OBJECTIVES: Assumptions related to modelling the long-term treatment effect on survival for immuno-oncology (IO) therapies where treatment-stopping rules apply has been a key subject of discussion during National Institute for Health and Care Excellence (NICE) oncology appraisals. This study reviewed treatment-effect waning methods included in NICE technology appraisals of IO therapies where a treatment-stopping rule was applied. Furthermore, this study assessed how assumptions proposed during NICE appraisals compare with more mature survival data.
METHODS: Published NICE appraisal documents—including company submission, Evidence Review Group report, appraisal consultation(s), and final appraisal determination for IO therapies across oncology indications were reviewed up to December 2021. From each appraisal, treatment waning information were extracted and evaluated. A targeted literature review was also conducted for those appraisals which included treatment waning assumptions to identify relevant publications with more mature survival data for the associated trial(s) following the NICE appraisal.
RESULTS: In total, 14 NICE appraisals of IO therapies were identified where both a treatment-stopping rule and treatment-effect waning assumption were applied. Across all relevant submissions, the application of treatment waning assumptions varied which include waning argued to be considered within distributions selected for survival extrapolation, gradual waning of treatment effect after a certain timepoint, or setting the hazard ratio to 1 at a prespecified timepoint. Longer follow-up data were identified for a subset of the appraisals which allowed for the comparison between the observed treatment effect and the waning assumptions applied in each respective appraisal.
CONCLUSIONS: This review provides insights into the assumptions used to model long-term treatment effect in IO economic models developed for NICE appraisals and how these alternative assumptions fare in comparison with more mature data. These results may be used to inform the methodology for incorporation of treatment-effect modelling for use in future NICE submissions.
