Mixture Cure Models: A New Paradigm to Model Long-Term Remission (LTR) for Chimeric Antigen Receptor (CAR) T-Cell Therapies?

OBJECTIVES: CAR-T therapies hold the promise of a cure for some patients with haematological cancer. However, capturing it in clinical trials is not always feasible due to time constraints. For cost-effectiveness-based health technology assessments (HTAs), such benefits need to be modelled to understand the impact over patients’ lifetime. The purpose of this research is to understand the methods used by companies to model long-term remission and their potential impact on reimbursement recommendations by the National Institute of Health and Care Excellence (NICE).

METHODS: We reviewed all NICE Technology Appraisals (TAs) for CAR-T therapies, investigating approaches for modelling LTR. We extracted details including LTR modelling approach, trial follow-up durations and NICE outcomes.

RESULTS: Ten submissions were identified: 6 were accepted via Cancer Drug Fund (CDF), 1 was rejected, 3 were terminated. All 6 accepted submissions used mixture cure models. All (6/6) of those implemented cure assumption on overall survival (OS) and 83% (5/6) to event- or progression-free survival. Half (3/6) implemented the cure assumption to non-CAR-T comparators, too. From the 6 CDF recommendations, 1 was terminated by the company and two were fully recommended upon reassessment with substantially longer data than in the original submission.

In the only rejected assessment (TA894 for axicabtagene ciloleucel in follicular lymphoma), the company applied an LTR assumption other than a mixture cure model, assuming 25% of all patients treated would be cured at 5 years. Mixture cure models were described as preferred but unfeasible because of data immaturity, also due to a better prognosis for follicular lymphoma patients compared to the indications of all other CAR-T NICE submissions (89.9% 5-year survival).

CONCLUSIONS: Mixture cure models have emerged as the standard approach to model LTR and were widely accepted by NICE, as shown by the high acceptance rates. Negative recommendations were associated with data immaturity and short trial follow-up durations.