ASID: A Bayesian Adaptive Subgroup-Identification Enrichment Design

Alzheimer"s disease;Biomarker;Clinical trial;Hierarchical Bayesian Model;Targeted therapies;not listed

Constantine, Florica Joanne

Johns Hopkins University

关键词: Alzheimer";    s disease;    Biomarker;    Clinical trial;    Hierarchical Bayesian Model;    Targeted therapies;    not listed;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/40729/CONSTANTINE-THESIS-2017.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

It has been observed and is generally accepted that patients with a given disease may respond differently to the same treatment. Hence, it is sensible to believe that there are subgroups of patients, delineated by their biomarker profiles, wherein certain treatments are better choices than others. However, it is difficult to predict a priori which patients are good candidates for a given treatment; an ideally designed trial would adaptively find and update subgroups of patients at an interim analysis point. We propose a method that does exactly this: ASID, for Adaptive Subgroup- Identification and enrichment Design. ASID finds predictive biomarkers, estimates which patient subgroups have differential treatment effects, and modifies the trial recruitment criteria at an interim analysis point. Moreover, ASID is based on a hierarchical Bayesian model. In this work, motivated by an Alzheimer’s Disease clinical trial, we derive and analyze ASID, and compare it to an alternative adaptive enrichment design built around a linear regression model as well as to a random forest based model (GUIDE). Via numerical simulations, we demonstrate the superiority of ASID.