Making better use of evidence to reflect heterogeneity and uncertainty in survival predictions within cost-effectiveness models

The application of survival analysis to estimate the rate at which disease or clinical events occur within cost-effectiveness models has become increasingly sophisticated. Relatively little attention has been paid to how survival analysis methods can be used to reflect underlying disease and clinical processes. This thesis uses three case studies to develop and demonstrate approaches that more explicitly link decision-analytic model structures and accompanying survival analyses to the underlying disease and treatment processes of interest. This allows for more comprehensive use of evidence, and explicit assessment of the effects of heterogeneity and uncertainty. The developed approaches allow disease and treatment mechanisms driving heterogeneity in event risk and relative treatment effects to be reflected within cost-effectiveness analyses. This allows cost-effectiveness results to robustly reflect differences between patients and inform transparent optimised recommendations. The implementation of different decision modelling approaches across the case studies reveals the importance of more explicitly linking the model structure and survival analyses. Partitioned survival models, by directly modelling overall survival, disconnect overall survival from other modelled disease and treatment processes. This can limit the potential to use evidence on intermediate outcomes to inform overall survival extrapolations, and limit exploration of how uncertainty in the extrapolation period impacts on decision uncertainty. State transition models underpinned by multi-state survival analysis can allow a fuller use of evidence and exploration of uncertainty, but can introduce practical challenges and technical uncertainties. The choice of when to implement each approach requires a model conceptualisation process that considers the anticipated importance of heterogeneity in survival outcomes, the nature of direct and external survival evidence, and the key areas of uncertainty. The importance of this work has been recognised by a broad range of decision makers and has directly informed NICE Technology Appraisal recommendations; clinical guidelines and methods guidance.