terms of overall deaths (Huchcroft et al. 1996). Cancer survival rates have increased dramatically as a result of progresses made in treatment. Forty years ago few children survived cancer, now approximately seventy-five percent are cured. Nonetheless, this year in Canada, over 1300 children will develop cancer and approximately 240 will die from it (in contrast, an estimated 129,200 new cases of adult cancer and 61,800 deaths from adult cancer occurred in Canada in 1996) (Huchcroft et al. 1996, Gibbons et al. 1994, Stats Canada 1996). More than 13,000 potential years of life were lost to childhood cancer in Canada in 1993, representing 1.5% of the potential years of life lost due to all cancers  (Stats Canada 1996). The emotional, physical and economic burden of childhood cancer is immense. Therefore, understanding the underlying mechanism(s) involved may enable preventive strategies to be developed and implemented.  Incidence Childhood cancer is uncommon. The average annual incidence for all childhood cancers (age 0 -19) in Canada between 1985 - 1992 was approximately 16/100,000, whereas for adults, the annual incidence rate over the same period was 350/100,000 (Huchcroft et al. 1996). The distribution of childhood cancer rates varies by age group. The highest incidence occurs in the first five years of life; in Canada infants under one have an annual rate of 23/100,000 and children aged 1 - 4, a rate of 21.5/100,00. The lowest rates occur in children aged 5 - 1 4 . Both 5 - 9 year olds and 1 0 - 1 4 year olds have an annual incidence of 11/100,000. Lastly, the incidence climbs to 18/100,000 in the 15 - 19 age category (Huchcroft et al. 1996).  Among Canadian children, leukemia is the most common cancer with an age standardized incidence rate (ASIR) of 4.7/100,000. Leukemia accounts for approximately 25% of all new cases and 25% of mortality. Cancers of the brain and central nervous system are the second most common form of childhood cancer (ASIR = 2.8/100,000) accounting for 21% of all new cases and 29% of deaths. The third most common childhood neoplasms are the lymphomas (ASIR = 2.5/100,000) and contribute approximately 18% of all new cases and 7% of deaths. Hence these three types of cancer represent two thirds of all childhood cancers. Neuroblastoma is the fourth most 10  frequently reported childhood cancer, with an ASIR of 1.1/100,000 and contributes 7% of new cases and 12% of deaths. Soft tissue sarcomas, renal and bone tumours represent the fifth, sixth and seventh most common childhood cancers, respectively. (Stats Canada 1996, Huchcroft etlal. 1996)  It has been observed that for each type of cancer there are subtypes which differ in cell-of-origin, sex ratio and age-incidence curves. Within the childhood leukemias, acute lymphoblastic leukemia (ALL) is the most common histological type accounting for 80% of cases. It has a peak incidence between the ages two and four years and exhibits a male:female ratio of approximately 1.20:1. Among the childhood central nervous system cancers, astrocytoma is the most common histological type accounting for approximately 50% of all brain cancers. Astrocytoma has a peak incidence between five and nine years of age and rates vary little by sex. Non-Hodgkin's lymphoma (NHL) accounts for 31% and Hodgkins Disease (HD) for 37% of all childhood lymphomas. The lymphomas occur two and a half times more frequently in boys. Hodgkins disease is rare before the age of five, then steadily increases in frequency until the mid-twenties. Non-Hodgkins lymphoma shows a peak incidence among boys in the early teenage years and exhibits a bimodal distribution for girls with peaks at age two and eleven.  Overall, cancer is more common among boys with a sex ratio of approximately 1.20:1. This male preponderance is primarily due to excess cases of acute lymphoblastic leukemia, non-Hodgkins lymphoma, soft tissue sarcoma and Hodgkin's Disease. Most of the childhood cancers occur in under five year olds (e.g. neuroblastoma, leukemia, eye  11  and renal cancers and soft tissue sarcomas); however, lymphomas and bone tumours occur more frequently in teenage children (Huchcroft et al. 1996, Chow et al. 1996, Stats Canada 1996, McBride and Gallagher 1987, Cartwright and Staines 1992, McBride 1991, Gurney et al. 1995, Stiller et al. 1995, Hoar Zahm and Devesa, 1995).  Time Trends The annual incidence of childhood cancer has increased over the past few decades. In the United States, the incidence among children less than fifteen was 12.7 per 100,000 in 1973 and 14.1 per 100,000 in 1988 representing a 4% increase (Bleyer 1993, Gold and Sever 1994). Most of this increase was associated with acute lymphoblastic leukemia, brain and central nervous system tumours, and not due to changes in the lymphomas, Wilms tumour, soft tissue and bone sarcomas. Similar trends have been seen in England, Canada, Scandinavia and Germany (Blair and Birch 1994, Blair and Birch 1994a and Gurney 1996) and Australia (McWhirter et al 1996). The reasons for the rise in incidence may include: more complete case ascertainment by registries, second primary cancers occurring in survivors, earlier detection and better diagnosis (such as improvements in brain imaging) and, for acute lymphoblastic leukemia, a reduction in the occurrence of untyped leukemias (Chow et al. 1996, Gurney et al. 1994). However, data such as that from the American National Cancer Institute's Surveillance, Epidemiology and End results program (SEER) suggests increases in ascertainment are not likely to explain these trends and that environmental factors may be involved at an early stage in a child's life.  12  International Incidence In 1988, I A R C published a report describing the incidence of childhood cancers internationally (Parkin et al.1988). Prior to that, little information was available regarding geographic distributions of childhood cancer. Substantial variations in incidence occur between countries and within subtypes of disease. Age-adjusted rates vary approximately four-fold among the 28 registries reported in the Parkin report, with the greatest variation being among Asian populations. One must be cautious about drawing conclusions from such data, however. Diagnostic techniques may differ between countries, classification of neoplasms may not be uniform, thoroughness of surveillance may be questionable and there may be disparities in access to medical care (Chow et al. 1996). However, these international variations may also be suggestive of environmental factors involved in the development of childhood cancer (McBride 1998). Nearly all populations show a greater proportion of cancer among boys (ratio 1.4:1) and among whites than blacks (1.2:1). Age-adjusted rates for leukemia and lymphoma are similar in most of the developed countries. Acute lymphoblastic leukemia (ALL) is the most common form of cancer in many developed countries, with high rates among Costa Ricans, US Hispanic males in Los Angeles and Australians, and low rates among Bombay Indians, Israeli non-Jews and US blacks. Rates appear to be uniform among Asians. Acute non-lymphoblastic leukemia (ANLL) is rare in children but shows a different distribution than A L L ; rates are similar among whites and blacks in the US and vary considerably among Asians. Lymphomas have the highest incidence in Africa (most likely because of the high incidence of Burkitt's lymphoma) and Papua New Guinea. In several Asian countries and Latin America, the lymphomas are the second  13  most common malignancy after leukemia. Lower incidence is seen in the Philippines, Hungary and among Blacks in the US. Hodgkin's disease has a bimodal distribution in most countries, except Japan and the more aggressive forms of the disease occur in under-developed countries. Brain and central nervous system cancers are the second most common childhood malignancy in Europe, North America, Australia and Japan but rates show considerable variability in the Asian countries (Chow et al. 1996).  Mortality and Survival Although incidence has been rising, mortality has been declining since the late 1960s due to improved treatment and survival. The Surveillance, Epidemiology and End Results (SEER) Program has recently released the following 5-year relative survival rates (adjusted for non-cancer deaths) for US children under 15 years of age: for all sites, survival has increased from 55% in 1974-76 to 71% in 1986-92; over the same period of time, 5-year survival for both leukemia and non-Hodgkin's lymphoma has increased from 45 % to 72 %, and for brain and CNS cancer, the increase has been from 54% to 61%) (Stat Bite 1997). The past two decades have seen an overall 38% decrease in mortality rates for childhood cancer (Bleyer 1993). Rates for acute lymphoblastic leukemia, soft tissue sarcoma, Hodgkin's disease, and non-Hodgkin's lymphoma have plummeted by more than 50% (Bleyer 1993). It has been predicted that by the turn of this century, 1 in 900 North Americans will be a survivor of childhood cancer (Gibbons et al. 1994). Survivors of childhood cancer are at increased risk for developing second primary tumours, partly as a result of therapeutic exposure to radiation and/or  14  chemotherapy and partly as a result of genetic factors interacting with radiation treatment (Chow et all996).  2.1 (ii) Etiology and Carcinogenesis The etiology of childhood cancer is poorly understood. Unlike cancers in adults which are predominantly epithelial in nature and involve surfaces exposed to chronic environmental insults (e.g. cigarette smoking and lung cancer, sun exposure and skin cancer), most childhood cancers involve deep organ structures or tissues such as the bone marrow, lymph nodes, brain and kidney. Only 10% of childhood neoplasms are epithelial and therefore a substantial proportion of childhood neoplasms may be due to spontaneous mutations or genetic predispositions (Huchcroft et al. 1996, Preston-Martin 1989). The most important contribution in the research on childhood cancer has been the increasing knowledge of genetic events associated with malignant transformation. For instance, tumour suppressor genes that were initially described with retinoblastoma have now been associated with Wilm's tumor and rhabdomyosarcoma. Forty to fifty percent of retinoblastoma cases are genetic in origin. Children of survivors of hereditary retinoblastoma have an increased risk for developing this disease as compared to children born to survivors of unilateral retinoblastoma with no family history of the disease (Chow etal. 1996).  Studies of pediatric cancer have also contributed to knowledge of familial cancer associations and syndromes. A number of autosomal dominant single-gene disorders predispose children to having central nervous system cancers. Neurofibromatosis and  tuberous sclerosis have been associated with gliomas, and nevoid basal cell carcinoma syndrome with medulloblastomas (Li and Fraumeni 1982, Kuijten and Bunin 1993). L i Fraumeni syndrome is an inherited condition known to predispose to sarcomas in children, premenopausal breast cancer and a variety of other neoplasms including brain tumours and leukemia in afflicted persons (Li and Fraumeni 1969). Familial tendencies exist suggesting hereditary susceptibility: siblings have at least a three fold increased risk for brain tumours and elevated risk for other tumours (Farwell and Flannery 1984) and the risk for astrocytic glioma has been associated with a history of seizures and epilepsy in first or second degree relative (Robison et al. 1995). It has been suggested that if there is a genetic predisposition to brain tumours, several genes may be involved (Kuijten and Bunin 1993); a gene may be specific for glioma ormedulloblastoma, as brain neoplasms within a family are generally concordant with respect to these histologies. Future studies should include detailed family health information, including cancer history, to show patterns of inheritance.  Furthermore, a number of conditions have been associated with childhood leukemia. Children with Down syndrome have a twenty to thirty fold increased risk of developing acute leukemia (Mili, Khoury et al. 1993, M i l i , Lynch et al. 1993), and children exhibiting the following autosomal recessive diseases which are associated with chromosomal instability: Bloom's syndrome, Fanconi's anemia and ataxia telangietasia (Chow et al. 1996). Twin studies (Miller 1971) show 20% concordance for acute leukemia before six years of age; the co-twin usually develops leukemia within months of the first case. Risk is highest if the first twin is affected before the age of one. Among 16  identical twins older than six years and fraternal twins and siblings, the risk of developing leukemia is two to four times that of the general population.  As childhood cancer is so rare, many of the published studies have been unable to include sufficient numbers of cases to provide vital information on exposure periods and specific exposures associated with increased risk for individual histologic types. Large, and possibly collaborative, studies are necessary to overcome these limitations. In the United States, the Children's Cancer Group (CCG) and the Pediatric Oncology Group (POG) diagnose and treat more than 90% of the approximate 8,000 annual cases of childhood cancer. The C C G has established an epidemiology committee and numerous studies have been initiated to investigate individual histologies as well as exposures associated with childhood cancer (Robison et al. 1995).  Research in perinatal carcinogenesis has proposed three mechanisms for the development of childhood cancer:  1. Multigenerational Transmission: Carcinogenic transformation of cells is generally believed to encompass multiple genetic events whereby normal growth, differentiation and development have gone awry (Yamasaki et al. 1992, Tomatis 1979, Shields 1993, Tomatis et al. 1992). Carcinogenesis involves the following steps: initiation, promotion and progression (Autrup 1993). Initiation of a tumour involves the direct effect of a carcinogen on cell D N A , thereby resulting in mutations which are passed on to all daughter cells. Promotion entails a clonal expansion of the  17  mutated cells. Cells containing one or a few genetic mutations may not result in overt cancer; however, several subsequent genetic insults may produce the tumour phenotype. Therefore, one or more genetic insults occurring in either somatic and/or germ cells in-utero may not be expressed until other events occur postnatally. Animal studies have shown that carcinogens can be transmitted multi-generationally either through exposure in utero or by parental exposure prior to conception. Skin tumour initiation following postnatal exposure has been demonstrated by painting 12O-tetradecanoylphorbol-13-acetate (TPA) on the skin of second-generation mice whose parents were exposed to 7, 12-dimethyl-benz(a)anthracene (DMBA) (Loktionov et al. 1992, Likhachev et al. 1989). Further studies demonstrated that if TPA was not painted on the offspring, no tumours developed (Yamasaki et al. 1992). The transplacental treatment of Patas monkeys with the liver carcinogen diethylnitrosamine (DEN) produced no hepatic tumours, but when the offspring were treated with the liver tumour promo tor phenobarbital, at age four, they developed hepatocellular carcinoma (Rice 1986). Also, lung tumours have occurred in the offspring of x-irradiated parents treated postnatally with urethane (Vorobtsova and Kitaev 1988). Hence, the in-utero exposure of a carcinogen may be the genetic damage equivalent to initiation which will only produce a phenotype with a promoting agent. In humans, evidence for multi-generational transmission is less conclusive. The two research areas most frequently quoted are the carcinogenic effects to the offspring of atomic bomb survivors and nuclear power plant employees. Offspring of atomic bomb survivors have not experienced an increased cancer risk (Miller 1994) whereas the offspring of a well-publicized report from the Sellafield  18  nuclear power station workers did show elevated risks for leukemia and nonHodgkin's lymphoma (Gardner et al. 1990). The "Gardener hypothesis" that an excess of childhood leukemia near a nuclear reprocessing plant is caused by paternal exposure to ionizing radiation has been challenged and rejected (Doll et al. 1994) based on observations that (i) no excesses of leukemia in young people have been observed in the rest of Cumbria outside Seascale even though 92% of births to Sellafield employees occurred outside Seascale and these fathers received 93% of the total preconception dose, (ii) only one of the fathers of ten children who developed acute lymphoblastic leukemia had any occupational exposure to ionizing radiation,-. : (iii) the excess cases in Seascale were not limited to children born there, but also included children born elsewhere but living in Seascale, and (iv) other studies of cancer among the offspring of exposed nuclear power plant operators have not found elevated risks (McLaughlin et al. 1993, Kinlen et al. 1993). Results of a recent large record linkage study of childhood cancer among offspring of radiation workers in Britain also did not support the Gardner Hypothesis (Draper et al. 1997).  2. Transplacental Carcinogenesis: Transplacental carcinogenesis was documented nearly sixty years ago, when it was demonstrated that pulmonary tumours could be induced in mice by exposing them in utero to dibenzanthracene (Law 1940). Shortly thereafter, it was reported that the administration of urethane to pregnant mice would induce lung tumours in their offspring (Larsen 1947). In animal systems, at least 50 low-molecular weight organic chemical compounds have exhibited transplacental carcinogenic activity (Rice 1984) and experiments with rats show the rat fetus to be  at least 50 times more susceptible to neoplastic effects than the adult animal (Rice 1986). Studying the transplacental transfer of a chemical in experimental animals is considerably easier than in humans where cord blood samples during the perinatal period and/or epidemiological investigations are the only acceptable methods. Cord blood samples of chemicals found in cigarette smoke have been shown to be identical in cord and maternal blood. More research is needed to examine metabolites of chemicals found in human cord blood. Results of animal research suggest that the embryo appears more resistant to the induction of tumours during early gestation but after organogenesis is complete, there is a steady increase in tumour production.. (Bolande 1994). The synthetic estrogen diethylstilbestrol (DES) was widely used in the 1940's by pregnant women with threatened miscarriage. Until the late 1960's only a dozen or so cases of clear cell adenocarcinoma of the vagina were reported in the literature. Then six cases were encountered in a couple of years at one Boston . Hospital (Herbst et al 1971). Follow-up of women treated with DES has identified 519 cases of clear cell adenocarcinoma. From these cases the following has been established: DES given at any stage during pregnancy is carcinogenic, the lowest dose associated with abnormalities is 1.5 mg/day or 135mg in total, most cancer is diagnosed between the ages of 15 and 27, vaginal ridges and adenosis are precursor lesions (epidemiological evidence indicates that in the female fetuses exposed during early gestation, 35-90% developed malformations of the urogenital tract), and the cancer may arise in the cervix as well as the vagina. (Mulvihill 1994, Bolande 1994). Hence DES is teratogenic as well as carcinogenic. The only transplacental physical carcinogen is ionizing radiation causing leukemia (Stewart et al. 1956).  20  Direct exposure of offspring to contaminants brought home by father. Far less research has been carried out in this area. Studies of the effects of environmental exposures on children generally focus on diet and exposure via consumption of pesticide or heavy metal-laden food and drink (Thomas 1995). What should be kept in mind when considering direct exposure of the offspring is that children are growing and developing; their metabolic rates are more rapid than those of adults, and their ability to activate, detoxify and excrete xenobiotic compounds may not compare to adults (Thomas 1995). Furthermore, children spend large amounts of time on the floor and/or ground putting hands and objects in their mouths, therefore exposure to pesticides and other chemicals used in the home and garden would be greater in children. In the context of this thesis, chlorophenates brought home from work on the father's clothing and/or skin, although at a lower concentration than found occupationally, may be sufficient to adversely affect growth and development. In light of the above concerns, researchers at a recent conference on Preventable Causes of Cancer made several recommendations regarding appropriate types of research to be undertaken to better understand children's cancer susceptibility (Carroquino etal 1998).  Paternal Occupational Exposures and Childhood Cancer Research into male mediated parental carcinogenicity is in its early stages. The first study by Fabia and Thuy in 1974 (Fabia and Thuy 1974) found an association between father's employment in a job involving exposure to poly cyclic aromatic hydrocarbons and childhood malignancy. Subsequent research has challenged this original observation (Hakulinen et al. 1976, Zack et al. 1980). Since these early studies a number of researchers have investigated paternal occupational exposure and childhood cancers. O'Leary et al. (1991), Savitz and Chen (1990), Gold and Sever (1994), McBride (1998) and Colt and Blair (1998) provide comprehensive reviews of the epidemiological data on parental occupation and childhood cancer. Two review papers specifically related to pesticide exposure and risk of childhood cancer have recently been published (Daniels et al. 1997, and Hoar Zahm and Ward 1998). A l l review papers conclude by suggesting that the literature demonstrates consistent associations between certain parental occupational exposures and childhood cancer; however, improved methods of exposure assessment are imperative if the etiologic agents associated with specific sites are to be uncovered. Often, exposure assessments are very crude and are based on job title alone as attained from birth certificates, medical records, death certificates, interviews or questionnaires. Added to this poorly defined occupational exposure assessment is the fact that exposure to mixtures of substances occurs in many workplaces. Inadequate measures of exposure may fail to detect real effects by shifting the point estimate towards the null. Often, in occupational studies, maternal exposure data is not collected - this is important because of the potential toxicant transfer to child during pregnancy and breast feeding. Investigating timing of exposure: preconceptual, gestational, and postnatally,  22  has also been highlighted as a desirable feature of studies examining the relationship between parental exposure and risk of childhood cancer. However, few studies are sufficiently large to delineate effects. Advances in biomolecular techniques such as the identification of markers for exposure in affected children, and different germline mutations will advance the understanding of relationships between parental occupational exposures and childhood cancer (Gold and Sever 1994, Chow et al. 1996, Gufferman 1998).  Leukemia and brain and central nervous system cancers are the most common childhood cancers. Childhood leukemia has been observed in the offspring of men employed as painters, mechanics, motor vehicle drivers, service station attendants, machinists, (Fabia and Thuy 1974, Lowengart et al. 1987, Buckley et al. 1989, Hemminki et al. 1981, Vianna et al. 1984, Olsen et al. 1991), nuclear power station employees (Gardner et al. 1990) and physicians and social service workers (Van Steensel-Moll et al. 1985) and with exposures to ionizing radiation (Dyer 1993, Gardner et al 1990) - although this has been hotly debated and refuted (Doll et al. 1994, Draper et al. 1997), wood dust (McKinney et al. 1991), solvents, paints, dyes and pigments, cutting oil (Lowengart et al. 1987, Buckley et al. 1989), plastics, polystyrene, polyethylene, petroleum products and pesticides (Lowengart et al. 1987, Buckley et al. 1989, Meinert etal. 1996).  Brain and central nervous system tumours have been observed in the offspring of men exposed to pesticides, chemical solvents, electro-magnetic fields and paints (Gold et  23  al. 1979, Peters J et al. 1981, Wilkins and Sinks 1990). Occupations at risk include: working in construction, agriculture, pig and chicken farming, fishery and forestry (Cordier et al. 1997, Wilkins and Koutras 1988, Kuijten et al. 1992, Kristensen et al. 1996) in the pulp and paper industry (Kwa and Fine 1980, Kuijten et al. 1992, Johnson et al. 1987), in carpentry (Olsen et al. 1991), metal and motor vehicle related jobs, in structural work jobs in the construction industry, in machine repair, in electrical/electronic assembling, installing and repairing occupations (Wilkins and Koutras 1988, Wilkins et al. 1991, Wilkins and Sinks 1990, Kuijten et al. 1992, Hemminki et al. 1981, Cordier et al. 1997) and in those working in the. aircraft industry, (Peters, J et al. 1981). Later studies of workers in the aerospace industry found no statistically significant associations (Olshan et all986, Kuijten et al. 1992).  Paternal Non-Occupational Exposures and Childhood Cancer Certain paternal lifestyle factors have been explored in relation to childhood cancer, usually in their role as confounders and not the exposure of interest. Studies examining paternal smoking and childhood cancer have had conflicting results. However, a couple of recently published reports show convincing evidence that paternal smoking in fact does play a role in the development of childhood cancer (Sorohan et al. 1997, Ji et al. 1997) . Sorohan and colleagues used data originally collected for a case-control study looking at the association between radiographs and childhood cancer (Oxford Survey of Childhood Cancers deaths from 1953-1955) to examine the relationship between paternal smoking and childhood cancer. A statistically significant (p < 0.001) trend between paternal smoking and risk of childhood cancer was observed. The authors suggest that i f 24  their results are accurate, approximately 15% of all childhood cancers in their series could be attributed to paternal smoking. Ji and colleagues performed a population based case-control study in Shanghai, China to investigate the role of paternal cigarette smoking in the etiology of childhood cancer (Ji 1997). Their results indicated that paternal cigarette consumption did play a role in the subsequent development of childhood cancer: children whose fathers smoked more than five-pack years prior to their conception had almost twice the risk of cancer (OR = 1.7, 95% CI, 1.2 - 2.5) than children whose fathers never smoked. Elevated risks were seen for A L L (OR = 3.8, 95% CI, 1.3 - 12.3), for lymphoma (OR = 4.5, 95% CI, 1.2 - 1.6.8) and for brain tumours (OR = 2.7, 95%) CI, 0.8 - 9.9). Alcohol consumption is a lifestyle variable that has.been associated with certain adult neoplasms (Blot WJ 1992). In a study investigating alcohol consumption and childhood cancer (Sorohan et al. 1993), no association was observed (RR = 1.05, 95% C I , 0.86 - 1.28). Other exposures possibly associated with childhood leukemia include household and garden pesticide use, and the burning of incense . (Buckley et al. 1989).  Maternal Exposures and Childhood Cancer To date, the only maternal prenatal exposures that have been conclusively associated with childhood cancer are diethylstilbestrol (Herbst et al. 1971) and ionizing radiation (Stewart et al. 1956). It is difficult to examine maternal occupational exposures, because in the past, women's inclusion in the workforce has been sporadic. Of interest to this thesis is pesticide exposures. In one study, maternal preconception occupational exposure to pesticides was associated with increased risk for acute myeloid leukemia,  25  (OR = 2.09, 95% 0:1.04 - 4.17), and the risk increased with increased cumulative pesticide exposure hours (p for trend = 0.02). Postnatal exposure to household rodenticides resulted in elevated risks (OR= 1.80, 95% CI, 1.11 - 2.89) (Steinbuch, 1994 - dissertation). The following occupations have also been associated with elevated risks for childhood neoplasms: pharmacy, medicine, dentistry, chemical processing, textiles and service as well as food industries (Lowengart et al. 1987, Shu et al. 1996, Olsen et al. 1991, Van Steensel-Moll et al. 1985). Maternal occupational exposures include: benzene, gasoline, soldering and pesticides (Shu et al. 1988, Steinbuch, 1994, Robison et al. 1995). Non-occupational exposures postulated as being potentially carcinogenic to the offspring of exposed women include: 7V-nitroso compounds such as tobacco (although recent studies refute this observation), cured meats, marijuana, and incense. As well antinausea medication, kerosene, pesticides (including the use of pest strips), herbicides and fertilizers have been associated with increased risk (Shu et al. 1996, Robison et al. 1995, Sorahan etal. 1997, Leiss and Savitz, 1995, Infante-Rivard 1997, Steinbuch 1994, Sarasua and Savitz, 1994, Lowengart et al. 1987, Buckley et al. 1989, Pershagen et al. 1992, Golding et al. 1990, Kuijten et al. 1990, Kuijten and Bunin 1993, Bunin et al. 1994). However, reviews of non-occupational maternal and perinatal risk factors associated with childhood cancer have concluded that the evidence linking advanced maternal age, use of neurally active medications, prenatal alcohol consumption, history of adverse reproductive outcomes (miscarriage and stillbirth), birth order (first born), and elevated birth weight with an increased risk for childhood cancer is inconclusive (McBride 1998, Chow et al. 1996).  