BACKGROUND: Pesticide exposures have recently been linked to prostate cancer, but accurate exposure assessment to date has been challenging. Additionally, historical exposures have rarely been examined. The utility of a geographic information system (GIS)-based model for assessing residential exposure to pesticides is examined in a population-based case-control setting among groups easily recruited as control subjects. METHODS: Historical pesticide and land-use data were used to generate exposure measures for two distinct pesticides previously linked to prostate cancer risk for control series and prostate cancer cases in three rural California counties. Simple estimates of residential exposures for different exposure periods are compared between case and control groups and the value of complete residential histories is examined. RESULTS: Residential exposure to methyl bromide based on current address resulted in an overestimation of exposure for distant exposure periods, whereas exposures to organochlorines were similar regardless of availability of historical residence information. A response bias was detected in Medicare controls such that unexposed elderly control subjects were characterized by a higher response rate. CONCLUSIONS: The frequency and amount of application of pesticides seem to affect the bias introduced into GIS-based exposure assessments. Inclusion of subjects' complete residential histories into the computation of exposure estimates seems to reduce bias from this source, but it may also introduce an additional bias through control self-selection. The use of randomly sampled controls from Medicare and residential parcels listings independent of subject response seems to result in the opportunity for relatively unbiased estimates of pesticide exposures.
BACKGROUND: Pesticide exposures have recently been linked to prostate cancer, but accurate exposure assessment to date has been challenging. Additionally, historical exposures have rarely been examined. The utility of a geographic information system (GIS)-based model for assessing residential exposure to pesticides is examined in a population-based case-control setting among groups easily recruited as control subjects. METHODS: Historical pesticide and land-use data were used to generate exposure measures for two distinct pesticides previously linked to prostate cancer risk for control series and prostate cancer cases in three rural California counties. Simple estimates of residential exposures for different exposure periods are compared between case and control groups and the value of complete residential histories is examined. RESULTS: Residential exposure to methyl bromide based on current address resulted in an overestimation of exposure for distant exposure periods, whereas exposures to organochlorines were similar regardless of availability of historical residence information. A response bias was detected in Medicare controls such that unexposed elderly control subjects were characterized by a higher response rate. CONCLUSIONS: The frequency and amount of application of pesticides seem to affect the bias introduced into GIS-based exposure assessments. Inclusion of subjects' complete residential histories into the computation of exposure estimates seems to reduce bias from this source, but it may also introduce an additional bias through control self-selection. The use of randomly sampled controls from Medicare and residential parcels listings independent of subject response seems to result in the opportunity for relatively unbiased estimates of pesticide exposures.
Authors: Justine L E Allpress; Ross J Curry; Carol L Hanchette; Michael J Phillips; Timothy C Wilcosky Journal: Int J Health Geogr Date: 2008-04-30 Impact factor: 3.918