Appendix 15
Application of optimal sampling strategies for substantiating disease freedom Matthias Greiner International EpiLab, Danish Institute for Food and Veterinary Research, Mørkhøj Bygade 19, DK-2860 Søborg, Denmark David Paton Institute for Animal Health, Ash Road, Pirbright, Surrey, GU24 ONF, UK Abstract We describe a method for optimising the application of a diagnostic test for surveillance purposes. The objective is to establish a procedure for maximising the herd level sensitivity by choosing a sample size (within herd) and a cutpoint (number of test positive animals required to classify a herd as positive) for each herd size under the constraint that the herd specificity is greater than a specified minimum value. The second objective is to use such herd size-specific sensitivities to assess the performance of a surveillance system. This includes sample size (among herds) calculations and the estimation of the confidence, i.e. the probability to detect at least one infected animal in a population, given a specified level of prevalence. The methods are illustrated using the example of post-vaccination surveillance for foot-and-mouth disease. 1. Introduction The methodology for substantiating freedom from disease can be derived from the concept of diagnostic sensitivity. For example, Martin et al. (2005) define the term “Surveillance component sensitivity” as the probability that a given surveillance component detects at least one truly infected herd, given that infected herds in the population and infected animals in infected herds occur at levels specified by the among-herd and within-herd “design prevalence”, respectively. The surveillance sensitivity is sometimes referred to as “confidence” or “power” and depends on the sensitivity and specificity of the diagnostic procedure used, the number of reactors required to classify the herd as infected (herd cutpoint) and the sampling scheme for herds and animals within herds. The statistical methods used in the context of disease freedom are described in detail by Cameron and Baldock (1998a), Cameron and Baldock (1998b), Cannon (2001) and Cannon and Roe (1982). Paton et al. (2004) have recently described the rationale for surveillance for foot-and-mouth disease (FMD) to regain the status of a country or region as free without vaccination after an outbreak, which has been controlled by measures including vaccination. It is an essential requirement for surveillance programmes in this context that the presence of persistently infected (carrier) cattle in the population can be detected with high (say 95%) level of confidence. The design prevalence in this case refers to the presence of such carrier animals in infected, vaccinated or unvaccinated herds. While the level of the confidence is set by 112