26  Childhood Risk Factors Associated with the Development of Childhood Cancer Risk factors associated with the development of childhood leukemia include: exposure to pesticides and/or living on a farm (Infante-Rivard 1997, Savitz and Leiss 1995, Buckley et al. 1989, Bunin et al. 1994), ionizing radiation, electrical and magnetic field (EMF) exposure (London et al. 1991, Cartwright and Staines, 1992), higher social class with cases being more prevalent in the higher social classes, suggesting that lifestyle factors may be involved (Cartwright and Staines, 1992), viral infection (Cartwright and Staines 1992, Alexander 1992, MacMahon 1992), and consumption of hamburgers (Sarasua and Savitz 1994). The following risk factors have been associated'. with the development of childhood brain tumours: home exposure to pesticides (Gold et al. 1979, Davis et al. 1992), therapeutic doses of X-rays to the head, such as in the treatment for tinea capitis (Ron et al. 1988), head and neck trauma to the child (Howe et al. 1989), higher birth weight (Kuijten et al. 1990, Gold et al. 1979), consumption of Nnitroso compounds such as those allegedly found in hot dogs (Sarasua and Savitz 1994), and being treated with cranial irradiation for leukemia (Chow et al. 1996). Electromagnetic radiation has been associated with brain cancer in some studies (Wertheimer and Leeper 1979) but not others (Savitz et al. 1988). A recent review has declared the results of studies on E M F and childhood cancer to be inconclusive (McBride 1998).  27  2.2 Chlorophenols Description and Use Chorophenols are organic compounds produced either by chlorination of phenol or by hydrolysis of chlorobenzene (Rotluff et al. 1990). Chlorophenols have been used since the 1930s, primarily as pesticides and/or as intermediates in the production of pesticides (IARC 1986). Nineteen chlorinated phenol structural isomers are available commercially, however only five are of considerable importance. They include 2,4-dichlorophenol, 2,4,5-trichlorophenol, 2,4,6-trichlorophenol, 2,3,4,6tetrachlorophenol, and pentachlorophenol (IARC 1986). 2,4-dichlorophenol and 2,4,5-trichlorophenol are most often utilized as intermediates in the manufacture of herbicides such as 2,4-dichlorophenoxyacetic acid (2,4-D) and ' 2.4.5- trichlorophenoxyacetic acid (2,4,5-T) (IARC 1986). The predominant use of 2.4.6- trichlorophenol is as an unisolated intermediate in the production of 2,3,4,6-tetrachlorophenol and pentachlorophenol (IARC 1986). Also, it has been used directly as a bactericide, germicide, glue and wood preservative, and as an anti-mildew . treatment (IARC 1986). 2,3,4,6-Tetrachlorophenol has predominantly been used to control sapstain fungus in wood. The sapstain fungus causes wood to absorb water thereby accelerating decomposition. Infected lumber becomes discoloured, and as a result, its financial value diminishes.  Of the chlorophenols, pentachlorophenol has received the most public and scientific attention. It has been used primarily for wood preservation since the 1930s. As such, it has been applied to sawn wood, telephone poles, railway ties, picnic tables, playground 28  equipment, log homes and other wooden structures that are subject to fungal decay, insect infestation and must endure harsh climatic conditions (Wood and White 1992). It has also been utilized as a microbial deterrent to preserve leather, burlap, cordage, paints and glues, as a molluscide to control snails that are the hosts of scistosomiasis (Meister 1990), and as an insecticide in termite control (IARC 1986). The sodium salt is used as a slimicide in pulp and paper manufacture (IARC 1986). Pentachlorophenol has been synthesized in several countries, including Canada, the United States, the United Kingdom, Germany, France and Japan. However, as a result of health concerns in the late 1970s and early 1980s, worldwide production has declined; in the United States it declined from approximately 23 million tonnes per year in the 1970s to approximately 12 million tonnes per year in 1987 (IARC 1991). Sweden banned the use of pentachlorophenol in 1977, Germany in 1987, and the U S A no longer uses pentachlorophenols for herbicidal and anti-microbial use nor for preservation of wood that is in contact with foodstuffs, domestic animals and livestock (IARC 1991). Pentachlorophenol use in Canada is regulated by the Pest Control Products Act. Since 1981, its use has been restricted to reduce human exposure through dermal contact, inhalation and diet. The British Columbian lumber industry ceased using pentachlorophenol in the late 1980s.  History of Chlorophenate use in British Columbian Sawmills A detailed description of the history of chlorophenate (the sodium salts of chlorophenols) use in the Western Canadian lumber industry was reported by Teschke et al. (1994a). The following is a summary of chlorophenate application processes and  formulation changes that occurred during the past fifty years. Chlorophenate fungicides (sodium tetra- and pentachlorophenate) were widely used in the sawmill industry between the 1940s and 1980s to prevent the growth of fungi that stain lumber surfaces. In the 1940s sawn wood was treated in diptanks; in the 1950s sprayboxes were introduced and, since then, both application methods were used. Early formulations contained primarily sodium pentachlorophenate. In this era, motivation behind changes in application methods or chemical formulations had more to do with increasing production efficiency and fungicidal efficacy than with health outcomes. However, in the mid 1960s health concerns began to emerge; tests on rats showed tetrachlorophenates to be less toxic than pentachlorophenates. Hence, a switch to sodium tetrachlorophenate occurred. Further health concerns arose in 1965 when mercury, which was used in the sodium tetrachlorophenate formulation, was identified as a neurotoxin. This resulted in the removal of mercury from the compound in 1970 and thereafter, numerous chlorophenate formulations came on to the market; all with higher concentrations of tetra- than pentachlorophenates. In the 1970s concerns over dioxin contamination emerged. During the next two decades, epidemiological research linked non-Hodgkins lymphoma, soft tissue sarcoma, multiple myeloma and nasal cancer to persons exposed to chlorophenols. This prompted foreign markets to restrict the importation of chlorophenate-treated softwood in the late 1980s. The B C lumber industry then replaced chlorophenates with other fungicides: copper-8-quinolinolate (Copper 8), 2thiocyanomethylthiobenzthiazole (TCMTB) and didecyldimethyl ammonium chloride (DDAC). Figure 1 describes the chronology of wood treatment in the B C sawmill industry (reprinted from Teschke et al. 1994a).  Figure 1. Changes in fungicide formulations and application methods from 1940-90.  Decade  Fungicide formulation  Major changes in application  1940  PCP in powder or pellet form  Crane diptanks, hand spraying or  with or without borax.  dipping, few sprayboxes  Product name: active ingredients Santobrite: PCP 90%, TCP 10% Permatox 10S: PCP 36%, TCP 57% 1950  As above  Linear sprayboxes more common  1960  TCP in water-based concentrate  Drive-through diptanks replace  from with P M A  remaining dock diptanks  Product name: active ingredients Permatox 100:TCP 18%, Borax 6%, Other CPs 5%, P M A 0.4% 1970  P M A removed from TCP concentrate Product name: active ingredients Permatox 100:TCP 18%, Borax 6%, other CPs 5% Diatox:TCP 19-24%, PCP 5% Woodbrite 24:TCP 16%, PCP 8%, Borax 2%  1980  Cross-chain sprayboxes replace linear, elevator diptanks replace drive-through  1987  T C M T B , Copper-8 and borax substitutes  1990  DDAC/IPBC and Azaconazole  Return to linear spray-boxes  substitutes Abbreviations for formulation ingredients: PCP= sodium pentachlorophenate, TCP = sodium tetrachlorophenate, PMA = Phenyl mercuric acetate, TCMTB = 2-thiocyanomethylthiobenzthiazole, Copper-8 = copper-8-quinolinolate, DDAC = didecyldimethyl ammonium chloride, IPBC = 3-iodo-2propynl butylcarbamate.  31  During the production of chlorophenols, polychlorinated diphenyl-p-dioxins (PCDDs) and polychlorinated diphenylfurans (PCDFs) are formed and chlorophenate formulations are contaminated with these compounds: the levels of which depend on the production process, production conditions and purification process used (Teschke et al. 1994, Rotluff et al. 1990). PCDDs and PCDFs are chlorinated aromatic compounds exhibiting similar chemical and physical properties. 75 Dioxin congeners and 135 furan congeners can be formed by substitution of one to eight chlorine atoms onto the rings (IARC 1986). Individual isomers vary considerably in biological activity and toxicity ~... (IARC 1986). The highly toxic dioxin, 2,3,7,8-tetrachlordibenzo-p-dioxin (TCDD) is rarely found in commercial pentachlorophenol preparations (IARC 1991). Hexa, hepta and octa-chlorinated polychlorinated dibenzodioxins and dibenzofurans are most often produced with the tetra and pentachlorophenates (Teschke et al. 1994a, IARC. 1986, IARC 1991).  Routes of Exposure Occupational Exposures: Workplace exposure to chlorophenols occurs during the production and use of chlorophenols and through contact with the treated materials. The primary route of exposure is skin contact (IARC 1980, Kelly et al. 1997). In the sawmill industry, workers frequently handle wet timber, which may enhance dermal absorption. Breathing chlorophenol vapours and aerosols is a potential route of exposure, although a worker's dose from inhalation is typically less than from skin contact (Kauppinen and Lindroos 1985, Kelly 1997). As the principle route of exposure to chlorophenols is 32  dermal, urine and plasma measurements are frequently used to assess exposure, rather than air samples.  Several studies have measured blood and urinary levels of pentachlorophenol in exposed and unexposed individuals. Urinary concentrations among individuals occupationally exposed range from 0.01 mg/1 to 9.68 mg/1; among individuals nonoccupational^ exposed values range from 0.01 - 0.2 mg/1, and among individuals with no known exposure from non detectable levels to 0.19 mg/1 ( IARC 1991). Studies of pentachlorophenol processing and production have reported mean, urinary concentrationsranging from 0.1-1.2 mg/1 (IARC 1991). In sawmills, concentrations of urinary chlorophenols reported were higher, ranging from 0.20 - 3.2 mg/1 among exposed workers. The highest concentrations were among workers dipping and spraying wood as well as loading newly treated lumber (Kaupinnen and Lindroos, 1985). Plasma concentrations have been reported to range from 0.004 - 0.075 mg/1 among individuals with no known exposure to 1.51-3.55 mg/1 among exposed sawmill workers (IARC 1991).  Non-Occupational Exposures: Chlorophenols are ubiquitous in nature because of their broad use as pesticides. As a result, there is continuous low level exposure. Nonoccupational exposure occurs primarily from eating food and drinking water containing residues of chlorophenols. Inhaling gases in homes where pentachlorophenol has been used for wood protection, and from the transport of aerial herbicide spraying is a secondary significant source of exposure. Pentachlorophenol has been observed in  33  surface waters throughout the world from levels that are barely detectable to 82 u/1 (micrograms per litre) (IARC 1991). Pentachlorophenol enters the food chain primarily through the soil (approximately 97%) and is consumed mainly in fruits, vegetables and grains. Average daily intakes for humans have been estimated at 15 - 20 micrograms per day (Hattemer-Frey and Travis, 1989). Soil samples taken around Finish sawmills showed pentachlorophenol levels of 45.6 mg/kg at 0 - 5 cm depth near the treatment basin and up to 0.14 mg/kg in the area for storing treated wood. Background pentachlorophenol levels were 0.012 mg/kg (Valo et al. 1984). A study investigated long-range atmospheric transport of herbicides by measuring atmospheric concentration of pentachlorophenol in three Canadian locations (Cessna et al. 1997). Ambient air concentrations ranged from 0.30 ng/m to 1.53 ng/m , thus suggesting another means by 3  3  which pentachlorophenol can potentially enter the food chain.  Distribution, Excretion and Metabolism Exposure to pentachlorophenol results in rapid absorption from the skin into the blood. More than 95% of pentachlorophenol in blood is bound to plasma proteins. Metabolism of pentachlorophenol in the body is minimal; 74% is excreted unchanged through the kidneys (Braun et al. 1979) and it has been estimated that the average amount excreted from unexposed persons is 4.3 nmol/day (Treble and Thompson, 1996). Human liver microsomes metabolize approximately 12% to glucoronide conjugate and a small amount undergoes oxidative dechlorination to tetrachlorohydroquine. The half-life for  34  urinary excretion of tetrachlorophenol has been estimated at between 2 and 5 days, and for pentachlorophenols at between 12 and 20 days (IARC 1986).  Toxicity Chlorophenols are toxic to both animals and humans. In animals, route of exposure plays little role in the resulting toxicity - oral, dermal, subcutaneous and intraperitoneal L D values vary insignificantly. Among laboratory animals, the L D ranges from 50 50  50  330 mg/kg (Seiler 1991). Among humans, the dermal L D is 150 to 350 mg/kg, the oral 50  L D is 50 to 140 and the lethal dose is 29 mg/kg body weight (Ahlborg and Thunberg, 50  1980). Symptoms of acute toxicity in animals and humans include tachycardia, hyperpyrexia, hyperthermia, nausea, abdominal pain, vomiting, convulsions, pulmonary edema and rapid death from cardiac arrest. Fatalities resulting from acute exposure to pentachlorophenol are rare but have been reported (Wood and White 1992). Toxicity occurs because pentachlorophenol stimulates cellular oxidative metabolism and-heat . production by uncoupling phosphorylation in the mitochondria and endoplasmic reticulum (Kelly 1997). Chronic exposure to chlorophenols results in skin irritation, rashes, chloracne (possibly as a result of chlorophenol contamination with dioxins and/or furans), chronic sinusitis, bronchitis and conjunctivitis. (Wood and White 1992).  Studies on the immune system are insufficient and equivocal. Immunological changes have been observed in workers exposed to pentachlorophenol, including marked T-cell suppression (IARC 1986). These observations are supported by a recent study that examined immune parameters of 188 male and female patients exposed for more than six 35  months to pentachlorophenol-containing pesticides (Daniel et al. 1995). The results indicated that pentachlorophenol exposure induces moderate to severe immune dysregulation, including decreased CD4/CD8 ratios, decreased CD4+ cell numbers in the blood and elevated serum neopterin levels. However, another study of 32 male factory workers whose job involved applying pentachlorophenol to wooden strip boards and a control group of 37 non-exposed workers from the marble manufacturing industry found no difference in immune function between exposed and unexposed workers (Colosio et al. 1993).  A series of case reports have linked aplastic anemia to pentachlorophenol exposure (Roberts 1981, Roberts 1983, Rugman and Cosstick 1990).  In non-occupationally exposed persons such as those living in pentachlorophenol treated log homes, several symptoms have been noted, including: dizziness, headache, respiratory tract infections and neurological and skin disorders (Seiler 1991).  Reproductive Effects Animals: Among rats, reproductive effects appear to occur above 13 g/kg body weight. In a study of Sprague-Dawley rats, the administration of pentachlorophenol at 0, 4, 13 and 43 mg/kg body weight per day throughout mating and pregnancy produced offspring with reduced crown-rump length and increased fetal skeletal variations at 13 mg/kg day, and at a dose of 43 mg/kg/day, pentachlorophenol was embryolethal (Welsh et al. 1987). In a study of rats fed pentachlorophenol at 3 mg/kg body weight, no effect on 36  reproduction, neonatal growth, survival or development was observed (Schwetz et al. 1978). Another study demonstrated a dose-gradient whereby the administration of 5 -50 mg/kg body weight per day at various intervals during days 6-15 of pregnancy showed a dose-related increase in the incidence of resorptions, subcutaneous edema, dilated ureters and anomalies of the skull, ribs and vertebrae. The most sensitive period was organogenesis (Schwetz et al. 1974).  Humans: Very little information exists regarding the reproductive effects to humans of chlorophenol exposure. A retrospective cohort study of 26,000 sawmill workers employed between 1950 and 1985 in British Columbia and exposed to tetra- and pentachlorophenol found no evidence of male infertility (Heacock et al. 1998). A nested case-control study of the offspring of the above mentioned cohort of sawmill workers showed no increased incidence of births that were low birth weight, premature or stillborn nor neonatal deaths (Dimich-Ward et al. 1996). However, an increased prevalence of congenital anomalies of the eye, particularly cataracts, genital organs and spina bifida or anencephaly was observed in the same cohort of children. Although in all retrospective cohort studies there exists some possibility of exposure misclassification, the initial study of cancer incidence and mortality among male sawmill workers used an exposure assessment that was shown to be valid and reliable (Teschke et al. 1996, Teschke et al. 1990, Hertzman et al.1988). Barring the above-mentioned studies, no other epidemiological studies on the male-mediated human reproductive effects of chlorophenol exposure have been published.  37  A cross-sectional study of 410 day care workers in Hamburg, Germany showed statistically significant reductions in birthweight and birth length among offspring whose parents were occupationally exposed to pentachlorophenol and lindane wood paneling as compared to unexposed workers (Karmaus and Wolf 1995). A case series describing two women who each experienced three successive miscarriages following home exposure to pentachlorophenol followed by successful pregnancies with removal of the putative exposure suggested a possible link between pentachlorophenol exposure and the miscarriages (De Maeyer et al. 1995). The first case had three uneventful pregnancies prior to purchasing an old piece of furniture that had been treated with pentachlorophenol. She then experienced three successive miscarriages. Her serum PCP level was 62 u/ml. Following removal of the furniture, the woman's serum level fell to 24 u/ml. Soon thereafter, she became pregnant and her serum level remained at 13 u/ml throughout the pregnancy. The second case also had three uneventful pregnancies, then purchased a house whose timber was treated with PCP and experienced three successive miscarriages. Her serum PCP measured 31 u/ml. Following treatment of all the wood in the house with polyurethane coating, her serum level fell to 24u/ml and she subsequently had a successful pregnancy. The epidemiological robustness of case series and crosssectional studies is less than that of case-control or cohort studies, hence results drawn from the above two studies should probably be taken with caution. They both, however, may warrant further research to follow-up on and corroborate their observations. Lastly, a study which questioned the wives of employees of a chemical company who were  38  potentially exposed to pentachlorophenol and 2,4,5-trichlorophenol and their dioxin contaminants, no adverse reproductive events were reported (Townsend et al. 1982).  Mutagenic Effects A recent review of the mutagenic potential of pentachlorophenols concluded that the conflicting results of genotoxicity tests did not permit a conclusive decision (Seiler 1991). Pentachlorophenol has given negative test results with Salmonella typhymurium in the absence of metabolic activation (Haworth 1983). However, in contrast to results of assays using bacterial strains, pentachlorophenol is able to induce changes in the D N A of eukaryote germinal and somatic cells. In vitro testing of tetra and pentachlorophenol produced forward mutations in strains of saccaromyces cerevisae (National Research Council of Canada 1981). In Chinese hamster V79 cells, pentachlorophenol did not induce gene mutation at the hprt locus, but in vitro it did increase the frequency of sister chromatid exchanges (IARC 1991). Chromosome analyses of cultured peripherallymphocytes taken from 22 men at a pentachlorophenol producing factory revealed an increased frequency of dicentric and acentric chromosomes, but no increased frequency of sister chromatid exchanges, compared to matched controls. (Bauchinger et al. 1982). A n increased frequency of chromatid types and unstable chromosome aberrations were also found in workers preparing dioxin-contaminated herbicides (Fahrig et al. 1978).  Carcinogenic Effects The International Agency for Research in Cancer (IARC) has determined that there is limited evidence for the carcinogenicity of occupational exposure to chlorophenols in 39  humans (IARC 1986). For pentachlorophenol specifically, IARC has determined that there is sufficient evidence in experimental animals to consider it a carcinogen, but inadequate evidence in humans (class 2B carcinogen) ( I A R C 1991).  Animal data: A study of mice fed technical grade pentachlorophenol at 100 or 200 mg/kg of diet for two years found a significant dose-related increased incidence of hepatocellular adenomas and carcinomas as well as adrenal phaeochromocytomas and haemangiosarcomas of the spleen and/or liver (US National Toxicology Program 1989).  Humans: In the late seventies and early eighties, population based case-control studies from Sweden reported significantly increased risks among males for soft tissue sarcomas, non-Hodgkin's lymphomas, Hodgkin's Disease and nasal cancers resulting from exposure to chlorophenols and/or chlorophenoxy herbicides. Specifically, in 1979, Hardell and Sandstrom (Hardell and Sandstrom 1979) reported an odds ratio (OR) of 6.6 (p< 0.001) based on 52 cases for exposure to chlorophenols and soft tissue sarcoma. Two years later, this was confirmed in a study of 110 cases: OR 3.3, 95% CI: 1.3-8.1 (Eriksson et all981). Regarding non-Hodgkin's lymphoma, a study of 169 cases (Hardell et al.1981) reported an OR of 4.3, 95 % CI: 2.7 - 6.9 for exposure to chlorophenols. In 1982, Hardell et al. published findings of an increased risk for nasal cancer with exposure to chlorophenols (OR 6.7, 95% CI: 2.8 - 16.2). Lastly, a study of 60 cases of Hodgkin's Disease showed that exposure to high-grade chlorophenol was associated with elevated risk: OR 9.8, 95% CI: 3 . 2 - 3 0 (Hardell and Bengtsson 1983).  40  Since these studies were published, several international case-control and cohort studies have been unable to reproduce the findings. Borderline or negative results have been reported. Smith and colleagues from New Zealand reported on a registry based case-control study of 82 cases that examined risk of soft tissue sarcoma with chlorophenol exposure (Smith et al.1984). The resulting odds ratio was 1.5 (95% CI: 0.5 - 7.8). Also from New Zealand was a registry based case-control study that investigated risk of non-Hodgkins lymphoma with exposure to chlorophenols. Based on 83 cases, the resulting OR was 1.3 with 95% CI: 0.6 - 2.7 (Pearce et al.1986). A population based case-control study of 704 cases of soft tissue sarcoma and non-Hodgkin's lymphoma from Washington State did not find elevated risks for either disease. The relativerisk (RR) in both cases was 0.99 with 95% CIs crossing unity (Woods et al.1987). A retrospective study of mortality in a cohort of 770 pentachlorophenol manufacturing workers in the United States reported an SMR of 95 and 95% CI of 71 - 125 for all cancers, based on 50 cases. SMRs for non-Hodgkins lymphoma and soft tissue sarcoma were not reported but for all lymphopoietic cancers the S M R was 140 and the 95% CI, 56-288 (Ramlow et al. 1996). As part of the International Register of Workers Exposed to Chlorophenols and Phenoxy Acids, a retrospective cohort study of 16,863 workers was performed (Saracci et al.1991). Risk for soft tissue sarcoma and non-Hodgkin's lymphoma were not significant; the standardized mortality ratio (SMR) for soft tissue sarcoma was 196; 95% CI 53-502 and for non-Hodgkin's lymphoma, 95; 95% CI 47 169. Another retrospective cohort study, this one based on 2074 Dutch men from the International Register of Workers Exposed to Chlorophenols and Phenoxy Acids, reported an S M R of 299 (95 % CI: 36 - 1,078) for exposure to chlorophenols and risk of  non-Hodgkin's lymphoma (Bas Bueno de Mesquita et al. 1993). Two nested casecontrol studies based on 11 sarcoma cases and 32 lymphoma cases gathered from ten countries found slightly elevated risks (1.29 and 1.26 respectively) for exposure to chlorophenols, but the confidence intervals included one (Kogevinas et al. 1995). Lastly, the retrospective cohort study of British Columbian sawmill workers exposed to tetra and pentachlorophenates (Hertzman et al.1997) demonstrated a borderline positive association between exposure and risk of non-Hodgkin's lymphoma (overall SIR = 1.20, 95% CI: 0.96-1.48, SIR by exposure gradient resulted in a chi square for trend with p = 0.04). However, no association was seen for soft tissue sarcoma (overall SIR = 1.17, 95% CI: 0.66-1.94, SIR by exposure gradient resulted in a chi square for trend with p = 0.21), lung (overall SIR = 1.11, 95% CI: 1.02-1.22, SIR by exposure gradient resulted in a chi square for trend with p = 0.15), and nasal cancer (overall SIR = 2.03, 95% CI: 0.95-3.83 SIR by exposure gradient resulted in a chi square for trend with p = 0.50).  42  R e p r o d u c t i v e Toxicology  The recognition that environmental and/or occupational exposure(s) to a toxicant can cause reproductive harm has only occurred during the passed few decades. For example, in the early 1960s a number of pregnant women were given the drug thalidomide to curb morning sickness. Their offspring were born with phocomelia, a condition affecting the upper limbs. In the late 1970s, it was reported that the daughters of women given the drug diethylstilbestrol during early pregnancy were at increased risk for developing clear cell vaginal adenocarcinoma (Herbst 1971). During the same decade, male infertility was observed after occupational exposure to the nematocide dibromochloropropane (DBCP) (Whorton et al. 1977). Since these early studies, there has been widespread concern that environmental and occupational exposures to both men and women could affect reproductive outcome.  Successful reproduction depends on a physiochemical balance within and between maternal, paternal and fetal systems. A chemical or physical insult to the reproductive system at any time could contribute to: couple infertility (frequency 10 - 15%), spontaneous abortion (frequency 15 - 30%), stillbirth (frequency 2 - 4%), low birth weight (frequency 5 - 10%), preterm births (frequency 5 - 10%), congenital malformations (frequency 1 - 3%), childhood cancer (frequency 0.01 - 0.02%) (Lemasters and Selevan 1993, Hatch and Marcus 1991). Although occupational exposures do not account for all adverse reproductive outcomes, they most probably play a role. Unfortunately, little research has been conducted on the reproductive toxicology and epidemiology of many workplace substances. Hence, of concern to many scientists  is the lack of reliable and valid data available on reproductive effects of exposures. There are over 100,000 chemical and physical agents used in the workplace but reproductive toxicity has been assessed for only 5 % of them (Gold, Lasley and Shenker 1994). The National Institute for Occupational Safety and Health (NIOSH) includes reproductive problems in its top ten work-related disorders yet, only four workplace exposures have been regulated on the basis of reproductive effects: lead, DBCP, ethylene oxide and ionizing radiation (Hatch and Marcus 1991). The rapid rate of introduction of substances into the workplace is such that scientists do not have adequate time to perform all the necessary scientific investigations. To study all of the workplace exposures for reproductive cause and effect relationships would require an army of epidemiologists and unlimited resources (Baird, 1992). One way to cost-effectively investigate reproductive outcome is to take advantage of data that has already been collected and to use record linkages of databases that are available (Baird 1992). This thesis project used previously collected paternal exposure data as well as complete offspring cohort information and linked these data sets to the British Columbia Cancer Registry.  