AUTHOR=Um Maryse Michèle , Castonguay Marie-Hélène , Mahamad Amine Khalie , Giguère Julie , Morin Isabelle , Dufour Simon 

TITLE=Repeatability of a Commercially Available ELISA Test for Determining the Herd-Level Salmonella enterica subsp. enterica Serovar Dublin Status in Dairy Herds Using Bulk Milk

JOURNAL=Frontiers in Veterinary Science

VOLUME=Volume 7 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2020.00401

DOI=10.3389/fvets.2020.00401

ISSN=2297-1769

ABSTRACT=The objective of the current research was to estimate the repeatability of an ELISA test when used on fresh and frozen samples. Each of 845 bulk milk collected samples was subdivided into 3 aliquots and analyzed using PrioCHECKTM Salmonella Ab Bovine Dublin. ELISA percent positivity results (PP%) were compared between aliquots submitted to the initial analysis and a second analysis conducted 24 h later. The third aliquots were either preserved for 13 to 14 d (n=413) or 25 to 28 d (n=432) at -20 °C prior to analysis and results were compared to the initial analysis. Finally, the repeatability of the test for determining sample status using cut-off 15% or 35% was evaluated. There was excellent concordance between the two initial values and with values obtained after 13-14 d- and 25-58 d-freezing. The corresponding concordance correlation coefficients were 0.96, 0.97, and 0.94, respectively. Bland-Altman plots showed differences of PP% of 0.1 percentage points on average between the initial and second fresh samples. Freezing for 13-14 d and 25- 28 d led to overestimation of the initial values by 0.1, and 0.4 percentage points (statistical significance), respectively. Regarding the classification of samples, greater disagreement was observed between 25-28 d-frozen and initial samples values when using the cut-off 15% (kappa=0.76) compared to 35% (kappa=0.90). 
Our study showed that PrioCHECKTM has good repeatability and that frozen bulk milk samples could generate reliable results. However, the larger variability at lower PP% should be considered when setting up a threshold.