Real-world utilization, effectiveness, and safety of GLP-1 receptor agonists for alcohol use disorder - PROJECT SUMMARY Glucagon-like peptide-1 receptor agonists (GLP-1RAs), which are indicated for type 2 diabetes, obesity, and cardiovascular disease, have been suggested as potentially effective treatments for alcohol use disorder (AUD). However, evidence from observational studies and randomized controlled trials has been limited. One cohort study found that GLP-1RAs may be associated with lower risk of subsequent hospitalization for AUD or purchase of medications for AUD (MAUD) compared to other T2DM medications. However, since this study was conducted in Denmark, the findings are not generalizable to patients with AUD in the United States. While trials of GLP- 1RAs for AUD are currently underway, these small trials will not provide direct comparative evidence across MAUD, and are likely to underrepresent older, rural, and racially and ethnically minoritized populations. These limitations have stimulated interest in using real-world data, causal inference methods, and novel machine learning subgroup identification approaches to generate evidence to guide the selection of optimal AUD treatment regimens in clinical practice. Therefore, we will use administrative claims data and linked electronic health record and laboratory result data from OptumLabs® Data Warehouse (OLDW), which includes health insurance claims for all Medicare fee-for-service enrollees and for commercial and Medicare Advantage enrollees across the United States, to generate evidence relevant to the use of GLP-1RAs in real-world clinical practice. We have identified 13,237 patients with AUD in OLDW who were prescribed GLP-1RAs from 2011- 2023. In Aim 1, we will follow the target trial emulation framework, which uses counterfactual theory to compare the effects of sustained treatment strategies, to evaluate the comparative effectiveness and safety of GLP-1RAs with both on-label (acamprosate, disulfiram, naltrexone) and off-label (gabapentin, topiramate, and varenicline) MAUD. These analyses will address the lack of direct comparative evidence between GLP-1RAs and other MAUD and will complement ongoing GLP-1RAs trials by providing evidence across more diverse and generalizable populations and practice settings with longer observation periods. In Aim 2, we will examine trends in GLP-1RA utilization among subgroups of patients with AUD, with a focus on potential differences in access to and use of these medications by key demographic and clinical characteristics, including race and ethnicity, rurality, gender, age, and co-morbidities. These analyses are essential to inform our understanding of potential inequities in GLP-1RA use among patients with AUD. In Aim 3, we will use traditional and novel machine learning subgroup methods to identify phenotypes of AUD patients more likely to experience beneficial or harmful outcomes with GLP-1RAs or other MAUD in real-world clinical practice. Identifying phenotypes of AUD patients will help guide clinical decisions, support higher quality AUD care, and inform the design of future clinical trials of GLP-1RAs among patients with AUD.
