OBJECTIVES: Results from the CheckMate 142 Study with 5-year follow-up data suggested that nivolumab plus ipilimumab (nivo+ipi) benefits both overall survival (OS) and progression-free survival (PFS), demonstrating improvements over nivolumab (nivo) in patients with MSI-H metastatic CRC (mCRC). This study evaluated the comparative effectiveness of nivo and nivo+ipi as a second line or later (2L+) therapy in MSI-H mCRC using a Bayesian borrowing approach to improve precision of relative effect estimates and overall study power. Quantitative bias analysis (QBA) was used to support study results by assessing the impact of multiple sources of bias.

METHODS: Three cohorts from CheckMate 142 were used, with 2L+ nivo (Cohort 1) and 2L+ nivo+ipi (Cohort 2) being the two cohorts of interest for comparison, and Cohort 3 (1L nivo+ipi) acting as the borrowing source for Cohort 2. A pre-processing matching step was used to balance differences in baseline characteristics across cohorts. Effect estimates for OS and PFS of nivo+ipi compared to nivo were estimated via Bayesian, parametric analyses. QBA was used to assess robustness of study findings for patients with uncertain MSI-H status, unmeasured confounding, missing data, and for target cohort differences due to borrowing.

RESULTS: 55, 119, and 35 patients were retained after propensity score matching in Cohort 1, 2, and 3, respectively. OS and PFS posterior medians of the hazard ratio (HR) ranged as follows: for OS, from 0.462 (95% CI: 0.295, 0.732) to 0.469 (95% CI: 0.293, 0.757), and for PFS, from 0.596 (95% CI: 0.395, 0.915) to 0.597 (95% CI: 0.389, 0.933), when using a power prior discount parameter ranging from 1 to 0.

CONCLUSIONS: Based on this analysis, 2L+ nivo+ipi patients have had significantly better OS and PFS compared with nivo. QBA results suggest effect estimates are generally or very robust against all four potential sources of bias assessed.