Authors
G Van Maele-Fabry, J L Willems.

Title
Occupation related pesticide exposure and cancer of the prostate: a meta-analysis.

Source
Occupational and Environmental Medicine: 60(9):634-642, Sep 2003.

Summary
Genetic predisposition, age and ethnic origin are considered as major risk factors for prostate cancer. Evidence regarding environmental and occupational risk factors is inconclusive. In prior studies, agricultural exposures were found to be associated with prostate cancer but specific agents were not clearly identified. The evidence is strong for pesticides as potential risk factors but not conclusive. Prior meta-analyses of epidemiological studies in farmers were suggestive of slightly increased risks of prostate cancer.

The present meta-analysis focused on identifying the relationship between pesticides and prostate cancer in both farming and other pesticide related occupations. Literature searches using Medline and POLTOX databases for the period 1995 to September 2001 identified 43 case-control, cohort, or proportionate mortality ratio (PMR) studies. Studies were excluded if they did not report original results, reported less than 5 cases, used data resulting from accidental exposures, analyzed other occupational groups not included in the current analysis, and were not published in English. In studies with multiple publications, only the most relevant was included. Both agricultural occupational groups (farmers, farm workers, pesticide applicators and loaders, agricultural pesticide applicators, crop duster pilots, and flaggers for crop duster pilots) and non-agricultural pesticide exposed groups (nursery and greenhouse workers, chemical lawn care workers, golf course workers, park maintenance workers, and landscape maintenance workers) were included in the meta-analysis. Following exclusions, a total of 22 studies (11 cohort studies, 7 case-control studies, 4 PMR studies) and resulting 25 rate ratio (RR) estimators were included in the meta-analysis.

A structured abstract form was created for the studies and rate ratio estimates with their 95% confidence intervals (CI) were tabulated separately by the two authors. Data was evaluated for homogeneity among studies and publication bias. The meta-rate ratio (1.13, 95% CI=1.04-1.22) obtained by applying a random effects model (because of heterogeneity among studies) on all 22 studies of prostate cancer in pesticide related occupations was suggestive of a weak association. No significant difference in pooled RR was observed after excluding proportionate mortality ratio studies which are generally considered to be of low quality. The meta-rate ratio was consistent with three prior meta-analyses of the literature published before 1995. The sensitivity analysis performed to determine the effect of studies with extreme RR estimates, high precision, the inclusion of prior redundant studies or studies that reported data for less than 5 cases on the pooled RR estimate indicated no significant change.

The possible sources for heterogeneity among studies were explored by subset analysis of following factors: study design (cohort studies, case-control studies, PMR studies), geographic location (US/Canada, Europe, or other), outcome data (cancer incidence or mortality), source of reference population (for the cohort studies: national/regional rates; for case-control studies: hospital, cancer, or general population controls), presence of a healthy worker effect (manifested as lower than expected all cancer mortality or incidence) and occupational category (pesticide applicators and farmers). Homogeneity was not revealed by using study design, outcome data, source of reference population, or occupation category as stratification variables. Stratification by geographic location and study design markedly reduced heterogeneity. Overall pooled rate ratios from Europe were lower than those derived from US/Canada. The possible reasons suggested were risks associated with American lifestyle factors (fat intake, obesity), genetic susceptibility, endocrine disruption (as both androgens and estrogens influence growth and regulation of prostate gland) and their interactions. Stratifying on presence of healthy worker effect revealed a markedly decreased heterogeneity in the group without healthy worker effect (HWE). However, the HWE for prostate cancer risk is expected to be minimal as patients are usually elders and the HWE tends to decline with increasing age. The pooled rate ratio estimate showed a significantly increased risk (no dose response relationship) in pesticide applicators but not in the broader category of farmers. No evidence of publication bias was observed.

The inclusion of other pesticide exposure groups, exploration of sources of heterogeneity, assessment of publication bias and the consistency of the results with prior meta-analyses contribute to the strengths of this analysis. The major weaknesses of the studies included in present meta-analysis were inadequate control for confounding and inadequate exposure assessment. The limited knowledge of risk factors for prostate cancer and inadequate information on potential confounders such as genetic predisposition restricted the control for confounding to age and calendar time in the studies.

Though the evidence was suggestive of a weak association between pesticide exposure groups and prostate cancer, more focused research is essential for identifying the specific agents related to prostate cancer. Future epidemiologic studies should focus on improving the data quality and exposure assessment strategies. Documenting the details such as frequency and duration etc., for quantifying exposure is particularly essential to address the causality and the dose-response relationship between occupational pesticide exposure and prostate cancer.