Time-to-event analysis for CER in Progressive Diseases: The Case of Parkinson's Disease
METHODS: We used the Disease Analyzer database (IMS HEALTH), containing basic medical data from approximately 20 million patients in Germany. The primary outcome was the therapy change rate from initial treatment to levodopa estimated by Kaplan-Meier analyses. A Cox proportional hazards model was used to estimate the relationship between time-to-levodopa and confounders for a maximum follow-up of 10 years (until Dec 2011). Adjusted hazard ratios(HR) and 95% confidence intervals(CI) are presented for change-to-levodopa rate. In addition to our empirical data we discuss applications of time-to-event analyses in PD using other databases and additional outcomes relavant for CER and health service research.
RESULTS: A representative sample of de-novo patients diagnosed with PD was drawn (n=108,885). 71.8% of patients received levodopa as a first-line treatment. 29,708 patients started with other anti-PD substances: 13.3% with dopamine agonists (DA), 3.6% with amantadine, 5.9% with anticholinergics, and 0.8% with MAO-B inhibitors. 29.0% of patients not starting with levodopa switched to levodopa within 5 years. Compared with PD patients starting with anticholinergics, patients starting with MAO-B-Inhibitors or DAs showed significantly lower proportions of levodopa free patients after 5 years (35% and 55%, respectively). Compared to MAO-B inhibitors, the HR for switching to levodopa was 0.38 (CI 0.34-0.43; p <.001) for anticholinergics and 0.74 (CI 0.67-0.83; p<.001) for non-ergot DA.
CONCLUSIONS: Comparative time-to-event analysis using administrative data is a fruitful approach for CER in health service research and pharmacoepidemiology . A systematic screening for its application in progressive chronic diseases may be of great practical value especially where head-to-head comparisons are unavailable.