A sounder of swine: The importance of clustering in the design, analysis, and interpretation of clinical trials

Abstract

Clustering of animals at the level of pens, rooms, barns, or farms leads to statistical nonindependence of individual pigs. Failure to consider clustering when determining sample size will result in clinical trials that are too small to detect meaningful differences between intervention groups when clustering is controlled in the analysis. Failure to control clustering in the analysis will lead to inappropriately narrow confidence intervals and increases the probability of a false-positive finding. Thus, failure to consider clustering in trial design and analysis results in research that could misinform decision making on the use of interventions.