Activity Number:
|
178
- Recent Development on the Analysis of Time-to-Event Data
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
|
Sponsor:
|
Lifetime Data Science Section
|
Abstract #313641
|
|
Title:
|
Mixture Cure Rate Models with BART
|
Author(s):
|
Xiao Li* and Rodney Sparapani and Brent Logan
|
Companies:
|
Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin
|
Keywords:
|
Mixture cure rate models;
BART
|
Abstract:
|
In many clinical studies with survival outcomes, a fraction of patients may be considered cured by the therapy in terms of achieving long-term survival or disease control. Such data can be analyzed using mixture cure rate models that assume the population contains a mixture of cured and uncured patients. Cure status is a latent variable for those censored patients. Recently, flexible survival prediction models using machine learning techniques have been used to improve predictive performance while minimizing assumptions about the functional form of the relationship between covariates and outcomes. We propose an extension to a Bayesian machine learning technique called Bayesian Additive Regression Trees (BART) to address mixture cure models. Two BART models are used to simultaneously model the cure probability, a probit BART; and the survival distribution in uncured individuals via a log-normal accelerated failure time BART model. The performance of the BART mixture cure rate model is evaluated on simulated data. We also apply our model to a real study of patients undergoing a potentially curative therapy of hematopoietic stem cell transplantation and compare the results with others
|
Authors who are presenting talks have a * after their name.