Abstract
INTRODUCTION: MET exon 14 skipping in patients with advanced non-small cell lung cancer (aNSCLC), can be targeted with MET inhibitors including tepotinib, capmatinib, savolitinib, and crizotinib. Matching-adjusted indirect comparison (MAIC) methodology was used to compare outcomes data between agents and to address bias from differences in baseline characteristics. METHODS: Patient-level data from the VISION study (tepotinib) were weighted for comparison with aggregate data from the GEOMETRY mono-1 (capmatinib), NCT02897479 (savolitinib) and PROFILE 1001 (crizotinib) studies in patients with aNSCLC, using baseline characteristics prognostic for overall survival (OS) in VISION. Overall response rate (ORR), OS, progression-free survival (PFS), and duration of response (DOR) were compared. Patients were stratified by line of therapy: overall (all lines), previously treated, and treatment-naïve. RESULTS: Improvements in ORR and all time-to-event endpoints were predicted for tepotinib compared with crizotinib and savolitinib in the different populations, although comparisons with savolitinib were hindered by considerable differences in baseline patient populations. Tepotinib appeared to be associated with prolonged PFS and OS compared with capmatinib in previously treated patients (PFS HR 0.54; 95% CI 0.36-0.83; OS HR 0.66; 95% CI 0.42-1.06) and the overall populations (PFS HR 0.60; 95% CI 0.43-0.86; OS HR 0.72; 95% CI 0.49-1.05), with smaller improvements in DOR. The ORR comparisons between tepotinib and capmatinib identified a swing of up to ± 6 percentage points in the weighted tepotinib ORR depending on the population studied (treatment-naïve vs. previously treated patients). CONCLUSIONS: The MAIC identified potential differences in efficacy endpoints with the different MET inhibitors, and predicted prolonged PFS and OS with tepotinib compared with capmatinib and crizotinib. Although MAIC cannot balance for unobserved factors, it remains an informative method to contextualize single-arm studies, where head-to-head trials are unlikely to be feasible.