215P Identifying predictors of overall survival among TMB-low cancer patients treated with immune checkpoint inhibitors

Dichotomizing tumors into tumor mutational burden (TMB)-high and TMB-low is common practice but may not determine which patients will benefit from immune checkpoint inhibitors (ICI). This study examines the impact of somatic mutations on overall survival (OS) for TMB-low metastatic cancer patients (MCP) treated with ICI. Genomic (MSK-IMPACT) and clinical data from 1,172 TMB-low (<10mut/Mb) MCP were collected. OS accordingly to single-genemutations was assessed by Kaplan-Meier (KM) method. For genes exhibiting a correlation with OS, we conducted a Cox multivariate analysis (CMA) stratified by TMB, sex, age, microsatellite instability (MSI), and histology. KM and CMA were also performed to investigate the relation between insertions and deletions (InDels) and OS. Linear regression (LR) investigated the relation between TMB and InDels burden. Decision tree modeling (DTM) aimed to predict whether patients that received ICI harbored OS shorter or longer than the median OS of 15 months. The discriminative features (DF) for DTM were TMB (continuous), gene mutations, sex, and histology. Mutations in 25 genes were related to OS. CMA confirmed 9 genes associated with superior (VHL) and inferior (CTNNB1, DAXX, H3C2, HLA-A, IGF1R, KMT2D, SMARC4, TP53) OS (P < 0.05). MSI, age, and gender did not affect OS. Melanoma, bladder, and RCC endowed patients with better OS (P < 0.01). TMB and InDels exhibited a R-squared of 0.38 - low (or absent) relation. The InDels burden was unrelated to OS (P = 0.21). VHL mutation was a predictor of better OS in uni- and multivariate levels for RCC (N=149; P < 0.01). At DTM, RCC histology was the first DF (gini 0.492). For RCC patients, VHL mutation was the second (gini 0.349). For non-RCC patients, melanoma histology was the second. At the third layer, a TMB threshold of 5.095 mut/Mb for VHL wild-type patients and TP53 mutation for non-RCC and non-Melanoma patients were the following DF. Tumor histology, mutated genes, and a better adjusted TMB cutoff (5.095 mut/Mb) are important for predicting ICI treatment outcomes in TMB-low patients. Among RCC patients, VHL mutations predict better OS. Individualized profiling is needed for a tailored ICI treatment strategy.