Bayesian Causal Machine Learning for Cure Models

2026-06-09

AI Summary

This study addresses a key challenge in survival analysis: disentangling whether a treatment acts by increasing the probability of cure or by delaying failure time among uncured individuals, along with identifying heterogeneous effects across subpopulations. Existing causal machine learning methods struggle to distinguish these mechanisms. To bridge this gap, the authors propose BartCure, a novel approach that, for the first time, decomposes the causal effect on restricted mean survival time into “stochastic cure” and “stochastic delay” components, linking them to principal stratum causal effects and stochastic intervention effects. Built upon Bayesian additive regression trees (BART), BartCure formulates a structured causal model enabling full posterior inference. Simulation studies demonstrate its competitive performance in estimating average effects and its robustness in detecting the direction of effect heterogeneity. Applied to the CALGB 40101 breast cancer trial, BartCure successfully uncovers clinically meaningful subgroup-specific treatment effects.

Abstract

In survival studies, treatments can benefit patients through different mechanisms: a treatment may increase the probability of being cured or delay failure among patients who are not cured. Quantifying which mechanism is dominant, and whether it varies across subpopulations, is clinically important, yet there is limited work in the causal machine learning literature addressing this problem. Standard causal survival learners target finite-horizon survival or restricted mean survival time, while many cure models capture cure structures without estimating causal effects. In this work, we define meaningful causal effects in the presence of a cured subpopulation and introduce BartCure, a Bayesian causal machine learning approach for estimating them. The causal effects we recommend decompose the causal effect on restricted mean survival time into a stochastic cure and stochastic latency component, and we relate these new effects to both stochastic intervention effects and causal effects in principal strata. In simulations, BartCure is competitive for estimating average effects and is especially effective at conservatively detecting the direction of treatment-effect heterogeneity. We apply BartCure to estimate average and subgroup causal effects and to identify treatment effect heterogeneity in the CALGB 40101 breast cancer trial.

Problem

• cure models
• causal effects
• treatment-effect heterogeneity
• survival analysis
• Bayesian causal machine learning

Innovation

• Bayesian causal machine learning
• cure models
• treatment effect heterogeneity
• restricted mean survival time
• BartCure