Health technology assessment bodies like the National Institute for Health and Care Excellence (NICE) in the UK are interested in hard outcomes, like long-term overall survival (OS), which are not always available in oncology clinical trials. We reviewed the methods used by immuno-oncology (I-O) manufacturers to extrapolate OS in their economic evaluations, and the critiques of these methods highlighted by NICE.

Technology appraisals of I-O therapies published between 2011 and 2020 were identified from the NICE website (excluding terminated appraisals), based on the Cancer Research Institute classification of immunotherapies. The methodology to extrapolate OS and NICE’s critiques of these methods were extracted from Final Appraisal Documents.

OS extrapolation was included in the majority of appraisals (58/75). The most widely used extrapolation method was a standard parametric model (42/75), followed by spline models (6/75), external sources (3/75), cure models (3/75), only trial data (2/75), response-based analysis (1/75) and trial data/published sources (1/75). Where NICE rejected the extrapolation method (16/75), the main critiques asserted that OS estimates were implausible and inconsistent with the clinical trial evidence or the real-world data (5/16), followed by scepticism of assumptions implying future mortality rates of patients will be similar to or lower than those of the general population (4/16). Other issues flagged by NICE included long extrapolation used despite median survival not being met or short follow-up-data (2/16), improper application of the hazard function/relative risk in OS analysis (2/16), trial issues including censoring (1/16), use of parametric models instead of observed data (1/16) and disregard of covariates (1/16).

The number of I-O therapies assessed by the NICE has increased exponentially in recent years. Uncertainty in the long-term OS impacts the cost-effectiveness of I-O treatments. Issues with plausibility and mortality risk assumptions can lead to a rejection of OS extrapolation methods by NICE.