Predicting Survival Probability in Clinical Trials Beyond Follow-Up Periods

• Activity Number: 372
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jerry Cheng* and John Kostis and Javier Cabrera
• Companies: Rutgers University and Rutgers University and Rutgers University
• Keywords: Survival analysis ; prediction ; parametric models ; model averaging

Abstract

The expected life expectancy in clinical trials may not be well estimated due to the lack of data or censoring beyond their follow-up periods. Making survival prediction and assessing its accuracy is a complex problem affected by many factors. Among them, the lengths of follow-up, fractions of the survived, etc. In this study, we will implement a general method to extrapolate survival curves. We will build predictive parametric models such as exponential, Weibull, log-logistic, and log-normal models. We also model hazard rate by utilizing referenced survival information from similar studies or mortality information from nationwide life tables. These models enable us to extrapolate survival estimates beyond the study observation period. Then we will apply the model averaging concept to count for model uncertainty. Therefore instead of using a single model, we will use all the modeling results by apply appropriate weights to predict long term survivability.