Tree-based ensemble methods for individualized treatment rules

ABSTRACT

There is a growing interest in the development of statistical methods for personalized medicine or precision medicine, especially for deriving optimal individualized treatment rules (ITRs). An ITR recommends a patient to a treatment based on the patient's characteristics. The common parametric methods for deriving an optimal ITR, which model the clinical endpoint as a function of the patient's characteristics, can have suboptimal performance when the conditional mean model is misspecified. Recent methodology development has cast the problem of deriving optimal ITR under a weighted classification framework. Under this weighted classification framework, we develop a weighted random forests (W-RF) algorithm that derives an optimal ITR nonparametrically. In addition, with the W-RF algorithm, we propose the variable importance measures for quantifying relative relevance of the patient's characteristics to treatment selection, and the out-of-bag estimator for the population average outcome under the estimated optimal ITR. Our proposed methods are evaluated through intensive simulation studies. We illustrate the application of our methods using data from Clinical Antipsychotic Trials of Intervention Effectiveness Alzheimer's Disease Study.

KEYWORDS:

• Individualized treatment rule
• weighted classification
• random forest
• predictive biomarkers
• treatment selection

Acknowledgments

Authors would like to thank CATIE-AD study group led by Drs. Lon Schneider and Pierre Tariot for providing the clinical trial dataset used in this paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

National Institutes of Health [grant number R01 GM106177-01].

Notes on contributors

Kehao Zhu

Kehao Zhu is a biostatistician at Axio Research, LLC. He graduated from Department of Biostatistics at University of Washington with Master of Science in December 2016. His research interests include biomarker data analysis.

Ying Huang

Ying Huang , PhD, is an associate member in the Biostatistics, Bioinformatics, and Epidemiology program at the Fred Hutchinson Cancer Research Center and an affiliate associate professor in the Department of Biostatistics at the University of Washington. Dr. Huang's major areas of research are statistical methods for design and analysis of biomarker studies aimed towards disease screening, diagnosis, treatment selection, and surrogate endpoint identification, with special focus on cancer and infectious diseases related research. She is the principal investigator of a National Institute of Health (NIH) funded R01 award on “Statistical Methods for Selection and Evaluation of Biomarkers”. Dr. Huang received her PhD degree in biostatistics from the University of Washington.

Xiao-Hua Zhou

Xiao-Hua Zhou , PhD, is PKU endowed professor at Peking University and a national “the Thousand Talents Plan” distinguished expert. He is also a professor emeritus in the Department of Biostatistics at University of Washington. He is a fellow of the American Association for Advancement of Science (AAAS) and fellow of American Statistical Association. He has made some important contributions to medicine and public health by developing new statistical methods. Specifically, he has developed new statistical methods for (1) a variety of cost-related issues, (2) the analysis of quality of life data with some deaths, (3) the accuracy of diagnostic tests, and (4) causal analysis of encouragement design studies (EDS) and causal inferences in broken clinical trials, such non-compliance and truncation by death. These statistical problems originate from collaborative research that he had been doing with his medical investigators. He has published over 230 statistical methodology and medical papers and is either the corresponding author or senior author on most of them; many of them have been published in top statistical journals.