Presenter: Aoqi Xie

Supervisory Committee: Clement Ma (Supervisor), Wendy Lou, Aya Mitani

Chair: Tony Panzarella

Date and Time: Tuesday, August 20, 2024, 09:00-11:00pm EST

Location: 155 College Street, Health Sciences Building, Room 734

Abstract: Given substantial public resources allocated to health research, it is crucial to ensure the resources are used effectively so that it maximizes the benefit. A pilot study is a small-scale study that aims to test for feasibility to ensure a high likelihood of success for a large-scale trial. Sample size re-estimation allows modification of the final sample size based on interim results, ensuring sufficient power are maintained. Sequential Multiple Assignment Randomized Trials (SMART) provide a framework for testing dynamic treatment regimens, which are pre-specified rules for rerandomizing participants based on their responses. However, there is a gap for combining these two methods within Sequential Multiple Assignment Randomized Trials. In this study, we aim to extend the fixed SMART design to include an Internal Pilot Study for feasibility and unblinded Sample Size Re-estimation (SSR). The SMART trial proceeds in two phases. In Phase I, we conduct an internal pilot study to assess trial feasibility using a one-stage exact test. Following this, we compare three methods for SSR based on interim efficacy data from the internal pilot: (1) Traditional; (2) Adjusted Effect Size (AES); and (3) the Cui, Hung, Wang (CHW) method. In Phas II, the large-scale main trial proceeds with re-estimated sample size. To evaluate our proposed design, we assess the statistical properties of our proposed adaptive design with internal pilot and SSR methods and the fixed design (without the feasibility test or SSR) in terms of expected re-estimated sample size, type I error rate, and statistical power using a simulation study. The internal pilot study saves the interim sample size (half of the total planned sample size) compared to the external pilot study design, and prevents infeasible study from continuing till the end. Results show that all SSR methods have the same or higher power than the fixed design, across all scenarios. All methods recover the power compared to the fixed design when the true effect size is smaller than planned. The CHW method generally performs better with sufficient power and requires a smaller sample size. By integrating an internal pilot study, we enhance feasibility and study design reliability before full-scale deployment. The use of SSR addresses key challenges in maintaining power and ethical standards, allowing trials to adapt to emerging data, optimize resources, and accelerate treatment development. The improved design will be instrumental in supporting the ethical and cost-effective execution of SMART.