Abstract
Floor pen trials are an efficient way to evaluate the effectiveness of potential Salmonella control interventions in broiler chickens. When treatments are allocated at the pen level, and outcomes are measured at the individual bird level, floor pen studies are considered to be cluster randomized trials. Estimating the sample size required to achieve a desired level of statistical power for a cluster randomized trial requires an estimate of the intra-cluster correlation (ICC) as an input. In this study, ICCs were estimated for the untreated challenged control group from 40 broiler chicken Salmonella pen trials performed using a seeder bird challenge model. The ICCs for ceca Salmonella prevalences ranged from 0.00 to 0.64, with a median of 0.17. The ICCs for ceca Salmonella log(10)(MPN/g + 1) ranged from 0.00 to 0.52, with a median of 0.14. These findings indicate that the effect of pen-level clustering is substantial in Salmonella floor pen trials, and it must be considered during both the study design and analysis. In a multivariable regression analysis, ICCs for ceca Salmonella prevalences were associated with the challenge status of sampled birds, age of birds at the time of challenge, and Salmonella serovar. ICCs were lower for studies in which a combination of direct (seeder) and indirect (horizontal) challenged birds were sampled, and for studies in which birds were challenged on the day of hatch or at one day of age. ICCs were higher for studies in which Salmonella Heidelberg was used as the challenge strain. These findings may be useful for investigators that are planning pen trials to evaluate Salmonella control interventions in broiler chickens. Choosing study design elements associated with a lower ICC may improve efficiency by leading to a larger effective sample size for the same number of experimental units.