OBJECTIVES : Standard parametric survival models (SPMs) often fail to reflect complex hazard behaviours for oncology therapies with potential long-term survivors. Survival extrapolation is a key driver of cost-effectiveness; thus, an extrapolation methodology with both statistical and clinical validity should inform reimbursement decisions. Mixture cure models (MCMs) have been proposed as an alternative to SPMs where a proportion of the patient cohort may be considered “cured”. This study aims to review how MCMs have been implemented in oncology cost-effectiveness models (CEMs) submitted to the National Institute for Health and Care Excellence (NICE).

METHODS : NICE oncology technology appraisals (TAs) final appraisal documents were searched without time limitation through June 18, 2020. Keywords searched were: “mixture cure model” and “cure fraction”. Committee papers were then screened for further information and commentary from the committee.

RESULTS : Six TAs were identified (478, 545, 552, 554, 559, 589). One manufacturer implemented MCM after committee consultation (552). None directly used the cure fraction and survival curves for the ‘cured’ portion to inform a defined “cured” health state. Manufacturers rationales for adopting MCMs included: implausible long-term survival modelled with SPM (554, 559), Kaplan Meier (KM) plateaus (478, 545, 554), and MCM projected survival validated by clinical expert opinion (545, 554, 559, 589). The main concerns raised by Evidence Review Groups (ERGs) were: differences in cure fraction estimates across different survival functions and endpoints, KM plateaus informed by small number of patients at risk without showing evidence of “cure”, and clinically implausible projections. MCMs were applied in the final CEM used for decision making (478, 545, 552, 554) or were considered alongside other extrapolation methods (559, 589).

CONCLUSIONS : NICE may accept the use of MCMs in cases where SPMs fail to project clinically plausible long-term survival, providing the underlying evidence is considered mature enough to estimate stable cure fractions and believable survival estimation.