PBI36 REVIEW AND COMPARISON OF OVERALL SURVIVAL EXTRAPOLATION IN HEALTH TECHNOLOGY ASSESSMENTS FOR CHIMERIC ANTIGEN RECEPTOR T-CELL THERAPIES (CAR-TS)

Abstract

The US Institute for Clinical and Economic Review (ICER) and the UK National Institute for Health and Care Excellence (NICE) recently evaluated chimeric antigen receptor T-cell therapies (CAR-Ts) for pediatric B-cell acute lymphoblastic lymphoma (B-ALL), and diffuse large B-cell lymphoma (DLBCL) in adults. We reviewed and compared assessments of CAR-Ts from ICER and NICE, focusing on the extrapolation of overall survival (OS). One ICER evaluation and three NICE technology assessments (TAs) of the cost effectiveness of CAR-Ts in B-ALL and DLBCL were identified. Methods used for extrapolation of OS were compared. For both B-ALL and DLBCL populations in ICER’s assessment, standard parametric survival curves were extrapolated for 5 years, after which survivors were considered cured and followed general population mortality. In NICE TA559 and TA554, mixture-cure models were used for modelling OS. Here, a cure fraction was derived; cured patients followed general population mortality, whereas patients who were not followed fitted parametric survival curves. The Committee considered this method appropriate, but also found the cure fraction, a key driver of results, to be highly uncertain. The Committee for NICE TA567 considered a 1-knot spline extrapolation appropriate but insisted more long-term data were needed to decide the point after which patients are assumed to follow general population mortality. In all three NICE TAs, CAR-Ts were recommended for the Cancer Drugs Fund due to uncertainties in OS data. While ICER’s assessment yielded favorable cost-effectiveness results, nine of 13 voters stressed the significant uncertainty around long-term clinical benefits. Both ICER and NICE noted the uncertainty with extrapolating short-term survival data beyond trial period for CAR-Ts but appeared to accept the use of mixture-cure and spline models over traditional parametric survival modelling to represent their potential curative nature. Longer follow-up data should be collected to confirm long-term survival benefit of CAR-Ts.