Evaluation of antiparkinsonian drugs in pharmacy records as a marker for Parkinson's disease

Aim: The aim of this study was to determine whether use of antiparkinsonian drugs in pharmacy records can be used as a marker for patients with Parkin son's disease (PD). Method: Data were obtained from the Rotterdam Study, a community-based pros pective cohort study among people aged 55 years or older who were all scree ned for PD. For 5510 persons, of whom 74 had PD, pharmacy records were avai lable. Stepwise logistic regression analysis was used to evaluate whether a ge, sex and use of the antiparkinsonian drugs amantadine, anticholinergics, dopamine agonists, levodopa and selegiline, were predictive variables for PD. For each individual a probability for having PD was calculated. Sensiti vity, specificity and positive predictive value (PPV) were calculated at di fferent cut-off values based on calculated probabilities. Results: More than 90% of the users of levodopa, bromocriptine, selegiline, and users of at least two different antiparkinsonian drugs had PD. Age, us e of amantadine, anticholinergics, bromocriptine, levodopa, and selegiline were predictive variables for PD. After application of different cut-off va lues, sensitivity was at most 66.2%, and specificity was at least 99.8%. A PPV of higher than 90% was obtained at higher probabilities. Conclusion: Based on the high PPV of our predictive model, antiparkinsonian drugs can be used as a reliable marker for PD in pharmacy records. Because sensitivity is low, pharmacy records cannot be used to estimate prevalence of PD.