Nonparametric machine learning for precision medicine with longitudinal clinical trials and Bayesian additive regression trees with mixed models

Abstract

Precision medicine is an active area of research that could offer an analytic paradigm shift for clinical trials and the subsequent treatment decisions based on them. Clinical trials are typically analyzed with the intent of discovering beneficial treatments if the same treatment is applied to the entire population under study. But, such a treatment strategy could be suboptimal if subsets of the population exhibit varying treatment effects. Identifying subsets of the population experiencing differential treatment effect and forming individualized treatment rules is a task well‐suited to modern machine learning methods such as tree‐based ensemble predictive models. Specifically, Bayesian additive regression trees (BART) has shown promise in this regard because of its exceptional performance in out‐of‐sample prediction. Due to the unique inferential needs of precision medicine for clinical trials, we have proposed the BART extensions explicated here. We incorporate random effects for longitudinal repeated measures and subject clustering within medical centers. The addition of a novel interaction detection prior to identify treatment heterogeneity among clinical trial patients and its association with patient characteristics. These extensions are unified under a framework that we call mixedBART. Simulation studies and applications of precision medicine based on real randomized clinical trials data examples are presented.

An integrated Bayesian framework for multi-omics prediction and classification

“…Another extension opportunity entails considering an ensemble learner for each individual layer to further improve prediction although at the expense of interpretability 12 or use the capability of conformal prediction 19 to further improve uncertainty quantification. Direct incorporation of random effects for repeated measures and clustered observations offers another promising avenue to integrate longitudinal biomarkers 20,21 . Finally, and relatedly, it is not straightforward to incorporate any type of prior immunological knowledge (e.g., regulatory networks) into the Bayesian paradigm.…”

The Role of Cytoreductive Nephrectomy in Metastatic Renal Cell Carcinoma: A Real-World Multi-Institutional Analysis

“…Adjusted analyses were conducted via Inverse Probability of Treatment Weighing (IPTW) based on the generalized propensity score, 13 with propensity scores estimated via Bayesian Additive Regression Trees. 14 Covariates in the propensity score model were age, gender, race, insurance at diagnosis and IMDC risk group. Where covariates were missing, we used a missing category in the propensity score model.…”

Stan and BART for Causal Inference: Estimating Heterogeneous Treatment Effects Using the Power of Stan and the Flexibility of Machine Learning

“…However, their primary aim was to produce consistent estimates in the presence of unmeasured, group-level confounders and as such, their approach addressed a different issue. Another related BART extension was described in Spanbauer and Sparapani [ 32 ]. This approach incorporated random effects for longitudinal repeated measures into the BART model as well as subject clustering within groups.…”