B0407
Title: New approaches to design and monitoring of SMARTs
Authors: Erica Moodie - McGill University (Canada) [presenting]
David Stephens - McGill University (Canada)
Armando Turchetta - McGill University (Canada)
Abstract: Sequential multiple assignment randomized trials (SMARTs) provide high-quality data to learn about multi-stage tailored treatment strategies. While the number of SMARTs is steadily increasing, there have been few improvements in planning and monitoring. For example, the standard, frequentist formulae require assumptions on interim response rates and variance components which, if mis-specified, yield inadequate power. A Bayesian "two priors" approach is discussed that relies on fewer assumptions, integrates pre-trial knowledge, and focuses on the minimal detectable difference rather than the standardized effect size. A new approach to monitoring enrolment is also pointed out.
