Effects of Agricultural Work and Other Proxy-derived Case-Control Data on Parkinson's Disease Risk Estimates

Abstract

This study examined the effects on Parkinson's disease risk estimates of exposure misclassification in proxy-derived data on agricultural work, pesticide use, rural living, well water drinking, head trauma, smoking, and family history of Parkinson's disease or essential tremor. The data were collected in 1989 as part of a population-based case-control study of Parkinson's disease in Calgary, Canada. Nondemented cases (n = 130) were selected from a case register of Calgary residents with neurologist-confirmed Parkinson's disease. For each case, two matched (sex and age +/- 2.5 years) community controls were selected by random digit dialing. Forty cases and 77 controls were randomly selected as index respondents. The cases, controls, and one proxy respondent (spouse or offspring) for each index respondent were interviewed using a structured questionnaire. The data were analyzed using conditional logistic regression. Incorporation of proxy-derived data for 30% of the cases or controls, or both, resulted in considerable misclassification of exposure for some variables and, in most cases, attenuation of the odds ratio. The results indicate that pooling dichotomously classified data derived in part from self- and proxy respondents may result in biased estimates of Parkinson's disease risk associated with agricultural, family history, and head trauma factors.