Primary Submission Category: Applications in Health and Biology
Unveiling Heterogeneous Treatment Effects in the BEST-CLI Trial: Harnessing Causal Machine Learning to Personalize Care for Chronic Limb-Threatening Ischemia
Authors: Tien Tran, Alik Farber, Matthew Menard, Kenneth Rosenfield, Niteesh Choudhry, Zafar Zafari,
Presenting Author: Tien Tran*
Randomized controlled trials (RCTs) are foundational for determining the efficacy of treatments, yet their insights often overlook heterogeneous treatment effects (HTEs) across subpopulations. Machine learning (ML) methods, such as causal forests, have emerged as powerful tools to identify these HTEs, enhancing healthcare decision-making by personalizing treatment and optimizing resource allocation. This study applies causal forests to the BEST-CLI trial, a multicenter RCT comparing open surgical bypass (OSB) with endovascular therapy (EVT) for patients with chronic limb-threatening ischemia.
Leveraging individual-level data from the trial, we estimated subgroup-specific treatment effects to explore variability in clinical outcomes. Our analysis revealed significant heterogeneity in the benefits of OSB versus EVT based on key patient characteristics, including age, comorbidities, and anatomical considerations. These findings underscore the potential of causal machine learning to refine clinical guidelines by tailoring treatment recommendations to individual patient profiles. By incorporating ML-derived insights into RCT analyses, this study demonstrates a methodological advance in translating trial data into actionable, patient-centered care strategies.