Selected characteristics of the study population are displayed in according to lifetime exposure-days category. In this table, “lowest exposed” refers to those in the lowest exposure tertile (> 0–20 lifetime exposure-days), whereas “other exposed” refers to those in the middle and highest tertiles (> 20 lifetime exposure-days). Overall, two-thirds of the study participants were from Iowa. More than two-thirds of the cohort reported corn farming. Study subjects were also predominantly white and male. Just over half reported being never smokers. Close to 55% reported that the highest level of schooling attained was no more than a high school diploma. Approximately 40% reported a history of cancer in first-degree relatives.
Characteristics of applicators by fonofos exposure in the AHS (1993–1997) [no. (%)].
The unexposed group was generally younger, less likely to use alcohol, and slightly less likely to report a family history of cancer, used fewer pesticides in general, and planted fewer acres than either exposed group. Both Iowa participants and corn farmers were over-represented in the exposed categories. Based on these differences between the unexposed group and either exposed group, the lowest exposed group may represent the exposed group more closely.
With the intensity-weighted metric, risk estimates for all cancers combined were not different from the null, regardless of the reference group used (). Colon cancer risk estimates were elevated, but only when using the unexposed as the reference, and the relationship was not monotonic. Leukemia risk estimates were elevated regardless of the reference group used. When the unexposed group was the reference, the RR was 2.67 (95% CI, 1.06–6.70) in the highest exposure category, and the test for linear trend was significant (ptrend = 0.04). When the lowest exposed group was the reference, the corresponding RR was 2.03 (95% CI, 0.58–7.05). The linear trend test was not significant. Fonofos intensity-weighted exposure-days were not related to the risk of any other examined cancer.
RRsa (95% CIs) for selected cancers by fonofos intensity-weighted exposure-days among AHS (1993–1997) applicators, using unexposed and lowest-exposed applicators as the reference groups.
Results were similar using the lifetime exposure-days metric (not shown). For example, using the unexposed group as the reference, leukemia RRs increased monotonically to 2.24 (95% CI, 0.94–5.34) in the highest tertile (ptrend = 0.07). When the lowest exposed tertile was used as the reference, the risk estimates increased monotonically with increasing exposure category to 2.18 (95% CI, 0.57–8.40) in the highest tertile. The test for linear trend was not significant.
To account for the effect of misclassification due to the inclusion of exposure that occurred too recently to affect cancer risk, we repeated the analyses excluding 39 cancer cases and 1,389 cancer-free subjects who either reported first using fonofos during the 1990s or did not provide this information. The results were similar to those presented here (not shown). The results were also similar after repeating the analyses among Iowa participants only (not shown). Additionally, to control for pesticide use in general, we repeated the analyses adjusting for lifetime exposure-days to all pesticides instead of the most correlated pesticides, and the results did not differ from those presented here (not shown). To evaluate the effect of missing information, we repeated the analyses while allowing subjects with missing information on covariates to influence the outcome by assigning them an unspecified category. Once again, the results were largely the same (not shown). Finally, the results were similar when we separately examined fonofos days of use and years of use after categorizing each into none, low, and high categories for exposure, using the median to distinguish between low and high (not shown).
We further investigated leukemia by separately examining chronic lymphocytic (eight exposed cases), chronic myelogenous (two exposed cases), acute myelogenous (five exposed cases), and all other leukemias (three exposed cases) (not shown). Acute lymphocytic leukemia could not be evaluated (no exposed cases). Although the CIs were wide and the point estimates were not significant, relative to the unexposed, the age-adjusted risk estimates in low- and high-exposure categories were elevated for all examined subtypes. In the high-exposure category, point estimates ranged from 1.75 for acute myelogenous to 3.65 for chronic myelogenous leukemia.
We also adjusted leukemia risk estimates using data on use of gasoline, solvents, and paint, which were collected among private applicators using the take-home questionnaire (not shown). Although the subset of otherwise eligible applicators who provided the aforementioned information was small (eight exposed cases), adjusting for these exposures did not weaken the leukemia risk estimates. Finally, controlling for animal exposures using information on the number of livestock (other than poultry) or whether applicators butchered animals, provided veterinary services to livestock, or worked in swine or poultry containing areas, did not affect risk estimates (not shown).
shows prostate cancer risk relative to the unexposed using both metrics and stratified by family history of prostate cancer in first-degree relatives. We generated uniform exposure categories based on the exposure distribution among prostate cancer cases. In the group with no prostate cancer family history, risk was not associated with exposure regardless of the metric. In those with a family history of prostate cancer, the risk estimates increased, and significant linear trends were observed using either metric. Using the lifetime exposure-days metric, we observed a significant dose–response relationship (ptrend = 0.02), which resulted in a RR of 1.77 (95% CI, 1.03–3.05) in the highest exposure category. The interaction term, defined as the cross-product of family history of cancer and category of lifetime exposure-days (treated as a continuous variable), was significant (RR = 1.28; 95% CI, 1.07–1.54). With the intensity-weighted exposure-days metric, risk in the highest category was 1.83 (95% CI, 1.12–3.00). The test for linear trend was significant (ptrend = < 0.01), as was the interaction RR of 1.27 (95% CI, 1.07–1.51).
Prostate cancer RRs (95% CIs) among AHS (1993–1997) participants by family history of prostate cancer using fonofos lifetime and intensity-weighted exposure-days metrics.
When the analysis in was repeated using the lowest exposed group as the reference, the results were similar but less pronounced due to decreased statistical power (not shown). Risk was related to fonofos use only in those with a family history of prostate cancer. Point estimates increased monotonically with lifetime exposure-days to 1.24 (95% CI, 0.61–2.51) in the highest category. The interaction RR was 1.25 (95% CI, 0.83–1.89). Point estimates generally increased with intensity-weighted exposure-days to 1.68 (95% CI, 0.83–3.39) in the highest category. The interaction RR was 1.27 (95% CI, 0.85–1.89). Linear trend tests were not significant using either metric.
When the risk of the other examined cancers (all cancers combined, melanoma, leukemia, lymphohematopoietic cancers, lung cancer, and colon cancer) was similarly stratified, no discrepancies were observed comparing those with and without a family history of the specific cancer (not shown).
When we did further analyses to disentangle the effects of prostate cancer family history and fonofos exposure, we observed that the age-adjusted main effect for ever compared with never fonofos exposure was 0.97 (95% CI, 0.80–1.17), whereas for family history of prostate cancer, it was 1.67 (95% CI, 1.35–2.07). The observed joint effect of the two exposures was 2.63 (95% CI, 1.96–3.53).