Association of high tissue TMB and atezolizumab efficacy across multiple tumor types.

12000 Background: PD-L1 expression has limitations as a biomarker for checkpoint immunotherapy (CI). Prior atezolizumab (atezo) monotherapy studies suggest improved efficacy in high tissue tumor mutational burden (tTMB-H) cohorts. We report a large retrospective analysis associating tTMB with neoantigen load (NAL) and CI efficacy across multiple studies, tumor types and lines of therapy. Methods: Tissue TMB was evaluated by the FoundationOne (F1) assay across 7 atezo monotherapy studies: NSCLC n= 342 (FIR, BIRCH, POPLAR, OAK), metastatic urothelial carcinoma (mUC) n=400 (IMvigor210, 211), and other advanced solid tumors n=245 (PCD4989g). Pooled data yielded a biomarker evaluable population (BEP) of 987 patients (pts); 175 (17.7%) had TMB-H defined as ≥16 mutations/megabase (mut/Mb). Efficacy endpoints were overall response rate (ORR) and duration of response (DoR). Survival analysis is ongoing. Neoantigen load (NAL) was calculated by whole-exome sequencing (WES) and RNA-Seq. Results: tTMB was associated with efficacy across tumor types and lines of therapy. In the BEP, ORR was 16.4% (95% CI 14.2, 18.9), vs. 29.7% (95% CI 23.1, 37.1) in tTMB-H (≥16 mut/Mb, N=175), and 13.5% (95% CI 11.3, 16.1) in tTMB-Low (<16 mut/Mb, N=812). DoR benefit was also observed: median DoR=29.0 mths (95% CI 18.6, NA) in tTMB-H vs. 16.6 mths (95% CI 13.8, 23.1) in BEP and 13.8 mths (95% CI 12.5, 17.4) in tTMB-Low cohorts. This association was not seen in control cohorts of randomized studies OAK, POPLAR, and IMvigor211. Pooled ORR in controls was 14.9% in BEP, 14.4% in tTMB-H and 15.1% in tTMB-Low. Survival analysis by tTMB is ongoing. tTMB-H also identified a population independent of PD-L1 status. tTMB by F1 was positively correlated with WES-based NAL in mUC (Pearson=0.85, N=218) and NSCLC (Pearson=0.78, N=70). NAL was associated with atezo ORR in mUC (p=2.7 x 10-9). Conclusions: High tTMB (≥16 mut/Mb) is associated with improved atezo response and DoR across NSCLC, mUC and melanoma and lines of therapy. tTMB-H by F1 may serve as a surrogate biomarker for NAL and may complement PD-L1 expression in providing predictive value. Ongoing studies are prospectively evaluating both tissue TMB and blood-based TMB for efficacy with CI. Clinical trial information: NCT02008227 NCT01903993 NCT02031458 NCT01846416 NCT02951767 NCT02302807 NCT01375842.