Hertzman and colleagues (1994) studied 127 Canadian PD patients identified by their physicians and compared them with cardiac-disease controls and healthy controls drawn from electoral rolls. The latter control group was chosen to reduce the impact of potential recall bias. Exposure was ascertained in face-to-face interviews. The main focus of the study was on pesticides, but there was one exposure question about occupational exposure to solvents. Exposure was defined only as ever or never exposed before reported onset of disease. When cases were compared with controls from the electoral rolls, a moderate association between occupational exposure to any solvents before disease onset was found for men (OR=2.16, 95% CI=1.07–4.37). No association was found for men in comparison with the cardiac-disease controls or for women in comparison with either control group. The validity of the association—men with PD compared with electoral-list controls—is limited for three reasons. The first is a difference in findings depending on the control group; the association only with the electoral-list controls suggests that the finding is a result of recall bias from underreporting of exposure by this group of controls. The second is the use of the electoral list as the source of the healthy control group. Noncitizens were excluded, and this exclusion was not applied to the case group. If citizenship status is related in some way to the probability of being exposed to solvents, the comparison between cases (including noncitizens) and controls (excluding noncitizens) may be biased. It is difficult, however, to predict the direction of the bias in terms of either underestimating or overestimating the effect of solvents on PD occurrence. The third is that the focus of this study was the relationship between pesticides and PD, not solvents and PD. As a result, the method describing solvent-exposure assessment is only briefly presented, and the analysis is limited.
Seidler and colleagues (1996) studied 380 PD patients from nine neurology clinics in Germany. The UK PD Society Brain Bank clinical diagnostic criteria were used to screen subjects for eligibility. Potential cases with dementia or secondary parkinsonian syndromes were excluded. Cases were compared with two control groups that were population-based and were recruited with the random-route procedure in which the interviewer contacts every second household, starting with the patients. In addition to the natural matching on residence, the controls were matched on sex and age ±3 years. The investigators elicited information on exposure to “solvents”-never, in free time, or at work. Controls were asked to report exposure at any time at least a year before interview, but cases reported exposure at any time before diagnosis. Because the cases had average illness duration of 3.7 years, that discrepancy meant that controls did not have to remember as far back. A detailed occupational history was also collected from each study subject, from which a panel of experts constructed a job-exposure matrix. Conditional logistic regression was used to analyze results, and smoking and education status were included as covariates in the model.
Seidler and colleagues (1996) reported positive associations regardless of whether the self-reported exposure was to solvents at work or in free time. Contrary to expectations, free-time exposure resulted in higher odds ratios than did work exposure. When the exposure assessment using the job-exposure matrix was used, no association was found. It is generally accepted that although a job-exposure matrix is itself based on information obtained from self-reports, it provides a more accurate measure of occupational exposure. The discrepancy in findings between the two types of exposure variables (self-report vs job-exposure matrix) suggests recall bias through possible underreporting of exposure in