OBJECTIVES: Duration of treatment effect is a common uncertainty in health technology assessments (HTA), but there is limited guidance about waning methods. This research aimed to investigate the implications of different waning methods used in past National Institute for Health and Care Excellence (NICE) technology appraisals (TA) for predicted survival (life years (LYs)), using published CheckMate-057 (NCT01673867) data.

METHODS: Four waning methods were applied to overall survival data for the intervention. Two methods assumed full treatment effect until Year 5, after which all treatment effect was lost, modelled by 1) applying the comparator hazard ratio (HR) to the intervention (NICE ID1510) or 2) equalizing the intervention hazard of death to the comparator (NICE TA581). Two other methods linearly waned treatment effect between Year 2 and Year 5 by 3) incrementally applying the comparator HR to the intervention or 4) incrementally equalizing the intervention hazard of death to the comparator. LYs for the intervention were calculated over a lifetime using an intervention curve choice based on best statistical fit (lognormal). Results using alternative comparator curve choices (implicated in the waning for some methods) were assessed.

RESULTS: Intervention LYs varied substantially across the four scenarios and comparator curve choices, ranging from a lower estimate of 1.41 (Method 4; Gompertz) to an upper estimate of 2.01 (Method 2; log-logistic). Intervention LYs from Method 2 and 4 varied within a mean range of 0.46 LYs based on comparator curve choice. Method 1 and 3 were independent of comparator curve choice.

CONCLUSIONS: Choice of waning method can have a substantial impact on estimated LYs and some methods were influenced by comparator curve choice. Given this, further methodological guidance and transparency on the implementation of waning methods would be valuable for future HTA and cost-effectiveness assessments.