Survival Analysis for Economic Evaluations Alongside Clinical Trials—Extrapolation with Patient-Level Data: Inconsistencies, Limitations, and a Practical Guide

Abstract

Background. In health technology assessments (HTAs) of interventions that affect survival, it is essential to accurately estimate the survival benefit associated with the new treatment. Generally, trial data must be extrapolated, and many models are available for this purpose. The choice of extrapolation model is critical because different models can lead to very different cost-effectiveness results. A failure to systematically justify the chosen model creates the possibility of bias and inconsistency between HTAs. Objective. To demonstrate the limitations and inconsistencies associated with the survival analysis component of HTAs and to propose a process guide that will help exclude these from future analyses. Methods. We reviewed the survival analysis component of 45 HTAs undertaken for the National Institute for Health and Clinical Excellence (NICE) in the cancer disease area. We drew upon our findings to identify common limitations and to develop a process guide. Results. The chosen survival models were not systematically justified in any of the HTAs reviewed. The range of models considered was usually insufficient, and the rationale for the chosen model was universally limited: In particular, the plausibility of the extrapolated portion of fitted survival curves was very rarely explicitly considered. Limitations. We do not seek to describe and review all methods available for performing survival analysis—several approaches exist that are not mentioned in this article. Instead we seek to analyze methods commonly used in HTAs and limitations associated with their application. Conclusions. Survival analysis has not been conducted systematically in HTAs. A systematic approach such as the one proposed here is required to reduce the possibility of bias in cost-effectiveness results and inconsistency between technology assessments.