Tumor-Type-Specific TMB Cutoffs for Improved ICI Outcome Prediction: Large-Scale Analysis with Real-World Targeted NGS Data

Purpose Tumor mutational burden (TMB) is a potential biomarker for predicting response to immune checkpoint inhibitors (ICIs). However, its clinical utility is limited by methodological inconsistencies. This study aimed to evaluate the predictive value of TMB for ICI outcomes using next-generation sequencing (NGS) data. Materials and Methods We retrospectively analyzed 9,459 patients with cancer who underwent tumor-only targeted NGS. TMB-high (TMB-H) cutoffs were defined using an interquartile range (IQR)-based method and validated by comparing the overall survival (OS) and progression-free survival (PFS) in ICI-treated cohorts against both The Cancer Genome Atlas whole-exome sequencing-derived TMB and the universal 10 mutations per megabase (mut/Mb) cutoff. We also examined programmed cell death-ligand 1 (PD-L1) expression and subclonality to address response heterogeneity. Results IQR-based TMB-H was significantly associated with longer PFS in the ICI-treated cohort (hazard ratio (HR)=0.85, 95% confidence interval (CI): 0.73–0.98, p=0.02), NGS before ICI subgroup (HR=0.86, p=0.049), and pre-ICI subgroup (HR=0.80, p=0.03). In contrast, the universal 10 mut/Mb cutoff showed no statistical significance. Subgroup analysis revealed significant PFS benefit in bladder (p=0.014), bowel (p=0.013), and uterine cancers (p=0.006). In lung cancer, patients with both TMB-H and very high PD-L1 expression (≥90%) had the longest PFS (HR=0.64, 95% CI: 0.44–0.93, p=0.021). Among the TMB-H samples, high subclonality was associated with worse OS in non-hypermutated cases (p=0.032). Conclusion Real-world TMB cutoffs derived from distribution-based methods offer improved predictive value for ICI outcomes. Integration of the PD-L1 expression and subclonality status further refines the predictive utility of TMB, improving precision in ICI treatment.