Harmonisation of the diagnostic performances of serological ELISA tests for C. burnetii in ruminants in the absence of a gold standard: Optimal cut-offs and performances reassessment.

None of the three ELISA tests currently available in Europe for the serological diagnosis of Coxiella burnetii (C. burnetii) infections in ruminants can be considered as a gold standard. The difference in their diagnostic performances (e.g., from 39 % to 87 % for sensitivity in sheep - Lurier et al. 2021) affects the agreement between the test results obtained in different veterinary laboratories and limits the comparability of prevalence estimates for surveillance. In order to harmonise these tests, we tried to adjust their cut-offs so that there is a maximum of agreement between them. Using the results of the three ELISAs applied to 1258 cattle, 1474 goat and 1432 sheep serum samples collected in France, we identified the cut-offs that maximised Hubert's kappa, an agreement coefficient applicable to more than two tests, using a differential evolution algorithm for optimisation. We then evaluated the effect of changing the cut-offs. In particular, we estimated the sensitivities and specificities of each test at the optimal cut-offs using a latent class model. While the manufacturer's cut-offs of the three tests were 30, 40 and 40 respectively, regardless of the species, the cut-offs that maximise the agreement are 44.7, 30.8 and 84.3 for cattle, 6.6, 18.8 and 49.8 for goats, and 26.2, 50.9 and 88.6 for sheep. By using these cut-offs instead of those provided by the manufacturers, the observed proportion of disagreement between the tests is reduced by approximately half, and the diagnostic performances and apparent prevalence are more consistent from one test to another. The use of these species-specific cut-offs allows for better harmonisation of the tests. However, before implementing these new cut-offs, it is essential to assess the measurement uncertainty around them to ensure that the analytical performances of the tests are maintained.