Parametric analysis and model selection for economic evaluation of survival data
Article type: Research Article
Authors: Nemes, Szilárd
Affiliations: AstraZeneca, BioPharmaceuticals R&D, Late-stage Development, Respiratory & Immunology Biometrics, Gothenburg, Sweden | E-mail: szilard.nemes@astrazeneca.com
Correspondence: [*] Corresponding author: AstraZeneca, BioPharmaceuticals R&D, Late-stage Development, Respiratory & Immunology Biometrics, Gothenburg, Sweden. E-mail: szilard.nemes@astrazeneca.com.
Abstract: Health technology assessments of interventions impacting survival often require extrapolating current data to gain a better understanding of the interventions’ long-term benefits. Both a comprehensive examination of the trial data up to the maximum follow-up period and the fitting of parametric models are required for extrapolation. It is standard practice to visually compare the parametric curves to the Kaplan-Meier survival estimate (or comparison of hazard estimates) and to assess the parametric models using likelihood-based information criteria. In place of these two steps, this work demonstrates how to minimize the squared distance of parametric estimators to the Kaplan-Meier estimate. This is in line with the selection of the model using Mean Squared Error, with the modification that the unknown true survival is replaced by the Kaplan-Meier estimate. We would assure the internal validity of the extrapolated model and its appropriate representation of the data by adhering to this procedure. We use both simulation and real-world data with a scenario where no model that properly fits the data could be found to illustrate how this process can aid in model selection.
Keywords: Parametric survival, model selection, survival extrapolation, FIC, mean squared error
DOI: 10.3233/MAS-241506
Journal: Model Assisted Statistics and Applications, vol. 19, no. 2, pp. 123-131, 2024
Health economics: choosing the right models to analyze of survival data
What is it about?
"It's tough to make predictions, especially about the future" - Yogi Berra Sill, when evaluating interventions that impact survival, we need to estimate long-term benefits based on current data. This involves analyzing trial data up to the maximum follow-up period and using parametric models to make projections. Typically, these models are compared visually to the Kaplan-Meier survival estimate, and likelihood-based information criteria are used to assess the models. However, this work suggests a different approach, focusing on minimizing the difference between the parametric estimators and the Kaplan-Meier estimate. This method aims to ensure that the extrapolated model accurately represents the data. The process is demonstrated using both simulated and real-world data, including a scenario where no suitable model was found, to show how this approach can assist in model selection.
Why is it important?
Eliminates subjective decision making with objective and measurable framework.