PURPOSE: Extrapolation of survival curves for cost-effectiveness analyses often rely on a single parametric distribution fitted to relatively short-term trial data. Extrapolations can be very sensitive to the choice of distribution, even when the distributions fit equally well to the short-term data, causing structural uncertainty. Sensitivity analyses to address parameter uncertainty do not account for this structural uncertainty. Further, the selection of the parametric distribution is frequently scrutinized by decision-makers, especially for therapies exhibiting a long tail in survival, e.g. immuno-oncology therapies. Methods available to address structural uncertainty in survival extrapolation include Bayesian Model Averaging (BMA) which pools results from different distributions and use of external evidence to provide constraints. This workshop aims to 1) answer the question, how BMA and external data can be utilized in survival extrapolations, and 2) debate the value of this approach. 

DESCRIPTION: This workshop will present how BMA and external data can be utilized to reduce uncertainty in survival extrapolations, based on practical examples. First, an overview of current practices and related uncertainties will be presented from an industry perspective (15 minutes). The concept of BMA will be introduced based on an illustrative case study using BMA for several indications (15 minutes). The BMA approach will also be discussed within the context of HTA appraisals. Further, appropriate external data sources and how these can be incorporated in the BMA will be presented (20 minutes). The workshop will conclude with a debate, involving the audience, to discuss the suitability of the BMA approach to reduce uncertainty (10 minutes). Throughout the workshop, the audience will be encouraged to participate in the discussion either in-person or via polling, evaluating the suitability of the presented approaches. The workshop will provide illustrative examples and will be relevant for a broader audience including data analysts, industry representatives, and decision-makers alike.