Design and analysis of veterinary vaccine efficacy trials

Vaccination-challenge tests that involve all-or-none responses and do not require a direct comparison between vaccinates and controls can be completely characterized by the binomial distribution. Consumer and producer risks associated with binomial distribution based tests can be adjusted by altering the number of animals involved and the criterion for acceptance. Clinical signs or other outcomes measured on an ordinal or ranking scale should generally be analyzed by nonparametric statistical procedures. Parametric statistical tests are the most appropriate for data measured on an interval scale if the necessary assumptions are met concerning the population sampled. The use of in vitro potency tests in quality control procedures for inactivated vaccines depends on the demonstration of a significant dose-response efficacy relationship in the host animal.