Work

Within-trial data borrowing for sequential multiple assignment randomized trials

Year: 2024

Type: article

Abstract: The Sequential Multiple Assignment Randomized Trial (SMART) is a complex trial design that involves randomizing a single participant multiple times in a sequential manner. This results in the branching nature of a SMART, which represents several distinct groups defined by different combinations of treatments, response statuses, etc. A SMART can then answer various scientific questions of interest, eg, the optimal dynamic treatment regime (DTR) for treating a chronic illness, what intervention to offer first, and what intervention to offer to nonresponders (or suboptimal responders). However, the analysis of a SMART can suffer from low precision, as the potentially widely branching structure can lead to reduced sample sizes in some groups of interest. In this paper, we propose a novel analysis method for a SMART in which dynamic borrowing is used to borrow strength across groups with similar expected outcomes, thus providing increased precision for the estimation of the expected outcomes of DTRs. We apply our method to a SMART evaluating various weight loss strategies using a binary endpoint of clinically significant weight loss and show by simulation that our method can improve the precision of the estimated expected outcome of a DTR, aid in the identification of the optimal DTR, and produce a clustering analysis of DTRs embedded in a SMART.

Source: Biostatistics

DOI: 10.1093/biostatistics/kxaf003

Authors: Ales Kotalik, David M Vock, Nancy E Sherwood, Brian P Hobbs, Joseph S Koopmeiners

Institutions: University of Minnesota, The University of Texas at Austin

Language: English

Topic: Advanced Causal Inference Techniques

Subfield: Statistics and Probability

Field: Mathematics

Domain: Physical Sciences

Open Access status: green

Funders: National Institutes of Health, National Heart, Lung, and Blood Institute

Award: G9000090923