Administrative Supplement: Construction and Application of Research-Grade Cohorts from EHR for Metabolic Dysfunction-Associated Cancer Research

ABSTRACT This application is being submitted in response to "Administrative supplements to advance development of transdisciplinary and large-scale population science and cancer control research projects”.كWe propose a one- year supplement to build scalable infrastructure and analytic tools to support transdisciplinary research on metabolic dysfunction-associated cancers (MDAC)—a growing category of malignancies such as breast, colorectal, prostate, and lung cancers, driven by metabolic abnormalities beyond obesity. While obesity has traditionally been the focus of cancer prevention efforts, emerging evidence highlights that metabolic disturbances such as hyperglycemia, dyslipidemia, insulin resistance, and systemic inflammation independently contribute to cancer risk. These mechanisms have prompted growing interest in the potential of GLP-1 receptor agonists (GLP-1RAs)—a class of anti-diabetic agents with pleiotropic effects on weight, insulin sensitivity, and inflammation—as novel interventions in cancer prevention. However, critical gaps remain in understanding their long-term impact on cancer outcomes, particularly for patients with type 2 diabetes and metabolic dysfunction. Existing real-world datasets such as administrative claims and cancer registries lack detailed metabolic data and are limited by reporting lags. While multi-institutional electronic health record (EHR) database offers greater clinical granularity, it also suffers from partial observation bias due to fragmentation of health systems. To address this, we propose leveraging the EPIC COSMOS database, a nationwide EHR platform covering over 295 million U.S. patients, to advance data science tools that enable robust MDAC research. Our Aim 1 will adapt a machine learning–based algorithm previously developed to identify a cohort of “loyal patients” in COSMOS who receive care in ZIP codes with near-complete market penetration. We will validate MDAC incidence captured in COSMOS against two state cancer registries to ensure completeness and reliability. In Aim 2, we will use the validated “loyal patient” cohort and emulate a target trial to evaluate the effect of newer GLP-1RAs (e.g., semaglutide, tirzepatide) on MDAC incidence among patients with type 2 diabetes. This supplement will lay the foundation for future extramural grant submissions by establishing EPIC COSMOS as a research-grade platform for large-scale, population-based cancer control studies and pharmacoepidemiologic evaluations focused on modifiable metabolic risk factors and cancer outcomes.