OBJECTIVES: NICE has approved many therapies for treating Non-Small Cell Lung Cancer (NSCLC) in recent years. In economic models, extrapolating time-to-event data beyond the follow-up time point is essential for evaluating long-term outcomes. Various methods are recommended in NICE Technical support documents 14 and 21 for extrapolation of time-to-event data. We aimed to review the extrapolation modeling approaches used in Technology Appraisals (TAs) of NSCLC therapies published by the NICE after 2016.

METHODS: We searched TAs on the NICE website with search terms “NSCLC” or “non-small cell lung cancer,” status of “published,” and the publication year of 2016-2023. The published documents were reviewed and summarized, focusing on the survival modeling approach, fitted statistical distributions, long-term treatment effect, and NICE's final recommendation on the extrapolation approach.

RESULTS: Twenty-four TAs were identified, where Atezolizumab (N = 5) and Nivolumab (N = 4) were the most frequently evaluated therapies. Most of the TAs (N = 20) used the conventional parametric modeling approach in the base case, and while only four TAs used the flexible parametric modeling approach, i.e., Spline based, to extrapolate the overall survival (OS) and event-free survival (EFS) beyond the clinical trial follow-up. Exponential (N=6) and log-logistic (N=6) for overall survival and Generalized gamma (N=5) and Log-logistic (N=4) for event-free survival were considered as best-fit distributions after assessing the lower AIC/BIC criterion and visual inspection. The modeling of long-term treatment effects beyond the follow-up period of the trial, including the assumptions of “Waning” (N=5) and “Cure” were explored in the most recent TAs (n=3). In most TAs, the extrapolation approach and fitted distributions were uncertain but fit for decision-making.

CONCLUSIONS: The most widely used extrapolation approach, i.e., parametric, has limitations in capturing the complex hazard function accurately. The NICE assessment emphasizes using flexible survival modeling techniques with adjusting background mortality.