OBJECTIVES: Mixture cure models (MCMs) can differentiate between survival of cured and uncured patients. Published cost-effectiveness analyses apply the aggregated survival from MCMs, so do not differentiate between cured and uncured survival. This study compares a standard partition survival framework (PSF) with three health states based on Weibull models for progression-free survival (PFS) and overall survival (OS) to PSFs with three- and five health states in which Weibull MCMs are implemented for PFS and OS.

METHODS: The five-state MCM PSF included the following health states; PFS cured, progressed disease (PD) cured, PFS uncured, PD uncured, death. In the five-state model, cured patients were assumed to have general population utility values and no treatment and disease monitoring costs from the point of cure. PSFs were compared based on lifetime incremental costs and effects.

RESULTS: The incremental life years, QALYs and costs for the three-state PSF based on standard Weibull models were 1.59, 1.33 and 4,488, respectively. For the three-state PSF with MCMs the corresponding numbers were 1.65, 1.39 and 4,629 and for the five-state PSF with MCMs 1.65, 1.42 and 1,816, respectively. The MCMs predicted cure rates for active and placebo arms of 36.7% and 25.7% for PFS and 43.6% and 34.0% for OS, respectively. The difference between the three-state PSFs with and without MCMs is explained by the longer predicted survival tails. The difference between the three-state and five-state PSFs is explained by the differentiation in utilities and costs of cured versus uncured patients.

CONCLUSIONS: If cure is clinically plausible and there is a difference in utilities and costs of cured versus uncured patients, the five-state PSF based on MCMs more accurately describes the decision problem than a standard three-state PSF. For the five-state PSF, it is important to validate the difference between the PFS and OS cure rates are patients cured after progression.