Impact of targeted therapies in cancers classified as TMB-high by liquid biopsy: Insights from the Gustave Roussy molecular tumor board

Background: Tumor mutational burden (TMB) has emerged as a biomarker for immunotherapy response, but its predictive value for targeted therapies remains uncertain. In genomically unstable tumors, high TMB may reflect passenger rather than driver events, potentially limiting the efficacy of targeted agents. Methods: We retrospectively analyzed adult patients with advanced solid tumors who underwent plasma-based comprehensive genomic profiling (STING master protocol, Foundation Medicine) and had a TMB ≥ 16 mut/Mb. All cases were reviewed by the Gustave Roussy Molecular Tumor Board for orientation toward early-phase trials, phase II-III studies, or MTB-validated off-label targeted therapies. Alterations were classified as driver or non-driver based on integrated molecular and clinical evidence, supported by reference databases such as IntOGen and DriverDB. Results: Fifty-six patients were included (mean age 61.6 y; 71% male). The most frequent tumor types were lung (41%), colorectal (11%), and bladder (11%). Brain (20%), bone (38%), and liver (5%) metastases were common; 72% had ≥ 2 metastatic sites. Median plasma TMB was 20 mut/Mb (range, 16-1,450). Patients had received a median 3 prior lines (range 1-6). The most frequent alterations involved KRAS (21%), MET (11%), and ERBB2 (11%). Targeted therapies included small-molecule inhibitors (41%), TKIs (27%), ADCs (14%), bispecific antibodies (5%), and PROTACs (5%). Overall, 13 patients (23%) achieved partial response (PR) and 21 (38%) stable disease (SD), resulting in an overall response rate (ORR) of 23% and a disease control rate (DCR) of 61%. Among driver cases (n=43), 12 PR and 17 SD were observed vs 1 PR and 4 SD in non-driver cases (n=13). Median PFS was 3.7 months (95% CI 2.8-4.8). The prior-line PFS was 5.6 months (95% CI 4.0-11.0). A PFS ratio >1/3, indicating clinical benefit compared with the previous treatment line, was achieved in 63.3% of evaluable patients (31/49). Presence of a driver alteration was associated with improved PFS (HR 0.43; 95% CI, 0.22-0.82; p = 0.011). No correlation was found between TMB and PFS (Spearman ρ = 0.10; p = 0.45), and no difference was seen across TMB groups (<20, 20-50, ≥50 mut/Mb). Conclusions: Only patients with bona fide driver alterations derived benefit from targeted therapy, irrespective of TMB level. High TMB reflects genomic instability rather than therapeutic sensitivity, underscoring the importance of molecular curation to guide precision oncology decisions. Comparative analyses with TMB-low cases are ongoing to refine the predictive versus confounding role of TMB in targeted therapy response.