Extending inferences of efficacy and selection of transformative diabetes therapies to diverse populations - ABSTRACT Type 2 diabetes mellitus (T2D) is a common chronic disease associated with morbid and costly complications. New medication classes – glucagon-like peptide-1 receptor agonists (GLP1RA) and sodium-glucose cotransporter-2 inhibitors (SGLT2i) – have proven cardiovascular, renal, and glycemic benefits in randomized clinical trials (RCTs) in individuals with T2D. RCTs provide high quality evidence of efficacy yet leave several evidence gaps relevant to routine clinical care. First, while a well-conducted RCT has strong internal validity, the applicability of the results in real-world care – or external validity – can be uncertain, particularly for populations underrepresented in the RCT. Unfortunately, minoritized race and ethnicity groups and rural individuals bear excess burden of T2D and T2D complications but are underrepresented in recent RCTs of GLP1RA and SGLT2i. Second, most Phase 3 RCTs for GLP1RA and SGLT2i utilize a placebo control. As a result, there are little to no head-to-head data comparing efficacy and harms of GLP1RA and SGLT2i for cardiovascular and kidney endpoints – potentially valuable data for routine diabetes care. Finally, RCTs are designed to estimate an average treatment effect in the trial population but are typically underpowered to identify heterogeneous treatment effects or subgroups in which one treatment or another has particular benefit or risk of harm. Moreover, prediction models to guide T2D treatment selection may not generalize well across populations – similar to the problem of RCT external validity. The overall goal of this proposal is to bring together real-world data from diverse T2D populations from three health systems and data from pivotal RCTs of GLP1RA and SGLT2i to address each of these evidence gaps using biostatistical tools known as transportability methods. Transportability methods use weighting to balance participant characteristics between a clinical research study population (e.g., an RCT) and a target population (e.g., individuals with T2D from a population underrepresented in an RCT). This allows inference of what the study’s results would have been had the target population participated in the study. Importantly, these methods avoid several critical threats to validity of conventional comparative effectiveness research methods. In Aim 1, transportability methods will be used to extrapolate the efficacy and harm of GLP1RA and SGLT2i from recent landmark RCTs to representative real-world samples of T2D patients and to populations underrepresented in RCTs. In Aim 2, extensions of transportability methods will be used to estimate head-to-head effects of GLP1RA and SGLT2i for glycemic, cardiovascular, and kidney outcomes. Aim 3 will focus on developing prediction models to guide individualized selection of GLP1RA versus SGLT2i and utilize transportability methods to improve prediction model performance across diverse populations. In completing the Aims, the study will create a roadmap that describes best practices for designing and executing transportability analyses and will provide insights into how RCT and real-world data can be integrated to generate timely evidence relevant to all individuals with T2D.
