Bayesian latent class analysis for evaluating diagnostic performance of three serological tests of bovine brucellosis in Pakistan.

Bovine brucellosis is a major zoonotic disease in Pakistan, yet diagnostic tests used in national surveillance have never been validated under field conditions. This cross-sectional study estimated the sensitivity (Se) and specificity (Sp) of three commonly used bovine brucellosis serological tests, viz. Rose Bengal plate test (RBPT), Serum agglutination test (SAT), and indirect ELISA (iELISA), and the true prevalence of the disease in the study area using Bayesian latent class analysis (BLCA). Serum samples from 1000 animals (507 cattle, 493 buffaloes) across four districts (Faisalabad, Jhang, Chiniot and T.T. Singh) were tested in parallel. Since all three assays target antibodies against smooth lipopolysaccharide (S-LPS), full conditional dependence (CD) was modeled, with an acceptable fit attained only under informative priors (posterior predictive p-value = 0.250; DIC = 211.73). RBPT showed the lowest Se (0.705), while SAT (0.938) and iELISA (0.952) performed substantially better. All three tests achieved high Sp (RBPT = 0.977; SAT = 0.993; iELISA = 0.997). True prevalence ranged from 7.5% (Faisalabad) to 25.3% (Jhang). SAT + iELISA in parallel achieved the highest performance index (PI = 1.961; Se = 0.971, Sp = 0.990). Species-level analysis revealed higher SAT sensitivity in buffaloes (0.969) than cattle (0.849), with a notable prevalence divergence in Jhang (cattle = 39.6%; buffaloes = 10.7%). It was concluded that a single diagnostic test alone is unreliable for screening in similar settings. SAT + iELISA in parallel is recommended for field screening and in series for culling decisions. Significant district-level variance in prevalence and species-specific diagnostic performance indicates that uniform provincial strategies may be inappropriate and that control programs should be designed at the district level and calibrated separately for cattle and buffaloes.