Multi-armed Bandit Models for the Optimal Design of Clinical Trials: Benefits and Challenges

Sofia Villar, Jack Bowden, James Wason

Research output: Contribution to journalArticle (Academic Journal)peer-review

98 Citations (Scopus)
47 Downloads (Pure)

Abstract

Multi-armed bandit problems (MABPs) are a special type of optimal control problem well suited to model resource allocation under uncertainty in a wide variety of contexts. Since the first publication of the optimal solution of the classic MABP by a dynamic index rule, the bandit literature quickly diversified and emerged as an active research topic. Across this literature, the use of bandit models to optimally design clinical trials became a typical motivating application, yet little of the resulting theory has ever been used in the actual design and analysis of clinical trials. To this end, we review two MABP decision-theoretic approaches to the optimal allocation of treatments in a clinical trial: the infinite-horizon Bayesian Bernoulli MABP and the finite-horizon variant. These models possess distinct theoretical properties and lead to separate allocation rules in a clinical trial design context. We evaluate their performance compared to other allocation rules, including fixed randomization. Our results indicate that bandit approaches offer significant advantages, in terms of assigning more patients to better treatments, and severe limitations, in terms of their resulting statistical power. We propose a novel bandit-based patient allocation rule that overcomes the issue of low power, thus removing a potential barrier for their use in practice.
Original languageEnglish
Pages (from-to)199-215
JournalStatistical Science
Volume30
DOIs
Publication statusPublished - 29 Jul 2015

Bibliographical note

Published at http://dx.doi.org/10.1214/14-STS504 in the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Keywords

  • stat.ME

Fingerprint

Dive into the research topics of 'Multi-armed Bandit Models for the Optimal Design of Clinical Trials: Benefits and Challenges'. Together they form a unique fingerprint.
  • Jack Bowden fellowship transfer

    Gaunt, L. F.

    1/08/1531/03/18

    Project: Research

  • MRC UoB UNITE Unit - Programme 1

    Davey Smith, G.

    1/06/1331/03/18

    Project: Research

  • IEU Theme 3

    Windmeijer, F., Tilling, K. M. & Tilling, K. M.

    1/06/1331/03/18

    Project: Research

Cite this