Improving the VAlidity of LInked real-world DAta sTudiEs for Diabetes Mellitus research (VALIDATE-DM)

PROJECT SUMMARY/ABSTRACT Individuals with diabetes and their prescribers often choose among various glucose-lowering treatments (GLTs), but information on the comparative effectiveness and safety of GLTs is limited. Most randomized controlled trials (RCTs) only compare one GLT with placebo or another GLT in highly selective populations. Real-world data (RWD), such as healthcare claims and electronic health records (EHRs), have been used to fill this knowledge gap. The value of RWD is further highlighted in the 21st Century Cures Act, which emphasizes the use of RWD and real-world evidence (RWE) to accelerate medical product innovation. RWD will likely continue to be used, if not increasingly so, in comparative effectiveness and safety research of GLTs as a complement to RCTs. However, individual RWD sources have well-known limitations. For example, claims data lack certain variables (e.g., HbA1c) critical for diabetes research, while EHRs do not fully capture out-of- system care. Linking complementary RWD sources helps overcome some of these limitations, but linked data are susceptible to numerous biases if not analyzed properly. Given the ubiquity of and ease of access to RWD for diabetes research, it is imperative to ensure valid findings from RWD studies. The proposed study will focus on linked RWD for comparative effectiveness and safety research of GLTs, which is expected to be increasingly common but with under-appreciated design and analytic challenges. Using a large claims-EHR linked database and realistic examples, the study will address the following aims: (1) Develop a systematic approach to investigate potential biases when using linked RWD for comparative effectiveness and safety research of GLTs, (2) Develop and evaluate a comprehensive design and analytic approach that maximizes internal validity in comparative effectiveness and safety research of GLTs that uses linked RWD, and (3) Apply novel generalizability and transportability methods to improve external validity and extend findings from linked RWD studies to target population(s) of interest. RWD studies complement and do not replace RCTs. Sometimes they are the only way to fill our knowledge gap. The path forward is to not discredit RWD studies but to keep improving their validity and generalizability. The proposed study is timely and critically needed. It addresses a key topic in the 2021 NIDDK 5-Year Strategic Plan for Research – leveraging existing data for clinical research and minimizing biases associated with RWD. The study will help the diabetes researchers maximize the potential of linked RWD, avoid or minimize pitfalls associated with linked RWD studies and generate robust RWE in diabetes research. The developed framework can be generalized and applied to other linked data, such as linked claims-registries and linked EHR-registries.