To elucidate factors associated with adverse reproductive outcomes, researchers need to understand male and female reproductive function and the effects of toxic exposure(s). It is not in the scope of this thesis to describe the female reproductive system. A brief summary of male reproduction, opportunities for reproductive damage, and methods to assess reproductive toxicology is provided. The discussion then focuses on issues of male reproductive epidemiology.  44  Male Reproduction: Spermatogenesis Human spermatogenesis takes approximately ninety days and is a cyclical process beginning at puberty and continuing through life. Male germ cells are apparent by the sixth week of development but remain undifferentiated until puberty, at which point they differentiate into spermatogonia in the seminiferous tubules of the male testis. These spermatogonia divide mitotically to replenish the stem cell population and provide daughter cells that become primary spermatocytes containing 46 somatic chromosomes and the sex chromosomes X and Y . These primary spermatocytes undergo two meiotic divisions which halve the chromosome number. For each spermatocyte, the first meiotic division results in two secondary spermatocytes (23,X and 23,Y) and the second division in four spermatids (23,X, 23,X, 23,Y and 23,Y). These spermatids then travel through the efferent ducts to the epididymus where they mature, gain motility and the ability to fertilize. This process, called spermiogenesis, takes approximately six days and transforms spermatids into spermatozoa (Lemasters 1992).  Male Reproductive Toxicology Far more epidemiologic and experimental animal research has been performed on the relationship between maternal exposures and reproductive outcomes than on paternal exposures. However, recent laboratory and epidemiologic research suggests that paternal exposures may be more important than previously suspected. The field of male-mediated developmental toxicity is relatively new and refers to "effects of exposures and other factors relating to the male parent that result in toxicity to the conceptus and abnormal development" (Olshan and Mattison, 1994). In 1992, the first conference focussing on 45  this subject was held, bringing together researchers from epidemiology, toxicology, genetics and reproductive biology. Since then, research on male reproductive health has flourished, new methods of study have been developed, and international collaborations have been initiated. Many of the new developments were presented at a conference on hazardous substances and male reproductive health that was held in New York City, in May 1998.  Human Studies Three mechanisms of male-mediated developmental toxicity have been suggested: (1) direct germ effect, (2) seminal transfer of the toxicant or direct exposure of the toxicant to the pregnant woman and (3) direct contact with offspring via exposures brought home from work. The first mechanism has received the most attention.  Ejaculated semen allows the evaluation of sperm quality and quantity: azoospermia (absence of sperm), oligospermia (fewer than 20 million sperm per milliliter of semen), decreased motility and velocity, altered morphology and changes in D N A content can be measured. As well, it is possible to assess the effect of an exposure during different stages of spermatogenesis by timing semen samples after an acute exposure to reflect the stages of interest. For instance, to assess damage to spermatogonia, one should examine semen 10-13 weeks after exposure, for spermatocytes, 3 - 8 weeks post exposure and to evaluate effects on spermatozoa, samples should be taken 1 - 3 weeks after the exposure (Lemasters and Selevan 1993). Several experimental and pharmaceutical drugs are known to affect spermatogenesis. Over 100 compounds have been evaluated by semen analyses and are shown to have detrimental effects on human sperm quality or quantity  46  (Wyrobek 1993). From the list of occupational exposures DBCP, ionizing radiation, lead, Kepone, Carbaryl, carbon disulphide, ethylene dibromide and dinotrotoluene have shown evidence of adverse affects on human sperm quality (Wyrobek 1993).  With respect to genetic damage, sperm cells are thought to be more vulnerable than oocytes because their constant renewal provides ample opportunity for the introduction of genetic abnormalities into large numbers of gametes. It is possible for subtle lesions to be passed on to successive generations. Exposures to mutagens during meiosis when the chromosome number is being halved may result in structural or numerical aberrations that can be replicated and transmitted to succeeding cell generations. Sperm mutations can fertilize and may result in sex chromosomal or numerical aneuploidy at birth. For instance i f a sperm cell with 24 chromosomes fuses with a normal one during fertilization, a zygote with 47 chromosomes results (trisomy), similarly, i f a sperm cell with 22 chromosomes fuses with a normal one during fertilization, a zygotewith 45 chromosomes results (monosomy). Sex chromosomal aneuploidies such as Klinefelter syndrome ( X X Y ) and Turner syndrome (XO) are predominantly associated with paternal chromosomes. (Colie 1993, Wyrobek 1993). Sperm may carry either a final lesion (e.g. an abnormal number of chromosomes) or a prelesion that may be transformed into a chromosomal abnormality only after fertilization (e.g. D N A adducts). Furthermore, the frequency of abnormal reproductive outcomes may depend on the capacity of the fertilized egg to repair D N A prelesions in the sperm (Wyrobek 1993).  47  Two techniques have been developed to provide reliable and efficient cytogenetic analyses of human sperm. The first is the hamster technique, which involves the fusion of human sperm with enzymatically prepared hamster eggs (Rudak et al. 1978). Using this technique, a small fraction of human sperm from healthy men was shown to carry chromosomal aneuploidies and aberrations (Wyrobek 1993). The second technique that has been developed probes the chromosomal content of human sperm and is called fluorescence in situ hybridization (FISH). It uses chromosome specific D N A probes to mark the locations of specific chromosomes and can detect numerical aneuploidy in sperm for one, two or three chromosomes at a time (Wyrobek, 1996). This assay is very promising and will be used to investigate exposure-related induction and persistence of genetic damage in sperm.  Determining the most appropriate animal on which to model'human reproductive response is necessary to further our understanding of male-mediated developmental toxicity. In particular, the identification of an appropriate animal model would be useful for quantifying heritable risk to the developing embryo and offspring. Mouse and human sperm show high concordance for several cancer chemotherapeutic drugs, whereas rat germ cells are a better model for DBCP induced toxicity (Wyrobek 1993). The morphology of rabbit sperm is much closer to human than rats or mice. Rabbits need not be sacrificed to collect sperm samples hence longitudinal investigations of induction and persistence of damage in the same animal can be conducted. On the other hand, rabbits are more expensive to keep than rodents (Wyrobek 1993).  48  There has been less scientific interest in the second and third proposed mechanisms of male-mediated developmental toxicity. Male exposure to chemical agents could result in the substance(s) entering the seminal fluid and thus may affect the ovulated egg, fertilization, or embryonic development. Research has demonstrated seminal transfer of methadone, trimethraprin, phenytoin, cyclophosphamide and sulfamethoxazole (Colie 1993). Direct exposure of the pregnant woman either via skin contact or the handling of contaminated clothing should be considered an atypical maternal exposure. The third proposed mechanism, that of postnatal exposure, has not been the focus of investigations.  Animals The most commonly used animal studies employed to assess genetic effects on male germ cells include the specific locus, heritable translocation, and dominant lethal tests. (Russell and Shelby 1985). Using the Specific Locus Test, a treated test-stock male mouse heterozygous for seven recessive loci producing visible phenotypes is crossed with an untreated homozygous female mouse. A germ-cell mutation at one of these loci in the male will produce offspring with a visible mutant phenotype. Evaluations of the effects of chemicals have demonstrated that many substances capable of inducing mutations in mouse germ cells do so after meiosis (Russell et al. 1990). With regard to the size of mutation, the specific locus test has demonstrated large lesions in post meiotic cells and small ones in premeiotic cells (Russell 1990). The heritable translocation test measures inherited chromosomal rearrangements in living male F, offspring (Russell and Shelby, 1985). The dominant lethal test examines genetic defects occurring in the germ cell that allows fertilization but results in embryonic death (either before or after  49  implantation). Structural and/or numerical chromosome abnormalities in the germ cells of the treated male are suspected of being the cause of these deaths. A list of over sixty chemicals evaluated using these tests has been reported (Olshan and Faustman 1993).  The antineoplastic drug cyclophosphamide has been used as a model to elucidate mechanisms of male-mediated reproductive toxicology in rodents. In a series of studies on the reproductive effects of cyclophosphamide, a nine-week treatment of cyclophospahamide to male rats resulted in both pre- and post-implantation loss as well as congenital malformations (Hales and Robaire 1994). The effects were dose-related and depended on the length of treatment. The authors suggested that pre-implantation losses reflected an effect on early spermatids and spermatocytes and the malformations, an effect on the spermatogonia. Although litter size is decreased after paternal exposure to cyclophosphamide (because of pre and post-implantation losses), some "apparently normal" rats survive. Further studies revealed that the offspring of treated males produced litters with increased incidences of spontaneous losses and malformations (Hales and Robaire 1994). Finally, these reproductive effects were reversible with cessation of treatment. These results suggest that exposure to cyclophosphamide not only affects the sperm cells of the treated male causing adverse reproductive outcome, but that mutations in the male rat's germ cells are transmissible to subsequent generations.  Epidemiological Studies Epidemiological studies of male reproductive function have increased considerably in the last decade, in part due to reports in both the scientific literature and the popular press 50  suggesting numerous adverse reproductive outcomes. Reported have been reductions in semen quality over time, progressive increases in testicular carcinoma, altered sex ratios (a decrease in the number of males at birth), and increasing birth defects of the male genitalia (e.g. hypospadius and cryptorchidism)(Carlsen et al. 1992, Adami et al. 1994, Irvine et al. 1996, Auger et al. 1995, Guillette et al. 1995, Allan et al. 1997). One hypothesis for these observations is that environmental chemicals, including pesticides and dioxins, may mimic estrogens or block androgen receptors ("endocrine disruptors") during fetal life, thereby causing a variety of male reproductive disorders, all of which may have a common etiology (Sharpe and Skakkebaek 1993).  Although the controlled setting of a laboratory provides opportunity for research at the cellular level, it is not always feasible or possible to obtain a sufficient number of semen samples from humans for the purpose of epidemiological investigations. Hence, epidemiologists study the effects of occupational and/or environmental exposures to males by examining reproductive endpoints, such as fertility, spontaneous abortion, birth defects and childhood cancer. A method used to pinpoint the specific component of the male reproductive process affected by the toxicant is to analyse data by windows of exposure; for example, an exposure any time prior to ninety days before conception would suggest a chronic mutagenic effect on the sperm stem cells, and an exposure during the three months prior to conception could affect spermatogenesis. Epidemiological investigations have demonstrated that exposures to the nematocide dibromochloropropane (DBCP) impaired spermatogenesis (Whorton et al. 1977). A sufficient insult to the male stem cells could compromise spermatogenesis causing a  51  reduction in the production of functional sperm and subsequent impairment in fertility, temporary sterility or permanent sterility.  A number of situations have provided a natural setting for long term studies of reproductive outcomes. For instance, the atomic explosion at Nagasaki and Hiroshima during World War II, the spraying of Agent Orange during the Vietnam War, the explosion at a nuclear power plant in Chernobyl in 1986 and survivors of childhood cancer all provide opportunities to study multigenerational effects of exposure(s) on reproduction. The offspring of atomic bomb survivors at Hiroshima and Nagasaki have shown no genetic effects of radiation: the frequency of stillbirths, neonatal deaths, chromosomal abnormalities, congenital malformations and cancers was not elevated (Miller 1994). This suggests that the affected stem cells were phagocytized early on and did not progress through spermatogenesis. Results of one of the studies emanating from the Chernobyl accident suggests that the incidence of thyroid cancer among children in the Ukraine following Chernobyl is at least five times that of unexposed children (Sobolev 1997). These results suggest direct damage to the children themselves.  Childhood cancer survivors provide another group by which to study multigenerational effects of exposure. Cases have been exposed to many types of therapy including ionizing radiation, antimetabolites, alkylating agents, antibiotics and alkaloids. Unlike a natural disaster or an occupational exposure, scientists can access records containing information on exact timing of exposure and dosage. As it is anticipated that one in every 900 North Americans aged 1 6 - 3 4 years will be a survivor of childhood  cancer by the turn of the century, there will be ample research opportunities to assess the long-term effects of chemical and physical agents on reproduction. At present, the limited research completed has suggested no increased incidence or prevalence of adverse reproductive outcomes, although the number of survivors of reproductive age is relatively small (Mulvihill 1994). A recently published article reported on risk of cancer among 5,847 offspring of 14,652 survivors of childhood cancer in Scandinavia (Sankila et al.1998). In all, 44 malignant neoplasms were diagnosed among the offspring (86,780 person years). After removing the 22 cases of hereditary cancers, there was no increased risk for cancer among the offspring of childhood cancer survivors (SIR 1.3, 95% CI 0.8 2.0). A study from Sweden that examined effect of maternal, and paternal cancer on,. cancer in offspring reported no increased risk if the mother had cancer, a slightly elevated, risk if the father had cancer, but a statistically significant increase in risk if both parents had cancer (RR 1.39, 95% CI 1.23 - 1.54) (Hemminki and Vaittinen 1997). Both of the above studies suggest that genetic events may play an important role in. cancer susceptibility among offspring.  Occupational settings provide another venue by which the reproductive effects of exposure in males can be estimated. Studies examining infertility have reported elevated risks with exposure to the nematocide dibromochloropropane and to agricultural work (Whorton et al. 1977, Strohmer et al. 1993) but not to sodium borates or chlorophenols (Whorton et al. 1994, Heacock et al. 1998). Spontaneous abortions have been associated with exposure to mercury, rubber, vinyl chloride and pesticides used in the cotton industry as well as to the nematocide dibromochloropropane (Savitz et al. 1994)  and congenital anomalies, are associated with men working in forestry and logging, as plywood mill workers, painter, printers and janitors (Olshan et al. 1989 and Olshan et al. 1991) and men occupationally exposed to chlorophenates (Dimich-Ward et al. 1996). A number of studies examining paternal occupational exposures and childhood cancer have been conducted. They will be discussed in the next section of this chapter.  2.4 Summary Does the literature review justify embarking on a study to answer the three research questions? Information on the carcinogenic effects of chlorophenols, particularly pentachlorophenol, suggests that it is a class.2B carcinogen; that is there is sufficient animal evidence but insufficient human evidence. Studies on the reproductive effects of chlorophenols in both animals and humans are sparse. In rats, the following was reported: (i) increasing doses of pentachlorophenol throughout mating and pregnancy^ produced adverse reproductive outcomes above 13mg/kg/day, (ii) dose-related adverse reproductive outcomes were apparent and (iii) at 43 mg/kg/day, pentachlorophenol was embryolethal. In humans, paternal occupational exposure to tetra- and pentachlorophenates in B C sawmills did not increase the risk for infertility, nor births that were of low birth weight, premature or stillborn. There was an increased prevalence of congenital anomalies of the eyes, genital organs and spina bifida and anencephaly. The risk for development of congenital anomalies was related to timing of exposure or to level of exposure. For instance increased risk for the development of eye defects occurred with exposure to chlorophenates during preconception, spermatogenesis, teratogenesis and throughout pregnancy, whereas congenital anomalies of the genital  organs was only related to father's cumulative exposure during pregnancy. Anencephaly and spina bifida, however, only showed elevated risks with peak exposures. A case series reported increased risk for miscarriage with home exposure to pentachlorophenol. These few studies may suggest that there are windows of time and chlorophenol exposure levels that affect reproduction differently. For instance, it may be that a peak exposure affects spermatogenesis such that germ cells are phagocytized or that a germ cell mutation results in pre- or post implantation loss. This type of effect may have produced the miscarriages among women in pentachlorophenol-treated homes reported by de Maeyer and colleagues in 1995. In these instances, a carcinogenic exposure would diminish the opportunity for any observable negative reproductive outcome. Experimental research is developing techniques to accurately assess cytogenetic damage in humans and researchers are investigating animal models that reflect human reproductive responses to chemicals. Epidemiological research of paternally-mediated carcinogenicity is necessary to substantiate experimental results. Numerous studies of childhood cancer and paternal occupational exposures have associated agricultural work and/or exposure to pesticides with both leukemia and brain neoplasms. Hence both will be considered cancers of a priori interest. Many of the published reports used a casecontrol design with crude exposure classification schemes and had limited power to detect differences. This thesis used both cohort and nested case-control designs. It contained the largest offspring cohort of chlorophenol exposed fathers gathered to date and had sufficient power to detect meaningful results for all cancers combined at a relative risk of 1.44 and greater. Analyses by individual sites and histologies would require higher risks to achieve the same power. The retrospective assessment of the  55  fathers' occupational exposures has been shown to be valid and reliable (Hertzman et al. 1988, Teschke et al.1990). The thesis research will contribute substantially to what is known about the reproductive effects of chlorophenols and to male-mediated carcinogenicity. Lastly, the research responds to the Report of the Royal Commission on New Reproductive Technologies that advocates for the use of existing data sets to create new information about reproductive occupational health.  56  CHAPTER THREE: METHODS 3.0 Introduction This thesis evolved from an opportunity to utilize previously collected data and apply it to a novel situation. The methodologies and results of two studies investigating a large sawmill worker cohort potentially exposed to chlorophenate fungicides provided employee identifying and job history information as well as an offspring cohort. Computerized record linkage of the offspring cohort to the British Columbia Cancer Registry was then performed and analyses carried out to investigate risk of childhood cancer among the offspring of chlorophenate exposed sawmill workers. The completed studies whose data sets, were utilized are:  •  1. A study of mortality and cancer incidence among a cohort of sawmill workers exposed to chlorophenate wood preservatives (Hertzman et al. 1997), and  2. A study of the reproductive effects of paternal exposure to chlorophenate wood preservatives in the sawmill industry (Dimich-Ward et al. 1996).  The opportunity to pursue research combining the data sets available from the sawmill worker cohort study with the British Columbia cancer registry was timely and relevant to the current interest in male mediated developmental toxicology. No studies have investigated the cancer outcomes of children whose father's were occupationally exposed to chlorophenates. To my knowledge, the offspring cohort is the largest cohort of children with documented paternal chlorophenate exposure in the world. Results of 57  this study should have far reaching implications beyond individuals exposed to chlorophenates in the sawmill industry, but potentially, for those applying phenoxy acid herbicides, using chlorophenates as chemical intermediates, or involved in processes in which dioxins are a by-product, such as pulp and paper production (Rotluff et al. 1990 and Kelly et al. 1992). In addition, this study presents a rare opportunity to examine not only the association between worksite exposure and childhood cancers but, more specifically, the effects of paternal exposure, an issue that has been little studied. The results should provide information on appropriate methodologies for studying paternal occupational exposures and childhood cancer and will address the Royal Commission on New Reproductive Technology's recommendation to use existing data bases to determine occupational exposures that adversely affect reproductive health. (Proceed With Care pp 288).  3.1 Funding Process In June of 1992, permission was sought to access the names of the offspring cohort from the B C Vital Statistics Agency in order to link them to the B C Cancer Agency's cancer registry. Permission was obtained from Mr. R.J. Danderfer, Executive Director of the Division of Vital Statistics, subject to the following conditions: all information be kept confidential, use of name-based information be restricted only to the proposed thesis research, and that the working file be destroyed upon completion. In August of 1992, an estimate of cost and availability of a computer programmer to perform the offspring cohort - cancer registry linkage between 1969 - 90 was provided by R. Gallagher, Head of Cancer Control Research. During the spring of 1993, an application was made for a  58  studentship award through the British Columbia Health Research Foundation (BCHRF). Two studentships were awarded, B C Studentship #ST 22(93) and #59(95), each for one year. Later in 1993, a ten-page proposal was developed for submission to peer-reviewed granting agencies. As part of the proposal, person-years available for study were calculated. At that point, follow-up of the offspring was expected to occur only until 1990 (the thesis study was able to follow the offspring cohort until 1993) so a total of 220,310 person-years were estimated as being available for analysis from the offspring of sawmill workers employed at all fourteen sawmills. Using a spreadsheet for sample size calculations, "Occupational cohort study with rare outcomes based on Poisson (Beaumont)", developed by Dr. Martin Schechter from the department of Health Care and Epidemiology at U B C , power calculations were performed. The proposed study would be able to detect a relative risk of 1.45 for all childhood cancers combined, a relative risk of 1.9 for all leukemias, and a relative risk of 2.1 for all brain cancers, with 80% power. Individual histologies would be detectable at higher relative risks. Appendix A provides a table of the power calculations as calculated at the proposal stage of the thesis. Ethics approval for the proposed study of childhood cancer among the offspring of chlorophenate exposed sawmill workers was obtained from the U B C ethics committee. Operating grant applications were prepared and submitted for peer review to the National Health and Research Development Program (NHRDP), the National Cancer Institute of Canada (NCIC) and the B C H R F in their autumn 1993 competition. Both N H R D P and B C H R F applications were successful, although only the B C H R F award was accepted; #234(94-1).  59  3.2 Power Calculations Prior to embarking on the study, new sample size calculations were performed using person-year data for the years 1969 - 1993 (i.e. the actual study period) and from the eleven chlorophenate-using sawmills (i.e. the mills used in the analysis). Results indicated that this study would be able to detect a relative risk of 1.44 for all childhood cancers combined, 1.84 for all leukemias and 2.17 for all brain cancers with 80% power (virtually the same as the previous power calculations, as indicated above). Calculating minimum detectable relative risks by sex produced the following results. Among the girls, this study has 80% power to detect a relative risk of 1.66 for all cancers combined, 2.30 for all leukemias and 2.70 for all brain cancers. Among the boys, this study has 80% power to detect a relative risk of 1.63 for all cancers combined, 2.17 for all leukemias and 2.62 for all brain cancers. These calculations address the situation where all person-time is used in the analysis and not where a portion of person-time is analysed e.g. assessing risk by window of exposure. Hence, in the sub-cohort analyses the power of the study will be diminished and this is a considerable limitation to the study.  3.3 Data Sources The British Columbia Cancer Registry The British Columbia Cancer Registry was established January 1, 1969 and receives notification of all new cases of cancer from pathology departments of provincial hospitals, death certificates and cancer treatment centers. The registry is estimated to be 95 - 98% complete (personal communication with R. Gallagher). Ascertainment of childhood cancers may even be better (i.e. close to 100%) as all children with cancer are  60  treated in specialized institutions that send their pathology reports to the B C Cancer Registry (personal communication with R. Gallagher). As well, there is reciprocal reporting between provinces. Few British Columbian children would go to the United States for treatment because of prohibitive costs. Hence, there is good reason to be confident that virtually all childhood cancer cases resident in British Columbia after 1969 appear in the B C Cancer Registry.  Computerized Record Linkage: Background The initial development of computerized record linkage was reported in. 1959 (Newcombe et al. 1959). A formal theory was proposed ten years later (Fellegi and Sunter 1969). Since then, computerized record linkage has been refined for use with a generalized computer system (Howe and Lindsay 1981) and has gained widespread acceptance as a major tool for use in epidemiologic research involving the follow-up of large populations (Howe 1985). Record linkage refers to the comparison of two or more records containing identifying information to determine the probability that they refer to the same person. There are two types of computerized record linkage: direct (or administrative) which involves the linkage of two files containing unique identifying information, such as Social Insurance Numbers, and probabilistic, which involves using non-unique identifying information, such as full name, full date and place of birth. The sawmill worker study used probabilistic linkage matching on full name and birth date between persons in the cohort file with persons in the death and cancer incidence outcome files. Probabilistic linkage is an iterative process that compares individual identifying items such as full name and date of birth from one record with those of  61  another. The process produces a "linkage odds" which is a ratio of the frequency of observing the outcome among linkable pairs to observing the outcome among unlinkable pairs (Howe and Lindsay 1981). By summing the odds in favour of a match, a "weight" is produced. The weights computed are specific for each identifying variable. Summing the weights for individual identifying variables provides a total weight that is a measure of the relative odds in favour of a true link between two comparisons in the data sets. After weights have been assigned for each matched pair, a decision is made to determine threshold values that divide the data into three sets: weight values above which can be considered definite links, a range of weights that are considered possible links ("grey zone"), and weights which will not be accepted as true links. Possible links are then resolved manually by checking identifying data that was not used in the computerized linkage, such as place of birth.  