A Review of the Use of Clinical Expert Opinion to Inform Survival Curve Extrapolations in Single Technology Appraisals of New Cancer Drugs By NICE, UK

OBJECTIVES: For immature survival data, extrapolation approaches are required to estimate the full survival benefit over a lifetime time horizon. Various extrapolation approaches are available; each will result in different survival estimates. Clinical expert opinion can be used to assess the plausibility of extrapolated curves.

METHODS: Single Technology Appraisals (STAs) for new cancer drugs, published by the National Institute for Health and Care Excellence (NICE), UK, were reviewed. The publicly available Applicant Submissions and Evidence Review Group (ERG) reports were accessed at www.nice.org.uk. Pre-specified data was extracted using a data collection instrument programmed in Microsoft Forms® and analysed in Microsoft Excel®.

RESULTS: A total of 21 STAs, published from 26 May 2021 to 30 March 2022 inclusive, were reviewed. In 20, survival data of interest here (i.e. overall survival (OS) and/or time to disease progression data) had been extrapolated. Applicants had sought clinical opinion, to inform extrapolations, in 17 STAs (85%) for the intervention and in 16 STAs (80%) for the primary comparator. For the intervention, the applicants’ chosen base case extrapolations aligned with the clinical opinion in 16 of the 17 relevant STAs (94%). For the primary comparator, the applicants’ chosen base case extrapolations aligned with the clinical opinion in 15 of the 16 relevant STAs (94%). For the intervention, the ERG-preferred extrapolation aligned with the applicant base case extrapolation in 6 out of the 17 STAs (35%). For the primary comparator, the ERG-preferred extrapolation aligned with the applicant base case extrapolation in 7 out of the 16 STAs (44%).

CONCLUSIONS: Immature survival data for OS and/or time to disease progression were extrapolated in the majority of the STAs here. In these STAs, the applicants’ base case extrapolations were routinely aligned with clinical opinion. However, the ERG-preferred extrapolations often differed to those chosen by the applicant.