|
Activity Number:
|
490
- Topics in Personalized/Precision Medicine - I
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
|
Abstract #311042
|
|
|
Title:
|
Machine Learning of Survival Model with Treatment Switch
|
|
Author(s):
|
Xiaolong Luo* and Mike Branson
|
|
Companies:
|
Bristol Meyers Squibb and Bristol Myers Squibb
|
|
Keywords:
|
Deep Learning;
Clinical Trial;
Survival Analysis;
GAN (generative adversarial network)
|
|
Abstract:
|
Treatment switch and crossover are common as time-dependent covariates in clinical trials with time to death as the primary endpoint. Typical intent-to-treat analysis compares randomized treatment groups without addressing inter-current events such as drop-off and treatment crossover that can significantly impact on the primary endpoint. Their dependency on safety profile makes it difficult to handle with standard Cox regression with time dependent covariates. In this paper, we propose a neuron network model to approximate the complex dependency between endpoint and time dependent predictive features and use tensorflow to assess the predictive model with cross validation. Simulation will be used to evaluate the modeling robustness. We will apply the method to a practical clinical trial setting and explain its application in clinical trial simulation based on GAN (generative adversarial network).
|
Authors who are presenting talks have a * after their name.
