Conference in 17th International Symposium of Veterinary Epidemiology and Economics (ISVEE 17)

November 14^th, 2024

Gómez-Buendía A., Denwood M., Saxmose-Nielsen S., Grau A., Nacar J., Minguez O., Romero B. and Alvarez J.

The sub-optimal performance of diagnostic tests can undermine the success of bovine tuberculosis (bTB) eradication programs. Reported sensitivity and specificity estimates range from very low (40%) to very high (99%) due to the wide disparity in study designs, animal populations, and reference tests used in their evaluation. Given the lack of a perfect reference test, Bayesian Latent Class Models (BLCMs), which are not reliant on a gold standard, are best suited for assessment of tests. We applied BLCMs to determine the performance of two versions of the interferon-gamma (IFN-γ) assay for the diagnosis of bTB along with the effect of certain covariates (age, region, year, herd, production type, and results of other tests). IFN-γ tests were performed in over 250,000 animals from 1089 infected herds, representing all bTB-positive herds in Castilla y León, Spain, from 2017 to 2020. All animals were tested with the single intradermal tuberculin skin test and with one of two IFN-γ kits available in Spain (IDvet or Bovigam). Numerous models, including Hui-Walter extensions adjusting for covariates, were run. IFN-γ sensitivity estimates varied depending on the BLCM used but were consistently higher (>60%) than values for the Se of the skin test (<50%). The opposite was true for the specificity, with the skin test being the most specific (~100%), while IFN-γ estimates ranged from 96.7% (IDvet) to 98.5% (Bovigam), depending strongly on the BLCM applied. Estimates varied substantially depending on populations considered (defined by region or year of testing) and covariates included (e.g., higher Se of the skin test in younger animals while lower in the IFN-γ), which complicates the estimation of true test performance across populations. This work illustrates the challenges associated with managing extensive databases from national eradication surveillance programs

	Servicio de Micobacterias (MYC). Centro de Vigilancia Sanitaria Veterinaria (VISAVET). Universidad Complutense (UCM).
	Department of Veterinary and Animal Sciences. University of Copenhagen (UCPH).
	Consejería de Agricultura y Ganadería. Junta de Castilla y León.
	Consejería de Sanidad. Junta de Castilla y León.