A Criterion-based Approach for the Systematic and Transparent Extrapolation of Clinical Trial Survival Data

Background: Trial data often does not cover a sufficiently long period of time to truly capture time-toevent endpoints, however, Health Technology Assessment (HTA) bodies often require overall survival (OS) and progression-free survival (PFS) estimates. Often, significant survival effects are found beyond the time period observed in clinical trials, thus, extrapolation of trial results is required for health economic and HTA evaluations. Objectives: This paper looks at different techniques that can be used to extrapolate trial data, as well as criteria that should be used to select the most appropriate technique. Using these insights a formal decisionmaking criteria will be established, allowing users to follow a systematic approach to extrapolating survival estimates. The techniques are then applied to a metastatic breast cancer (MBC) example. Methods: A criterion-based guide was devised to allow the accurate extrapolation and justification of survival estimates in a MBC study comparing eribulin (Halaven) monotherapy with treatment of their (patient’s) physician’s choice (TPC). Parametric and piecewise models are used to extrapolate survival estimates, and statistical as well as visual tests are used to decide the most appropriate modelling technique. Results: In the case study presented, the optimal model was identified as the Accelerated Failure Time (AFT) Parametric model using a Gamma distribution with a treatment covariate for OS, and the KaplanMeier survival estimates for PFS. Conclusions: Survival estimates must be extrapolated to a time point such that the benefits of a therapy can be clearly demonstrated. A systematic approach combined with a formal decision-making structure should be used to minimize the potential for bias as well as making the process transparent. 1Eisai Co. Ltd., Woodcliff Lake, NJ, USA; 2Curo Consulting, Marlow, Buckinghamshire, UK; 3University of Glasgow, Glasgow, UK *Corresponding author [email protected]