OBJECTIVES: Differences in computational sample size formulas indicate that randomized cluster samples require more patients to demonstrate the same effect as studies that use Individual Patient Recruitment (IPR) formulas.  We compared the differences in randomized cluster sampling and IPR formulas through a simulation study by varying the cluster size and Intra-Cluster Correlation Coefficient (ICC) to determine the magnitude of sample size differences. METHODS: The sample size formula for cluster sampling included two terms: 1) estimate of cluster size, and 2) estimate of ICC. Four Mean/Standard Deviation ratios were used reflecting the effect size, three ICC values, and three cluster sizes. Sample size was calculated for non-cluster and cluster formulas for 80% and 90% power. Sample size calculation results between cluster and IPR formulas were compared. RESULTS: Differences between cluster and IPR designs found that under sampling in IPR formulas vary from 5-15% and are largest when effect sizes are smallest. The IPR samples were smaller than cluster samples for the same effect size and power. Sample size using the cluster formula was smallest when ICC was small (0.15), at 80 percent power and cluster size of 5 patients per group. Cluster sample size was largest when ICC was large (0.25), at 90 percent power and cluster size of 20. CONCLUSIONS: In the research environment where prospective observational methods are used to gather “real world” data, studies that are conducted using cluster sampling, but powered with IPR formulas, are underpowered by as much as 15%.  Ethical implications must be considered in prospective studies that require patient informed consent if the study is underpowered. If the prospective study involves risk the equipoise argument may be violated and place patients at risk (assuming there is a study treatment regimen), as the study may not be conclusive because of low power.