Abstract
BACKGROUND: Levodopa (LD) is the most effective drug to treat Parkinson's disease (PD). The recently concluded multinational Parkinson's Real-World Impact Assessment (PRISM) trial revealed highly variable prescription patterns of LD monotherapy across six European countries. The reasons remain unclear. METHODS: In this post hoc analysis of PRISM trial data, we used multivariate logistic regression analysis to identify socio-economic factors affecting prescription practice. We applied receiver-operated characteristics and split sample validation to test model accuracy to predict treatment class (LD monotherapy vs. all other treatments). RESULTS: Subject age, disease duration, and country of residence were significant predictors of treatment class. The chance of receiving LD monotherapy increased by 6.9% per year of age. In contrast, longer disease duration reduced the likelihood of receiving LD monotherapy by 9.7% per year. Compared to the other countries, PD patients in Germany were 67.1% less likely and their counterparts in the UK 86.8% more likely to receive an LD monotherapy. The model classification accuracy of treatment class assignment was 80.1%. The area under the curve to predict treatment condition was 0.758 (95% CI [0.715, 0.802]). Split sample validation revealed poor sensitivity (36.6%), but excellent specificity (92.7%) to predict treatment class. CONCLUSION: The relative lack of socio-economic variables affecting prescription practice in the study sample and limited model accuracy to predict treatment class suggest the presence of additional, country-specific factors affecting prescription patterns that were not assessed in the PRISM trial. Our findings indicate that physicians still avoid prescribing LD monotherapy to younger PD patients.