In this study we found that increased symptom count was associated with cumulative lifetime use of pesticides, particularly insecticides and fumigants. Increased symptom count was also associated with a history of pesticide poisoning or events involving high personal pesticide exposure. Significantly, however, associations with cumulative use persisted even after excluding individuals with a history of pesticide poisoning or high exposure events. Recent pesticide use, within the year before reporting symptoms, was not related to symptom count after accounting for cumulative exposure, and adjustment for recent use did not affect the association of cumulative use with symptom count.
Although poisoning by high exposures to organochlorines, fungicides, and fumigants as well as organophosphates is well documented, and carbamates, pyrethroids, and herbicides are also neurotoxic (Ecobichon 1996
; Keifer and Mahurin 1997
), questions remain concerning the effects of moderate exposure to pesticides other than organophosphates. One study of moderate exposure found that dichlorodiphenyltrichoroethane (DDT) was associated with increased symptom prevalence (van Wendel de Joode et al. 2001
), as did one study of fumigants (Anger et al. 1986
), but not another (Calvert et al. 1998
). We found that symptom count was related to all classes of pesticides examined, although associations with herbicides and fungicides appeared to be due to confounding by insecticide use. Organophosphate, carbamate, and organochlorine insecticides were independently associated with increased risk. The relative neurotoxicity of specific chemicals or chemical classes may differ for acute high-level and chronic moderate exposure. For example, the stronger effects that we observed for organochlorines may be related to their long biologic half-lives (Ecobichon 1996
Few previous studies were able to distinguish between effects of acute and chronic exposure because the two are often correlated. Two studies with sufficient information to make the distinction found that in farmworkers who applied pesticides increased symptom prevalence was associated with acute but not chronic exposure (London et al. 1998
; Ohayo-Mitoko et al. 2000
). In contrast, our results suggest that at moderate levels cumulative lifetime exposure has a greater impact on symptom prevalence than exposure during the year before reporting symptoms. This disparity may be due to the higher level of exposure experienced by farm workers compared with licensed applicators.
The role of pesticide poisoning in the apparent effects of cumulative use is still a question. We confirmed previous reports that a history of pesticide poisoning is associated with increased symptom prevalence (Kamel and Hoppin 2004
). A notable finding in our study is that a history of events involving high personal pesticide exposure conferred equally great risk, even in the absence of diagnosed poisoning. Some studies have not differentiated exposed individuals with a history of pesticide poisoning from those without. Two studies that excluded poisoned individuals found no relationship of moderate organophosphate exposure to symptom prevalence (Ames et al. 1995
; Fiedler et al. 1997
), although a study of DDT that excluded poisoned individuals did find an association (van Wendel de Joode et al. 2001
). We found dose-related associations of symptom count to cumulative exposure to all insecticides, organophosphates, and organochlorines whether or not we excluded individuals with a history of pesticide poisoning or those who had experienced high-exposure events, indicating that moderate exposure itself is associated with increased risk.
Our findings were similar regardless of whether we considered summary measures of all symptoms, individual symptoms, or symptom groups defined a priori
. These results are consistent with previous studies suggesting that moderate pesticide exposure is associated with a wide range of symptoms, reflecting cognitive, sensory, and motor dysfunction and affecting both the central and peripheral nervous systems (Kamel and Hoppin 2004
). Pesticide exposure may be associated with some fundamental disorder, such as depression or neurologic disease, which then influences the experience or perhaps the reporting of multiple symptoms. Similarly, confounding by head injury, which was related to pesticide exposure in the AHS cohort, might explain some of the increase in symptoms. However, our findings were not affected by excluding individuals with depression, neurologic disease, or head injury. The earliest manifestation of neurotoxicity after moderate pesticide exposure may in fact be an increase in many symptoms, not restricted to particular aspects of neurologic function. A similar increase in a wide range of symptoms is associated with solvent exposure in mild cases of chronic solvent-related toxic encephalopathy (van der Hoek et al. 2000
; White and Proctor 1997
Confounding by demographic factors does not appear to explain our results. There was a strong inverse association of symptom prevalence with age. The basis of this association is unclear; it may represent participants’ understanding of the symptom questions or reporting proclivities rather than a real relationship. Other explanations are possible. Symptomatic individuals may have left farming at an early age and thus never entered our cohort, representing a type of healthy worker effect. Younger applicators used more pesticides in the year before enrollment, the period for which symptom prevalence was reported; however, adjusting for recent use did not affect associations with cumulative use. In any case, because symptom prevalence decreased and cumulative exposure increased with age, confounding by age or age-related conditions like heart disease or diabetes cannot explain the positive associations we observed with cumulative exposure; moreover, excluding individuals with the latter conditions did not affect our results. This point is particularly important in interpreting results for organochlorine pesticides. Secular trends in use mean that older applicators are more likely to have used these chemicals, but this cannot account for the association with symptoms because age was inversely related to symptom count. We adjusted for other potential confounders, including education, so these are also unlikely to account for our results. We had no information on personality traits that may have affected symptom reporting, and so could not adjust for these, but they are unlikely to have covaried with exposure, particularly in a way that could account for the dose–response relationships we observed.
Potential bias is also a concern. The present analysis was based on the subset of private applicators who completed the take-home questionnaire. Although these are only 44% of the private applicators enrolled in the AHS, they are clearly representative of the cohort as a whole: Applicators who did or did not complete the take-home questionnaire were similar for every lifestyle or demographic characteristic except age; for health outcomes, including experience of pesticide-related health symptoms; and for farm characteristics and tasks and a variety of measures of pesticide exposure (Tarone et al. 1997
). These results mitigate concerns regarding selection bias. Because symptoms were self-reported, another concern is potential recall or reporting bias. However, the fact that only some pesticides were associated with symptoms suggests that recall bias does not account for our findings. Risks associated with insecticide exposure were dose-related, further suggesting that bias does not explain our results. Moreover, our findings are biologically plausible, because we found the greatest risk for insecticides, which are designed to be neurotoxicants.
An important strength of our study is its large size. Further, because farming practices are considerably different in Iowa and North Carolina, the AHS cohort represents a diverse farming population (Alavanja et al. 1996
). We used internal comparisons of more and less exposed individuals from the same population, thereby reducing potential confounding. The primary strength of the study is, however, the availability of detailed exposure information. Although the present analysis is limited by its cross-sectional design, data on symptoms and pesticide exposure were collected in separate portions of the questionnaires, some completed at different times, minimizing potential bias. Exposure data were reported by the applicators themselves, but farmers in general and AHS cohort members in particular report pesticide use reliably (Blair et al. 2002
; Hoppin et al. 2002
In conclusion, we found that prevalence of neurologic symptoms was associated with cumulative lifetime exposure to pesticides, particularly organophosphate and organochlorine insecticides and fumigants. These associations were present in individuals with no history of pesticide poisoning or high exposure events and were independent of recent exposure. Thus, they are likely due to chronic moderate exposure. Although the neurotoxicity of high-level exposure is accepted, more attention to the risks associated with moderate exposure may be required.