Hybrid sample size calculations for cluster randomised trials using assurance.

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Health Research Methods

Background:

• Sample size determination for cluster randomized trials (CRTs) is challenging due to the need for accurate intra-cluster correlation coefficient (ICC) estimation.
• Traditional power calculations are sensitive to ICC inaccuracies, potentially leading to under- or over-powered trials.
• Imprecise ICC estimates often arise from studies with few clusters, complicating sample size planning.

Purpose of the Study:

• To propose a hybrid Bayesian assurance and frequentist approach for sample size determination in CRTs.
• To incorporate uncertainty in key parameters like ICC, standard deviation, and coefficient of variation of cluster size.
• To demonstrate the approach using a CRT design for post-stroke incontinence.

Main Methods:

• Utilized Bayesian assurance as an alternative to traditional power, incorporating prior distributions for key parameters.
• Specified prior distributions for standard deviation, ICC, and coefficient of variation of cluster size.
• Applied the method to a CRT for post-stroke incontinence, comparing results with standard power calculations.

Main Results:

• Bayesian assurance allows sample size calculation using elicited prior distributions for ICC, unlike power calculations that use single point estimates.
• The proposed approach avoids sample size misspecification when prior distributions differ significantly, even with similar medians.
• Accounting for uncertainty in all nuisance parameters did not substantially increase the required sample size.

Conclusions:

• Bayesian assurance enhances understanding of trial success probability and provides more robust sample sizes against parameter uncertainty.
• This method is particularly beneficial when reliable parameter estimates are difficult to obtain.
• The hybrid approach offers a more reliable alternative for sample size determination in CRTs.