Estimating sensitivity and specificity of a PCR for boot socks to detect Campylobacter in broiler primary production using Bayesian latent class analysis

Abstract

The present study compares three different assays for sample collection and detection of Campylobacter spp. in broiler flocks, based on (i) the collection of faecal samples from intestinal organs (caecum), (ii) individual faecal droppings collected from the bedding and (iii) faecal material collected by socks placed on the outside of a pair of boots (boot socks) and used for walking around in the flock. The two first methods are examined for Campylobacter using a culture method (ISO-10272-2:2006), while the boot socks are tested using PCR. The PCR-assay is a genus specific multiplex PCR with primers targeting 16S rDNA in Campylobacter and primers targeting Yersinia ruckerii. Sixty-seven broiler flocks from Austria and 83 broiler flocks from Denmark were included in this prospective study and 89 of these were found to be positive in at least one method (AT: 49 samples, DK: 40 samples) whereas 61 of these were negative in all assays. In Austria samples for the three assays were collected simultaneously, which facilitates a direct comparison of the diagnostic test performance. In Denmark, however, boot socks and faecal droppings were collected three days before slaughter while caecum samples were collected at slaughter. The results were evaluated in the absence of a gold standard using a Bayesian latent class model. Austrian results showed higher sensitivity for PCR detection in sock samples (0.98; Bayesian credible interval (BCI) [0.93-1]) than for culture of faecal droppings (0.86; BCI [0.76-0.91]) or caecal samples (0.92; BCI [0.85-0.97]). The potential impact of Campylobacter introduction within the final three days before slaughter was observed in Denmark, where four flocks were tested negative three days before slaughter, but were detected positive at the slaughterhouse. Therefore the model results for the PCR sensitivity (0.88; BCI [0.83-0.97]) and cultural ISO-method in faecal samples (0.84; BCI [0.76-0.92]) are lower than for caecal samples (0.93; BCI [0.85-0.98]). In our study, PCR detection on boot sock samples is more sensitive than conventional culture. In view of the advantage of rapid results before slaughter and low costs for sampling, especially in combination with existing Salmonella surveillance systems (just another pair of boot socks needed), this method-matrix combination could be a valuable surveillance tool in the broiler primary production.

Keywords: Bayesian latent class analysis; Boot sock PCR; Campylobacter; Diagnostic accuracy; Surveillance tool; Validation.