Abstract:
|
Sequential multiple assignment randomized trials (SMARTs) is a valid and powerful design to study Dynamic Treatment Regimes (DTRs). However, some practical challenges may hinder the implementation of SMART including: 1) high dropout rates requires a large sample size in the initial stage to ensure sucient power for comparing DTRs; 2) it is time-consuming and challenging to manage and monitor a sequential multi-stage trial with a large number of participants. To handle these issues, we propose a SMART with EnRichment (SMARTer) design, which can not only improve design eciency but also make the trial more practical with limited time and resource. Specifically, at each stage of the original SMART, we enrich the study sample with new patients who have received previous stage treatments in a naturalistic fashion without randomization, and only randomize them among the current stage treatment options. One special case of the SMARTer is to synthesize separate independent single-stage randomized trials at each stage. We show SMARTer data allows for unbiased estimation of DTRs as SMART does. Furthermore, we show analytically the eciency gain of the new design over the SMART is signicant
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.