Computerized Record Linkage: Generating the Offspring Cohort Tracing the offspring cohort was the most time-and resource consuming aspect of the project. Offspring were identified via probabilistic linkage based on the Generalized Iterative Linkage computer system (Howe and Lindsay 1981). Machine-readable records of British Columbia (BC) marriages and births are maintained under the jurisdiction of the B C Division of Vital Statistics (DVS). The probabilistic linkage was performed at the D V S in Victoria using the computer program Automatch (Data Star Inc.). The process of ascertaining the offspring cohort involved three linkages: (i) sawmill cohort to provincial marriage files, (ii) provincial marriage to birth files, and (iii) sawmill cohort directly to provincial birth files. The sawmill cohort register contained the following  information: full name, date of birth, country of birth, wife's first name (in some instances), location, job title(s), duration of employment and exposure estimates. Marriage records contained information on husband and wife's full names, age at marriage, place of birth, place of marriage, parent's full names and previous marital status. Birth records contained information from the "Physicians Notice of Birth (PNOB)" which includes child's date and place of birth, gestational age, sex, birthweight, mother's and (if available) father's full name, age, marital status and address. The sawmill cohort file was first linked to the marriage file to obtain valuable information on mother's full name (mother's full name appears on all birth certificates;-whereas father's -  full name is sometimes incomplete). The cohort file with any additional information from -. the marriage file was then linked to the birth files. The matching criteria used for the probabilistic linkage was fathers' full name and date of birth, and, when available, wife's first name and date of birth. This produced an offspring cohort containing all cohortlinked children born between 1952-1988 in British Columbia who were registered in the Division of Vital Statistics.  The linkage provided weights for matches that the U B C research team then assessed and decided upon a weight above which the link was considered a definite match and below which was not considered a match. A "grey zone" of values between these weights was assessed manually, by deciding upon rules for matches which could be evaluated using cohort information that was not linked with the marriage or birth file. In order to evaluate the validity of this procedure, a sample of 60 "grey zone" cases were  63  manually resolved by the research team at U B C , then these resolutions were checked at the DVS. Fewer than 1% invalid links were found.  Because provincial birth records were used in the linkage, children born outside the province were excluded from the study. This exclusion has the potential to create a migration bias. However our cohort mortality data (based on linkage to the Canadian Mortality Data Base) indicates that 96% of the deaths occurred in British Columbia, hence one would assume that most workers resided in British Columbia throughout their working lives and their children were born in this province. As well, the in-migration rate in this province has exceeded the out-migration rate during the past four decades. Children born in British Columbia but who move out will cause a conservative bias since they will be retained in the denominator. There is no reason to believe that there would be a change in cancer risk for out-migrators. Reciprocal reporting for children treated outside the province occurs, hence cancer cases living in British Columbia but being treated outside the province will be accounted for. Also excluded were children born prior to their father achieving one full year of employment. The rationale for this is that the duration of spermatogenesis and gestation combined is one year, hence only pregnancies and births occurring thereafter would definitely have the potential to involve paternal chlorophenate exposure.  Computerized Record Linkage: Generating the Cancer Offspring Cohort Linkage of the offspring cohort file to the Cancer Registry was carried out by W. Threlfall, a computer programmer, and this candidate at the Cancer Control Research 64  Unit of the British Columbia Cancer Agency (BCCA). Two computer programs were used to link the offspring cohort to the British Columbia Cancer Registry. The first was written by a computer programmer at the B C C A , based upon the original algorithm by Dr. Geoffrey Howe (Howe and Lindsay 1981) which is derived from the Fellegi and Sunter algorithm (Fellegi and Sunter 1969). The program was successfully used to link the sawmill worker cohort to the Cancer Registry for the study on mortality and cancer incidence among chlorophenate-exposed sawmill workers (Hertzman et al. 1997). The second program, Automatch by Datastar Inc., became available to the Cancer Control Research Unit at the time the first linkage was being carried out. To verify the programmer's linkage, Automatch was used to link the offspring cohort to the Cancer Registry. Included in the linkage were all children born in British Columbia between January 1, 1952 and December 31, 1988 less than twenty years of age at diagnosis and diagnosed between. 1969 - 1993. Probabilistic linkage was used to link full name (first, middle, last) and date of birth (day/month/year).  Manual verification of all links was carried out using supplementary cohort information to resolve uncertainties. The process involved the following: the computer generated weights for individual fields (e.g. last name or month of birth) and summed them across all fields for each match to provide a total weight for each potential link. Supplementary information, such as address or Health Unit associated with address, also appeared on the output, however it was not assigned a weighting value. In some instances, this supplementary information was used to help verify a link. For instance, i f there was uncertainty about assigning "definite" to the link, the address of the case, as  65  indicated on the Cancer Registry file, was compared to the mill location as indicated on the offspring cohort file as well as the Health Unit and/or School District. If the locations were similar between files, the potential match was deemed a definite match. The output of the linkage contained the following information: (1) from the Cancer Registry: B C C A ID number, full name, date of birth, date of diagnosis, death certificate number (if applicable), death date, primary cause of death, secondary cause of death, address. (2) from the Offspring Cohort File: sawmill worker ID (which indicated mill location), full name, date of birth, School District and Health Unit.  The linkage identified one hundred sixty-seven potential matches among the boys and girls. These potential matches were linkages between the two data files with sufficient matching information for the computer to identify the persons in both files as a potentially "true" match, i.e. the same person. Weights were assigned to these potential matches and then the candidate and the computer programmer, W. Threlfall, independently reviewed results of the linkage to determine true linkages, uncertain linkages and definite non-linkages. Twenty-eight links were identified as potentially "true" matches among the boys and thirty links as potentially "true" matches among the girls based on concordance with full names and dates of birth and, to a lesser degree, place of residence. To resolve the linkages that could not be confirmed by examining the output, supplementary information was obtained by reviewing charts, pathology reports and the death file from the Cancer Agency. Further investigation of the 28 male matches revealed that four were twenty or twenty-one years of age and therefore ineligible, three  66  were recorded in the cancer registry but did not, in fact, have cancer, two were diagnosed before January 1, 1969, hence excluded. The histology of one was not coded as malignant (ie recorded in the cancer registry with a '73 " at the end of the four-digit histology code) but coded with a '71" at the end of the histology code, referring to a neoplasm with uncertain behaviour and therefore was excluded from the analysis. Hence, in total there were eighteen "true" linkages. Among the girls, 30 potentially true matches were identified. Of these, one had a father who worked at a non-chlorophenate using mill and therefore was ineligible, one was diagnosed prior to 1969, hence excluded and six whose histologies were not malignant (five cases of in-situ carcinoma, '72" and one of uncertain behaviour, " / l " ) , thereby leaving twenty-two "true" linkages. The data manager from the sawmill worker cohort study (R. Hershler) then reviewed the information used to accept or reject the linkages. Once the true matches were finalized, B C C A registry data was viewed to elucidate site and histology of cancer.  Several studies have shown associations between certain congenital anomalies and childhood cancer. For instance, children with Down syndrome are at increased risk for leukemia. As a result, it was decided that the data file developed for use in the nested case-control analysis should contain data from the Physicians Notice of Birth (PNOB) regarding adverse outcomes at birth, i.e. birth defects, in order to identify relationships between cancer cases and birth defects. Studies are also emerging that have examined risk for cancer in the parents of childhood cancer cases (Olsen et al. 1995, Hemminki and Viattinen 1997) and in the offspring of childhood cancer survivors (Sankila et al. 1998). Hence, also investigated was whether or not the fathers of the childhood cancer cases had  67  been diagnosed with cancer. Unfortunately, the data available did not permit the assessment of maternal cancer.  To ensure confidentiality, no name-based information left the Cancer Control Research Unit. Data leaving the Unit and used in the analysis identified cases by sawmill worker ID only.  3.4 Study Design  Previous Research The method of retrospective exposure assessment developed for the study of mortality and cancer incidence among a cohort of sawmill workers exposed to chlorophenate wood preservatives (Hertzman et al. 1997) was as follows:  Validity and Reliability of Methods For eighteen months prior to embarking on a full retrospective cohort study of mortality and cancer incidence among British Columbian sawmill workers, a method of retrospectively reconstructing exposures was tested for reliability and validity in feasibility studies (Hertzman et al 1988, Teschke et al 1989). This was an important aspect of the study because no quantitative exposure data had been routinely collected during the study period, i.e. 1950 to 1985. The only exposure related data available was job title information.  68  In the first phase of the study evaluating the validity and reliability of the retrospective exposure assessment, 10 randomly selected workers (raters), each having more than five years seniority, were interviewed separately. The sawmill selected for this pilot test began applying chlorophenates in the 1950s using both diptanks and sprayboxes and, at the time of interviewing, employed approximately 200 male workers. The task of the raters was to estimate the frequency and duration of exposure to chlorophenates in all 59 jobs at the mill. The categories of frequency included daily, several times a week, several times a month, several times a year or never. If frequency was rated as more than never, an estimate of duration was requested according to the following categories: more than seven hours per day, 5-7 hours per day, 3-5 hours per day, 1-3 hours per day, and up to one hour per day. Frequency was converted into days of exposure per year (daily equaled 250 days per year, several times a week referred to 150 days, several times a month represented 36 days, several times a year equaled 3 days and never was rated as zero). Duration was calculated from the midpoint of the stated range, i.e. 5 - 7 hours per day was assigned a duration of six hours. Inter-observer agreement between workers' exposure estimates was measured by intraclass correlation coefficients (ICC) between individual raters and for all raters. The ICC proved to be excellent for all raters (0.92) and good for single raters (0.54).  In the second phase of the study, validity was assessed by examining urinary chlorophenate levels of all workers in the mill and comparing them with results of the raters' exposure estimates. A detailed description of the procedure and urinary analysis can be found in Hertzman et al. 1988. At the time of sampling, each worker completed a 69  self-administered questionnaire about his job and personal protective equipment used. As climatic conditions were suspected to affect exposure and absorption of chlorophenates, samples were taken in the summer (hot and dry) and autumn (cool and wet). Samples taken in the summer differed significantly from those taken in the fall (mean level 239 micrograms/1 and 111 micrograms/1 respectively). Both seasons produced chlorophenate levels that were approximately log-normally distributed. For each of the 172 workers (86% of the workforce) who participated in the study, a personal exposure estimate was calculated based on job titles he held. Pearson correlation coefficients tested associations between raters' estimates of exposure (frequency, duration and an index combining the two) as assessed in the first phase of the study and the logarithm of the measured urinary chlorophenates. Correlation coefficients were always greater than 0.65. Comparison of urinary chlorophenates with worker raters showed errors were distributed heterogeneously, therefore any error was expected to be random, not biased. These results indicated that the retrospective exposure assessment was acceptably reliable and valid and, therefore, the study team felt confident to embark on the larger study.  The second study (Teschke et al. 1990) was undertaken to assess whether estimates of exposure (frequency, duration and route of exposure for every job in the sawmill) by nine senior workers (raters) correlated with those of industrial hygienists. Three pairs of industrial hygienists were interviewed. They included one pair who inspected sawmills for a regulatory agency, one pair who worked predominantly in the sawmill industry and another pair who represented industries other than sawmills. The mill elected for study  70  was not the same sawmill evaluated in the first study. It employed 268 workers in 73 job titles and used sprayboxes to apply sodium tetra and penta-chlorophenate. The nine raters and the hygienists answered questions regarding frequency, duration and route(s) of exposure. Reliability was assessed for the senior raters and the hygienists. High group intraclass correlation coefficients resulted: 0.71 for the senior workers and 0.57, 0.67 and 0.81 for the three groups of industrial hygienists. To assess validity, workers were asked to supply urine samples and to complete a self-administered questionnaire about their job(s) and personal protective clothing worn as well as answering the same questionnaire given to the senior raters and hygienists. 225 workers (84% of the workforce) participated. Results of the validity analyses indicated that the proportion of variance explained (R ) did not differ significantly between industrial hygienists (0:24, 2  0.26, 0.22), senior workers (0.24) and self-reporting volunteers (0.17). This suggests that senior workers' job exposure estimates should be similar in quality to industrial hygienists and possibly better than those provided by the individual workers' self reports. Hence, results of both feasibility studies supported the selection of methods proposed for the retrospective cohort study.  Personal Identifying Information The main data collection was undertaken between 1988 - 1990. The protocol involved sending coders with portable computers into fourteen sawmills to extract both personal identifying and job history information from company personnel records. A l l males employed for at least one year between January 1, 1950 and December 31, 1985 in one of 11 large non-cedar softwood lumber mills using chlorophenates (located on  71  Vancouver Island and the lower mainland) were included in the cohort. The mills had each used chlorophenates for extended periods of time, with the majority beginning its use in the 1940's and 1950's and continuing until the 1980's. Sawmill workers who had an otherwise similar work environment in three sawmills that did not use chlorophenates (located in the interior of the province) were also included. Cleaning the data involved considerable human and financial resources. It was performed with an automatic errorchecking program that coders ran at the end of each day at the mill and at U B C on the mainframe computer. The final study cohort included 26,487 male sawmill workers: 23,829 in chlorophenate-using mills and 2,658 in non-user mills.  Retrospective Exposure Reconstruction Once the sawmill cohort was assembled, and the method to retrospectively assess chlorophenate exposure in the sawmills was shown to be reliable and valid, a full cohort study was undertaken. It involved the following steps. Initially, mill records were reviewed and interviews with experienced workers, managers, engineers, and chlorophenate distributors were conducted to extract information about technology and formulation changes, and locations in the milling process where chlorophenates were used, thereby reconstructing each mill's history of chlorophenate application. Discrete time periods within which exposure conditions remained constant (exposure constant time periods - "ECTPs") were then established. Each mill experienced approximately four ECTPs throughout the study period. To ensure comparability between mills and to reduce the data file to a manageable level, all recorded job titles from a given period (sometimes as many as 700) were collapsed into a standardized format resulting in no  72  more than 100 job titles containing similar opportunities for exposure. The collapsed titles were re-attached to individual worker files. Thus, workers' employment histories could be tracked over time and easily compared. Subsequently, job titles that existed for each time period in every mill were rated by 10-15 experienced workers (who had worked in the mill during the ECTP in question) for frequency and duration of exposure to chlorophenates. The fifteen estimates of chlorophenate exposure for each job title in individual ECTPs were pooled to obtain an estimate of exposure for each job title. A n index of cumulative chlorophenate exposure for each worker was devised based on his job history and categorized as < 120 hours, 120- 1,999 hours, 2,000 - 3,999 hours, 4,000 - 9,999 hours and > 10,000 hours.  Present Study: Cohort Design and Analysis Initially, a cohort design was used that enabled rates of childhood cancer to be calculated among the offspring cohort whose fathers were occupationally exposed to chlorophenates (i.e. fathers working at the eleven mills where chlorophenates were used) and compare these rates to rates of the British Columbia population. Adjustment was accomplished by indirect standardization wherein the expected numbers of outcome events were based on "person-years at risk" of the children, categorized by individual calendar year (1969 - 1993) and five-year age categories (0 -4, 5 - 9, 10 - 14, 15 -19), then multiplying the person-years at risk by the respective incidence rates for B C children at those specific age and calendar combinations. Standardized incidence ratios (SIRs) were estimated from the ratio of the observed number of childhood cancer cases  73  in the offspring cohort and the expected number of cases from the B C population. This analysis did not consider any person time contributed by the fathers, nor did it incorporate any measures of paternal chlorophenate exposure. Hence, results will indicate risk associated with father's employment at chlorophenate-using sawmills but not risks associated with varying levels of exposure. Dose-response analyses will be incorporated into the internal analysis.  Person - Y e a r Calculations For each year from 1952 (the first year for which birth data was available for this study) to 1988 (when chlorophenate use was ceased), the number of births occurring among sawmill workers employed at chlorophenate-using mills was provided by gender . 1  Person-year contributions were tabulated from each child's entry into the cohort (i.e. their birth year) until cancer diagnosis, termination of the follow-up period (twenty years of age), or study end (December 31, 1993). Once a child was diagnosed with cancer, he/she was censored and no further person-years were contributed by that person.  No information is available with respect to childhood deaths so it was impossible to censor person-years as a result of accidental or other causes of death. However, deaths among children are infrequent, therefore it is unlikely that a significant reduction in person-years would occur as a result. Also, no information is available regarding  Although the cohort analysis planned only to examine offspring of chlorophenate-exposed sawmill workers, birth data was available for the offspring of children whose father's worked at non-user mills so person-year calculations and SIR analyses were also carried out for the children of sawmill workers employed at both chlorophenate-using and non-user mills, i.e. all sawmills. 1  74  emigration of the sawmill worker families from B C . However, in-migration has greatly exceeded out-migration during the past four decades, and data from the sawmill worker cohort suggested that 96% of all sawmill workers died in B C , hence it is unlikely that many families moved out of the province. Also, the families that did leave B C are likely to be short-term sawmill workers with less exposure to pentachlorophenols. The inability to censor children who die or leave the province would artificially increase the person-years at risk of the offspring cohort, resulting in a slightly higher number of "expected" cases, and a potential reduction in the SIRs. Person-year calculations were computed using Excel version 4.0 (Microsoft Excel). The B C Cancer Registry commenced in 1969, hence gender-specific sums of person-year contributions in five year age categories (0 - 4, 5 - 9, 1 0 - 1 4 , 15 - 19 as well as the age groups 0-14 and 0 19) were tabulated from then each year until December 31,1993.  Person - Y e a r Calculations until Offspring Cohort all Achieve 20 Years of Age Results of the cohort analysis identified a few increased risks for childhood cancer, based on small numbers and having confidence intervals that included unity (see Results). A n attempt was made to determine whether further follow-up of the cohort would strengthen these observations. Person-years left in the cohort were calculated based on all children achieving twenty years of age (final year of follow-up would be 2007). 61,423 person years are left if the follow-up is continued until 2007. Adding this to the 283,579 person years already accounted for from 1969 - 1993 results in a total of 345,002 person years of follow-up from 1969 - 2007. Dividing the person years already accounted for (283,579) by the total person years that would occur between 1969 - 2007  75  (345,002) results in 82% of the potential person years already having been accounted for. Further follow-up would mainly involve offspring in their late childhood and teen years, hence it is unlikely that new cases of the common childhood cancers (leukemia and brain) would be identified. Appendix B provides a crude estimate of the expected number of new cases and SIRs that might occur if the cohort were followed until 2007. Hence, the effort involved in following up the offspring cohort until all children attain 20 years of age, in all likelihood, will not alter the results of this study.  Calculation of Provincial Incidence Rates Numerator data: The British Columbia Cancer Registry provided total number of malignant cases (those with "/3" at the end of the histological classification) by sex, age at diagnosis (0 - 19), year of diagnosis (1969 - 93), site and histology according to the I C D - 0 (International Classification of Diseases - Oncology, WHO Publication, 1  st  edition 1976) and ICD-0(2) (International Classification of Diseases - Oncology, W H O Publication, 2 edition 1990). The ICD-0 system is easily converted to the ICD-0(2) nd  system and vice versa.  Denominator data: Mid-year population by sex, five year age categories (0 - 4, 5 - 9, 10 -14 and 15-19) and individual calendar year periods (1969 - 93) were abstracted from Statistics Canada census data for the province of British Columbia. When available, postcensal population estimates were used. Census data has generally been collected every five years and estimates of the population between censuses are projections based on previously collected data. Postcensal estimates are corrected based on data collected  76  subsequently, thus provide a more accurate estimate of the yearly British Columbian population. The B C Cancer Registry also calculates childhood cancer incidence for their annual report however, the report is based on the last complete year of data which is usually two years behind. Hence, the 1996-97 report used 1995 cancer figures. These incidence rates are not corrected and may therefore disagree slightly with the ones calculated for this thesis.  Incidence rates per 100, 000 population were calculated by sex, individual calendar year periods, 5-year calendar periods, 5-year and 0-19 year age categories for all cancers combined and for the following sites and histologies.  77  Table 1. Cancer sites and histologies (ICD-O and I C D - 0 ( 2 ) classifications) used to calculate provincial incidence rates.  Site Leukemia  169  Histology C42  A l l malignant leukemias  Acute lymphoblastic (ALL) Brain & Nervous  191  System  Lymphoma  196  C71  9821/3  A l l malignant brain  Astrocytoma  9400/3  Ependymoma  9391/3  C77  A l l malignant lymphomas  Hodgkin's disease  9656/3  non-Hodgkin's lymphoma  9630/3  Bone  170  C40  A l l malignant bone  Connective tissue  171  C49  A l l malignant connective tissue  Eye  190  C69  A l l malignant eye  Liver  155  C22  A l l malignant liver  Kidney  189  C64  A l l malignant kidney  Ovary  183  C56  A l l malignant ovary  Cervix  180  C53  A l l malignant cervix  Skin  173  C44  A l l malignant non-melanoma skin  Thyroid  193  C73  A l l malignant thyroid  78  Analysis: Standardized Incidence Ratios (SIR) Risk for developing childhood cancer was assessed by site and/or histology using the standardized incidence ratio (SIR). Hence for the SIR analysis, all sites and histologies identified through the offspring cohort - B C cancer registry linkage had SIRs calculated. The following histologies had SIRs calculated: A L L , astrocytoma, ependymoma, Hodgkin's disease and non-Hodgkin's disease. Gender-specific SIR calculations were performed on an Excel spreadsheet (Microsoft Excel 4.0). Ninety-five percent confidence intervals (95% CIs) were calculated by the method of Breslow and Day (Breslow and Day, 1980).  Nested Case-Control Design and Analysis The initial analysis (external cohort) produced largely negative results. A few elevated risks were seen but these were based on small numbers and their confidence intervals always included unity. To minimize differences associated with using an external reference population, an internal analysis was proposed which would ascertain whether the extent and timing of exposures affect the relative odds of developing childhood cancer. Information in the sawmill worker and offspring databases would permit both a nested case-control analysis using odds ratios (ORs) as the risk measure, and an internal cohort analysis using the Mantel-Haenszel rate ratio estimator (RR,^). Which would be better - an internal cohort analysis or a nested case-control one? A cohort analysis would require generating considerably more data including a new cumulative exposure variable for all father-child combinations (19,675 children were born to the exposed sawmill worker cohort). Exposure would be tabulated from father's  79  first employment until cancer diagnosis, or the child reached twenty years of age, or study end. As well, windows of exposure variables would also have to be generated for many more father-child units, resulting in a considerable expenditure in time and money. A less resource-consuming approach would be a nested case-control study, whereby the case population would represent all cases of childhood cancer occurring among offspring of the sawmill worker cohort between January 1, 1969 and December 31,1993 and the control population, a random sample of children who did not develop cancer. The linkage identified forty cases of childhood cancer; with five controls per case selected, new data on cumulative exposure and windows of exposure would need to be generated for only 240 persons. In light of the largely negative results of the SIR analysis, and reduced time and energy involvement needed for a nested case-control study, this approach was adopted. The estimated odds ratios (ORs) were calculated using logistic regression.  