Blinded and unblinded sample size reestimation procedures for stepped-wedge cluster randomized trials

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Epidemiology

Background:

• Accurate sample size estimation is vital for stepped-wedge (SW) cluster randomized trials (CRTs).
• Misspecification of nuisance parameters can lead to overpowered (costly) or underpowered (false negatives) trials.
• This impacts the reliability and efficiency of SW-CRT designs.

Purpose of the Study:

• To propose and evaluate blinded and unblinded sample size reestimation (SSRE) methods for cross-sectional SW-CRTs.
• To address the challenge of nuisance parameter misspecification in SW-CRT sample size calculations.
• To enhance the statistical power and cost-effectiveness of SW-CRTs.

Main Methods:

• Development of blinded estimators for variance parameters in SW-CRTs using the Hussey and Hughes linear-mixed model.
• Detailed procedures for both blinded and unblinded SSRE applicable at any time point within a SW-CRT.
• Performance evaluation of SSRE procedures through simulations and analysis of two trial design scenarios.

Main Results:

• Underspecification of key variance parameters by 50% was addressed by SSRE procedures.
• SSRE increased statistical power by up to 41% compared to conventional SW-CRT designs.
• Empirical power achieved exceeded desired levels, demonstrating the effectiveness of SSRE.

Conclusions:

• Proposed SSRE methods are effective in correcting for nuisance parameter misspecification in SW-CRTs.
• Incorporating SSRE into SW-CRT designs can improve statistical power and trial efficiency.
• Researchers should consider implementing SSRE in future stepped-wedge cluster randomized trials, despite practical considerations.