OBJECTIVES: Chemotherapy is widely used in previously-treated advanced/metastatic NSCLC, however limited evidence is available in recently identified biomarker-driven populations, such as METex14 skipping. To estimate comparative effectiveness versus tepotinib (a selective MET inhibitor), comparisons were made using RWE in METex14 skipping NSCLC, and MAIC of published data in non-oncogenic-driven NSCLC.

METHODS: To estimate the effectiveness of chemotherapy, seven real-world datasets were pooled in a common data model (TOGETHER study), providing information from 68 chemotherapy-treated patients to compare with the 149 previously-treated patients from VISION (NCT02864992), using propensity scoring. Outcomes for progression-free survival (PFS) and overall survival (OS) were compared. Additionally, TOGETHER and VISION were reweighted to match published evidence predating the introduction of immunotherapies from 2014 (Garon et al.) and 2000 (Fossella et al.).

RESULTS: A variety of chemotherapy regimens were received, with the majority being platinum-, taxane-, or antifolate-based regimens. In the TOGETHER propensity scoring comparison, for PFS tepotinib had longer median and 24-month Restricted Mean Survival Time (8.3 vs 5.2 months and 11.1 vs 7.9, respectively) and OS (19.3 vs 12.5 and 16.6 vs 14.1, respectively). However OS comparisons were confounded by post-progression usage of immunotherapies and MET inhibitors in TOGETHER. Reweighting both VISION and TOGETHER using MAIC to match published data gave consistent PFS estimates to TOGETHER, but VISION comparative OS improved, as published chemotherapy OS was lower (median 9.3 months in Garon et al. 2014, 6.0 months in Fossella et al. 2000) than in TOGETHER (chemotherapy median 11.0).

CONCLUSIONS: In comparisons using both TOGETHER and published studies, tepotinib provided longer PFS than chemotherapy, with OS differences seemingly reduced by subsequent therapy use. Although cross-study comparisons can be complex, using a combination of individual and aggregate level data, hypotheses (such as the impact of subsequent therapies) can be tested, overcoming the limitations of a single data source.