OBJECTIVES: Several methods are available for extrapolating survival data beyond the observation period, although the acceptability of these methods by HTA bodies is uncertain. This study aims to evaluate current and future trends in methods for survival extrapolation in HTA across countries.

METHODS: Nine oncology products were selected based on pre-defined criteria. HTA reports and survival extrapolation methods were identified within Australia, Canada, France, Portugal, Sweden, the Netherlands and the UK. A survey of 10 experts provided expert commentary regarding the future trends for the acceptability of extrapolation methodologies in these countries.

RESULTS: Ninety-two submissions were analysed in total, data on survival extrapolations were available in 61% of them. Parametric models were considered in 100% submissions with available data, even if not included in the base case scenario, while non-parametric models were omitted from 38% of submissions. Non-parametric methods were chosen as the base case for at least one survival curve in 35 submissions, with response driven models used in 4, mixture-cure models applied in 3, and all other non-parametric approaches applying various piecewise models. When non-parametric methods were used in the base case, 47% submissions received major comments from authorities, whereas when only parametric methods were applied to the base case, major comments were received 39% of the time. These results are consistent with the expert survey, where 8/10 experts indicated parametric models as preferred over innovative methods and more acceptable by authorities. All respondents noted this will likely change in the future with increased awareness of innovative methods.

CONCLUSIONS: These results illustrate that more innovative non-parametric survival extrapolations are not well established in HTA, and may take some time to be regularly accepted by authorities. Increased use of these methods, however, will likely increase their acceptability.