Web-based Sample Size Calculator for Cluster-Randomized and Stepped-Wedge Designs

Cluster-randomized and stepped-wedge are pragmatic trial designs that have become increasingly popular in recent years. Due to feasibility or logistical constraints, individual-level randomization is often not possible, and interventions must be implemented at the cluster (e.g., site, clinic) level. Power/sample size calculations are used to identify whether a proposed design s feasible for detecting a clinically meaningful effect of an intervention. Tools that perform these calculations are thus essential in the planning of an effective study and for assessing various design options.

Power calculations for cluster-randomized and stepped-wedge designs incorporate the correlation between multiple observations in the same cluster. They also require additional consideration such as the number of clusters, and individuals per cluster. There are trade-offs when evaluating each of these two designs, and often one is considered when the other is not feasible and/or does not provide sufficient power. With a free, web-based applet, we unify the power/sample size calculations for these two clustered study designs in a single application, allowing for easy comparison and evaluation of alternative designs.

Using an R Shiny application, we implement methodology developed for cluster-randomized and stepped-wedge designs for sample size/power calculations. We incorporate recent extensions, such as cluster auto-correlation, washout effects, and hybrid designs. The application will use a guided step-by-step process, where users will specify the parameters of their trial design. Users will provide inputs related to the outcome of interest and study design, such as number of clusters, individuals per cluster, desired power, type I error rate, outcome distribution, effect size, and the intraclass correlation coefficient. Outputs will include a visualization of the study design, and a summary statement describing the design, assumptions, and power/sample size values.

The R Shiny calculator will be hosted online as a web applet that can be used by clinicians and statisticians to help plan their trial design. A range of examples will be presented to demonstrate the use of the calculator. Documentation for the methods and references be provided. Code for the application and power calculations will be shared using Github, where users can provide feedback and request modifications or extensions.

With this online calculator, we aim to increase the accessibility of current and emerging sample size methodology for researchers who are considering pragmatic design alternatives to answer their research question.