|
Activity Number:
|
29
- Statistical Issues Specific to Therapeutic Areas, Power and Sample Size Calculations, and Trial Monitoring
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
|
Abstract #317889
|
|
|
Title:
|
A Flexible Ensemble Learning Method for Survival Extrapolation
|
|
Author(s):
|
Meijing Wu* and Ran Dai and Yabing Mai and Weili He
|
|
Companies:
|
Abbvie Inc. and University of Nebraska Medical Center and Boehringer Ingelheim and AbbVie
|
|
Keywords:
|
Survival extrapolation ;
health technology assessment
|
|
Abstract:
|
Survival extrapolation to estimate long-term survival from short-term clinical trial data beyond trial follow-up has become an important part for cost-effectiveness analyses of oncology products. It is widely used in health technology assessment (HTA). The common strategy for survival extrapolation is to fit one or two parametric models based on experience, or select a model from a predefined model collection based on some goodness-of-fit statistics. The main challenge in survival extrapolation is that the result is very sensitive to model misspecification. In this paper, we propose a new framework, namely the Ensemble Learning for Survival Extrapolation (ELSE). Instead of selecting one best model from the model collection, ELSE builds an ensemble model based on the survival extrapolation results from all the models in the model collection. Extensive numerical studies are performed to compare the performance of ELSE with the model selection procedure based on AIC. We also apply this method to cancer data from an open-source research project as a demonstrating example.
|
Authors who are presenting talks have a * after their name.
