Most previous studies that have exam-ined the survival of patients with parkinsonism have recruited them from specialist centres. No previous study has ever reported cause specific mortality. We report on the mortality of a cohort of 220 parkinsonian patients recruited between 1970 and 1972 from 40 primary health care practices all over England and Wales and matched to 421 controls. At 20 years of follow up, 195 cases (88-6%) and 295 controls (70.1%) were no longer alive (P < 0.001). The median age at death for cases was 77-6 (range 53-8-97-3) and 83'5 (range 55.0-100.1) for controls (P < 0.001). The all cause hazard ratio for cases compared with controls was 2-6 (95% confidence interval (95% CI) 2.2-3.2) controlling for age, sex, and geographical region. There was little difference between men and women. Differences for cause specific mortality also emerged. Both ischaemic heart disease (2.3, 95% CI 1.5-3.4) and cerebrovascular disease (3'6, 95% CI 2.2-6.1) showed significantly increased hazard ratios. Possible reasons for these findings are discussed in terms of (a) competing causes of death, (b) a secondary effect of drug treatment, and (c) common aetiological factors for both parkinsonism and cardiovascular disease. (7 Neurol Neurosurg Psychiatry 1995;58:293-299) Keywords: Parkinson's disease; aetiology controls and may have underestimated the relative mortality as such controls are more likely to have other serious diseases. Almost all of these studies have recruited their cases from specialist neurology units. This is problematic as specialist units may, in general, recruit more atypical and severe cases of neurological diseases.'7 Only one study has been based on the follow up of a representative population based sample.'6 This study importantly reported that parkinsonian cases had a 2*5-fold greater relative risk of dying than the general population, similar and only slightly improved on the relative mortality seen for the prelevodopa era.4 This result was based on follow up for only a 3-5 year period.Cause of death among parkinsonian patients may provide a useful insight into aetiology. If they are also more likely to die of another disease, there may be a common genetic or environmental link between these disorders. No previous study has ever examined cause specific mortality of Parkinsonian cases compared with the general population.We have examined both total and cause specific mortalities of a cohort of parkinsonian cases and matched population based controls recruited from primary health care facilities followed up for a 20 year period. This avoids previous methodological problems concerning ascertainment bias and the choice of an appropriate comparative group. Our aims were (a) to determine whether the mortality of parkinsonian cases had improved since the introduction of levodopa; (b) to determine whether any possible clue to the aetiology of parkinsonism might be found if these cases were more or less likely to die of other diseases.