Targeting low-dose CT to reduce mortality from lung cancer: a machine learning-based analysis of heterogeneous treatment effects in the NLST
Principal Investigator
Name
Aaron Baum
Degrees
PhD
Institution
Icahn School of Medicine at Mount Sinai
Position Title
Assistant Professor
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-324
Initial CDAS Request Approval
Jun 30, 2017
Title
Targeting low-dose CT to reduce mortality from lung cancer: a machine learning-based analysis of heterogeneous treatment effects in the NLST
Summary
The NLST showed that, on average, participants who received low-dose helical CT scans had 15-20 percent lower risk of dying from lung cancer than participants who received standard chest X-rays. However, the average study result may mask important heterogeneous treatment effects (HTEs), or systematically different outcomes among different types of study subjects. It is well recognized that traditional subgroup analyses will typically fail to identify such HTEs, because they are under-powered, are susceptible to multiple testing errors, and generally only consider one factor at a time rather than combinations of factors that are typically thought to generate HTEs. Yet detecting HTEs is critical to practicing clinicians and policy makers who must weigh the risk and benefit of an intervention for patients in order to guide personalized, safe, and cost-effective care.
This project aims to identify patient subgroups with greater and lesser lung cancer-related mortality benefit from low-dose helical CT scans. We will employ unsupervised machine learning approaches, including causal forest modeling, to estimate heterogeneous treatment effects in the NLST. These methods have been shown to identify otherwise undiscovered and clinically-meaningful relationships between interventions, outcomes, and subgroups while avoids multiple hypothesis testing concerns. Based on our previous work, we expect that estimating HTEs with this data-driven method will contribute toward safely and rationally applying the study's important scientific insights to clinical practice.
This project aims to identify patient subgroups with greater and lesser lung cancer-related mortality benefit from low-dose helical CT scans. We will employ unsupervised machine learning approaches, including causal forest modeling, to estimate heterogeneous treatment effects in the NLST. These methods have been shown to identify otherwise undiscovered and clinically-meaningful relationships between interventions, outcomes, and subgroups while avoids multiple hypothesis testing concerns. Based on our previous work, we expect that estimating HTEs with this data-driven method will contribute toward safely and rationally applying the study's important scientific insights to clinical practice.
Aims
Apply unsupervised machine learning approaches, including causal forest analysis, to the NLST data to test the hypothesis that the overall average treatment effect in the trial masked important heterogenous treatment effects in benefit from the low-dose CT intervention.
Collaborators
Juan Wisnivesky, MD, DrPh Icahn School of Medicine at Mount Sinai
Joseph Scarpa, PhD, Icahn School of Medicine at Mount Sinai
Liangyuan Hu, PhD, Icahn School of Medicine at Mount Sinai