Masters candidate in Biostatistics, Katelyn Tessier, will present:
“Evaluation of Adaptive Experimental Design in Sequential, Multiple Assignment, Randomized Trials (SMARTs): A Simulation Study”
Plan B Adviser: David Vock
Abstract: A Sequential, Multiple Assignment, Randomized Trial (SMART) is a relatively new clinical trial design that aims at individualizing interventions to participants through adaptive intervention. Adaptive experimental design is another aspect of clinical trials that includes various modifications of either trial or statistical procedure such as treatment duration, evaluation of clinical responses, randomization, sample size, and interim analyses (Mahajan & Gupta, 2010). Adaptive experimental design, through dropping intervention(s) from future randomizations based on interim analyses, combined with SMART design is the main focus of this paper. Several simulations were performed in R version 3.3.1 to investigate the effect of the combination of these two methods. These simulations varied in possible sample size, number of interim analyses, and data generating mechanisms to observe the impact on power, average number randomized, and average percentage randomized to different treatments and the optimal embedded intervention. Based on these simulations with interim analyses, when the effect sizes were large, the average number randomized in the trial was substantially less than the possible sample size, and the power was smaller than the expected power with no interim analyses. As the number of interim analyses increased, the power was less, especially in the case with moderate effect sizes. Additionally, when interim analyses were included, the average percentage randomized to different treatments was greater than 50%, and the average percentage randomized to the optimal embedded intervention was greater than expected. The implementation of SMART design combined with interim analyses to reduce future randomization to inefficacious treatments has not been used in clinical trials. However, based on the results of these simulations, it provides a promising new approach for clinical trials and research.
Refreshments will be served prior to the presentation.