Statistical aspects of fish vaccination trials.

: Vaccine trials are performed to measure the efficacy of vaccines or vaccination strategies. The statistical aspects pertain both to design and analysis of such trials. In experimental trials, the design should ensure that the groups to be compared are equal in all respects except for the factor to be assessed. This is achieved by proper randomisation procedures. The diagnosis of cases should as far as possible be carried out without knowledge of vaccination status (blinding). If the administration of a vaccine per se is considered to influence the outcome, a similar administration of placebo in the control group should be considered. The required sample size will depend on a specified significance level, a specified power, and on the magnitude of the effect one wants to detect. If the cumulative numbers of cases in vaccinated and unvaccinated groups are registered at the end of the observation period, the effect can be measured by the risk ratio, risk difference or the relative percent survival. If the estimation is based on a fish-time approach and the presumption of constant rate of outcome over time is fulfilled, the rate ratio or the rate difference can be estimated. If the rate is not constant, the life table would be a method of choice. In the statistical analysis of the vaccination effect it is important to consider the assumptions on which the statistical tests are based. Proper choice of the experimental and statistical units in the trials is crucial. If a trial is set up with the individual fish as the statistical unit, adjusting for dependence (cluster effect) between the units may be indicated. If possible, the vaccine effect should be given by the point estimate and the confidence intervals. The statistical aspects of fish vaccination trials are based on general principles for controlled clinical trials, but they are also influenced by characteristics connected to the fish population and the experimental conditions.