CYP2D6 polymorphisms and pesticide exposure in Parkinson's disease

The objective of this dissertation was to develop and examine new methods to estimate gene-environment interaction (GEI) effects in disease etiology using Parkinson's disease (PD) as a model. We conducted a case-only study of GEI effects in PD, evaluated the reproducibility of a published CYP2D6 assay and designed a novel CYP2D6 assay for use with buccal DNA, and, designed a simulation study to determine the relative accuracy of case-only and case-control risk estimates of GEI.; In a case-only study of 21 Caucasian PD patients, we examined interaction effects of pesticide exposure and functional polymorphisms in the CYP2D6 gene associated with defective pesticide metabolism in 7–10% of Caucasians. Exposure was ascertained via a questionnaire administered in personal interviews, at which time self-collected DNA samples were obtained from patients using buccal swabs. Several subjects reported pesticide exposure; however, no CYP2D6 poor metabolizers (PM) were identified, thus GEI evaluation was not possible. However, using two historical control groups with characteristics similar to our subjects, we found a statistically significant association between gardening and PD (OR = 5.83, 95% CI 1.70, 21.93) and an increased but not statistically significant risk for ever versus never pesticide exposure (OR _combined = 1.52, 95% CI 0.50, 4.46; OR_{ControlGroup1} = 1.39, 95% CI 0.43, 4.33; OR_{ControlGroup2} = 1.67, 95% CI 0.51, 5.24).; We demonstrated failure of a published polymerase chain reaction (PCR) assay for simultaneous detection of five most common CYP2D6 PM alleles in Caucasians to produce results using a non-blood source of DNA and designed a new PCR assay for use with buccal cells to simultaneously detect CYP2D6*4 and *5, the two most common PM alleles in Caucasians.; We compared accuracy of case-only and case-control risk estimates of Type 2 GEI effects in a simulation study using populations with known risks. Case-only estimates were shown to approximate known risks more consistently and accurately under most conditions provided gene and exposure of interest are independent in the population. We demonstrated the influence of sample size on accuracy of both estimates, and showed that inadequate study size can lead to such large random error as to result in inaccurate estimates despite efforts to eliminate bias.