Abstract 741: Assessing overall survival relative to LORIS response prediction in patients receiving immunotherapy in a large community-based health system

Abstract Purpose: The use of immune checkpoint inhibitors (ICI) has revolutionized outcomes in several cancers. Despite markers such as tumor mutational burden (TMB) and PD-L1 expression, predicting which patients will respond remains a challenge. Recently, a tool was developed using machine learning to predict clinical outcomes from ICI across cancer types known as LORIS (logistic regression-based immunotherapy-response score) that provides a probability of response from 0-100% based on six factors. LORIS values <27% were deemed ‘unlikely to respond’ and >70% were ‘likely to respond’ in the originating study. This project evaluates the association of LORIS scores to overall survival (OS) with the aim of understanding its utility in clinical-decision making in a community health-based cancer center. Methods: We performed a retrospective analysis of 136 patients at the Avera Cancer Institute that received at least one dose of an ICI and possessed all components to calculate LORIS, including age, cancer type, previous therapy, albumin, blood neutrophil-to-lymphocyte ratio (NLR) and TMB, in addition to outcomes. LORIS values were calculated according to the most recent values available prior to initiation of therapy. The association between OS and LORIS was evaluated by Cox proportional hazards model. With no scores >70% in our population (the threshold for likely response), patients were categorized as ‘unlikely to respond’ or ‘not unlikely to respond’ based on a score of ≤27% or >27%, respectively, and assessed via Kaplan-Meier. Results: The median age was 67 years (range 36-87) and 48.5% were female. Lung cancer represented 47% of patients, followed by 15.4% melanoma, 14.7% gynecologic cancer and less than 5 patients each with GI, GU, and head and neck cancers; the remaining were considered “other” per LORIS. The median TMB was 6.1, median albumin was 3.8, and median NLR was 3.8. In contrast to the original LORIS datasets where 72% of patients were previously treated, our population was greater at 84% indicating a more heavily pre-treated group. Only 20 patients (14.7%) had a LORIS ≥27%. The hazard ratio was 0.09 (95% confidence interval 0.017-0.51) and LORIS was statistically significant for impact on OS (p=0.006). The median OS for those unlikely to respond (n=116) was 13.7 mo (9.9 - 18.7) vs. not reached (NR) (15.9 - NR) for those not unlikely to respond (n=20). Conclusion: Higher scores were associated with longer OS, confirming its ability to predict long-term outcomes. This was upheld despite having few patients with >27% scores and no patients >70%, possibly due to the greater portion that previously received systemic therapy. Future studies will need to address how LORIS can be applied to individual patient scenarios to support clinical decision-making. Citation Format: Cheryl P. Jones, McKenna Deaton, Bing Xu, Rachel Elsey, Crystal Hattum, Mason Arndt, Victoria Moe, Alex Mohama, Casey Williams, Tobias Meissner. Assessing overall survival relative to LORIS response prediction in patients receiving immunotherapy in a large community-based health system [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 741.