Journal: Cancer science
This study evaluated tumor mutational burden (TMB) as a real-world predictor of pembrolizumab efficacy in a large Japanese cohort with advanced solid tumors, using data from 63,952 patients in a national genomic–clinical database.
Among 1,899 patients treated with pembrolizumab who had comprehensive genomic profiling (FoundationOne CDx, OncoGuide NCC Oncopanel, or GenMine TOP), TMB was categorized as:
- TMB-high: ≥10 mut/Mb
- TMB-low: <10 mut/Mb
Key findings:
- Distribution of TMB: Of 1,899 patients, 946 were TMB-high and 953 were TMB-low.
- Objective response rate (ORR): In TMB-high patients, ORR exceeded 30% and was significantly higher than in TMB-low patients (16.8%, p < 0.001), supporting TMB as a clinically useful predictive biomarker in routine practice.
- Borderline TMB: Patients with “borderline” TMB (10 to <13 mut/Mb) had more modest responses (ORR 20.8%), indicating that values just above the conventional cutoff may be less robustly predictive.
- Excluding hotspot mutations: When hotspot mutations were excluded from TMB calculations, ORR discrimination between groups improved, suggesting that removing recurrent driver alterations refines TMB’s predictive accuracy.
Overall, the data support TMB as a meaningful biomarker for pembrolizumab response in an Asian real-world setting and indicate that adjusting TMB by excluding hotspot mutations may be particularly important for patients with TMB values near the standard threshold.