Bayesian estimation of sensitivity and specificity of a PCR method to detect Coxiella burnetii in milk and vaginal secretions in sheep and goat samples

Abstract

Coxiella burnetii is a gram-negative and polymorphic rod bacterium that causes Q fever, a common zoonotic disease distributed worldwide. Widespread occurrences of the disease outbreaks indicate the importance of coordinated animal and human health efforts to control these outbreaks. Different tests are available to determine the C. burnetii infection status of a flock, but false negative responses may occur, as infectious animals can shed bacteria in milk intermittently, especially during an asymptomatic infection. In this study, a Bayesian latent class model was implemented to estimate the sensitivity (Se) and specificity (Sp) of a PCR method for the detection of C. burnetii in milk samples (PCR_M) and vaginal swabs (PCR_V) from Iranian sheep and goats. A nested PCR assay was conducted to detect infected animals among 170 milk samples and 170 vaginal swabs from goat flocks and 170 milk samples and 170 vaginal swabs from sheep flocks. We implemented a Bayesian latent class model to estimate the Se and Sp of a PCR method for the detection of C. burnetii in milk samples and vaginal swabs from sheep and goats. Estimations were based on the cross-classified results of PCR_M and PCR_V from the sheep and goat subpopulations. Positivity was 17.6 and 33.5%, respectively, for PCR_M and PCR_V samples among sheep. In goats, the apparent prevalence was 32.9 and 56.4% in PCR_M and PCR_V samples testing positive, respectively. This indicated the lower sensitivity of PCR_M. The Se of PCR_V was significantly higher than Se of PCR_M, which corresponded to a higher rate of vaginal-positive, milk-negative PCR samples. In contrast, Sp of PCR_V was lower than Sp of PCR_M, representing the higher false-positive rate of vaginal swabs. The PCR_V outperformed PCR_M in terms of identifying latently infected sheep and goats; however, neither method could identify all latently infected sheep and goats, thus the combination is recommended to maximize our ability to identify infected animals. The true prevalence of C. burnetii infection was higher in Iranian goats than sheep.

Keywords: Bayesian estimation; Coxiella burnetii; PCR; sensitivity; specificity.