The data for this study on HL come from the CCSPH, a population-based incident case–control study conducted among male residents in six Canadian provinces (Ontario, Quebec, Alberta, British Columbia, Manitoba, and Saskatchewan). Men aged 19 years or older with a first diagnosis of soft-tissue sarcoma (STS), NHL, multiple myeloma (MM), and HL between 1 September 1991 and 31 December 1994 were included. Each province was given a target number of cases for each cancer site to balance the number of cases by province. The study size was determined based on a priori sample size calculations, and cases were ascertained from provincial cancer registries, except in Quebec where hospital ascertainment was used. Cancer registries and hospitals provided information to confirm the diagnosis, including pathology reports, although subjects for whom pathological material was unavailable remained in the study. Deceased individuals were not eligible as cases or controls, and surrogates for deceased individuals were not contacted. Proxy respondents were allowed for subjects eligible for the study. Subjects who were known to be HIV-positive were excluded. For controls, men aged 19 and older were selected at random from the provincial health insurance records (Alberta, Saskatchewan, Manitoba, Quebec), computerized telephone listings (Ontario), or voters’ lists (British Columbia). Controls were frequency-matched by age (±2 years) to cases within each province.
The postal and telephone interview questionnaires were modified versions of the telephone interview questionnaires used by studies focusing on pesticide exposure and cancer in Kansas and Nebraska [13, 14]. The postal questionnaire for the CCSPH was sent to all cases and controls to capture information on demographic characteristics, medical history, and other variables that could be potential confounders. Based on the results of a pilot study , a cumulative total exposure of 10 h per year to any pesticide combination was determined to be an appropriate cutoff for intensive exposure to pesticides. Both occupational and non-occupational (home, garden, hobby) uses of pesticides were considered in this classification of cumulative exposure. Participants who reported 10 or more hours of cumulative pesticide exposure per year on the postal questionnaire and a 15 % random sample of the participants who reported less than 10 h were sent a document listing pesticides and then interviewed by telephone to collect more detailed information about each pesticide used. The information collected on each pesticide included where the pesticide was used (work, home, or garden), the number of acres sprayed or treated if used on the farm, the year the pesticide was first used, how many years and days/year it was used, and how many of these days the pesticide was personally handled. The pesticides included in the questionnaire were those that had ever been registered for use in Canada and reviewed by IARC; pesticides that were recently banned or restricted in Canada; and pesticides commonly used in Canada for specific purposes. A more detailed discussion of the recruitment procedure, study design, and data collection has been published elsewhere .
Exposure to multiple pesticides
We examined the risk of HL by the total number of pesticides ever used (0, 1, 2–4, and 5+ pesticides). Similar analyses were conducted for the number of pesticides used grouped by class, that is, herbicides, insecticides, and fungicides. Analyses were also conducted to examine the number of phenoxyherbicides used to further investigate the mixed findings of other studies and urea herbicides to confirm the findings of other studies.
The risk of HL was also evaluated according to the number of possibly carcinogenic pesticides used. These were pesticides considered possibly carcinogenic or higher by the International Agency for Research on Cancer (IARC) , suggestive evidence of carcinogenic potential or higher (2005), known/likely to be carcinogenic (1996), or possible human carcinogen or higher (1986 classification) by the United States Environmental Protection Agency (EPA) Office of Pesticides Program (OPP), or EPA Integrated Risk Information System (IRIS) [18, 19]. A more stringent level of carcinogenic pesticides (probably carcinogenic) was also used to evaluate and compare the risk of HL. A pesticide was considered probably carcinogenic if it was classified as probably carcinogenic to humans or higher by IARC , or likely to be carcinogenic to humans or higher (2005 classification), known/likely to be carcinogenic (1996 classification), or probable human carcinogen or higher (1986 classification) by EPA OPP, or the EPA IRIS [18, 19]. The list of pesticides classified as possibly and probably carcinogenic used in the analysis is presented in “Appendix 1”.
The risk of HL was also examined by the number of pesticides used with a similar mode of action in target species. These groups of pesticides were chosen for investigation if there was an adequate number of exposed individuals and there was more than one pesticide belonging to that group. Based on these criteria, the following modes of pesticidal action presented by Costa  were used for insecticides: (A) acetylcholinesterase inhibitor; (B) sodium channel activator; (C) GABA receptors-gated chloride channel inhibitor and for herbicides; and (D) auxin growth regulator (please refer to “Appendix 2” for the pesticides found in each group). Fungicides could not be grouped according to common modes of action due to small numbers.
These analyses have been restricted to the self-reported “ever/never” use of individual pesticides and various pesticide groupings. Since use and subsequent exposure to pesticides in the home and garden on a seasonal basis may be different from occupational settings, we repeated the above analyses evaluating home use and work use separately.
The risk of HL from reported use of various pesticides was evaluated using unconditional logistic regression. Interactions between age and the pesticide variables were also tested in each model for class, carcinogenicity, mode of action because of the bimodal age-incidence pattern for HL suggesting the possibility of two different etiologies. Those models with significant interaction terms were then stratified and analyzed by subjects younger than 40 and older than or equal to 40. Models for the multiple pesticides were adjusted for age, as a continuous variable, and province since controls were age-frequency-matched to cases in each province. Variables that might be potential confounders of the relationship between multiple pesticide exposure and the risk of HL were investigated. Based on a review of the literature, possible a priori confounders included ethnicity, diagnosis of cancer in an immediate family member, smoking, level of education, and use of a proxy respondent. The variable for ethnicity used in this study included the following categories: North American, Scandinavian, Eastern European, Western European, Asian, British, and other. The variable used for smoking was a binary variable for having ever smoked. Individuals with missing data for a variable of interest were placed into a “missing” category for that variable and were included in all models. Confounders that changed the odds ratios by at least 10 % were included in the final multivariate model. Trends for the number of pesticides used within each pesticide class were examined using the Cochran-Armitage exact trend test. All analyses were conducted using SAS version 9.2 (SAS Institute Inc.).
The University of Toronto Health Sciences Research Ethics Board reviewed and approved the protocol for these analyses.