OBJECTIVES: Health economic evaluations typically require extrapolation beyond the clinical trial period to assess long-term treatment outcomes. The objective of this review was to describe the survival models used for extrapolations, and the justifications provided by investigators, in National Institute for Health and Care Excellence (NICE) submissions.

METHODS: A targeted search was conducted to identify NICE technology appraisals in oncology, published between 01 January 2018 and 16 June 2023, where a pharmacological intervention was evaluated and extrapolation methods were reported. From eligible appraisals, we extracted the types of survival models along with justifications provided (if any), and the impact of the chosen extrapolation method on base-case and scenario analyses results.

RESULTS: Of the 389 records identified, 166 were eligible; with 150 appraisals using cost-utility analyses, 15 reported cost-effectiveness analyses and 1 cost-comparison analysis. Non-small-cell lung cancer (19%), breast cancer (12%), and lymphoma (12%) were the most commonly evaluated indications. Partitioned-survival models (72%) were most commonly used in economic analyses, while Markov models (13%), semi-Markov (7%), and hybrid models (6%) were utilized less frequently. Overall survival (80%) and progression-free survival (72%) were the most commonly extrapolated outcomes and the choice of survival models ranged from parametric to piecewise and mixture cure models. The majority of appraisals determined the suitability of models through visual inspection (79%), goodness-of-fit statistics (83%), and clinical validity (71%). Scenario analyses reported that the choice of extrapolation methods had important impacts on incremental cost-effectiveness ratios, ranging from -83% to 2,136% of the base-case model.

CONCLUSIONS: Our targeted review found that the choice of survival models varied across NICE appraisals and can have an important impact on cost-effectiveness results. Therefore, the justification provided for the choice of extrapolation approaches should be clearly stated. This consideration should be extended to all economic literature to better facilitate decision-making and increase transparency and credibility.