Evaluation of an ELISA for the diagnosis of bovine tuberculosis using milk samples from dairy cows in China

Abstract

Bovine tuberculosis (bTB) is an important zoonotic disease in developing countries including China. Lack of rapid and simple detection methods for surveillance affects efficient control and eradication of this disease. The objective of this study was to determine the feasibility of using an individual and bulk milk sample ELISA based on the fusion protein of four specific antigens (CFP10/ESAT-6/MPB70/MPB83) of Mycobacterium bovis for the rapid detection and surveillance of bTB in dairy cattle herds and to identify individual animal factors that affect its results. The study population consisted of 388 dairy cows from four herds in Henan province, China. The serum sample ELISA (cut-off value 30%) and milk sample ELISA (cut-off value 10%) were shown to have good agreement, while, there was fair agreement between a blood sample IFN-γ in vitro release assay and the milk sample ELISA (cut-off value 10%) (Cohen's kappa statistic of 0.492 and 0.306, respectively). The sensitivity and specificity for the milk sample ELISA using a S/P% cut-off value of 10%, as estimated using a Bayesian latent class analysis model, was 48.4% (95%CI: 31.0 - 85.2) and 97.1% (95%CI: 92.2 - 99.7), respectively. Although both ELISAs had comparable sensitivities and specificities, the milk sample ELISA had a greater discriminatory ability, as measured by the area under the curve (ROC), than the serum sample ELISA with the difference of their AUCs being - 0.069 (95%PCI: -0.121 to -0.021). Further, it was estimated that the lower limit of detection at an animal level lacto-prevalence might be 1.1% with this milk ELISA using bulk milk samples. Additionally, a mixed linear regression model, incorporating herd as a random effect variable, was constructed to investigate the association between natural log-transformed bTB sample-to-positive ratio (LnS/P%) and milk data. Across the model, natural log-transformed Somatic Cell Count (SCC) and parity had a positive effect on the LnS/P% of bTB antibody response, and milk urea nitrogen had a curvilinear association with LnS/P%. In conclusion, a milk ELISA for bTB antibodies, especially bulk tank milk antibodies, could be used as an efficient tool for rapid screening in surveillance, which might be followed by confirmatory testing with IFN-γ assay in dairy cattle, during the implementation of bTB control programmes. Relevant milk factors should be considered to adjust the S/P% with the aim of classifying the infection status of animals and herds as accurately as possible.

Keywords: Antibody surveillance; Bayesian latent class analysis; Bovine tuberculosis; Bulk-tank milk; Linear regression models; ROC.