Estimating heterogeneous survival treatment effects of colorectal cancer screening approaches: A causal machine learning analysis
Principal Investigator
Name
Wang Zhe
Degrees
M.P.H
Institution
Weifang Medical University
Position Title
Postgraduate Student
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-824
Initial CDAS Request Approval
Sep 14, 2021
Title
Estimating heterogeneous survival treatment effects of colorectal cancer screening approaches: A causal machine learning analysis
Summary
These important issues warrant further investigation of the overall and heterogeneous survival effect of colorectal cancer in relation to screening using more flexible and rigorous statistical approaches. In this study, we apply recent advances in statistical machine learning (ML) to precisely identify and evaluate TEH in relation to covariates presented in the PLCO data. Specifically, we use the Bayesian additive regression trees (BART) model, which is a decision tree ensemble in the generative probabilistic model framework that has gained wide popularity and influence over the past years in statistical and public health research . The BART model has been shown to have better prediction performance than many alternative ML techniques, including random forests, boosting, neural net- works, and parametric methods in a wide variety of comparative study settings including those where the proportion of outcome events is small In this study, we reanalyzed PLCO data using ML techniques to generate insights into personalized screening strategies for improving LC survival as well as aid in the planning of clinical trials.
Aims
There may be heterogeneity in PLCO population response to screening. The purpose of this analysis is to
Identifying subgroups in the PLCO population that would actually benefit from screening could improve the rate of early colorectal cancer diagnosis and improve prognosis.
Collaborators
Wang suzhen,Weifang Medical University
Shi fuyan,Weifang Medical University