Estimation of the relative sensitivity of the comparative tuberculin skin test in tuberculous cattle herds subjected to depopulation

Abstract

Bovine tuberculosis (bTB) is one of the most serious economic animal health problems affecting the cattle industry in Great Britain (GB), with incidence in cattle herds increasing since the mid-1980s. The single intradermal comparative cervical tuberculin (SICCT) test is the primary screening test in the bTB surveillance and control programme in GB and Ireland. The sensitivity (ability to detect infected cattle) of this test is central to the efficacy of the current testing regime, but most previous studies that have estimated test sensitivity (relative to the number of slaughtered cattle with visible lesions [VL] and/or positive culture results) lacked post-mortem data for SICCT test-negative cattle. The slaughter of entire herds ("whole herd slaughters" or "depopulations") that are infected by bTB are occasionally conducted in GB as a last-resort control measure to resolve intractable bTB herd breakdowns. These provide additional post-mortem data for SICCT test-negative cattle, allowing a rare opportunity to calculate the animal-level sensitivity of the test relative to the total number of SICCT test-positive and negative VL animals identified post-mortem (rSe). In this study, data were analysed from 16 whole herd slaughters (748 SICCT test-positive and 1031 SICCT test-negative cattle) conducted in GB between 1988 and 2010, using a bayesian hierarchical model. The overall rSe estimate of the SICCT test at the severe interpretation was 85% (95% credible interval [CI]: 78-91%), and at standard interpretation was 81% (95% CI: 70-89%). These estimates are more robust than those previously reported in GB due to inclusion of post-mortem data from SICCT test-negative cattle.

Figure 1. Overall mean relative Single Intradermal Comparative Cervical Tuberculin test sensitivity (rSe) estimates for each herd.

Estimates are shown for severe and standard test interpretation, derived from logistic regression models fitted with and without herd-level effects. The solid lines represent the mean estimates, and the dashed lines represent the 95% credible intervals. The red lines are from the model including herd-level effects, and the blue lines are from the model excluding herd-level effects.