Unlike the SIR analysis which included only the offspring of chlorophenate exposed sawmill workers, the proposed nested case-control analysis was originally planned to include offspring of workers from both chlorophenate using and non-user mills. However, between the time of the study inception and the execution of the internal analysis, a decision was made to analyse only the offspring of sawmill workers who had been employed at chlorophenate-using sawmills. The main reason is that only one case of cancer was identified among the offspring of sawmill workers employed at the nonuser mills. Using this one case in the unexposed category of a nested case-control analysis would result in unstable estimates. As well, demographic differences exist 80  between the two types of mills. The three interior (non-user) mills had later start-up dates and less follow-up time than the eleven coastal (chlorophenate-using) mills. Differences due to geography, climate and milling processes may vary considerably between the two groups of mills so selecting controls from the whole pool of offspring would not provide as homogeneous a group as selecting controls from only the offspring of chlorophenate-exposed sawmill workers. For these reasons, a study of birth defects among the sawmill worker cohort also only selected cases and controls from the offspring of workers employed at chlorophenate-using sawmills (Dimich-Ward et al. 1996).  Case and Control Selection (1) Cases: A l l cases of cancer (age 0 - 1 9 inclusive) appearing in the B C Cancer Registry between 1969 - 1993 and identified as "true cases" in the offspring cohort - B C Cancer Registry linkage (described in section 3.3) were included. (2) Controls: Five controls per case were randomly selected based on sex and year of birth using incidence density sampling. Incidence density sampling involves a random selection of controls from the person-time experience that generated the cases. This requires controls to be selected from the set of persons "at risk" at the time of case diagnosis, i.e. all persons alive and at risk for, but free of disease, at the time the case was diagnosed (Checkoway et al. 1989 ppl76). Incidence density sampling allows one to estimate the incidence rate ratio without the need for a rare disease assumption (Miettinen, 1976). As an example of incidence density sampling in an occupational nested case-control study of bladder cancer, a case was diagnosed at age 61. Controls 81  would be randomly selected among individuals from the same workplace still alive and "at risk" for bladder cancer at age 61. Cumulative exposure for the controls would be calculated from their entry into the study until age 61. In this study, it is not possible to tell i f a control was still alive at the time of his/her corresponding case's diagnosis (apart from data provided on the PNOB, no further health information on the child was available in the available data sets). As childhood death is uncommon, it is assumed that the likelihood of a control being alive at the time of case diagnosis is high so that matching on year of birth is the same as matching on age at diagnosis.  Very little has been published with respect to appropriate inclusion and exclusion criteria for control selection in occupational nested case-control studies using workers' offspring as the sampling frame. In a paper presented at the joint meeting of the University of British Columbia and the University of Washington's Epidemiology and Occupational Hygiene programmes (January 1997), Marion and Ward described a survival analysis approach to control selection in a study of reproductive outcomes at birth among offspring of sawmill workers. They suggest that critical events occur only once in survival analysis, thus when a birth-related abnormality such as a birth defect has occurred, the subsequent experience of that parent should not be used. They recommended the following: (1) once a case has occurred, subsequent births to the same parent should be excluded as either cases or controls, (2) siblings of controls can serve as controls unless an affected birth (or disease diagnosis) has intevened, 82  (3) multiple births are acceptable, by considering each multiple birth as a single event, and, (4) siblings of a case can never be a control for that case, but earlier births to the parent of an affected child are permitted to be controls for earlier cases. The same criteria were applied to the control selection for this thesis. The control selection resulted in two instances where siblings served as controls (but not for the same case) and one instance where the sibling of a case was selected as a control. The birth of the case (1972) occurred before the birth of the control (1974) but the diagnosis of the case occurred after the birth of the control (1975).  Indices of Exposure The two research questions to be answered in the nested case-control analysis were: (1) Does risk for developing childhood cancer increase with increasing paternal cumulative exposure to chlorophenates? and, (2) Do windows of exposure exist that affect the relative odds of disease development? Data on exposure among the fathers of the forty cases and two hundred controls was provided by R Hershler, (data manager for the sawmill worker cohort study) in a FoxPro database (Microsoft Corporation, Redmond, Washington, USA). Much of the information in the database had been generated for the birth defects study (Dimich-Ward et al. 1996). Included was data from the PNOB, such as birth date, gender, birthweight, gestation, total number of births, total number of pregnancies, complications, birth defects, mother's and father's age. Dependent variables of interest created for this study included (see also figure 2):  83  (1) Window 1 = paternal occupational exposure to chlorophenates during the period from commencing employment until 90 days prior to conception (exposures during this period could alter sperm stem cells). (2) Window 2 = paternal occupational exposure to chlorophenates during the ninety days prior to conception (exposures during this period could alter spermatogenesis). (3) Window 3 = paternal occupational exposure to chlorophenates during pregnancy (paternal exposures during this period could result in transplacental transmission of the chemical). (4) Window 4 = paternal occupational exposure to chlorophenates from birth of offspring until diagnosis (exposures during this period would indicate a direct effect of the chemical on the child). (5) CumExp = paternal occupational exposure from commencing employment until cancer diagnosis of the offspring.  Figure 2 . Time line describing cumulative exposure to chlorophenates and windows of exposure.  CumExp  ->  Window 1  Window 2  Window 3  Window 4  _t  1,  Father starts work  Spermatogenesis Pregnancy Child is born Cancer diagnosis.  o  1  2  1  3  1  4  84  The data was imported from the FoxPro database into Stata Release 5.0, (StataCorp, 1997) and reviewed to ensure the information provided was internally consistent. A few variables needed recoding to accommodate Stata, including case-control status and sex (recoded from 1 and 2 to 1 and 0). A l l cumulative exposure variables and age variables were presented as continuous data. These variables were kept in the database but, also, new categories of exposure and age were created. The analyses were also run using SPSS, Statistical Package for Social Sciences, version 7.5 (SPSS, Inc.) to verify the results.  To answer the question regarding effects of cumulative exposure, one could either fit continuous data or categorical data into the logistic regression model. Both methods were attempted. A model incorporating cumulative exposure as continuous assumes that the risk associated with exposure is linear. To avoid this assumption, the analysis was initiated using categories of cumulative exposure. By using categories, there is no assumption of a monotonic/linear change with increasing exposure. Several considerations were involved in deciding the appropriate number of categories. One could adopt the categories of cumulative exposure devised for the study of mortality and cancer incidence among the sawmill worker cohort (Hertzman et al. 1997) and subsequently used in the fertility study (Heacock et al. 1998). In these studies, cumulative exposure was divided into five categories: < 120 hours, 120 - 1,999 hours, 2,000 - 3,999 hours, 4,000 - 9,999 hours and > 10,000 hours. However, spreading forty cases into five categories would provide cells with small numbers. For instance, only one case would fall into the lowest exposure category. Clearly five categories is too many with only forty cases. Two categories would not permit modeling of a dose-  85  response relationship. Three categories of cumulative exposure (low, medium and high) were thus created to allow estimation of risk associated with increasing exposure. Selecting the categories was somewhat arbitrary: criteria included having at least five cases in each cell, having paternal chlorophenate exposure in each cell (i.e. the lowest category not only containing offspring whose fathers' had zero exposure hours), and attempting to create cells that would reasonably represent low, medium and high exposure categories. "Tertiling" the distribution would not have been appropriate. Each "window of exposure" contained cells with no exposure; Window 1 had twenty cases where the father had no cumulative chlorophenate exposure, in Window 2, there were 132 instances of no exposure, in Window 3 there were 124 births whose fathers had no exposure, and in Window 4, there were 57 instances of no exposure and another 57 of less than one hour of exposure (see tables 12-15). In creating the categories of high exposure, an attempt was made to include predominantly offspring whose fathers' chlorophenate exposure showed an exponential increase at the upper end of the exposure range. Hence, the high category contains fewer children than the low and medium categories in an attempt to aggregate only the highest paternal exposures. The categories of low exposure often contain more children than the medium and high categories because they must be sufficiently large to contain all children whose father's had no exposure.  For the analysis of cumulative exposure, to obtain at least five cases in the lowest category, a minimum of 1250 hours of exposure was needed. The highest exposure category was selected to reflect the highest category from the previous studies, i.e.  86  greater than 10,000 cumulative hours. Hence, the medium exposure category contained cases whose paternal cumulative exposure to chlorophenates was between 1250 and 9,999 hours.  Trend was tested three ways; (1) by assigning a value to each of the three categories equal to the midpoint of the values in the exposure category, i.e. for the cumulative exposure analysis, the category 0 -1250 was assigned the value 625, (2) by assigning the mean value of hours of exposure in each category, i.e. the category 0 -1250 had a mean of 592 hours of exposure, and (3) by using 1, 2 and 3 to represent three equally spaced values. Models were fitted with the new trend variable as a linear effect. A l l three trend tests produced similar results, hence significance of the effect from only the first test is reported.  As brain cancer and leukemia were cancers of a priori interest and this study identified nine and eleven cases respectively, an analysis by these sites was also performed. The data was divided into low and high categories of exposure (case numbers were too small to permit three categories of exposure) and analyses run on these two categories. Trend testing was not, therefore, possible. Individual odds ratios and confidence intervals are reported.  Determining a Baseline Model to Assess Cancer Risk The dependent (dichotomous) variable was case-control status. Controls were frequency-matched to cases on the basis of sex and birth year, therefore these variables 87  were included in the model, although the associated effects cannot, of course, be estimated. The covariates of interest were mother's age, father's age, birthweight, gestation, total births and child's age (in four five-year categories). The importance of age, birthweight, gestation and total births was unknown, so a stepwise logistic regression analysis was carried out and variables with p-values greater than 0.05 were sequentially dropped from the model. This process resulted in the covariates child's age, mother's age, father's age, birthweight, gestation and total births being dropped from the model. Age is a covariate that is often controlled for in statistical analyses because of its role as a confounder. The potential for confounding by parental age was questioned. Parental age does increase risk for certain peri-and postnatal birth outcomes but little.has been published on its risk associated with childhood cancer. A comprehensive book chapter on cancers in children evaluated the literature on maternal age and risk of childhood cancer and concluded that there was insufficient evidence to implicate increasing maternal age with risk for childhood cancer (Chow et al. 1996). No evidence for associations between paternal age and risk of childhood cancer was found. As both maternal and paternal age were not significant predictors in the stepwise logistic regression, and the published literature does not identify parental age as a risk factor it was agreed that parental age did not need to be adjusted for in the model. Hence, the baseline model included only sex and birth year. Therefore, for each exposure category being examined, the model included the baseline model as well as the exposure category.  88  C H A P T E R 4.0 R E S U L T S  4.0 Offspring Cohort Description From the original cohort of 26,487 male sawmill workers, 21,390 births occurred after the father had achieved one full year of employment during the period 1952 - 1988. Children born prior to their father commencing employment in the sawmill were excluded from the study. Among the 23,829 men employed at chlorophenate-using mills, 19,674 births were identified, and from the 2,658 men employed at non-user mills, there were 1,716 births. In the families of men employed at the chlorophenate-using mills, forty cases of childhood cancer were diagnosed, and one case was diagnosed in the families of men employed at non-user mills. For 48.3% of the fathers, only one child was born during the study period. Figure 3 describes the studies whose data facilitated the childhood cancer study.  Characteristics of the offspring cohort are described in Table 2. Of the 21,390 children born to the sawmill worker cohort, there were 10,992 boys (51.4%) and 10,398 girls (48.6%o) for a birth ratio of 1.06:1, which approximates the provincial norm (1.05:1). Mill type did not affect the birth ratio. The offspring cohort was followed up for a diagnosis of childhood cancer for 283,579 person years; 259,919 person years among the offspring whose fathers' worked at chlorophenate-using mills and 23,660 person years among the offspring whose fathers' worked at non-user mills. Another 61,423 person years would be added if the offspring cohort were followed until they all reached twenty years of age (the year 2007). By December 31, 1993 (study end), 82% (283,579/345,002) of the total  89  potential person years had been accounted for: 83% in the chlorophenate-using mills and 73%) in the non-user mills. The non-user mills had later start-up dates and a younger workforce than did the chlorophenate using mills, hence their children were born later and did not contribute as much person time into the study by study end as did those of chlorophenate using sawmill workers. This study reports descriptive data from all offspring of the sawmill worker cohort but analyses were performed only on offspring of the chlorophenate-using sawmill workers.  90  Figure 3. Diagrammatic representation of the BC sawmill worker cohort study, and subsequent studies examining reproductive outcomes of the sawmill workers.  Cohort of male sawmill workers STUDY 1  Mortality and Cancer Incidence  (a) employed in 11 large non-softwood mills where  (Hertzman et al. 1997) Linkage  chlorophenates were used,  Chloro-using -> N=23,829  (b) employed in 3 mills that  Pyrs=583,190  Non-using N=2,658 Pyrs=41,280  did not use chlorophenates  Linkage to BC Division of Vital Statistics  Offspring Cohort (workers employed > 1 yr) |  -» STUDY 2  Linkage to BC Health Status Registry  Birth Defects (F/U 1952-88) (Dimich-Ward et al. 1996) Chloro-using  Non-using  N= 19,674  n/a  Defects=942 -  STUDY 3  No additional data required  CURRENT STUDY ^STUDY 4 Linkage to BC Cancer Registry  Fertility (F/U 1955-88) (Heacock et al. 1998) Chloro-using  Non-using  N=18,016  N=l,668  Pyrs=432,435  Pyrs=36,215  Childhood Cancer (F/U 1952-88) Chloro-using  Non-using  N=19,674  N=l,716  Pyrs=259,919  Pyrs=23,660  Cases=40  Cases=l  91  Table 2:  Characteristics of Offspring Cohort: Ages 0 - 20.  Chlorophenate Using Mills  Non-User Mills  All Mills  Births by Year  boys  girls  boys  girls  1952 - 1954  727  716  13  16  1,472  1955 - 1959  1,423  1,382  40  29  2,874  1960 - 1964  1,419  1,334  50  43  2,846  1965 - 1969  1,280  1,225  66  54  2,625  1970 - 1974  1,409  1,310  138  136  2,993  1975 - 1979  1,494  1,331  199  177  3,201  1980- 1984  1,486  1,461  240  217  3,404  1985 - 1988  866  811  142  156  1,975  1952 - 1988  10,104  9,570  888  828  21,390  133,924  125,995  12,378  11,282  283,579  (b) 1994-2007#  27,142  25,771  4,248  4,262  61,423  (c) 1969-2007  161,066  151,766  16,626  15,544  345,002  83%  83%  boys and girls  Person years* (a) 1969- 1993  s  Proportion already accounted for = a/c:  74%  73%  82%  * no censoring s  current follow-up of offspring cohort  * person-time left in study if all children followed until 20 years of age  92  4.1 Characteristics of Cancer Cases Table 3 presents results of the offspring cohort to British Columbia Cancer Agency linkage for girls. Among those aged 0 - 20, twenty-four malignant (histology code ending with '73") cancers were identified. Of the twenty-four neoplasms identified, one was diagnosed prior to 1969 (study commencement) and one was diagnosed in a girl whose father worked at a mill that did not use chlorophenols; hence these latter two cases were excluded from the analyses. The most common types of childhood cancer are leukemia and brain neoplasms. The linkage identified five leukemias, six brain cancers, two lymphomas, three reproductive (including two ovarian and one cervical) cancers, two bone cancers, two eye cancers, one liver and one skin cancer.  Table 4 presents results of the offspring cohort to British Columbia Cancer Agency linkage for boys. Among those aged 0 - 20, twenty malignant (ICD (O), histology code ending with '73) cancers were identified. Of the twenty neoplasms identified, two were diagnosed prior to January 1, 1969 (when follow-up commenced), therefore were excluded from the study. Six cases of leukemia, three of brain cancer, three of bone and connective tissue, two lymphomas, and one each of liver, thyroid kidney and non-melanoma skin cancer were identified.  Table 3: Results from linkage of offspring cohort to British Columbia Cancer Registry. A l l malignant neoplasms ("histology/3") identified among female offspring.  Name  Age at Diagnosis  LEUKEMIAS 1.  acute lymphoblastic leukemia  8  2.  acute lymphoblastic leukemia  4  3.  acute lymphoblastic leukemia  16  4.  acute myeloblastic leukemia  2  5.  lymphoid leukemia  3  6.  acute leukemia, NOS*  12  BRAIN CANCERS 8.  astrocytoma  9  9.  astrocytoma  10  7.  astrocytoma, NOS  16  10.  astrocytoma, NOS  6  11.  ependymoma  1  12.  ependymoma#  9  13.  medulloblastoma  3  94  Table 3, cont.: Results from linkage of offspring cohort to British Columbia Cancer Registry. All malignant neoplasms ("histology/3") identified among female offspring.  Name  Age at Diagnosis  LYMPHOMAS 14.  Hodgkin's disease  14  15.  lymphocytic lymphoma  7  REPRODUCTIVE CANCERS 16.  endocervix: adenocarcinoma  19  17.  ovary: papillary adenocarcinoma  9  18.  ovary: adenocarcinoma, NOS  14  BONE AND CONNECTIVE TISSUE CANCERS 19.  osteosarcoma NOS  18  20.  fibrous histiocytoma  19  E Y E CANCERS 21.  embryonal rhabdomyosarcoma  8  22.  retinoblastoma  2  LIVER CANCERS 23.  hepatocellular carcinoma  9  SKIN CANCERS 24.  basal cell carcinoma  12  * identified in linkage but diagnosis prior to 1969, hence not included in analyses. # offspring of sawmill worker from non-chlorophenate using mill; not included in analyses.  95  Table 4: Results from linkage of offspring cohort to British Columbia Cancer Registry. AH malignant neoplasms ("histology/3") identified among male offspring.  Name  Age at Diagnosis  LEUKEMIAS 1.  acute leukemia, N O S  12  2.  acute lymphoblastic leukemia  13  3.  acute lymphoblastic leukemia*  6  4.  acute lymphoblastic leukemia  5  5.  acute lymphoblastic leukemia  8  6.  acute lymphoblastic leukemia  0  7.  acute promyelocyte leukemia  0  BRAIN CANCERS 8.  astrocytoma  15  9.  malignant neoplasm  0  10.  PNET  3  LYMPHOMAS 11.  reticulosarcoma  19  12.  Hodgkin's disease, NOS  4  96  Table 4, cont.: Results from linkage of offspring cohort to British Columbia Cancer Registry. All malignant neoplasms ("histology/3") identified among male offspring.  Name  Age at Diagnosis  BONE AND CONNECTIVE TISSUE CANCERS 13.  Ewing's sarcoma, NOS  9  14.  osteosarcoma  15  15.  synovial sarcoma  18  16.  ganglioneuroblastoma*  0  LIVER CANCERS  17.  hepatoblastoma, NOS  19  SKIN CANCERS  18.  fibrosarcoma  6  KIDNEY CANCERS  19.  nephroblastoma  2  THYROID CANCERS * identified in linkage but diagnosis prior to 1969, hence not included in analyses. 20.  medullary carcinoma  7  97  Table 5 : Results from linkage of offspring cohort to British Columbia Cancer Registry. All in-situ ("histology/2") and uncertain behaviour ("histology/1") neoplasms identified among the offspring.  Name  Age at Diagnosis  IN-SITU (only identified among females) 1.  cervix: squamous cell carcinoma, NOS  19  2.  cervix: squamous cell carcinoma, NOS  17  3.  cervix: squamous cell carcinoma, NOS  18  4.  cervix: squamous cell carcinoma, NOS  16  5.  vulva: Bowen's disease  18  UNCERTAIN BEHAVIOUR 1. 2.  pineocytoma (male)  17  blood: immune thrombocytopenia (female)  19  In-situ and uncertain behaviour neoplasms were not included in any of the analyses.  98  Childhood cancer is more common among boys than girls with a M:F ratio of approximately 1.2:1. This study showed the opposite - cancer was more common among the girls with a M:F ratio of 1:1.2. These observations, however, may have been chance findings as only forty cases were identified in the study. Eleven cases (28%) occurred in children ages 0 - 4 , twelve (30%) in children 5 - 9 years, seven (17%) in the 10 - 14 age group and ten (25%) in 15 - 19 year olds. These proportions differ slightly, particularly in the 5 - 9 year olds, from those identified in both American and Canadian studies where approximately 30%- 35% of cases are expected to occur in the 0 - 4 year olds, 17 % - 18% in both 5 - 9 and 10 - 14 year olds and 30% - 35% in 15 - 19 year olds (Ross et al.1993 and Hutchcroft et al. 1996).  Table 5 lists the cases identified that were either in-situ or of uncertain behaviour. The objective of this study was to examine malignant cases of childhood cancer. However, the three malignant reproductive cancers among the girls (two ovary, one endocervix) prompted curiosity with respect to whether there was an increased risk for other reproductive neoplasms, such as in-situ cervical cancers. The SIR analysis resulted in a risk of 4/5.56 = 0.72, 95% CI; 0.20-1.84. No further analyses on other in-situ or uncertain behaviour neoplasms were performed.  4.2 Associations with Adverse Reproductive Outcomes Data was available to investigate birth defects among cases as well as cancer in the fathers' of childhood cancer cases. The data file produced for the case-control analysis provided information on adverse birth outcomes by ICD code, hence it was possible to verify i f any  99  cases also were diagnosed with abnormalities at birth. Among the girls, one who was diagnosed in 1991 with astrocytoma was born with the following three conditions: ICD1 7569 (other and unspecified anomalies of the musculoskeletal system), ICD2 7566 (anomalies of the diaphragm), and ICD3 7708 (other respiratory problems after birth originating in the perinatal period, e.g. apnea, cyanotic attacks, respiratory distress). None of the boys were diagnosed with an abnormal outcome at birth. There is no data in the published literature linking these abnormalities with risk of developing childhood cancer. Eight control children were born with one or more birth defect(s). Examining whether being born with a defect increased the risk of cancer among the offspring cohort produced an odds ratio of 0.62 (95%CI 0.39, 0.97). Hence, it is unlikely that a positive relationship exists between abnormalities diagnosed at birth and malignant diagnosis. However, these results may indicate that having a birth defect is protective against developing childhood cancer.  4.3 Associations with Cancer in Parents  Data gathered for the original sawmill worker cohort study permitted investigation of relationships between child and paternal cancers. Among the girls, one father died of lung cancer in 1989; his daughter had been diagnosed with malignant cervical cancer in 1985. Among the boys, one father died of kidney cancer in 1979; his son was diagnosed with Hodgkin's disease in 1978. A study on the effect of parental cancer on cancer in the offspring did not identify paternal kidney cancer as being a risk factor for any cancer among sons, although there was a slightly elevated risk for melanoma among the daughters. Lung cancer in fathers resulted in a slightly increased risk for all sites of cancer in both male and female  100  offspring (Hemminki and Viattinen 1997). Follow-up of the sawmill worker offspring cohort into adulthood would clarify the extent to which father-child-cancer relationships exist.  4.4 Cohort Analysis Tables 6 - 1 0 present results of the cohort (SIR) analysis. Although some risks were elevated, all confidence intervals include unity, thus suggesting that chance may have played a role in the results. For both sexes combined (Table 8), the overall risk, based on forty cases, was 1.04 (95% CI 0.74 - 1.41), risk for leukemia based on eleven cases was 0.97 (95% CI 0.48 - 1.77) and the risk for brain cancer, based on nine cases, 1.31 (95% CI 0.60 - 2.49). By separating the boys from the girls it becomes apparent that cancer risk is consistently higher in the girls. Table 6 shows that for all sites combined, the risk is 1.22 (95% CI 0.76 - 1.8), that the six cases of brain cancer provided a twofold increase in risk, SIR = 2.02 (95% CI 0.74 1.41), and the five cases of leukemia contributed a risk of unity. Other sites with a two-fold or greater increase in risk (and at least two cases) were eye and ovary. In every category of age, risk for brain cancer was elevated at least one and a half fold (Table 9). The highest risk overall and in brain cancer occurred in the five to nine year old group. Overall, risks among boys (Table 7) were less than one; all sites (eighteen cases) SIR = 0.89 (95% CI 0.53 - 1.41), leukemia (six case) SIR - 0.95 (95% CI 0.35 - 2.08) and brain (three cases) SIR = 0.77 (95% CI 0.16 - 2.25). Similar to girls, the highest risk by age group in the all sites category was among five to nine year olds (Table 10).  Analysis of the data combining both chlorophenate using and non-user mills (Appendix C) added one case of ependymoma to the offspring cohort (a nine-year-old girl). Adding this 101  one case to an analysis of all sawmills did not significantly alter any of the risks. The risk for ependymoma among the girls was elevated considerably, but the confidence interval still included unity (SIR = 6.97, 95% CI 0.84 - 25.16). When the results were analysed for only those aged 0-14 (Appendix D) there were no statistically significant associations.  