Presentation (CTI)

OBJECTIVES: Extrapolating survival data is central to health technology assessment (HTA), yet it often involves considerable uncertainty. Clinical experts are frequently consulted to assess the plausibility of extrapolated survival curves. However, until recently, there has been limited guidance on how to systematically obtain credible, reliable, and transparent expert judgement on survival outcomes via structured expert elicitation (SEE). This work introduces the key elements of the newly published NICE Technical Support Document (TSD 26) on this topic and explores how these recommendations are expected to shape future elicitations.
METHODS: Existing SEE frameworks, including the Sheffield Elicitation Framework (SHELF) and the MRC protocol, were reviewed to identify generic principles of good elicitation practice. These were adapted to address the specific challenges associated with survival data, such as censoring and the role of the hazard function, and incorporated into TSD 26 to support tailored recommendations for survival extrapolation.
RESULTS: TSD 26 presents 24 key recommendations, with several addressing issues unique to survival data. A central recommendation is the discussion of the underlying hazard trend with clinical experts. This allows for experts to provide qualitative insights to contextualise and validate quantitative survival estimates, ensuring alignment with clinical expectations for the disease area and patient population. While past applications of SEE often elicit survival at multiple time points, the TSD advises that this should only occur when conditional survival relationships are explicitly addressed.
CONCLUSIONS: Structured expert elicitation is increasingly used to inform long-term survival estimates but remains under-utilised compared to informal methods, partly due to a lack of specific guidance. The introduction of TSD 26 provides detailed, survival-focused guidance that supports transparent and robust expert input. This is expected to encourage wider adoption of SEE in NICE submissions and more broadly across the literature.