Projecting ‘time to event’ outcomes in technology assessment: an alternative paradigm

Abstract: One of the main considerations in NICE appraisals of novel therapeutic technologies is the estimation of life-time survival outcomes from limited data available from one or two clinical trials.  Generally this problem is addressed by selecting from a limited set of standard statistical functions, and then applying the ‘best fit’ option to project outcomes beyond the available Kaplan-Meier analysis results.  This approach is widely taught and accepted, but there are a number of problems associated with this paradigm (both theoretical and practical), which are outlined and illustrated using examples from various NICE appraisals and other HTA evaluations.

An alternative approach to projective modelling is described which involves focussing on the dynamics of risk, the quality and characteristics of the empirical data, and the primary objective of anticipating the likely pattern of risk beyond the available data.

This presentation draws heavily on a range of real-life examples, especially those involving the treatment of advanced or metastatic cancers, illustrating how a more pragmatic approach to projective estimation of life-time outcome measures can often give novel insights into the nature of the progression of a disease and the modes of action of the medical interventions available.