Methods and interpretation of retrospective comparative claims data studies: Analysis of patient-level data in a case study (part 2)

Background Comparative studies based on claims data may generate substantial evidence for the assessment of the effectiveness and safety of therapies in the real world. However, interpretation of study results is challenging due to methodological uncertainties associated with these studies. Objectives Based on part 1 of this manuscript which outlined methodological options for comparative claims data studies, the aim of this paper (part 2) is to demonstrate the impact of specific assumptions on the results of such a study. Materials and methods We analyzed the impact of specific methodological decisions using the example of patients with ulcerative colitis (UC) who started therapy with either vedolizumab or adalimumab. Data from the German health insurance fund AOK PLUS were used. Effectiveness was defined by a composite endpoint consisting of UC-related hospitalizations and surgeries, frequent corticosteroid usage, and discontinuation or switch of the index therapy. Results Our analysis showed that specific methods within the previously described seven dimensions substantially impact study results. Depending on the assumptions and the decisions, sample size as well as observational periods and, finally, study results vary substantially. In our example, based on the chosen composite endpoint, vedolizumab showed superior effectiveness compared to adalimumab. Conclusions Publications dealing with comparative claims data should outline methodological assumptions in the described seven dimensions transparently and, in addition, provide reasons why specific assumptions were used.