102  Table 6: Childhood cancer (ages 0 -19) among female offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  All cancers  Observed Cases  22  Expected Cases  SIR  (95% CI)  18.1  1.22  0.76- 1.8  5.0  1.00  0.32-2.33  3  3.1  0.98  0.20 - 2.86  Brain cancer  6  3.0  2.02  0.74 - 4.40  •  Ependymoma  1  0.3  3.91  0.10-21.78  •  Astrocytoma  4  1.6  2.46  0.67 - 6.30  All lymphomas  2  2.6  0.79  0.10-2.85  •  Hodgkin's disease  1  2.0  0.51  0.01 -2.84  •  Non-Hodgkin's lymphoma 1  0.6  1.69  0.04 - 9.41  2  1.0  2.03  0.45 - 13.39  Bone and connective tissue 2  2.0  1.02  0.12-3.68  Endocervix  1  0.1  9.62  0.24-53.58  Ovary  2  0.8  2.67  0.32 - 9.64  Liver  1  0.3  3.69  0.09 - 20.55  Skin Cancer  1  1.43  0.70  0.02 - 3.90  Leukemias •  acute lymphoblastic  Eye  103  Table 7: Childhood cancer (ages 0 -19) among male offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  Observed Cases  Expected Cases  SIR  (95% CI)  All cancers  18  20.3  0.89  0.53 - 1.41  Leukemias  6  6.3  0.95  0.35-2.08  •  4  4.4  0.92  0.25-2.36  Brain cancer  3  3.9  0.77  0.16-2.25  •  Ependymoma  0  0.5  0  -  •  Astrocytoma  0  1.9  0  -  All lymphomas  2  3.3  0.61  0.07-2.20  •  Hodgkin's disease  1  0.8  1.27  0.03-7.07  •  Non-Hodgkin's lymphoma 1  2.5  0.40  0.01-2.23  Acute lymphoblastic  -  Eye  0  0.5  0  Bone & connective tissue  3  2.3  1.31  Liver  0  0.2  0  Kidney  1  0.9  1.07  0.03 - 5.96  Skin  1  1.4  0.73  0.02 - 4.07  Thyroid  1  0.2  4.39  0.11 -24.45  0.27 - 3.82  -  104  Table 8: Childhood cancer (ages 0 -19) among all offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs.  Age-group and type  Observed cases  Expected cases  SIR  (95% CI)  11  11.6  0.95  (0.47- 1.70)  12  7.6  1.59  (0.82 - 2.78)  7  7.9  0.89  (0.36- 1.83)  10  11.3  0.88  (0.42- 1.62)  ~38A  1.04  (0.74-1.41)  4  5.0  0.80  (0.22 - 2.05)  3  2.8  1.08  (0.22-3.15)  3  1.7  1.79  (0.37- 5.23)  1  1.8  0.56  (0.01-3.12)  TO"  ~097  (0.48 - 1.77)  4  2.2  1.79  (0.49 - 4.58)  2  1.8  1.11  (0.13-4.01)  1  1.7  0.58  (0.01 - 3.23)  2  1.1  1.85  (0.22 - 6.68)  ~~9  6.8  1.31  (0.60 - 2.49)  AH cancers 0-4 5-9 10-14 15-19 0-19  ^0  Leukemia 0-4 5-9 10-14 15 - 19 T  T  0-19  Brain cancer 0-4 5-9 10-14 15 - 19 0-19  105  Table 9:  Age-group and type  Observed and expected cases, SIRs and 95% CIs, for the most common sites, by age group (girls).  Observed cases  Expected cases  SIR  (95% CI)  6  5.3  1.14  (0.42 - 2.49)  7  3.4  2.08  (0.83 - 4.28)  4  3.7  1.08  (0.29 - 2.76)  5  5.7  0.87  (0.28 - 2.03)  "22  18T  1.22  (0.76- 1.84)  3  2.3  1.32  (0.27 -3.85)  1  1.2  0.82  (0.02 - 4.57)  0  0.7  0  1  0.8  1.21  (0.03 - 6.74)  To~  (0.32 - 2.33)  All cancers 0-4 5-9 1 0 - 14 15 - 19 0-19  Leukemia 0-4 5-9 10-14 15 - 19 0-19  ~5  Brain cancer 0-4 5-9 10-14 15 - 19 0-19  2  1.0  2.01  (0.24 - 7.26)  2  0.8  2.38  (0.29 - 8.59)  1  0.7  1.46  (0.04-8.13)  1  0.5  2.21  (0.06-12.31)  ~~6  3.0  2.02  (0.74 - 4.40)  106  Table 10:  Age-group and type  Observed and expected cases, SIRs and 95% CIs, for the most common sites, by age group (boys).  Observed cases  Expected cases  SIR  (95% CI)  5  6.3  0.79  (0.26- 1.84)  5  4.2  1.19  (0.39 - 2.77)  3  4.2  0.71  (0.15-2.07)  5  5.6  0.90  (0.29-2.10)  203~  089  (0.53 - 1.41)  1  2.8  0.36  (0.01 -2.01)  2  1.6  1.27  (0.15-4.58)  3  1.0  2.97  (0.61-8.67)  0  1.0  0  ~~6  ~63  0-4  2  5-9  All cancers 0-4 5-9 10-14 15 - 19 7 8  0-19  Leukemia 0-4 5-9 10-14 15 - 19  ~035  (0.35 - 2.08)  1.3  2.01  (0.24 - 7.26)  0  1.0  0  10-14  0  1.0  0  15 - 19  1  0.6  1.59  (0.04 - 8.86)  0.77  (0.16-2.25)  0-19  Brain cancer  0-19  ~3  T9  107  4.5 Nested Case-Control Analysis Results of the nested case-control analysis are presented in Tables 11 - 24. A comparison of reproductive and workplace exposure characteristics of the parents of cases and controls can be found on Tables 11 and 12. Mothers and fathers of cases are slightly older and give birth to heavier babies with somewhat longer gestation than their control counterparts. Case fathers experienced longer hours within each of the windows of exposure as well as more hours of cumulative exposure than control fathers did. However, the standard deviations for each of the exposure categories exceeds the mean, thereby indicating the wide variation of values in each category. Logistic regression of these variables did not result in any statistically significant differences between cases and controls.  Table 11:  Characteristics of cases and controls. Cases (N=40)  Controls (N=200)  mean  sd*  range  mean  sd*  range  mother's age (years)  27.4  6.5  17-44  26.0  5.4  17-43  father's age (years)  30.3  6.8  18-50  29.6  5.6  20-46  total no. births  2.2  1.4  1-6  2.4  1.5  1 - 13  gestation (weeks)  40.0  1.4  36-43  39.6  1.9  32-44  birthweight (grams)  3,486  505  2,370 - 4,649  3,385  591  1,361 - 5,300  *sd = standard deviation  108  Table 12.  Paternal chlorophenate exposure characteristics of cases and controls. Cases (N=40) mean  sd*  Window 1 (hours)  3,323  4,666  Window 2 (hours)  95  127  Window 3 (hours)  307  Window 4 (hours) CumExp (hours)  Controls (N=200) range  mean  sd*  0 - 24,780  3,096  3,520  0-383  85  120  0-462  401  0 - 1,239  276  387  0-1,615  3,759  5,669  0-20,612  2,613  4,601  0-24,146  7,485  8,253 107-33,567  6,068  6,921  0-44,112  range  0 - 26,464  *sd = standard deviation  Window 1 = cumulative hours of chlorophenate exposure by the father from first employment until 90 days prior to conception. Window 2 = cumulative hours of chlorophenate exposure by the father during the 90 days prior to conception. Window 3 = cumulative hours of chlorophenate exposure by the father during pregnancy. Window 4 = cumulative hours of chlorophenate exposure by the father from birth until child's cancer diagnosis. CumExp = cumulative hours of chlorophenate exposure by the father from first employment until child's cancer diagnosis.  109  Table 13: Risk for developing childhood cancer (all sites) associated with cumulative hours of paternal chlorophenate exposure.  Category  Cumulative hours  Cases  Controls  OR* 95% CI  of exposure  Low Medium High  <1,250  6  33  1.0  1,250-9,999  23  127  1.0  0.4-2.7  >10,000  11  40  1.6  0.5-4.8  * adjusted for sex and year of birth p for trend = 0.28  To assess whether a dose response relationship existed, the data was analysed by levels of increasing paternal cumulative exposure. Table 13 shows results of the internal (nested casecontrol) analysis by three levels of increasing exposure to chlorophenates. The results indicate a slight increasing trend of cancer risk with exposure (OR = 1.6, 95% CI 0.5 - 4.8) to high levels of chlorophenates. The confidence intervals around the point estimates all contained unity, however, and trend test was negative (p for trend = 0.28). The selection of categories of low, medium and high cumulative exposure was somewhat arbitrary, although all cells contained at least five cases. The choice of exposure levels also reflected categories selected in the original cohort study of mortality and cancer incidence. Analyses (not reported) were also carried out with different cut-off points for the three levels of exposure. In each analysis, high levels of paternal cumulative exposure resulted in a non-significant increased risk.  110  Table 14: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposure from one year post commencing employment until 90 days prior to conceiving (Window 1).  Category  Cumulative hours  Cases  Controls  OR* 95% CI  of exposure  Low Medium High  <1,000  14  59  1.0  1,000-4,999  19  106  0.75  0.4- 1.6  >5,000  7  35  0.84  0.3 - 2.3  •adjusted for sex and year of birth p for trend = 0.88  Table 15: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposure during the 90 days prior to conception (Window 2).  Category  Cumulative hours  Cases  Controls  OR* 95% CI  of exposure  Low Medium High  <10  21  114  1.0  10-249  11  60  1.0  0.5 -2.5  >250  8  26  1.7  0.7-4.7  •adjusted for sex and year of birth p for trend = 0.30  111  Table 16: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposures during pregnancy (Window 3).  Category  Cumulative hours  Cases  Controls  OR*  95% CI  <10  19  106  1.0  10-699  13  60  1.2  0.6- 3.1  >700  8  34  1.3  0.6- 3.9  of exposure  Low Medium High  *adjusted for sex and year of birth p for trend = 0.55  Table 17: Risk for developing childhood cancer (all sites) associated with windows of cumulative hours of paternal chlorophenate exposure: exposure from birth until cancer diagnosis (Window 4).  Category  Cumulative hours  Cases  Controls  OR*  95% CI  <1  18  96  1.0  1 - 8,999  15  80  1.0  0.5 -2.6  >9,000  7  24  1.7  0.5-4.8  of exposure  Low Medium High  * adjusted for sex and year of birth p for trend = 0.42  112  To assess whether a relation existed by timing of exposure, the data was analysed by four windows of exposure, each reflecting a distinct period whereby reproductive damage could occur. Results of the analysis by windows of exposure (tables 14-17) showed increasing risk with increasing cumulative exposure, most notably at the highest level of exposure, for all windows except Window 1 where the risk was less than one in both the medium and high levels of exposure. Again, however, all confidence intervals included unity and the tests for trend were negative.  As elevated risks for brain cancer were identified in the cohort analysis, sub-cohort analyses were performed for brain cancer. Exposure was divided into only two levels because of the small number of cases (nine), and therefore trend testing was not possible. Tables 18 and 19 describe risk for developing brain cancer by cumulative exposure for both sexes combined and for the girls alone. Tables 20 through 23 describe risk for developing brain cancer by the four windows of exposure. For brain cancer, there is a nonsignificant increased risk in each high category of exposure. This observation is also apparent when analysing the data for girls alone. Table 24 examines risk for leukemia by cumulative exposure. No further analyses were performed on leukemia because of the non-elevated risk in this dose-response analysis. There were no sites that consistently showed high levels of paternal exposure to chlorophenates. Generally, the high levels of cumulative exposure, i.e. greater than 10,000 hours, occurred among children who were diagnosed in the teenage years, hence the father potentially had more time to accumulate exposure. No further sub-cohort analyses were performed on any site.  113  Table 18: Risk for developing childhood brain cancer associated with cumulative hours of paternal chlorophenate exposure.  Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <3,560  4  19  1.0  High  >3,560  5  16  1.5  0.35-6.85  •adjusted for sex and year of birth  Table 19: Risk among girls for developing childhood brain cancer associated with cumulative hours of paternal chlorophenate exposure.  Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <4,500  3  14  1.0  High  >4,500  3  6  2.5  0 . 3 7 - 16.43  •adjusted for year of birth  114  Table 20: Risk for developing childhood brain cancer associated with windows of cumulative hours of paternal chlorophenate exposure: exposure from one year post commencing employment until 90 days prior to conceiving (Window 1). Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <2,000  4  21  1.0  High  >2,000  5  14  2.1  0.45-9.65  •adjusted for sex and year of birth  Table 21: Risk for developing childhood brain cancer associated with windows of cumulative hours of paternal chlorophenate exposure: exposure during the 90 days prior to conception (Window 2).  Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <100  5  24  1.0  High  >100  4  11  2.1  0.43 - 9 . 8 8  •adjusted for sex and year of birth  115  Table 22: Risk for developing childhood brain cancer associated with windows of cumulative hours of paternal chlorophenate exposure: exposures during pregnancy (Window 3).  Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <400  5  24  1.0  High  >400  4  11  2.0  0.43-9.34  •adjusted for sex and year of birth  Table 23: Risk for developing childhood brain cancer associated with windows of cumulative hours of paternal chlorophenate exposure: exposure from birth until cancer diagnosis (Window 4).  Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <500  5  22  1.0  High  >500  4  13  1.4  0.30-6.16  •adjusted for sex and year of birth  116  Table 24: Risk for developing childhood leukemia associated with cumulative hours of paternal chlorophenate exposure.  Category  Cumulative hours  Cases  Controls  OR*  95% CI  of exposure  Low  <3,000  6  26  1.0  High  >3,000  5  19  0.8  0.16-3.57  *adjusted for sex and year of birth  117  C H A P T E R 5.0 DISCUSSION A N D C O N C L U S I O N 5.0 General Comments No statistically significant associations were found between paternal occupational exposure to chlorophenate fungicides in British Columbian sawmills between 1950 - 1988 and risk of childhood cancer. In the cohort analysis, overall risks were not elevated; however, when the cohort was disaggregated by gender, risks among girls, particularly for brain cancer, ovarian and eye cancer, were elevated, although confidence intervals always included unity. The nested case-control analyses also did not produce statistically significant results for all cancers combined. Risk appeared to increase with increasing levels of exposure; however, the small number of cases in each category of exposure provided unstable estimates and trend tests produced non-significant results. In this study, cases were ascertained using the best methods available at the time, hence should be reasonably complete, and paternal occupational exposure assessment of high quality. To my knowledge, this study contains the most extensive exposure assessment of any study published on risk of childhood cancer in relation to paternal occupational exposure. The exposure assessment allowed the evaluation of dose-response relationships and timing of exposure to be investigated. It is unlikely therefore, that new studies based on a retrospective assessment of paternal occupational exposure to chlorophenols will provide vastly different estimates of risk. However, this study is also the only one reported that was specifically undertaken to explore the relationship between paternal chlorophenol exposure and risk of childhood cancer. In terms of epidemiologic certainty, it is not a definitive study. Although some risks were elevated, none were statistically significant. Furthermore, the results of the linkage included many different sites; 40 cases were spread  118  over twelve different sites, and eight sites contained two or fewer cases. By histology, only acute lymphoblastic leukemia and astrocytoma had more than two cases. If chlorophenol exposure was associated with childhood cancer, I would have expected to observe a concentration of cases for one specific site or histology. For example, DES exposure lead not to an overall increased incidence of childhood cancer, or to an increase in the common childhood cancers, but to an increase a rare type and specific site of cancer in the offspring of exposed women. Future follow-up of the sawmill worker offspring cohort (discussed in section 5.2) and further studies using similar e.g. nested case-control designs, as well as different methods, e.g. case-control designs (discussed in section 5.3) would substantiate or refute the results obtained in this study.  This study is one of several that utilized a database containing personal identifying and exposure data created for 23,829 sawmill workers occupationally exposed to chlorophenate fungicides between 1950 and 1988 (Teschke et al. 1998). Concern over the potential carcinogenicity of chlorophenols and chlorophenoxy herbicides had been mounting since the early 1980s. In response, the International Agency for Research in Cancer (IARC) carried out risk assessment studies and established that there is limited evidence for the carcinogenicity of chlorophenols. IARC's most recent evaluation of chlorophenols (IARC 1998) continues to classify it as class 2B carcinogen (sufficient evidence in experimental animals but inadequate human evidence). In 1987 British Columbia banned the use of chlorophenols in the sawmill industry. The studies performed by a group of researchers in the University of British Columbia's Department of Health Care and Epidemiology have addressed many health-related issues of chlorophenate exposure and health outcome. The 119  original study of the mortality and cancer incidence among chlorophenate exposed sawmill workers (Hertzman et al. 1997), which was the most statistically powerful published at that time, found only borderline results. These results confirmed several others from around the world which have been published in the past fifteen years indicating borderline positive or negative associations with exposure to chlorophenols and/or chlorophenoxy herbicides. Overall, these results support the IARC classification and should alleviate concerns regarding the carcinogenicity of chlorophenols.  The reproductive studies performed by the sawmill worker researchers at U B C were among a handful of human studies investigating the reproductive effects of chlorophenol exposure. The fertility (Heacock et al. 1998) and childhood cancer studies (Heacock, PhD thesis 1998) were negative and the birth outcomes study found elevated risks for but a few outcomes; namely congenital anomalies of the eye, anencephaly or spina bifida, and congenital anomalies of the genital organs (Dimich-Ward et al. 1996): Therefore, regarding the reproductive toxicology of chlorophenols, the evidence to date indicates that it may be a teratogen. However, few epidemiological studies have been performed and further research is necessary to establish the validity of these results.  Based on the above, should chlorophenols be re-introduced into the British Columbian sawmill industry? No. Firstly, with respect to the adult carcinogenicity of chlorophenols, the epidemiological evidence, including results from the retrospective cohort study of British Columbian sawmill workers, suggests there may be a weak carcinogenic effect. Secondly, research into the reproductive toxicology of chlorophenols is sparse and there  remains uncertainty as to which, if any, reproductive endpoints are affected by paternal exposures. Lastly, because of their widespread use, chlorophenols are ubiquitous in nature and should be found throughout the food chain. They are carcinogenic in animals and, therefore, sufficient exposure in the food chain could eventually result in considerable human exposure. Adding occupational to environmental exposure may have detrimental effects on humans. Future studies are needed to clarify results previously obtained.  5.1 How do the results of this study relate to previous research?  How do these findings relate to previous research-on paternal occupational exposure and risk of childhood cancer, particularly associations between pesticides and cancer? Several studies have reported associations between paternal occupational exposures and risk of childhood cancer. From an epidemiological point of view, the question to be addressed, however, is how rigorous were the studies that produced excess risks? Two recently published review articles addressed the question of pesticide exposure and risk of childhood cancer (Daniels et al. 1997 and Hoar Zahm and Ward 1998). The paper by. Daniels etal. (1997) identified thirty-one studies published between 1970- 1996 that examined whether occupational or residential exposure to pesticides by either parent was associated with increased risk of cancer in their offspring. Overall, the authors reported a slightly positive association between childhood cancers and parental pesticide exposure with stronger risk estimates when pesticide exposure was measured in greater detail. The authors pointed out, however, that most of the studies were case-control designs and therefore potentially subject to exposure misclassification, insufficient sample size, biases in control selection and uncontrolled confounding. The main concern was lack of detailed  121  exposure information. Most occupational studies did not utilize direct measurement techniques to quantify pesticide exposure; rather they inferred exposure based on job title and industry, e.g. employment in agriculture. In this thesis research, brain cancer showed elevated, although not statistically significant, risks with paternal exposure to chlorophenate fungicides in both the cohort and nested case-control studies. The review paper by Daniels and colleagues (1997) identified only one of five studies on brain cancer and paternal occupation in agriculture as having an elevated risk and a confidence interval that excludes unity (OR = 2.4, 95% CI; 1.2 - 4.9). Studies investigating farm residence during pregnancy or childhood showed'significantly elevated risks for purchase of pesticides in one study (OR = 2.9, 95% CI;. 1.5- 5.6). Farm residence and pesticide purchase may be a proxy for pesticide exposure but not a very accurate one. Residential pesticide use has also been studied, but often not as the primary hypothesis. One study that did examine residential pesticide use as the primary hypothesis, and therefore carefully collected detailed information on timing, frequency and form of pesticide use found several statistically significant associations; home extermination of termites by the father (OR = 2.9, 95% CI; 1.3 - 7.1), use of pesticide bombs during pregnancy (OR = 6.2, 95% CI; 1.4 28.4), exposure to no-pest strips during pregnancy (OR = 5.2, 95% CI; 1.2 - 22.2) and childhood (OR = 3.7, 95% CI; 1.0 - 13.7), childhood use of lice shampoos (OR = 4.6 95% CI;1.0 - 21.3), and childhood contact of pesticides on pets (OR = 4.4, 95% CI; 1.4-14.3. The review paper by Hoar Zahm and Ward (1998) identified seventeen papers that specifically addressed the role of pesticides in the development of childhood brain cancer. They found significant risk elevations in nine studies, non-significant elevations in five  122  studies and deficits or no associations in three studies. The highest risks were reported for pesticide use in the home or garden or on pets.  This thesis research found no association between chlorophenol exposure and risk of childhood leukemia. Daniels et al. (1997) found positive associations for five of nine studies evaluating occupational exposures and leukemia risk; two investigating paternal occupation and three investigating maternal occupational exposures; only two of these studies had confidence intervals including unity. With respect to residential pesticide exposures two studies found significantly elevated risks during pregnancy and two" studies found elevated risks during childhood. Hoar Zahm and Ward (1998) reviewed eighteen studies of pesticide exposure and childhood brain cancer risk. Most of the studies showed elevated risks for parental occupational or home use of pesticides and five studies produced statistically significant results. In light of the findings of this thesis research and those identified by Daniels et al. (1997) as well as Hoar Zahm and Ward (1998), future research on the etiology of childhood cancer, and specifically the relationship between exposure to pesticides and childhood cancer, will require improved exposure assessment and large study populations.  5.2 Limitations of the Study (1) Study Power The large size of the offspring cohort provided sufficient power to detect true differences for all cancers combined at a relative risk of greater than 1.43; higher risks were required for individual sites and histologies. Although none of the results were statistically significant, 123  i.e. all confidence intervals included unity, some risks were elevated, thereby suggesting the possibility that further follow-up of the offspring cohort might provide the necessary persontime to tighten the confidence intervals around the point estimates. Eighty-three percent of the study's person years have been accounted for between 1969 - 1993. Complete followup of the offspring cohort i.e. until each child has attained twenty years of age, would occur in 2007, hence fourteen further years of follow-up. It is unlikely that a sufficient number of new cases of childhood cancer will develop in this time to produce statistically significant results (see Appendix B). Most of the time left in the study would involve children in their early and late teens and hence the a priori cancers of interest, i.e. leukemia and brain cancer would have passed their age incidence peak. Hence, follow-up of the offspring cohort until 2007 would probably not be efficient. Another option would be to link the offspring cohort to a mortality database. The rationale for this would be to pick up cancer cases diagnosed after 1952 (when the offspring births were first identified) and before 1969 (when the B C Cancer Registry was established, and hence when this study began). Most childhood cancer cases diagnosed before 1969 died and often with childhood cancer prior to 1970, incidence and mortality occurred in close approximation to one another. Follow-up of cases from January 1, 1952 (when births to the sawmill worker cohort were first counted) until December 31, 1968 (just prior to the commencement of the B C cancer registry) would add a total of 80,648 person years to the database of offspring whose fathers worked at chlorophenate-using sawmills (39,634 person years for the girls and 41,014 person years for the boys). This offspring cohort would be young; 83% under the age of ten, which is the age range where the a priori cancers of interest (leukemia and brain) would be predominantly diagnosed. The sawmill worker database contains complete personal identifying and  124  exposure data for the sawmill workers from 1950 - 1988. High exposures to chlorophenates may have occurred in the 1950s and 1960s due to the different technologies and formulations used (see section 2.2), hence the years 1952 - 1968 would probably be a more efficient time period to follow-up, in terms of identifying exposure-disease associations, than 1994 - 2007. The estimated cost of a mortality linkage for the offspring cohort would be in the vicinity of $3,000 (personal communication with W. Threlfall, 1996) which, given the trends shown here, would be a cost-effective exercise, and will be pursued at a later date. Mortality data obtained from the linkage would be considered as incidence data for the period 1952 - 68 (as both incidence and mortality occurred in close proximity in these years) then added to the data analysed for the 1969-93 follow-up. This new follow-up period would allow incidence calculations for both the cohort and nested case-control studies to be calculated over a longer period of time, i.e: 1952-1993, thereby producing more statistically powerful results, which would also be more conclusive. This new study would not include a mortality analysis, hence a mortality linkage from 1969-93 will not be performed.  Another method to increase the power of a study is to increase the number of study subjects. Our offspring database contains all cohort-linked B C births between 1952 - 1988 and registered in the Division of Vital Statistics. Barring incomplete ascertainment from the above-mentioned linkage, the sawmill worker offspring cohort is complete. Hence, no further children could be added to the offspring cohort. If underascertainment of births occurred, there is no reason to believe that it would have occurred preferentially in one of the exposure groups. One possibility to increase sample size is to include other similar 125  studies. At present, there are no other large cohorts of chlorophenate-exposed workers and their children, therefore the number of offspring of chlorophenate exposed workers can not be increased. Hence, without being able to increase the sample size or the follow-up time, the power of the study can not be increased. However, the Cancer Control Research Program at the British Columbia Cancer Agency is considering initiating a study to examine the reproductive outcomes of a cohort of more than 30,000 pulp and paper and sawmill workers who were employed in B C between 1950 -1992 (personal communication with G. Astrakianakis, research scientist, Cancer Control Research Program, June 1998). This proposed study would use data collected for a retrospective cohort study of all cause mortality and cancer incidence among the pulp and paper workers (Bandet al. 1997). Of the fourteen pulp and paper mills included in the B C Cancer Agency's study, at least ten of the mills also had sawmills at the site and detailed work history and exposure information was collected on all workers. Three sawmills included in the pulp and paper study were also included in the chlorophenate study (Hertzman et al. 1997) hence, seven sawmills in the pulp and paper study would have exposure data that could potentially be added to the database of health outcomes of sawmill workers. If a linkage study is initiated which traces the offspring cohort of the sawmills identified in the pulp and paper study, and then links the offspring to the cancer registry, this new sawmill worker offspring cohort could be added to the sawmill cancer offspring cohort already developed and a new larger study population will be available to re-address the issue of paternal employment in the sawmill industry and risk of childhood cancer.  