INTRODUCTION

Blood-based tumor mutational burden (bTMB) has been studied to identify patients with NSCLC who would benefit from anti-programmed cell death protein 1 (anti-PD-1) or anti-programmed death ligand 1 (anti-PD-L1) therapies. However, it failed to predict overall survival (OS) benefits, which warrants further exploration.

METHODS

Three independent cohorts of patients with NSCLC treated with immunotherapy were used in this study. A new bTMB algorithm was first developed in the two independent cohorts (POPLAR, N = 211, and OAK, N = 462) and further validated in the third National Cancer Center (NCC) cohort (N = 64).

RESULTS

bTMB-H (bTMB ≥ cutoff point) was not associated with favorable OS after immunotherapy regardless of the cutoff points in either the POPLAR and OAK or the NCC cohorts (p > 0.05) owing to its correlation with the amount of circulating tumor DNA, which was associated with poor OS. In the POPLAR and OAK cohorts, with allele frequency (AF) adjustment, a high AF bTMB (HAF-bTMB, mutation counts with an AF > 5%) was strongly correlated with the amount of circulating tumor DNA (Pearson r = 0.65), whereas a low AF bTMB (LAF-bTMB, mutation counts with an AF ≤ 5%) was not (Pearson r = 0.09). LAF-bTMB-H was associated with favorable OS (hazard ratio [HR] = 0.70, 95% confidence interval [CI]: 0.52-0.95, p = 0.02), progression-free survival (PFS; HR = 0.62, 95% CI: 0.47-0.80, p < 0.001), and objective response rate (ORR) (p < 0.001) after immunotherapy but not chemotherapy, with a cutoff point of 12 trained in the POPLAR cohort and validated in the OAK cohort. The LAF-bTMB algorithm was further validated in the NCC cohort in which LAF-bTMB-H was associated with OS (HR = 0.20, 95% CI: 0.05-0.84, p = 0.02), PFS (HR = 0.30, 95% CI: 0.13-0.70, p = 0.003), and ORR (p = 0.001).

CONCLUSIONS

We developed and validated a new LAF-bTMB algorithm as a feasible predictor of OS, PFS, and ORR after anti-PD-(L)1 therapies in patients with NSCLC, which needs to be prospectively validated.