Abstract:
The diagnostic performance of four serological assays for bovine brucellosis in African buffaloes,
namely Rose-Bengal test (RBT), complement fixation test (CFT), indirect enzyme-linked
immunosorbant assay (iELISA) and fluorescence polarisation assay (FPA) were evaluated and
compared in a case-control study. The study followed the OIE assay validation pathway for validation
of diagnostic tests applicable to wildlife species where there is a validated test available in a
taxonomically closely related species. Two uninfected and four infected herds were recruited and an
uninfected composite reference panel of 107 sera and infected composite reference panel of 93 were
selected using composite reference standards. Diagnostic sensitivity (DSe) and diagnostic specificity
(DSp) were calculated for individual tests and for different combinations of two tests in series and in
parallel. Cut-off points were adjusted using receiver operating characteristics (ROC) analysis. Using
these cut-off values, the index tests performed as follows: RBT DSe of 98.9% (95% CI 96.83% -
100%) and DSp of 98.1% (95% CI 95.6% - 100%), iELISA (cut-off >40.5%) DSe 98.9% (95% CI
94.2% - 100%) and DSp 100% (95% CI 96.6% - 100%), CFT (cut-off >0 iU/ml) DSe 74.2% (95% CI
64.1% - 82.7%) and DSp 100% (95% CI 96.6% - 100%) and FPA (cut-off >16 mP) DSe 97.9% (95%
CI 94.2% - 99.7%) and DSp 100% (95% CI 96.6% - 100%). Based on performance index and area
under the ROC curve, the iELISA performed best (198.9% and 1.0), followed closely by the FPA
(197.9% and 0.989) and the RBT (197.0%). The CFT s lower performance (174.2%, and 0.871) was
due to low DSe. Kappa values for test agreement between the index tests was above 0 for all
combinations, and varied from unweighted Kappa of 0.685 (95% CI 0.608 0.762) between FPA and
iELISA to 0.26 (0.136-0.383b) between CFT and RBT. Consideration of the indices for positive and
negative test agreement between the index tests supported the differential specificity of tests for
different immunoglobulin classes and higher in line with the findings in cattle. Positive predictive
value in herd C and E were 100% for the iELISA, CFT and FPA, 97.3% in herd C and 98.4% in herd E
for the RBT. Negative predictive values in herd C ranged from 89% for the CFT to 99.2% for the RBT
and in herd E 73.1% for the CFT to 98.7% for the RBT. Overall repeatability was satisfactory, except
for the FPA, which was considered the result of sample quality related to prolonged storage in a
freezer. The index tests were all found fit for use to detect or confirm brucellosis in populations and
individual animals. The values for DSe and DSp that were estimated will be of use in the
interpretation of serological results and determination of diagnostic strategies in different
circumstances. Different combinations of tests in series and parallel increased the DSp and DSe.
Using the RBT in combination with the CFT/FPA/iELISA interpreted in series or in parallel in relation
to the epidemiological setting and objective of testing is recommended.
