Heterogeneous treatment effect analysis based on machine‐learning methodology

Abstract

Heterogeneous treatment effect (HTE) analysis focuses on examining varying treatment effects for individuals or subgroups in a population. For example, an HTE‐informed understanding can critically guide physicians to individualize the medical treatment for a certain disease. However, HTE analysis has not been widely recognized and used, even given the explosive increase of data availability attributed to the arrival of the Big Data era. Part of the reason behind its underuse is that data are often of high dimension and high complexity, which pose significant challenges for applying conventional HTE analysis methods. To meet these challenges, a newly developed causal forest HTE method has been derived from the random forest machine‐learning algorithm. We conducted a systematic performance evaluation for the causal forest method against the conventional two‐step method by simulating scenarios with different levels of complexity for the analysis. Our results show that causal forest outperforms the conventional HTE method in assessing treatment effect, especially when data are complex (e.g., nonlinear) and high dimensional, suggesting that causal forest is a promising tool for real‐world applications of HTE analysis.

Artificial Intelligence for Quantitative Modeling in Drug Discovery and Development: An Innovation and Quality Consortium Perspective on Use Cases and Best Practices

“…From a regulatory perspective, AI/ML approaches can be leveraged across several areas to support overall drug development and regulatory efficiency. This includes, but is not limited to: causal inference 47 ; automation tools for bioequivalence assessment 48 or facilitating product specific guidance 49 ; business intelligence to predict submissions of abbreviated new drug applications 50,51 ; regulatory equivalence assessment for complex particle size distribution 52 ; and multivariate analysis methods to facilitate active pharmaceutical ingredient sameness assessment. 53 Of note, the value of adopting AI/ML approaches to mine large and heterogenous datasets has been shown in recent regulatory applications focusing on the assessment of heterogeneous treatment effect (HTE).…”

Causal Effect of Parent-Child Relationship on Oppositional Defiant Disorder Symptoms Based on Machine Learning Approach

“…The heterogeneity further detected in CF model is also quite critical, which focuses on evaluating the different causal effects across individuals or subgroups. With the heterogeneous analysis, doctors could be guided to conduct personalized treatments for speci c diseases [37]. Moreover, the heterogeneity is speci c helpful when evaluating the effect of public policy.…”

Rule ensemble method with adaptive group lasso for heterogeneous treatment effect estimation

“…However, this is challenging because real‐world data are complex, noisy, and highly dimensional. Recent studies on HTE have proposed machine learning (ML) methods due to their superior performance on increasingly complex and high‐dimensional data 11 . One common approach is to apply the ML method to build models for the control group $$$ {\mu}_0\left({\boldsymbol{x}}_i\right) $$$ and treatment group $$$ {\mu}_1\left({\boldsymbol{x}}_i\right) $$$, and using the estimates of each group to calculate the HTE 12 .…”

“…Recent studies on HTE have proposed machine learning (ML) methods due to their superior performance on increasingly complex and high-dimensional data. 11 One common approach is to apply the ML method to build models for the control group 𝜇 0 (x i ) and treatment group 𝜇 1 (x i ), and using the estimates of each group to calculate the HTE. 12 However, these models can have different structures (ie, the basis function).…”