Abstract Gastrointestinal (GI) toxicities are the most common drug-induced adverse events in human clinical trials. Drugs that cause GI toxicities often affect the proliferative stem and progenitor cell populations responsible for maintaining the cellular composition, self-renewing capacity, and barrier function of the intestinal epithelium. This disruption leads to impaired GI barrier integrity, culminating in a broad range of clinical symptoms, including diarrhea, inflammation, and increased risk of infection. The impact of GI toxicity is substantial and can cause dose-limiting clinical side-effects that may affect patient compliance and dosing adherence. Common nonclinical animal models, including rats, mice and dogs are often poor predictors of human GI toxicity and safety outcomes. Furthermore, in vivo animal testing poses limited screening throughput, which is particularly important in fields such as oncology, where matrices of double or triple drug combinations represent the standard of care. Additionally, existing cell-based model systems, such as Caco-2 cells, are of limited predictive value as they are tumor-derived and do not reprise native epithelium. Thus, there is a need for the development of assay systems that more accurately recapitulate human biology, have higher throughput testing capabilities, and ultimately improve clinical predictability. To this end, we have developed an assay to assess GI toxicity potential using the RepliGut Planar platform, which consists of a primary human stem cell-derived epithelium in a 96-well Transwell format. Quantitative dose-response curves were generated on proliferative and/or differentiated GI cultures for various test articles utilizing endpoints that assess GI proliferative capacity, cell viability, and epithelial barrier formation and maintenance. An initial set of marketed drugs (including diverse anti-proliferative, cytotoxic, or anti-inflammatory mechanisms) with known clinical incidence of diarrhea were over a 5-log dose range to enable modeling of the Toxic Concentration 50% (TC50). When benchmarked against the known human plasma concentrations (Cmax), this model was able to accurately predict clinical outcomes for diarrheagenic-drugs, as well as non-toxic negative controls, within 30-fold of the clinical Cmax. Overall, this model system provides straightforward and informative screening for GI toxicity that integrates sensitive and robust endpoints, enables screening to support early nonclinical development, and uniquely allows for mechanistic insights. Citation Format: James Levi, Reganne Lorichon, Lauren Boone, Maureen Bunger, Elizabeth Boazak, William R. Thelin. A sensitive and specific human primary stem cell-based in vitro assay for predicting the gastrointestinal toxicity risk of therapeutic agents [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2054.