Multiple Real-World Data Sources in a Bayesian Framework to Inform Long-Term Survival Estimates of Mosunetuzumab in Patients with Follicular Lymphoma.

Abstract

Patients with follicular lymphoma (FL) receiving third-line or later (3L +) therapy have long survival, which can make estimating long-term overall survival (OS) from trial data challenging. The objective of this study was to estimate long-term OS for mosunetuzumab from the GO29781 trial (NCT02500407) using multiple real-world databases (RWDs) in a Bayesian framework. Seven RWD sources for patients with FL receiving 3L + therapy and the expansion cohort in the GO29781 trial for mosunetuzumab were used. Hazard trends from the RWD sources were analyzed, and disease-wide pointwise OS and its corresponding uncertainty were estimated using Bayesian random-effects meta-analysis from the RWD sources. Pointwise OS obtained was used as an informative prior in Bayesian survival extrapolations to data from patients receiving mosunetuzumab. Results after adjusting for background mortality were compared to equivalent frequentist extrapolations using trial data only. Hazard patterns from RWD sources supported a constant or linearly decreasing hazard. Mean pointwise OS for patients with FL receiving 3L + therapy was estimated at 0.52 (95% credible interval, 0.29-0.85) at 8years. Bayesian extrapolations for mosunetuzumab produced median survival estimates of 11.6 (6.7-20.7)years to 17.0 (6.4-22.7)years depending on the distribution used, reducing uncertainty by 20% to 46% relative to the frequentist estimation. Multiple RWD sources can be synthesized to augment the credibility of data with short follow-up, long patient survival, and few events to effectively estimate long-term survival and reduce estimated uncertainty. This method can be applied to other indications with similar characteristics. NCT02500407, July 16, 2015.