Modeling the Accuracy of Two in-vitro Bovine Tuberculosis Tests Using a Bayesian Approach

Catalina Picasso-Risso, Andres Perez, Andres Gil, Alvaro Nunez, Ximena Salaberry, Alejandra Suanes, Julio Alvarez

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Accuracy of new or alternative diagnostic tests is typically estimated in relation to a well-standardized reference test referred to as a gold standard. However, for bovine tuberculosis (bTB), a chronic disease of cattle, affecting animal and public health, no reliable gold standard is available. In this context, latent-class models implemented using a Bayesian approach can help to assess the accuracy of diagnostic tests incorporating previous knowledge on test performance and disease prevalence. In Uruguay, bTB-prevalence has increased in the past decades partially because of the limited accuracy of the diagnostic strategy in place, based on intradermal testing (caudal fold test, CFT, for screening and comparative cervical test, CCT, for confirmation) and slaughter of reactors. Here, we evaluated the performance of two alternative bTB-diagnostic tools, the interferon-gamma assay, IGRA, and the enzyme-linked immunosorbent assay (ELISA), which had never been used in Uruguay in the absence of a gold standard. In order to do so animals from two heavily infected dairy herds and tested with CFT-CCT were also analyzed with the IGRA using two antigens (study 1) and the ELISA (study 2). The accuracy of the IGRA and ELISA was assessed fitting two latent-class models: a two test-one population model (LCA-a) based on the analysis of CFT/CFT-CCT test results and one in-vitro test (IGRA/ELISA), and a one test-one population model (LCA-b) using the IGRA or ELISA information in which the prevalence was modeled using information from the skin tests. Posterior estimates for model LCA-a suggested that IGRA was as sensitive (75–78%) as the CFT and more sensitive than the serial use of CFT-CCT. Its specificity (90–96%) was superior to the one for the CFT and equivalent to the use of CFT-CCT. Estimates from LCA-b models consistently yielded lower posterior Se estimates for the IGRA but similar results for its Sp. Estimates for the Se (52% 95%PPI:44.41-71.28) and the Sp (92% 95%PPI:78.63–98.76) of the ELISA were however similar regardless of the model used. These results suggest that the incorporation of IGRA for detection of bTB in highly infected herds could be a useful tool to improve the sensitivity of the bTB-control in Uruguay.

Original languageEnglish (US)
Article number261
JournalFrontiers in Veterinary Science
StatePublished - Aug 13 2019

Bibliographical note

Funding Information:
The authors thank the experts who provided insight for the prior distributions, and the veterinarians from the participant company for their willingness and time. Funding. This research was partially supported by Uruguayan Agency of Research and Innovation (ANII in Spanish language), Plan Nacional de Investigación en Salud Animal & Comité de Investigación Científica de la Universidad de la República (PLANISA–CSIC). JA was the recipient of a Ramón y Cajal postdoctoral contract from the Spanish Ministry of Economy, Industry and Competitiveness (MINECO) (RYC-2016-20422).

Publisher Copyright:
© Copyright © 2019 Picasso-Risso, Perez, Gil, Nunez, Salaberry, Suanes and Alvarez.


  • Uruguay
  • chronically infected
  • diagnosis
  • elisa
  • interferon-gamma release assay
  • latent class analysis

PubMed: MeSH publication types

  • Journal Article


Dive into the research topics of 'Modeling the Accuracy of Two in-vitro Bovine Tuberculosis Tests Using a Bayesian Approach'. Together they form a unique fingerprint.

Cite this