The prevalence of idiopathic Parkinson's disease (PD) in an area of Northwestern Italy (Socio-Sanitary District of Cossato, 61,830 inhabitants) was analyzed and the reliability of "drug tracer methodology" evaluated. The patients were identified by both conventional epidemiological methodology (general practitioners, consultant neurologist, computerized archives) and drug tracer methodology (registered levodopa prescriptions). The diagnosis of PD was clinically verified by examining all patients. Of 146 patients examined, 104 were affected by PD; the remaining by other types of parkinsonism. The crude point prevalence rate (October 20, 1991) was 168 per 100,000 population (95% confidence interval [95% CI] 138-204), with a progressive increase up to the 80-89 age group. Using "tracer" methodology, the estimated crude point prevalence rate was 196 (95% CI 163-235), with an overestimation of the prevalence in older patients and women and an underestimation in younger and less seriously affected cases. The uneven distribution of PD in Italy appears to be the result of, at least in part, methodological factors. "Tracer" methodology may cause biases in the evaluation of the epidemiological characteristics of PD, especially if the clinical diagnosis of cases identified by this method is not carefully verified. Nevertheless, it appears useful as an additional source of cases for descriptive surveys.