126  (2) C o n f o u n d i n g  Concerns with respect to confounding include the limited information on other occupational exposures to the father (e.g. wood dust), the complete lack of maternal occupational information, as well as limited information on "lifestyle" exposures to both parents, e.g. smoking and alcohol consumption, as well as home pesticide use. Apart from exposure to chlorophenols, sawmill workers are also exposed to wood dust which was classified as a known human carcinogen by IARC in 1995. The sawmill worker cohort study team previously examined dust measurement data from two British Columbian sawmills (Teschke et al. 1993). The results indicated no association between wood dust exposure and chlorophenol levels in urine, hence these factors should not be confounders of any effects detected in association with chlorophenate levels. The sawmill worker study team is presently re-visiting all fourteen sawmills from the original study (Hertzman et al. 1997) to re-construct technological history pertinent to wood dust exposure (Teschke et al. 1998) . Mortality and cancer incidence will be examined, however, there are no plans to investigate reproductive outcomes in relation to wood dust exposure. With respect to maternal occupational exposure data, none was available. It must also be kept in mind that female participation in the work force has grown rapidly in the past two decades, however, in the 1950s 60s and 70s there would have been limited participation in the workforce by women. Hence, occupational data may not have been as important as home and recreational exposure data. A number of lifestyle factors potentially associated with risk of developing childhood cancer have been identified, such as exposure to and/or consumption of 7V-nitroso substances and home use of pesticides. Lifestyle risk factors may be systematically different between geographical location of the sawmills (i.e mills in rural vs.  127  urban locations) but it is unlikely that these factors would be correlated with chlorophenate exposure within the mills using chlorophenates. One often cited reason for performing nested case-control studies is to capture information on potentially confounding factors, such as smoking history, that are not normally collected in occupational studies (Checkoway and Demers 1994). With respect to this study, a retrospective collection of potentially confounding variables would require considerable time and money. From a practical point of view, the resources necessary to interview the families (or surrogates) of the cases and controls to elucidate potentially confounding data, would most likely not be worth the effort for the forty cases of childhood cancer that were identified. As there are only two known maternal exposures that cause childhood cancer (DES and ionizing radiation) and no known paternal causes, it is highly unlikely that confounding played a significant role in the results of this study. For a factor to be a confounder it has to be associated with exposure and independently associated with disease, but not be a link in the causal chain between exposure and disease (Hennekens and Buring 1987). Confounding variables mask the detection of true differences either by diluting the results which causes an underestimation of the true risk (negative confounding) or by enhancing the results providing a more extreme effect than is actually the case (positive confounding). This study was negative therefore, if confounding occurred, it would have to be a protective factor to threaten the validity of the results. It is difficult to imagine a factor that would affect exposure in one direction and childhood cancer risk in another. Therefore, it is highly unlikely that confounding played a significant role in determining the results of this study. Also, variables from the Physicians Notice of Birth (e.g. maternal and paternal age,  128  birth weight, gestation), did not appear to have a confounding effect in the stepwise logistic regression performed to determine a baseline model.  (3) Bias The main type of bias encountered in this study was exposure misclassification. The original study (Hertzman et al. 1997) was designed to reduce quantifiable errors in exposure assessment (Hertzman et al. 1988, Teschke et al. 1990 and Teschke et al. 1996); yet there was still potential for a certain amount of exposure misclassification. However, any misclassification that did occur would likely have been non-differential and therefore would have biased the findings towards the null, and therefore obscured true dose-response relationships. This dampening of dose-response has been described in a paper examining the reliability and validity of a retrospective exposure assessment to tetra and pentachlorophenates in the sawmill industry (Hertzman et al. 1988). The paper describes a three by three classification table which compares, on one axis, low, medium and high levels of urinary chlorophenate concentrations provided by 132 workers, and on the other axis, low, medium and high levels of job exposure estimates by the same workers. This classification table indicated that 100 of 132 points had concordant job scores and urinary chlorophenate levels. Of the 32 points misclassified, two were discordant by two levels (i.e. ranked high by urinary chlorophenate level but low by job exposure score) and the other 30 points were misclassified by one category; medium-high or medium-low combinations. This misclassification resulted in a dampening of the dose-response relationship, particularly among those with high levels of exposure. In this thesis, the  129  elevated but non-significant dose-response relationships that occurred may have been stronger had misclassification of exposure not occurred.  Another source of bias in cohort studies is loss to follow-up. In the offspring cohort, no information was available on the vital status of the children, with the exception of data in the Cancer Registry which records date of diagnosis and last visit of case to B C Cancer Agency. No information on accidental death was available. Accidental deaths and childhood cancer are the two leading causes of mortality among children. However, the rate for these causes of death is so low that it is unlikely that many person-years of observation would be censored, nor should the risk of death for children of fathers with high chlorophenate exposures differ from those with low exposures.  Lastly there is a possibility that the linkage of the sawmill worker cohort file to the Division of Vital Statistics (DVS) marriage and birth files failed to identify all births born to the sawmill worker cohort who had been employed at least one year between 1952 1988. Evidence for this is suggested in the study of fertility in the sawmill worker cohort (Heacock et al. 1998). The Standardized Fertility Ratio (SFR) for the workers employed at chlorophenate-using mills was 0.74 and for the workers employed at mills not using chlorophenates, the SFR was 0.82. If fertility is uniquely related to exposure, then unexposed workers should have an SFR approximating unity. In the fertility study, this was not the case and it is therefore possible that 10% or more of the offspring cohort was not identified.  130  The offspring cohort was captured using probabilistic linkage, augmented with manual resolution of uncertain links. Hence, the methodology used to obtain the offspring cohort provided links between the sawmill worker cohort and all matching births registered at the D V S . It is possible that a small proportion of the births to the sawmill worker cohort were not registered at the DVS or that data entry errors resulted in "true" matches not being picked up in the probabilistic linkage. If underascertainment occurred equally at exposed and unexposed mills, then results of the analyses would have been biased towards the null, but if births were reported differentially by geographical location and potentially by mill type, bias may have occurred. -  (4) Chance The wide confidence intervals around the point estimates suggest that chance may have been responsible for the elevated risks. Because childhood cancer is rare, even with almost 260,000 person-years of follow-up, the power of the study was low for detecting small increases in risk. Although there appeared to be increased risks for certain cancer sites among the girls, notably brain, eye and ovarian cancer, the possibility that chance influenced these results cannot be ruled out. One must always be cautious when interpreting elevated risks based on small numbers and wide confidence intervals.  5.3 Opportunities for Future Research (1) Endocrine Disruptors and Male Reproductive Health A n area gaining considerable scientific momentum at the moment is that of environmental endocrine disruptors, (e.g. pesticides, dioxins, plasticizers) and their effects 131  on the reproductive capacity of both animals and humans (Langer and Sang 1997). Research from laboratory animals, from animals in their natural environments and from humans is suggesting that endocrine disruptors may be having a subtle but profound influence on reproduction. Reports from a recent conference entitled, "Hazardous Substances and Male Reproductive Health" (New York City, May 1998) identified several wildlife species that have been threatened by exposure to chemicals: Florida panthers are being born with cryptorchidism and are becoming demasculinized, the alligators in Lake Apopka have shown male reproductive anomalies (foreshortened penises and ovary-like organs) and a reduced population (only one in five eggs hatch, half the hatchlings die within days), and turtles have been observed with reversed reproductive systems (Smith 1998). At the same conference, new epidemiological studies from several countries confirmed previously reported increases in cryporchidism, hypospadius and testicular cancer, thus suggesting that common environmental factors may play a role in these adverse reproductive outcomes (Skakkebaek 1998, Ekbolm 1998, Erikson 1998). A n international collaboration has been initiated to assess the unresolved issue of declining semen quality (Swan 1998). New methods are being developed to assess genetic damage to the male reproductive system, such as biomarkers for spermatogenesis. Multidisciplinary, international collaborations will be necessary to fully understand the complex issues surrounding environmental endocrine disruptors and male-mediated reproductive toxicology.  132  (2) Future Childhood Cancer Studies As knowledge of male-mediated reproductive toxicology increases, so too will the understanding of the paternal contribution to childhood cancer. Because of the rarity of childhood cancer and hence limited number of cases, investigating the epidemiology and etiology of this disease will require collaborative research. Presently, two large international group of researchers, the Children's Cancer Group (CCG) and the Pediatric Oncology Group (POG) are involved in collaborative case-control studies to understand the epidemiology, and elucidate the etiology of childhood cancer. International studies will provide information on geographic, socio-demographic and race or ethnic group differences, thereby providing clues to environmental etiologic factors. Population-based epidemiologic studies with a high quality exposure, environmental and genetic data, along with detailed investigations of families with two or more members affected by childhood cancer, or those with heredity syndromes which increase childhood cancer risk will be of vital importance in clarifying the role of host and environmental factors. (Chow et al. 1996). Advances in molecular genetics/molecular epidemiology should be incorporated into studies to elucidate mechanisms of childhood cancer (Robison 1994). Markers which are both sensitive and specific need to be identified to be used to describe high risk families and children (Chow et al. 1996). The relationship between postnatal environmental exposures such as ionizing radiation, electromagnetic field radiation and pesticide exposures and risk of childhood cancer needs further investigation.  Because the etiology of childhood cancer is still a mystery, a multitude of risk factors need to be evaluated. As well, childhood cancer is rare. Examining rare diseases and 133  multiple risk factors is done most cost-effectively with case-control studies. However, case-control studies have been criticized for their lack of sufficiently detailed exposure information. On the other hand, occupational cohort or nested case-control studies, which gather extensive occupational exposure data, generally do not collect sufficient data on potential confounders, including paternal non-occupational and maternal information. Also, because of the rarity of childhood cancer, large cohorts with lengthy follow-up are required. What would the ideal design be to study paternal exposure and childhood cancer? M y first recommendation would be a large, possibly collaborative, case-control study that collects detailed paternal, maternal and childhood exposure data. This type of design is used by the POG and C C G collaborations. I would also recommend focussing on individual histologies as opposed to all cancers or even specific sites (e.g. all brain cancers) as there may be cell-specific effects for different chemicals. Case-control studies that include inexpensive, reliable and valid laboratory measures of past exposure would be ideal (Guffermanl 998). Secondly, i f a researcher finds him or herself in a position where paternal occupational exposure data has been collected, and the workplace chemical or physical agents are suspected of having adverse reproductive effects, he/she may wish to consider performing a linkage study to ascertain an offspring cohort and then link this offspring cohort to various childhood health databases. This second option would contain more detailed exposure data but would lack information on potential confounders.  5.4 Conclusion This large retrospective cohort study of cancer in the offspring of British Columbian sawmill workers employed between 1950 - 1988 failed to find a relationship between 134  paternal occupational exposure to chlorophenate fungicides and risk of childhood cancer, although the study power was limited. However, the elevated risks observed with exposure to chlorophenates and childhood brain cancer in both cohort and nested case-control analyses suggests that further investigation using reliable and valid exposure assessments combined with detailed family history information would clarify the association between chlorophenol exposure and risk of childhood brain cancer. This study identified the utility of using computer technology to link large data sets to examine rare diseases such as cancer. The promising research being undertaken in the fields of molecular genetics, along with collaborative research by clinicians and epidemiologists should provide fruitful information on the etiology of childhood cancer as we approach the new millenium.  135  Bibliography:  Adami HO, Bergstrom R, Mohner M , Zatonski W, Storm H, Ekbom A , Tretli S, Teppo L , Ziegler H , Rahu M , Gurevicius R and Stengrevics A. Testicular cancer in nine northern European countries. Int J Cancer 1994;59:33-8. Ahlborg U G and Thunberg T. Chlorinated phenols: occurrence, toxicity, metabolism and environmental impact. Crit Rev Toxicol 1980:7;l-35. Allan B B , Brant R, Seidel and Jarrell JF. Declining sex ratios in Canada. Can Med Assoc J. 1997;156(1):37-41. Alexander FE. Space-time clustering of childhood acute lymphoblastic leukemia: indirect evidence for a transmissible agent. Br J Cancer 1992;65(4):589-92. Annegers JF and Johnson CC. Studying paternal occupation and childhood cancer. Epidemiology 1992;3(l):l-2. Auger J, Kuntsmann J M , Czyglik F and Jouannet P. Decline in semen quality among fertile men in Paris during the past twenty years. N Engl J Med 1995;332(5):281-5. Autrup H . Transplacental transfer of genotoxins and transplacental carcinogenesis. Environ Health Perspect Suppl 1993;101(suppl2):33-8. Baird PA. Reproductive hazards and the workplace. Can Med Assoc J 1992;147(2): 15760. Band PR, Le N D and Krewski D. Cohort mortality study of pulp and paper mill workers in British Columbia, Canada. A m J Epidemiol 1997;146(2):186-94. Bas Bueno de Mesquita H, Doornbos G, van der Kuip D A M , Kogevinas M and Winkelmann R. Occupational exposure to phenoxyherbicides and chlorophenols and cancer mortality in the Netherlands. A m J Ind Med 1993;23:289-300. Bauchinger M , Dresp J, Schmid E and Hauf R. Chromosome changes in lymphocytes after occupational exposure to pentachlorophenol (PCP). Mutation Res 1982;102:83-8. Blair V and Birch J M . Patterns and temporal trends in the incidence of malignant disease in children: leukemia and lymphoma. Eur J Cancer 1994;30A(10):1490-8. Blair V and Birch J M . Patterns and temporal trends in the incidence of malignant disease in children: solid tumours of childhood. Eur J Cancer 1994;30A(10):1498-1511.  136  Bleyer W A . What can be learned about childhood cancer from 'Cancer Statistics Review 1973-1988'. Cancer Supplement 1993;71(10):3229-36. Bolande R. Prenatal carcinogenesis; an appraisal. Cancer 1994;74(6): 1674-9. Bonde JPE, Olsen JH and Hansen KS. Adverse pregnancy outcome and childhood malignancy with reference to paternal welding exposure. Scand J Work Environ Health 1992;18:169-77. Braun W H , Blau G E and Chenoweth M B . The metabolism/pharmacokinetics of pentachlorophenol in man, and a comparison with the rat and monkey. In Toxicology and Occupational Medicine, Deichman WB ed. 1979 Elsevier, New York, pp289-96. Breslow N E and Day N E . Statistical Methods in Cancer Research. Volume 1. The Analysis of Case-Control Studies. IARC Scientific Publication No. 32. International Agency for Research on Cancer, Lyon, France 1980. Buckley J, Robinson L, Swotinsky R, Garabrant D, LeBeau M , Manchester P, Nesbit M , . Odom L , Peters J, Woods W and Hammond G. Occupational exposures of parents of children with acute non-lymphoblastic leukemia: a report from the children's cancer study group. Cancer Research 1989;49:4030-7. Bunin GR, Buckley JD, Boesel CP, Rorke L B and Meadows AT. Risk factors for astrocytic glioma and primitive neuroectodermal tumor of the brain in young children: a report from the Children's Cancer Group. Cancer Epidemiol Biomarkers Prev 1994;3:197204. Carlsen E, Giwercman A , Keidrig N and Skakkebaek N E . Evidence for decreasing quality of semen during the past 50 years. Br Med J 1992;305:609-13. Carroquino M J , Galson SK, Licht J, Amler RW, Perera FP, Claxton L D , Landrigan PJ. The US E P A conference on preventable causes of cancer in children: a research agenda. Environ Health Perspect 1998;106(Suppl 3):867-73. Cartwright R and Staines A. Acute leukemias. Baillieres Clinical Haematology 1992;5(l):l-25. Cessna A J , Waite DT and Constable M . Concentrations of Pentachlorophenol in atmospheric samples from three Canadian locations, 1994. Bull Environ Contam Toxicol 1997;58:651-8. Checkoway H and Demers P. Occupational case-control studies. Epidemiologic Reviews 1994;16(l):151-62.  137  Checkoway H, Pearce N E and Crawford-Brown DJ. Research Methods in Epidemiology. Oxford University Press 1989. Chow W H , Linet M S , Liff J M and Greenberg RS. Cancers in children. In, Cancer Epidemiology and Prevention, Schottenfeld D and Fraumeni JF, eds. Oxford University Press, New York, 1996;ppl331-69. ColieCF. Male mediated teratogenesis. Reprod Toxicol 1993;7(1): 3-19. Colosio C, Maroni M , Barcellini W, Meroni P, and Alcini D. Toxicological and immune findings in workers exposed to pentachlorophenol (PCP). Arch Environ Health 1993;48(2):81-8. Colt JS and Blair A . Parental occupational exposures and risk of childhood cancer. Environ Health Perspect 1998;106(Suppl 3):909-25. Cordier S, Lefeuvre B , Fillippini G, Peris-Bonet R, Farinotti M , Lovicu G and Mandereau L. Parental occupation, occupational exposure to solvents and polycyclic aromatic hydrocarbons and risk of childhood brain tumors (Italy, France, Spain) Cancer Causes Control 1997;8:688-97. Daniel V , Huber W, Bauer K and Opelz G. Impaired in-vitro lymphocyte responses in patients with elevated pentachlorophenol (PCP) blood levels. Arch Environ Med 1995;50(4):287-92. Daniels JL, Olshan A F and Savitz D A . Pesticides and childhood cancers. Environ Health Perspect 1997:105(10): 1068-77. Davis JR. Childhood cancer and pesticide use in the home, garden and yard. Dissertation abstract, 1992, University of California, Berkeley. De Maeyer J, Schepens PJ, Jorens P G and Verstraete R. Exposure to pentachlorophenol as a possible cause of miscarriages. Br J Obstet Gynaecol 1995;102(12):1010-1. Dimich-Ward H , Hertzman C, Teschke K, Hershler R, Marion SA, Ostry A and Kelly S. Reproductive effects of paternal exposure to chlorophenate wood preservatives in the sawmill industry. Scand J Work Environ Health 1996;22(4):267-73. Doll R, Evans H and Darby S. Paternal exposure not to blame. Nature 1994;367:678-80. Draper GJ, Little M P , Sorahan T, Kinlen LJ, Bunch K J , Conquest AJ, Kendall G M , Kneale GW, Lancashire RJ, Muirhead CR, O'Connor C M and Vincent TJ. Cancer in the offspring of radiation workers: a record linkage study. Br Med J 1997;315:1181-8. Dyer C. Sellafield case closes after 50 experts give evidence. Br Med J 1993;307:7. 138  Ekbom A . Trends in the incidence of testicular cancer. Paper presented at the Hazardous Substances and Male Reproductive Health International Conference, New York Academy of Medicine New York City, May 1998. Erikson R L . Rising testicular cancer rates in the US army 1971 - 1994. Poster presented at the Hazardous Substances and Male Reproductive Health International Conference, New York Academy of Medicine New York City, May 1998. Eriksson M , Hardell L, Berg NO, Moller T and Axelson O. Soft-tissue sarcomas and exposure to chemical substances: a case-referent study. Br J Ind Med 1981;38:27-33. Fabia J and Thuy T. Occupation of father at time of birth of children dying of malignant diseases. Br J Prev Soc Med 1974;28:98-100. Fahrig R, Nilsson C A and Rappe C. Genetic activity of chlorophenols and chlorophenol impurities. In Pentachlorophenol : Chemistry, Pharmacology and Environmental Toxicology. Rao Rang ed. 1978. Plenum Press, New York. Farwell J and Flannery JT. Cancer in relatives of children with nervous system neoplasms. N Engl J Med 1984;311:749-53. Fellegi IP and Sunter A B . A theory of record linkage. J A m Stat Assoc 1969;1183-1210. Gardner M J , Snee M P , Hall AJ, Powell C A , Downes S and Terrell JD. Results of a casecontrol study of leukemia and lymphoma among young people near Sellafield nuclear plant in west Cumbria. Br Med J 1990;300:423-9. Gibbons L , Mao Y , Levy I and Miller A . The Canadian Cancer Control Program. Can Med Assoc J 1994; 151 (12): 1704-09. Gold E B , Lasley B L and Schenker M B . Introduction: Rationale for an update. In, Occupational Medicine: State of the Art Reviews. Reproductive Hazards. 1994;9(3). Gold E, Gordis L , Tonascia J and Sizklo M . Risk factors for brain tumours in children. 1979;109:309-19. Gold E B and Sever L E . Childhood cancers associated with parental occupational exposures. Occupational Medicine: State of the Art Reviews 1994;9(3):495-539. Golding J, Paterson M and Kinlen LJ. Factors associated with childhood cancer in a national cohort study. Br J Cancer 1990;62(2):394-408. Gufferman S. Methodologic approaches to studying environmental factors in childhood cancer. Environ Health Perspect 1998;106(Suppl 3):881-887. 139  Guillette L J Jr, Gross D A , Rooney A A and Percival HF. Gonadal steroidogenesis in vitro from juvenile alligators obtained from contaminated or control lakes. Environ health Perspect 1995;103 supple 4:31-6. Gurney JG, Davis S, Severson R K , Fang J-Y, Ross JA and Robison L L . Trends in cancer incidence among children. Cancer 1994;78(3):532-40. Gurney JG, Severson R K , Davis S and Robison L L . Incidence of cancer in children in the United States: sex-, race-, and 1-year age-specific rates by histologic type. Cancer 1995;75(8):2186-95. Gurney JG, Davis S, Severson R K , Fang JY, Ross J and Robison L. Trends in cancer incidence among children in the U.S. Cancer 1996;78(3):532-40. Hakullinen T, Salonen T and Teppo L. Cancer in the offspring of fathers in hydrocarbon related occupations. Br J Prev Soc Med 1976;30:138-40. Hales B F and Robaire B. The male-mediated developmental toxicity of cyclophosphamide. In, Male-Mediated Developmental Toxicity. Olshan A F and Mattison DR, eds. Plenum Press, New York 1994, pp 105-17. Hardell L and Sandstrom A. Case-control study: soft tissue sarcomas and exposure to phenoxyacetic acids or chlorophenols. Br J Cancer 1979;39:711-7. Hardell L , Eriksson M , Lenner P and Lundgren E. Malignant lymphoma and exposure to chemicals, especially organic solvents, chlorophenols and phenoxy acids: a case-control study. B r J Cancer 1981;43:169-76. Hardell L, Johansson B and Axelson O. Epidemiological study of nasal and nasopharyngeal cancer and their relation to phenoxy acid or chlorophenol exposure. A m J IndMed 1982;247- 57. Hardell L and Bengtsson NO. Epidemiological study of socioeconomic factors and clinical findings in Hodgkin's disease, and reanalysis of previous data regarding chemical exposure. B r J Cancer 1983;48:217-25. Hatch M and Marcus M . Occupational exposures and reproduction. In Reproductive and Perinatal Epidemiology, CRC Press 1991 :pp 132-40. Hattemer-Frey H A and Travis CC. Pentachlorophenol: environmental partitioning and human exposure. Arch Environ Contam Toxicol 1989;18:482-9. Haworth S, Lawlor T, Mortelmanns K , Speck W and Zeiger E. Salmonella mutagenicity test results for 250 chemicals. Environ Mutagenicity 1983;5(suupl 1):3-142. 140  Heacock H , Hogg R, Marion SA, Hershler R, Teschke K , Dimich-Ward H , Demers P, Ostry A , Kelly S and Hertzman C. Fertility among a cohort of male sawmill workers exposed to chlorophenate fungicides. Epidemiology 1998;9(l):56-60. Hemminki K , Saloniemi I, Salonen T, Partanen T and Vainio H. Childhood cancer and parental occupation in Finland. J Epidemiol Community Health 1981 ;35:11-5. Hemminki K and Vaittinen P. Effect of paternal and maternal cancer on cancer in the offspring: a population-based study. Cancer Epidemiol Biomarkers Prev 1997;6:993-7. Hennekens C H and Buring JE. Measures of Disease Frequency and Association. In, Epidemiology in Medicine. Mayrent SL, ed., Little, Brown and Company, 1987, pp54131. Herbst A L , Ulfelder H and Poskanzer DC. Adenocarcinoma of the vagina: association with maternal stilbestrol therapy with tumour appearance in young women. New Engl J Med 1971;284:878-81. Hertzman C, Teschke K , Dimich-Ward H and Ostry A. Validity and reliability of a method for retrospective evaluation of chlorophenate exposure in the lumber industry. A m JInd Med 1988;14: 703-13. Hertzman C, Teschke K, Ostry A , Hershler R, Dimich-Ward H, Kelly S, Spinelli JJ, Gallagher RP, McBride M and Marion SA. Mortality and cancer incidence among sawmill workers exposed to chlorophenate wood preservatives. A m J Pub Health 1997;87(1): 71-9. Hoar Zahm S and Devesa SS. Childhood cancer: overview of incidence trends and environmental carcinogens. Environ Health Perspect 1993;102(supple 6):177-84. Hoar Zahm S and Ward M H . Pesticides and childhood cancer. Environ Health Perspect 1998;106(Suppl 3):893-908. Howe G R and Lindsay J. A generalized record linkage computer system for use in medical follow-up studies. Comput Biomed Res 1981;14:327-40. Howe GR. Use of computerized record linkage in follow-up studies of cancer epidemiology in Canada. Natl Cancer Inst Monogr 1985;67:117-21. Howe GR, Burch JD, Chiarelli A M , Risch H and Choi B C K . A n exploratory case-control study of brain tumours in children. Cancer Res 1989;49:4349-52. Huchcroft S, Clarke A , Mao Y , Desmeules M , Dryer D, Hodges M , Leclerc J-M, McBride M , Pelletier W and Yanofsky R. This Battle Which I Must Fight: Cancer in Canada's Children and Teenagers. 1996, Ottawa: Supply and Services Canada. 141  IARC (1980) IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, vol. 25, The Lumber and Sawmill Industries (including logging) pp 49-97. IARC (1986): IARC monographs on the evaluation of carcinogenic risk of chemicals to humans. Volume 41. Some halogenated hydrocarbons and pesticide exposures: pp 319-46. IARC (1991): IARC monographs on the evaluation of carcinogenic risks to humans. Volume 53. Occupational exposures in insecticide application and some pesticides: pp 371-402. Pentachlorophenols. IARC (1998): IARC monographs on the evaluation of carcinogenic risks to humans.Volume 71. Some industrial chemicals including monomers. Infante-Rivard C. Home exposure to pesticides and incidence of childhood acute lymphoblastic leukemia. Paper presented at Canadian Society for Epidemiology and Biostatistics conference, May 1997, London, Ontario. Irvine S, Cawood E, Richardson D, McDonald E and Aitken J. Evidence of deteriorating semen quality in the United Kingdom: birth cohort study in 577 men in Scotland over 11 years. Br Med J 1996;312:467-71. Ji B-T, Shu X - O , Linet M S , Zheng W, Wacholder S, Gao Y-T, Ying D M and Jin F. Paternal cigarette smoking and the risk of childhood cancer among the offspring of nonsmoking mothers. JNCI 1997;89(3):238-44. Johnson C, Annegers JF, Frankowski RF et al. Childhood nervous system tumours - an evaluation of the association with paternal occupational exposure to hydrocarbons. A m J Epidemiol 1987;126:605-13. Karmaus W and Wolf N . Reduced birthweight and length in the offspring of females exposed to PCDFs, PCP and lindane. Environ Health Perspect 1995; 103:1120-5. Kauppinen T and Lindroos L. Chlorophenol exposure in sawmills. A m Ind Hygiene Assoc J 1985;46:34-8. Kelly S, Teschke K and Hertzman C. Pentachlorophenol and tetrachlorophenol. In, Environmental and Occupational Medicine. Rom W N , ed. 3 ed 1997. rd  Kelly S, Hertzman C, Teschke K and Weins M . Review of the published literature on polychlorinated dibenzo-p-dioxins and dibenzofurans. A report for the British Columbia Environment. April 1992.  