Methodological considerations in using claims databases to evaluate persistence with bisphosphonates for osteoporosis

Objective: To demonstrate that retrospective analyses of medication persistence require careful methodological approaches to assure accuracy and consistency across various types of databases. Bisphosphonates (BPs) are used as an example because of the availability of diverse dosing options that can create a unique set of challenges for persistence analyses. Methods: Reports of BP persistence were compared for methodological approaches, including data source, duration of follow-up, allowed gap for persistence, and presentation of results. Main outcome measures: Medication persistence. Results: Comparisons among reports of BP persistence for weekly and monthly formulations revealed inconsistent definitions and a variety of methods. Persistence analyses varied greatly, particularly in allowed gaps and adjustment for demographic and clinical characteristics that affected results. Persistence with weekly dosing was 179-249 days, with 24-78% remaining on treatment at 1 year. Analyses of persistence with monthly treatment was complicated by the variety of gap lengths (30-90 days). The studies reviewed had many limitations, including lack of an established threshold for efficacy, inadequacy of information in databases, and potential biases in case selection (treatment-naive or experienced). Conclusions: The limitations of published studies reveal the need for a more consistent approach to medication persistence analyses using claims databases to allow for comparison across reports. The analysis plan should present definitions of all terms, details of all methods and types of adjustments needed for demographic and clinical characteristics, as well as specify allowed gaps between refills. These approaches would improve clinical utility of data describing BP persistence and its impact on fracture risk.