It is a non-inferiority study design with cluster randomization. For example, the client wants to assign 1/3rd of the sample size to placebo and 2/3rd of the sample size to treatment and analyze the outputs for different Intraclass correlation coefficients (0.1 & 0.5) for different event numbers, sample sizes, and hazard ratios. The drug they are working on is to increase the Hb levels, and the response time is 4 weeks as the primary purpose for cluster randomization would be for operational ease for the investigators to handle the patients efficiently.
The clusters will be a whole clinic. The patients in the clinic will all be randomized into one treatment considering each site to be a cluster. The 2-1 randomization is on the site level, not the patient level. So, by the end, 2/3 sites are on one treatment, 1/3 sites are on another. All patients on one site will receive the same treatment.
Depending on the number of patients at the sites, each site will have different event numbers. For consistency, the user wants us to assume the same event rate across all the sites for now and to include the site id in the analyses.
It is a non-inferiority study design with cluster randomization. For example, the client wants to assign 1/3rd of the sample size to placebo and 2/3rd of the sample size to treatment and analyze the outputs for different Intraclass correlation coefficients (0.1 & 0.5) for different event numbers, sample sizes, and hazard ratios. The drug they are working on is to increase the Hb levels, and the response time is 4 weeks as the primary purpose for cluster randomization would be for operational ease for the investigators to handle the patients efficiently.
The clusters will be a whole clinic. The patients in the clinic will all be randomized into one treatment considering each site to be a cluster. The 2-1 randomization is on the site level, not the patient level. So, by the end, 2/3 sites are on one treatment, 1/3 sites are on another. All patients on one site will receive the same treatment.
Depending on the number of patients at the sites, each site will have different event numbers. For consistency, the user wants us to assume the same event rate across all the sites for now and to include the site id in the analyses.