Determination of the most appropriate method for extrapolating overall survival data from a placebo-controlled clinical trial of lenvatinib for progressive, radioiodine-refractory differentiated thyroid cancer

BACKGROUND: Cost-effectiveness models for the treatment of long-term conditions often require information on survival beyond the period of available data. OBJECTIVES: This paper aims to identify a robust and reliable method for the extrapolation of overall survival (OS) in patients with radioiodine-refractory differentiated thyroid cancer receiving lenvatinib or placebo. METHODS: Data from 392 patients (lenvatinib: 261, placebo: 131) from the SELECT trial are used over a 34-month period of follow-up. A previously published criterion-based approach is employed to ascertain credible estimates of OS beyond the trial data. Parametric models with and without a treatment covariate and piecewise models are used to extrapolate OS, and a holistic approach, where a series of statistical and visual tests are considered collectively, is taken in determining the most appropriate extrapolation model. RESULTS: A piecewise model, in which the Kaplan–Meier survivor function is used over the trial period and an extrapolated tail is based on the Exponential distribution, is identified as the optimal model. CONCLUSION: In the absence of long-term survival estimates from clinical trials, survival estimates often need to be extrapolated from the available data. The use of a systematic method based on a priori determined selection criteria provides a transparent approach and reduces the risk of bias. The extrapolated OS estimates will be used to investigate the potential long-term benefits of lenvatinib in the treatment of radioiodine-refractory differentiated thyroid cancer patients and populate future cost-effectiveness analyses.