|
Abstract:
|
The clinical management of an individual patient's condition may require subsequent changes in therapy based on the patient's response to an initial treatment. Such a predefined sequence of actions defines a treatment strategy. Sequential multiple assignment randomized trials (SMART) have been proposed to efficiently evaluate and compare treatment strategies. Traditional statistical approaches to study design do not account for costs. Given an effect size, the power to detect differences between treatment strategies is typically a function of the sample size, and the initial and re-randomization allocation proportions. Therefore, the same level of power will be achieved based on various combinations of initial and re-randomization allocations. Herein, among all possible combinations, we seek the design that minimizes the total cost. Motivated by an application for evaluating interventions for antibiotic-resistant bacteria, we illustrate the proposed methods in the context of a SMART design with two stages. Methods for both continuous and binary outcomes are described. A Shiny web application implementation of the proposed methods will be presented.
|
