We investigated the risk of cancer after the diagnosis of Parkinson's disease (PD) through a historical cohort study. We used the medical records-linkage system of the Rochester Epidemiology Project to identify all incident cases of PD in Olmsted County, Minnesota from 1976 through 1995. Patients with PD were matched by age (+/- 1 year) and gender to referent subjects from the same population. For 196 patients and 185 referent subjects, we ascertained the incidence of cancer through medical records abstraction between the date of diagnosis (or index date) and death, loss to follow-up, or end of study. The risk of cancer was higher among patients than in referent subjects (relative risk [RR] = 1.64; 95% confidence interval [CI] = 1.15-2.35; P = 0.007). The RR did not change noticeably after adjustment for smoking. The increased risk was significant for nonmelanoma skin cancer (RR = 1.76; 95% CI = 1.07-2.89; P = 0.03), but not for other more severe types of cancer; therefore, we cannot exclude the occurrence of a surveillance bias. Among PD patients, there was no relation between the risk of cancer and the cumulative dose of levodopa received or the use of other PD medications.