Abstract:Individuals often respond differently to identical treatments, and characterizing such variability in treatment response is an important aim in the practice of personalized medicine. In this article, we describe a non-parametric accelerated failure time model that can be used to analyze heterogeneous treatment effects (HTE) when patient outcomes are time-to-event. By utilizing Bayesian additive regression trees and a mean-constrained Dirichlet process mixture model, our approach offers a flexible model for the regression function while placing few restrictions on the baseline hazard. Our non-parametric method leads to natural estimates of individual treatment effect and has the flexibility to address many major goals of HTE assessment. Moreover, our method requires little user input in terms of tuning parameter selection or subgroup specification. We illustrate the merits of our proposed approach with a detailed analysis of two large clinical trials for the prevention and treatment of congestive heart failure using an angiotensin-converting enzyme inhibitor. The analysis revealed considerable evidence for the presence of HTE in both trials as demonstrated by substantial estimated variation in treatment effect and by high proportions of patients exhibiting strong evidence of having treatment effects which differ from the overall treatment effect.

What problem does this paper attempt to address?

This paper aims to solve the problem of how to analyze treatment effect heterogeneity (HTE) in the presence of censored data. Specifically, the author proposes a fully non - parametric Bayesian accelerated failure time (AFT) model for analyzing treatment effect heterogeneity when the patient outcome is survival time. By using Bayesian additive regression trees (BART) and mean - constrained Dirichlet process mixture models, this method not only provides a flexible way of modeling regression functions, but also imposes fewer restrictions on the baseline risk function. This method can naturally estimate individual treatment effects and is flexible enough to address multiple main objectives in HTE assessment, such as quantifying the heterogeneity of the overall treatment response, identifying important patient characteristics related to HTE, and estimating the proportion of patients who benefit from treatment. In addition, this method requires very little user input in terms of selecting adjustment parameters or specifying subgroups, thereby enhancing its convenience and effectiveness in practical applications. The main features of the model and method proposed in the paper include: - **Non - parametric AFT model**: The model assumes that there is a relationship between the log - failure time and covariates and treatment assignment, where the distribution of the residual term satisfies the condition of having a mean of zero. This allows the regression function to be directly interpreted as the expected log - failure time given treatment assignment and baseline covariates. - **Bayesian additive regression trees (BART)**: Used to flexibly model patient outcomes without making specific parametric assumptions or specifying the functional form of the regression model, and without the need to use pre - defined patient subgroups. BART can automatically detect non - linearities and interactions between covariates. - **Dirichlet process mixture model**: Used to model the residual distribution, providing a flexible non - parametric prior that allows for probabilistic modeling of unknowns and generates the complete posterior distribution of the desired response surface. Through these methods, researchers can directly address a series of inference targets without having to fit a series of separate models or adopt different procedures. The paper also demonstrates the advantages of the proposed method in evaluating treatment effect heterogeneity in the use of angiotensin - converting - enzyme inhibitors for the prevention and treatment of congestive heart failure in two large - scale clinical trials through detailed clinical trial analysis. The analysis results show that there is significant treatment effect heterogeneity in both trials, manifested as large estimated variation in treatment effects and strong evidence that a high proportion of patients show significant differences from the overall treatment effect.

Individualized Treatment Effects with Censored Data via Fully Nonparametric Bayesian Accelerated Failure Time Models

Bayesian Nonparametric Model for Heterogeneous Treatment Effects With Zero‐Inflated Data

Nonparametric failure time: Time‐to‐event machine learning with heteroskedastic bayesian additive regression trees and low information omnibus dirichlet process mixtures

Optimal Bayesian Hierarchical Model to Accelerate the Development of Tissue-Agnostic Drugs and Basket Trials.

Dynamic treatment effect phenotyping through functional survival analysis

Accelerated failure time models with error-prone response and nonlinear covariates

Doubly Robust Targeted Estimation of Conditional Average Treatment Effects for Time-to-event Outcomes with Competing Risks

Targeted Estimation of Heterogeneous Treatment Effect in Observational Survival Analysis

Recovering Sparse and Interpretable Subgroups with Heterogeneous Treatment Effects with Censored Time-to-Event Outcomes

Bayesian Dose Finding by Jointly Modelling Toxicity and Efficacy As Time-To-Event Outcomes

A flexible approach for causal inference with multiple treatments and clustered survival outcomes

Matching methods for obtaining survival functions to estimate the effect of a time-dependent treatment

A new method for clustered survival data: Estimation of treatment effect heterogeneity and variable selection

Some methods for heterogeneous treatment effect estimation in high-dimensions

Nonparametric analysis of delayed treatment effects using single‐crossing constraints

Semiparametric Piecewise Accelerated Failure Time Model for the Analysis of Immune-Oncology Clinical Trials

Optimal Statistical Inference for Individualized Treatment Effects in High-dimensional Models

Estimating Treatment Effect under Additive Hazards Models with High-dimensional Covariates

Semiparametric Bayesian Inference for Accelerated Failure Time Models with Errors-in-covariates and Doubly Censored Data

Some methods for heterogeneous treatment effect estimation in high dimensions

A nonparametric relative treatment effect for direct comparisons of censored paired survival outcomes