This paper will discuss the design and sample size requirement for a cluster randomized non-inferiority trial with two binary co-primary outcomes. A hypothetical study (the EXAMPLE Trial) will be considered.
Lets assume the EXAMPLE Trial will consist of two separate binomial non-inferiority two-sample trials. Trial 1: the Coronary Artery Disease known population (co-primary 1) and Trial 2: the Coronary Artery Disease unknown population (co-primary 2). A physician-month cluster randomization scheme will be used. That is, for each trial (trial 1 and trial 2) every month for a 12-month period, each physician participating in the EXAMPLE Trial will be allocated a randomized cluster of size 10. The physician will need to consent and enroll 10 patients each month for the 12-month period for each trial (trial 1 and trial 2). Each cluster will be specific to a treatment group (either EXPERIMENTAL or CONTROL).
The design and sample size method discussed by Bland (2003) and Donner and Klar (2000) will be used.
The EXAMPLE Trial will be declared a success if statistical significance is demonstrated at the pre-specified nominal alpha-level for both co-primary outcomes.
McCarthy, William F., "The Design and Sample Size Requirement for a Cluster Randomized Non-Inferiority Trial with Two Binary Co-Primary Outcomes." (July 2008). COBRA Preprint Series. Working Paper 36.