142  Kinlen L J Clarke K and Balkwill A. Paternal pre-conceptual radiation exposure in the nuclear industry and leukemia and non-Hodgkin's lymphoma in young people in Scotland. Br Med J 1993;306:1153-8. Kogevinas M , Kauppinen T, Winkelman R, Becher H, Bertazzi PA, Bas Bueno-de Mesquita H , Coggon D, Green L, Johnson E, Littorin M , Lynge E, Marlow D A , Mathews JD, Neuberger M , Benn T, Pannett B, Pearce N and Saracci R. Soft tissue sarcoma and non-Hodgkin's lymphoma in workers exposed to phenoxy herbicides, chlorophenols and dioxins: two nested case-control studies. Epidemiology 1995;6(4):396-402. Kristensen P, Andersen A , Irgens L M , Bye A S and Sundheim L. Cancer in offspring of parents engaged in agricultural activities in Norway: incidence and risk factors in the farm environment. Int J Cancer 1996;65(l):39-50. Kristensen P and Andersen A . A cohort study on cancer incidence in offspring of printing workers. Epidemiology 1992;3(1):6-10. Kuitjen R R and Bunin GR. Risk factors for childhood brain tumors. Cancer Epidemiol Biomarkers Prev 1993;2:277-88. Kuijten RR, Bunin GR, Nass C C and Meadows AT. Parental occupation and childhood astrocytoma: results of a case-control study. Cancer Res 1992;52:782-6. Kuijten RR, Bunin GR, Nass C C and Meadows AT. Gestational and familial risk factors for childhood astrocytoma: results of a case-control study. Cancer Res 1990;50:2608-12. Kwa S and Fine L. The association between parental occupation and childhood malignancy. J Occup Med 1980;22:792-4. Langer J and Sang S. Endocrine disruption: cause for concern, cause for action. Canadian Chemical News 1997: 16-7. Larsen CD. Pulmonary tumour injection by transplacental exposure to urethane. JNCI 1947;8:63-9. Law L W . The production of tumours by injection of a carcinogen into the amnionic fluid of mice. Sciencel940;91:96-7. Leiss J K and Savitz D A . Home pesticide use and childhood cancer: a case-control study. A m J public Health 1995;85(2): 249-52. Lemasters G K . Occupational exposures and effects on male and female reproduction. In Environmental and Occupational Medicine.2 Edition, Rom W N ed. 1992; ppl47-170. Little, Brown and Company. nd  143  Lemasters G K and Selevan SG. Toxic exposures and reproduction: a view of epidemiology and surveillance. In Scialli A R and Zinaman MJ, eds. Reproductive Toxicology and Infertility. McGraw-Hill 1993, pp 307-21. L i FP and Fraumeni JF, Jr. Soft tissue sarcomas, breast cancer and other neoplasms: a familial syndrome? Ann Intern Med 1969;71:747-52. L i FP and Fraumeni JF, Jr. Prospective study of a family cancer syndrome. J A M A 1982;247:2692-4. Likhachev A , Anisimov V , Loktionov A , Zabezhinski M , Napalkov N and Tomatis L . Increased tumour incidence and skin tumour promotion in two generations of decendents of 7,12-dimethylbenz(a)-anthracene treated pregnant mice. In Napalkov NP, Rice J M , Tomatis L and Yamasaki H (eds), Perinatal and Multigeneration Carcinogenesis, IARC Scientific Publication 96, International Agency for Research n Cancer, Lyon;1989:81-92. Loktionov A , Popovich I, Zabezhinski M et al. Transplacental and transgeneration carcinogenic effect of 7,12-dimethyl-benz(a)anthracene: relationship with ras oncogene activation. Carcinogenesis 1992; 13(1): 19-24. London SJ, Thomas DC, Bowman JD et al. Exposure to residential electric and magnetic fields and risk of childhood leukemia. A m J Epidemiol 1991;143:923-37. Lowengart R, Peters J, Ciconi C, Buckley J, Bernstein L, Preston-Martin S and Rappaport E. Childhood leukemia and parents' occupation and home exposures. JNCI 1987;79:3946. MacMahon B. Is acute lymphoblastic leukemia in children virus-related? A m J Epidemiol 1992;136(6):916-24. McBride M and Gallagher R. Cancer in British Columbia: Incidence and Mortality 19741984. Monograph No. 1;1987, and McBride M . Cancer in British Columbia: Incidence and Mortality 1984-1988. Monograph No. 3; 1991 B C Cancer Agency, Division of Epidemiology, Biometry and Occupational Oncology. McBride M . Childhood cancer and environmental contaminants. Can J Public Health 1998;89(Suppl 1):S53-S62. McKinney PA, Alexander FE, Cartwright R A and Parker L. Parental occupations of children with leukemia in West Cumbria, North Humberside, and Gateshead. Br Med J 1991;302:681-6. McLaughlin JR, King WD, Anderson TW, Clarke E A and Ashmore JP. Paternal radiation exposure and leukemia in offspring: the Ontario case-control study. Br Med J 1993;307:959-66. 144  McWhirter WR, Dobson C and Ring I. Childhood cancer incidence in Australia, 1982 1991 Int J Cancer 1996;65:34 - 8. Meinert R, Kaatsch P, Kaletsch U , Krummenauer F Miesner A and Michaelis J. Childhood leukemia and exposure to pesticides: results of a case-control study in northern Germany. Eur J Cancer 1996;32A(ll):1943-8. Meister RT. Farm Chemicals Handbook '90, Willoughby OH. Meister Publishing Company, pC221. Microsoft Excel, User's Guides 1 and 2: Spreadsheet with Business Graphics and Database, Version 4.0. 1992 Microsoft Corporation. Miettinen OS. Estimability and estimation in case-referent studies. A m J Epidemiol 1976;103:226-35. M i l i F, Khoury M J , Flanders WD and Greenberg RS. Risk of childhood cancer for infants with birth defects. 1. A record linkage study, Atlanta, Georgia 1968-1988. A m J Epidemiol 1993;137(6):629-38. M i l i F, Lynch CF, Khoury M J , Flanders WD and Edmonds L D . Risk of childhood cancer for infants with birth defects. 1. A record linkage study, Iowa 1983-89. A m J Epidemiol 1993;137(6):639-44. Miller RW. Deaths from childhood leukemia and solid tumors among twins and other sibs in the United States, 196-67. J Natl Cancer Inst 1971;46:203-9. Miller RW. Prenatal origins of cancer in man: epidemiological evidence. In, Transplacental Carcinogenesis. IARC Scientific Publications No. 4, Lyon, 1973. Miller RW. Genetic effects of atomic-bomb exposure. In, Male-Mediated Developmental Toxicity. Olshan A F and Mattison DR, eds. Plenum Press, New York, 1994;pp 205-209. Moller H . Work in agriculture, childhood residence, nitrate exposure, and testicular cancer risk: a case-control study in Denmark. Cancer Epidemiol Biomarkers Prev 1997;6:141-4. Mulvihill JJ. Reproductive outcomes among men treated for cancer. In, Male-Mediated Developmental Toxicity. Olshan A F and Mattison DR, eds. Plenum Press, New York, 1994; pp 197-203. Mulvihill JJ. Childhood cancer, the environment and heredity. In, Principles and Practice of Pediatric Oncology, 2 edition. Pizzo P A and Poplack DG., eds. JB Lippincott Company Philadelphia 1994;ppl 1-23. nd  145  National Research Council of Canada. Polychlorinated dibenzo-p-dioxins: criteria for their effects on man and his environment. N R C Associate Committee on Scientific Criteria for Environmental Quality, 1981. Newcombe, H B , Kennedy J M , Axford SJ and James AP. Automatic linkage of vital records. Science 1959;130:954-9. O'Leary L M , Hicks A M , Peters J M and London S. Paternal occupational exposures and risk of childhood cancer: a review. A m J Ind Med 1991;20:17-35. Olsen J, de Nully Brown P, Schulgen G and Moller-Jensen O. Parental employment at time of conception and risk of cancer in offspring. Eur J Cancer 1991 ;8:958-65. Olsen JH, Boice JD, Seersholm N , Bautz A , Fraumeni Jr. JF. Cancer in the parents of children with cancer. N Engl J Med 1995;333(24): 1594-9. Olshan A F , Baird P A and Teschke K. Paternal occupational exposure and the risk of Down syndrome. A m J Hum Genet 1989;44:646-51. Olshan A , Breslow N , Daling J, Weiss N and Leviton A. Childhood brain tumours and paternal occupation in the aerospace industry. JNCI 1986;77:17-9. Olshan A F and Faustman E M . Male-mediated developmental toxicity. Annu Rev Publ Health 1993;14:159-81. Olshan A F and Mattison DR. Preface, in Male-Mediated Developmental Toxicity. Olshan A F and Mattison DR, eds. Plenum Press, New York, 1994. Olshan A F , Teschke K and Baird PA. Paternal occupation and congenital anomalies. A m JInd Med, 1991;20:447-75. Ostry A. Discussion Paper: Can the sawmill worker's database be used for historical/socio-economic purposes? 1993, U B C . Document prepared for interdepartmental meeting. Parkin D M , Stiller C A , Draper GJ, Bieber C A , Terracini B and Young JL. International incidence of childhood cancer. IARC Scientific Publication No. 87. Oxford University Press 1988. Pearce N E , Smith A H , Howard JK, Sheppard RA, Giles HJ and Teague C A . NonHodgkins lymphoma and exposure to phenoxyherbicides, chlorophenols, fencing work and meat works employment: a case-control study. Br J Ind Med 1986;43(2):75-83. Pershagen G, Ericson A and Otterblad-Olausson P. Maternal smoking in pregnancy: does it increase the risk f childhood cancer? Int J Epidemiol 1992;21(l):l-5. 146  Peters J, Preston-Martin S and Y u M . Brain tumours in children and occupational exposures of parents. Science 1981;213:235-7. Preston-Martin S. Epidemiological studies of perinatal carcinogenesis.. In Napalkov NP, Rice J M , Tomatis L and Yamasaki H (eds), Perinatal and Multigeneration Carcinogenesis, IARC Scientific Publication no. 96. International Agency for Research n Cancer, Lyon; 1989:pp289-314. Proceed with Care: Final Report of the Royal Commission on New Reproductive Technologies. Chapter 13: Exposure to Harmful Agents in the Workplace and the Environment and Infertility 1993; pp269-302. Minister of Government Services. Ramlow J M , Spadacene N W , Hoag SR, Stafford B A , Cartmill JB and Lerner PJ. Mortality in a cohort of pentachlorophenol manufacturing workers, 1940 - 1989. A m J Ind Med 1996;30:180-94. Rice J M . Exposure to chemical carcinogens during pregnancy: consequences for mother and conceptus. Adv Exp Med Biol 1984;176:13-49. Rice J M . Effects of prenatal exposure to chemical carcinogens and methods for their detection. In Kimmel C A , Buelke-Sam J, eds. Developmental Toxicology. 1986, New York, N Y Raven Press: ppl91-212. Roberts HJ. Aplastic anemia due to pentachlorophenol. New Engl J Med 1981 ;305:16501. Roberts HJ. Aplastic anemia and red cell aplasia due to pentachlorophenol. South Med J.1983;76:45-8. Robison L L . General principles of the epidemiology of childhood cancer. In, Principles and Practice of Pediatric Oncology, 2 edition. Pizzo P A and Poplack DG., eds. JB Lippincott Company Philadelphia 1994;pp3-l 1. nd  Robison L L , Buckley JD and Bunin G. Assessment of environmental and genetic factors in the etiology of childhood cancers: the Children's Cancer Group epidemiology program. Environ Health Perspect 1995;103(suppl 6):111-6. Ron E, Modan B, Boice JD, Alfandary E, Stovall M , Chetrit A , and Katz L. Tumours of the brain and nervous system after radiotherapy in childhood. N Engl J Med 1988;31:1033-9. Ross JA, Severson R K , Robison L L , Pollack B H , Neglia JP, Woods W G , and Hammond GD. Pediatric cancer in the United States. Cancer Suppl 1993;71(10);3415-21.  147  Rottluff W, Teschke K , Hertzman C, Kelly S and Bert J. Sources of dioxins and furans in British Columbia. Can J Public Health 1990;81:94-100. Rudak E , Jacobs P A and Yanagamuchi R. Direct analysis of the chromosome constitution of human spermatozoa. Nature, 1978;274:911. Rugman FP and Cosstick R. Aplastic anemia associated with organo-chlorine pesticides: case reports and review of evidence. J Clin Pathol 1990;43:98-101. Russell L B , Russell W L , Rinchik E M and Hunsicker PR. Factors affecting the nature of induced mutations. In, Biology of Mammalian Germ Cell Mutagenesis, eds. Allen JW, Bridges B A , Lyon M F , Moses M J and Russell L B , 1990;34:271-85. Cold Spring Harbor Lab Press, New York. Russell L B and Shelby M D . Tests for heritable genetic damage and for evidence of gonadal exposure in mammals. Mutat Res, 1985;154:69-84. Sankila R, Olsen JH, Anderson H, Garwicz S, Glattre E, Hertz H, Langmark F, Lanning M , Moller T and Tulinus H . Risk of cancer among offspring of childhood-cancer survivors. New Engl J Med, 1998;338(9): 1339-44. Saracci R, Kogevinas M , Bertazzi P A et al. Cancer mortality in workers exposed to chlorophenoxy herbicides and chlorophenols. The Lancet. 1991;338:(8774):1027-32. Sarasua S and Savitz D. Cured and broiled meat consumption in relation to childhood cancer: Denver, Colorado (United States). Cancer Causes Control 1994;5:141-8. Savitz D A and Chen J. Parental occupation and childhood cancer: a review of epidemiologic studies. Environ Health Perspect 1990;88:325-37. Savitz D A , Sonnenfeld N L and Olshan A F . Review of epidemiological studies of paternal occupational exposure and spontaneous abortion. A M J Ind Med, 1994;25:361-83. Savitz D A , Wachtel H , Barnes F A , John Em, and Tvrdik JG. Case-control study of childhood cancer and exposure to 60-Hertz electric and magnetic fields. A m J Epidemiol 1988;128:10-20. Schwetz B A , Keeler P A and Gehrig PJ. The effect of purified and commercial grade pentachlorophenol on rat embryonal and fetal development. Toxicol Appl Pharmacol 1974:28;151-61. Schwetz B A , Quast JF, Keeler PA, Humiston C G and Kociba RJ. Results of two-year toxicity and reproduction studies on pentachlorophenol in rats. In, Pentachlorophenol, Ranga Kao ed, Plenum Press, New York, 1978 pp 301-9.  148  Seiler JP. Pentachlorophenol. Mutation research 1991;257:27-47. Sharpe R M and Skakkebaek N E . Are oestrogens involved in falling sperm counts and disorders of the male reproductive tract? Lancet 1993 ;341; 13 92-5. Sheilds PG. Inherited factors and environmental exposures in cancer risk. J Occup Med. 1993;35(1):34-41. Shu X - O , Ross JA, Pendergrass TW, Reaman G H , Lampkin B and Robison L L . Parental alcohol consumption, cigarette smoking, and risk of infant leukemia: a Children's Cancer Group study. JNCI 1996;88(1):24-31. Skakkebaek N E . Trends in male reproductive health: environmental aspects. Paper presented at Hazardous Substances and Male Reproductive Health International Conference, New York Academy of Medicine, New York City, May 1998. Smith A H , Pearce N E , Fisher DO, Giles HJ, Teague C A and Howard JK. Soft tissue sarcoma and exposure to phenoxyherbicides and chlorophenols in New Zealand. J Natl Cancer Inst 1984;73:1111-7. Sobolev B , Likhtarev I, Heidenreich WF and Jacob P. Thyroid cancer incidence among children in the Ukraine after the Chernobyl accident. Paper presented at the Canadian Society for Epidemiology and Biostatistics conference, London, Ontario, May 1997. Sorahan T, Lancashire R, Hulten M A , Peck I and Stewart A M . Childhood cancer and parental use of tobacco: deaths from 1953-1955. Br J Cancer 1997;75(1): 134-8. Stat Bite . Cancer survival in children. JNCI 1997;89(3):191. StataCorp. 1997. Stata Statistical Software: Release 5.0. College Station, T X : Stata Corporation. Statistical Package for Social Sciences, version 7.0, SPSS Inc. Statistics Canada. Cancer in Canada 1996. Steinbuch M . The role of environmental exposures in the etiology of childhood acute myeloid leukemia. Dissertation abstract 1994, Ohio State University. Stewart A , Webb J, Giles D and Hewitt D. Malignant disease in childhood and diagnostic irradiation in utero. Lancet 1956;ii:447. Stiller C A , Allen M B and Eatock E M . Childhood cancer in Britain: The national registry of childhood tumours and incidence rates 1978-1987. Eur J Cancer 1995;31A(12):202834. 149  Strohmer H , Boldizar A , Plockinger B, Feldner-Busztin M and Feichtinger W. Agricultural work and male infertility. A m J Ind Med 1993;24:587-92. Swan S. Is semen quality influenced by environmental factors? What can be learned from international collaboration? Paper presented at Hazardous Substances and Male Reproductive Health International Conference, New York Academy of Medicine, New York City, May 1998. Teschke K , Hertzman C, Dimich-Ward and Ostry A , Blair J and Hershler R. A comparison of exposure estimates by worker raters and industrial hygienists. Scand J Work Environ Health. 1990;15:424-9. Teschke K , Hertzman C and Morrison B. Level and distribution of employee exposures to total and respirable wood dust in two Canadian sawmills. A m J Ind Med 1993. Teschke K , Marion SA, Jin A , Fenske R A and van Netten C. Strategies for determining occupational exposures in risk assessment: a review and a proposal for assessing fungicide exposure in the lumber industry. A M J Ind Hygiene Assoc J. 1994;55(5):443- 9. Teschke K , Hertzman C, Fenske RA, Jin A , Ostry A and Liess W. A history of process and chemical changes for fungicide application in the Western Canadian lumber industry: what can we learn? Appl Occup Environ Hyg 1994a;9:984-93. Teschke K , Marion SA, Ostry A , Hertzman C, Hershler R, Dimich-Ward H and Kelly S. Reliability of retrospective chlorophenol exposure estimates over five decades. A m J Ind Med 1996;30:616-22. Teschke K, Ostry A , Hertzman C, Demers P, Barroetavena M , Davies H , Dimich-Ward H , Heacock H and Marion SA. Opportunities for a broader understanding of work and health: multiple uses of an occupational cohort database. Can J Public Health 1998;89(2): 132-6. Thomas RD. Age-specific carcinogenesis: environmental exposure and susceptibility. Environ Health Perspect 1995;103(suppl 6):45-8. Tomatis L. Prenatal exposure to chemical carcinogens and its effect on subsequent generations. Natl Cancer Inst Monog 1979;51:159-84. Tomatis L, Narod S and Yamasaki H. Commentary: Transgeneration transmission of carcinogenic risk. Carcinogenesis 1992;13(2): 145-51 Treble R G . And Thompson TS. Normal values for pentachlorophenol in urine samples collected from a general population. J Anal Toxicol 1996;20:313-17.  150  US National Toxicology Program. Toxicology and Carcinogenesis Studies of Two Pentachlorophenol Technical-grade Mixtures (CAS No. 87-85-5) in B6C3F, Mice (Feed Studies) (Technical Report Series No. 349; N I H Publ. No. 89-2804), Research Triangle Park, North Carolina, USA. Valo R, Kitunen V , Salkinoja-Salonen N and Raisanen S. Chlorinated phenols as contaminants of soil and water in the vicinity of 2 Finnish sawmills. Chemosphere 1984:13;835-844. Van Steensel-Moll H A , Valkenburg H A , and Van Zanen GE. Childhood leukemia and parental occupation: a register-based case-control study. 1985 A m J Epidemiol; 121:21624. ViannaNH, Kovasznay B, Polan A. et al. Infant leukemia and paternal exposure to motor vehicle exhaust fumes. J Occup Med 1984;26:679-82. Vorobtsova I and Kitaev E. Urethane-induced lung adenomas in the first generation progeny of irradiated male mice. Carcinogenesis 1988;9:1931-4. Welsh JJ, Collins T F X , Black T N , Graham SL and 0'Donnell,MW Jr. Teratogenic potential of purified chlorophenol and pentachloroanisole in subchronically exposed Sprague-Dawley rats. Food Chem Toxicol 1987:25;163-72. Wertheimer N and Leeper E. Electrical wiring configurations and childhood cancer. A m J Epidemiol 1979; 109:273-84. Whorton D, Kraus R M , Marshall S and Milbt TH. Infertility in male pesticide workers, Lancet 1977;2:1259-61. Whorton D, Haas JL, Trent L and Wong O. Reproductive effects of sodium borates on male employees: birth rate assessment. Occup and Environ Med 1994;51:761-67. Wilkins J III and Koutras R. Paternal occupation and brain cancer in offspring: a mortality based case-control study. A m J Ind Med 1988;14:299-318. Wilkins J III and Sinks T. Parental occupation and intracranial neoplasms in childhood: results of a case-control interview study. A m J Epidemiol 1990;132:275-92. Wilkins J III, McLaughlin J, Sinks T and Kosnik E. Parental occupation and intracranial neoplasms of childhood: anecdotal evidence from a unique occupational cancer cluster. A m J Ind Med 1991;19:643-53. Wood SD and White GL. Pentachlorophenol. In Environmental and Occupational Medicine. Rom W N , ed. 2 edition, Little, Brown and Company 1992; pp 1027-31. nd  151  Woods JS, Polissar L, Severson R K , Heuser LS and Kulander B G . Soft tissue sarcoma and non-Hodgkin's lymphoma in relation to phenoxyherbicide and chlorinated phenol exposure in Western Washington. J Nat Cancer Inst 1987;899-910. Wyrobek A J . Methods and concepts in detecting abnormal reproductive outcomes of paternal origin. Reprod Toxicol 1993;7, Supple 1:3-16 Wyrobek A J and Adler ID. Detection of aneuploidy in human and rodent sperm using FISH and applications of sperm assays of genetic damage in heritable risk evaluation. Mutation Res 1996;352:173-79. Yamasaki H , Lokionov A and Tomatis L. Perinatal and multigenerational effect of carcinogens: possible contribution to determination of cancer susceptibility. Environ Health Perspect 1992;98:39-43. Zack M , Cannon S, Lloyd D et al. Cancer in children of parents exposed to hydrocarbonrelated industries and occupations. A m J Epidemiol 1980; 111:329-36.  152  Appendix A: Power calculations completed for thesis proposal  Age  Person years  All Cancers  1969-74  61,467  16.1  15,278  Astrocytoma  Medulloblastoma  NHL  HD  3.0  1.4  0.6  0.8  0.9  22 .3  6.1  4.0  0.8  0.5  0. 6  14,983  12.6  3.5  1.3  0.5  0.6  0.7  16,678  10. 6  1.0  1.7  0.7  0.7  0. 8  14,528  20.9  2.2  0.8  0.3  1.3  1.5  52,299  19.8  2.8  1.2  0.5  0.7  0.8  13,199  21.7  6.8  1.8  0.7  0.3  0.4  12,866  9.1  2.4  1.2  0.5  0.2  0.3  12,295  11.4  1.0  1.2  0.4  0.5  0. 6  13,939  32 .5  1.8  0.9  0.4  1.3  1. 6  1930-84  53,344  22.4  3.0  1.8  0.7  0.7  0.8  0_- 4  14,984  21.2  6.1  2.4  0.9  0.2  0.2  5-  13,199  13.2  3.1  1.8  0.7  0.4  0.5  12,866  14.7  1.8  1.8  0.7,  1.1  1.3  15  12,295  37.7  1.36  1.5  0.6  1.1  1.3  1935-39  53,200  20.7  3.2  1.6  0.6  0.9  1.0  0-4  12,151  26.2  7.0  2.6  1.0  0.4  0.5  5- 9  14,934  12.1  2.7  1.6  0.6  0.5  0.6  fQ - 14  13,199  13.8  2.5  1.1  0.4  0.9  1.1  15 - 19  12,866  29.8  1.5  1.1  0.4  1.7  2.0  TOTAL  220,310  19.6  3.0  1.5  0.6  0.7  0.9  (43.2)  (6.6)  (3.3)  (1.3)  (1.5)  (2.0)  5 - 9 tO -  14  (5 - 19 1975-79 0-  4  $-  9  jO ~ 14 15  -  19  9  10 -  14  txpacted cases ()  •1S3  Appendix B: Crude estimate of SIRs for offspring cohort if they are studied until 2007. Assumptions needed were that the provincial childhood cancer incidence rates remained at the 1990-1993 levels and the rate of cancer development in the cohort between 1969 1993 remains constant between 1994 - 2007. The initial step involved tallying personyears left in the study for all offspring less than twenty years of age in 1993 until they reached twenty years of age. The next step involved dividing the number of incident cases identified in the study by the total person-years contributed from 1969 - 1993. This incidence rate, or incidence density, is a measure of the instantaneous rate of development of disease in the offspring cohort (Hennekens and Buring, 1987). This rate was then multiplied by the remaining person-years in the study (until the year 2007) to provide an estimate of the number of new cases expected, ("observed cases"). Then, the most recent estimate (1990 - 1993) of the provincial cancer incidence rates for the age group 0-19 was multiplied by the total person-years in the study between 1969 - 2007 to provide the "expected" number of cases. Incidence rates have been increasing since 1969 (incidence rates in B C for girls have increased from approximately 12.5/100,000 to 16.6/100,000 between 1969 and 1993). It is expected that they will continue to increase, hence using the 1990-93 estimates will probably slightly underestimate the incidence rates that will occur between 1994 and 2007. Using girls as an example, the following calculations provide the SIR and 95% CI if the offspring cohort is followed until all the girls reached twenty years of age. (1) Between 1994 - 2007 there would be 25,771 person-years left in the study.  154  Appendix B , cont. (2) Between 1969 - 1993, the girls whose fathers' worked in the chlorophenate-using mills contributed 125,826 person-years (altogether there would be 151,597 person-years between 1969 - 2007), and (3) 22 cases of childhood cancer were identified between 1969 - 1993. Hence: (1) Incidence Density (ID): 22/125,826 = 0.000174844 or 17.48/100,000. (2) Cases expected to be diagnosed between 1994 - 2007: ID x person-years available from 1994 - 2007 = 0.000174844 x 25,771 = 4.5 = 5 further cases. (3) Revised "expected cases" based on 1990 -1993 provincial incidence rates among girls aged 0 -19 and person-time contributed between 1969 - 2007: 16.63/100,000 x 151,597 person-years = 25.21 expected cases. (4) Revised "observed cases" based on another 5 cases occurring by study end: 22 + 5 = 27 observed cases. (5) Revised SIR and 95% CIs: 27/25.21 = 1.07 (0.71 - 1.56). Similar calculations were performed for the boys and for both sexes combined. For the boys, the revised SIR and 95% CIs are: 22/25.5 = 0.88 (0.55 - 1.32) For both sexes, the revised SIR and 95% CIs are: 48/50.4 = 0.96 (0.71 - 1.27)  The above is a crude estimation where adjustment for age and sex has not been accounted for. The point of the exercise is to show that 83% of all potential person-years are already accounted for in the follow-up from 1969 - 1993 and it is unlikely many more cases of cancer will occur before 2007. 155  156  Appendix C: Childhood cancer (ages 0 -19) among female offspring of sawmill workers employed in both chlorophenate-using and non-user mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  Observed Cases  Expected Cases  SIR  (95% CI)  All cancers  23  19.8  1.16  0.74-1.74  Leukemias  5  5.5  0.91  0.29-2.12  • ALL  3  3.4  0.87  0.18-2.54  Brain cancer  7  3.3  2.14  0.86-4.41  •  Ependymoma  2  0.3  6.97  0.84-25.16  •  Astrocytoma  4  1.8  2.24  0.61-5.73  All lymphomas  2  2.8  0.73  0.09 - 2.64  •  Hodgkin's disease  1  2.1  0.47  0.01-2.62  •  Non-Hodgkin's lymphoma 1  0.6  1.56  0.04-8.69  2  0.6  3.30  0.40-11.91  Bone and connective tissue 2  2.1  0.94  0.11-5.74  Ovary-  2  0.8  2.47  0.30-8.92  Liver  1  0.3  3.34  0.08-18.60  Eye  157  Appendix C, cont: Childhood cancer (ages 0-19) among male offspring of sawmill workers employed in both chlorophenate-using and non-user mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  All cancers  Observed Cases  18  Expected Cases  SIR  (95% CI)  22.3  0.81  0.48 - 1.28  Leukemias  7.0  0.86  0.32-1.87  ALL  4.9  0.82  0.22-2.10  Brain cancer  4.3  0.70  0.14-2.04  Ependymoma  0.5  0  •  Astrocytoma  1  2.1  0.48  0.01-2.67  All lymphomas  2  3.6  0.56  0.07 - 2.02  •  Hodgkin's disease  1  0.9  1.14  0.03 - 6.35  •  Non-Hodgkin's lymphoma 1  2.7  0.37  0.01 - 2.06  0  0.6  0  Bone and connective tissue 3  2.5  1.20  0.25 - 3.50  Liver  0.2  4.22  0.11-23.51  Eye  1  158  Appendix C, cont. Childhood cancer (ages 0 -19) among all offspring of sawmill workers employed in both chlorophenate-using and non-user mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  Observed Cases  Expected Cases  SIR  (95% CI)  All cancers  41  42.0  0.98  0.67-1.33  Leukemias  11  12.5  0.88  0.44-1.58  7  8.3  0.84  0.34-1.73  10  7.5  1.33  0.64 - 2.44  • ALL Brain cancer •  Ependymoma  2  0.8  2.53  0.31-9.13  •  Astrocytoma  5  3.9  1.30  0.42-3.03  All lymphomas  4  6.3  0.63  0.17-1.61  •  Hodgkin's disease  2  3.0  0.67  0.08 - 2.42  •  Non-Hodgkin's lymphoma2  3.3  0.6  0.07-2.17  2  1.2  1.67  0.20-6.03  Bone and connective tissue 5  4.6  1.09  0.35-2.54  Liver  0.5  3.70  0.45-13.36  Eye  2  159  Appendix D: Childhood cancer (ages 0-14) among female offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  Observed Cases  Expected Cases  SIR  (95% CI)  All cancers  17  12.3  1.38  0.80-2.21  Leukemias  4  4.2  0.96  0.26 - 2.46  • ALL  2  2.8  0.72  0.12-2.60  Brain cancer  5  2.5  1.98.  0.64-4.61  •  Ependymoma  1  0.2  4.11  0.10-22.89  •  Astrocytoma  3  1.3  2.30  0.47-6.72  All lymphomas  2  1.0  2.03  0.24 - 7.33  •  Hodgkin's disease  1  0.6  1.74  0.04 - 9.69  •  Non-Hodgkin's lymphoma 1  0.4  2.44  0.06-13.59  2  0.5  3.89  0.47-14.04  Bone and connective tissue 0  1.4  0  Ovary  2  0.4  5.55  0.67-20.04  Liver  1  0.2  4.80  0.12-26.74  Eye  160  Appendix D, cont: Childhood cancer (ages 0 -14) among male offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  Observed Cases  Expected Cases  SIR  (95% CI)  All cancers  13  14.7  0.88  0.47-1.50  Leukemias  6  5.4  1.12  0.41-2.44  • ALL  4  3.8  1.07  0.29 - 2.74  Brain cancer  2  3.3  0.62  0.07 - 2.24  Ependymoma  0  0.3  0  -  •  Astrocytoma  0  1.4  0  -  AH lymphomas  1  1.6  0.62  0.08 - 2.24  •  Hodgkin's disease  1  0.2  4.18  0.11 -23.28  •  Non-Hodgkin's lymphoma 0  1.4  0  -  0  0.5  0  -  Bone and connective tissue 1  1.2  0.81  Liver  0.2  0  Eye  0  0.10-2.92  161  Appendix D, cont. Childhood cancer (ages 0 -14) among all offspring of sawmill workers employed in chlorophenate-using mills: Observed and expected cases, SIRs and 95% CIs.  Type of cancer  Observed Cases  Expected Cases  SIR  (95% CI)  All cancers  30  27.1  1.11  0.75 - 1.59  Leukemias  10  9.5  1.05  0.50-1.93  • ALL  6.5  0.92  0.34 - 2.00  Brain cancer  5.8  1.21  0.49 - 2.49  Ependymoma  1  0.6  1.75  0.04 - 9.75  Astrocytoma  3  2.8  1.10  0.23 - 3.21  AH lymphomas  3  2.6  1.15  0.24 - 3.36  •  Hodgkin's disease  2  0.8  2.44  0.30-8.81  •  Non-Hodgkin's lymphoma 1  1.8  0.56  0.01-3.12  2  1.0  2.0  0.24 - 7.22  Bone and connective tissue 1  2.6  0.38  0.01-2.12  Liver  0.4  2.44  0.06-13.60  •  Eye  1  162  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items