OBJECTIVES: Traditionally, oncology therapies prescribed across multiple tumors received independent sequential approvals, by indication. The first tumor-agnostic FDA approval was in 2017 for microsatellite instability-high (MSI-H) tumors, based on objective response rate (ORR). Biomarker-driven therapies will continue to support this approach. Here, we explore whether sharing of information across tumors in an unanchored ITC could increase the precision of treatment effects for a specific tumor type.

METHODS: An unanchored Bayesian multilevel meta-regression model for ORR across multiple tumors is proposed, leveraging individual patient data (IPD) from an index trial and aggregate data (AD) from external trials. Efficacy of adagrasib versus sotorasib as monotherapy was compared for patients with previously treated KRAS^G12C-mutated advanced/metastatic solid tumors, including non-small cell lung cancer (NSCLC), colorectal cancer (CRC), and pancreatic ductal adenocarcinoma (PDAC), using IPD from KRYSTAL-1 (adagrasib) and AD from CodeBreaK100/200 (sotorasib). The model adjusted for differences in age, gender, performance status, and number of prior lines of therapy. Tumor-independent versus exchangeable models were compared. The exchangeable models allow for borrowing of treatment effects across tumors. The prior distribution for heterogeneity was informed based on pembrolizumab versus standard-of-care data in MSI-H solid tumors; sensitivity analyses explored non-informative priors.

RESULTS: Tumor-specific ORR odds ratios (OR) and 95% credible intervals (CrI) for adagrasib versus sotorasib were: NSCLC 1.60 (1.02-2.53); CRC 2.77 (1.06-8.21); and PDAC 2.15 (0.58-8.76). Using the exchangeable model, estimates became more precise, especially for PDAC where sample size was limited: OR NSCLC 1.87 (95%CrI: 1.21-2.84); CRC 2.08 (95%CrI: 1.22-3.93); and PDAC 2.02 (95%CrI: 1.14-4.05), suggesting that adagrasib is more efficacious vs sotorasib across these tumor types.

CONCLUSIONS: The proposed model allows information sharing across tumor types in the context of an unanchored ITC, increasing the precision of treatment effects where sample size may be limited.