Whirling Disease: Reviews and Current Topics
A Brief Critique of Methods of Sampling and Reporting Pathogens in Populations of Fish
Christopher J. Williams and Christine M. Moffitt
doi: https://doi.org/10.47886/9781888569377.ch18
Studies of the disease status and distribution of pathogens in natural and captive populations of fish and of the implications of fish health in fisheries management have increased in recent years. With this new focus is a desire to use retrospective analyses from a variety of sources in order to examine trends on a larger geographic scale. To compare data and to model various outcomes with accuracy, estimates of the prevalence and of the confidence limits are needed. For years, pathologists have considered the implications of sample size on estimations of pathogen prevalence in fish populations, and sampling protocols used for screening have been developed and adapted from recommendations by Ossiander and Wedemeyer (1973) and Simon and Schill (1984). However, these methods addressed samples in which individual fish were examined for the agent of interest.
Confounding the interpretation of the outcome of tests is the use of pooled samples from several fish, processed in a single assay, and low numbers of pooled samples. The American Fisheries Society Fish Health Bluebook (Thoesen 1994) recommends a pooling of samples for analysis of several specific pathogens, including most viral screenings, and for many analyses of parasites, such as Myxobolus cerebralis (the causative agent of salmonid whirling disease), in which whole heads or half heads from like size fish are combined and digested with pepsintrypsin to free spores for quantification.
Likelihood-based methods can be used to estimate apparent prevalence and to construct confidence intervals that can be used for samples with any combination of group sizes. This method of estimation extends the work of Worlund and Taylor (1983), whose method was only applicable for equal group sample sizes. Our analysis includes calculation of sampling distribution to assess variance and mean-squared error, to show how these parameters change, as a function of the number and size of pooled groups and the disease prevalence.
We examined the statistics surrounding the outcome of sample analysis, based on samples ranging from low prevalence to 90% prevalence and several pooling strategies for samples of 30 and 60 fish. We calculated estimates of apparent prevalence for samples collected and screened for M. cerebralis infections by Utah Division of Wildlife scientists. Our calculations assume perfect specificity and sensitivity of assays, but we provide one method for adjusting these values when specificity and sensitivity are known. When all pools were positive, the apparent prevalence was 100%, but the bounds of the confidence interval ranged from 31% to 100%. A positive pool indicates that at least one fish in the pool was positive; however, when a pool is negative, all fish in the pool are presumed negative. Interpretations of data sets that are based only on the outcome of positive pools may be misleading, as the percentage of pools positive, when any single pool is scored negative, is higher than the maximum likelihood estimates of apparent prevalence. The confidence interval bounding estimates are generally lower when larger numbers of groups are used in analysis and when samples have fewer fish per pool. In populations with higher prevalence, the use of pooled samples significantly reduces the confidence of estimates.