Multimodal immune profile score (IPS) signature as a predictive response biomarker to immune checkpoint inhibitors in tumor mutational burden-high (TMB-H) advanced pancancer cohorts.

e14590 Background: Immune checkpoint inhibitors (ICI) have altered the oncology treatment landscape with improved outcomes. Despite approved predictive biomarkers for ICI such as TMB, refining selection can identify more patients (pts) who may benefit, as most TMB-H pts still fail to achieve response, a critical translational gap. Here we introduce IPS, a validated multimodal DNA-/RNA-based molecular signature, as a predictor of response to ICI. Methods: To establish IPS as a robust predictor of response to ICI, we evaluated its performance in two distinct clinical settings of pancancer pts: a Tempus real-world data (RWD) cohort of TMB-H pts (cohort 1) and metastatic pan-cancer pts from the external AHN Moonshot Biorepository (cohort 2). Cohort 1 consisted of TMB-H advanced solid cancer pts treated with ICI for indications lacking a cancer-specific FDA label. Cohort 2 consisted of metastatic pancancer pts who received ICI therapy on-FDA label. Pts were categorized as IPS-H or IPS-L using a previously validated threshold. In cohort 1, real-world objective response rate (rwORR) was determined by clinician-abstracted longitudinal records, defined as the proportion of pts with documented complete/partial response per physician assessment. In cohort 2, response was assessed according to RECIST v1.1 criteria based on independent radiologic review. ORRs and exact 95% CI are reported in each cohort. Results: Higher response rates were seen in pts with IPS-H status across both cohorts. Among all TMB-H pts (n = 24), higher ORR in IPS-H was seen compared to IPS-L (76% vs. 29%; see table). In TMB-H RWD Cohort 1 (n = 17), 10 tumor types were represented with pancreatic ductal adenocarcinoma as most common. rwORR was 91% (95% CI: 59-100) in IPS-H (n = 10/11), compared to 33% (95% CI: 4-78) in IPS-L (n = 2/6). In observational cohort 2 (n = 17), 6 tumor types were represented; renal cell carcinoma (n = 6) and melanoma (n = 5) were most common. IPS-H pts maintained higher response and disease control rates than IPS-L pts (ORR 33% vs 25%, DCR 67% vs. 50%). Among the subset of TMB-H pts (n = 7), ORR was 50% in IPS-H pts (n = 3/6) and 0% in IPS-L pts (n = 0/1). Conclusions: IPS is a novel multimodal immune response signature that improves precision of prediction of response to ICI compared to TMB. Within the TMB-H population, IPS identifies a subset of pts (IPS-L) that will not respond to ICI therapy. Given regulatory reliance on ORR for accelerated approvals, IPS represents a potentially practice-changing biomarker for refining treatment selection. Additional clinical studies are warranted for prospective validation of IPS. ORR by IPS status among TMB-H Pts. Pt Group (TMB-H only) ORR in IPS-H, % (n/n) ORR in IPS-L, % (n/n) Combined cohort (N=24) 76% (13/17) 29% (2/7) Cohort 1: RWD (N=17) 91% (10/11) 33% (2/6) Cohort 2: Observational (N=7) 50% (3/6) 0% (0/1)