The ruminant trophoblast produces pregnancy-associated glycoproteins (BPAGs) starting at day 25 after conception and continuing until parturition. These glycoproteins are detectable in the blood and milk of pregnant females and can be used to diagnose and monitor pregnancy in cattle, small stock and buffaloes. The objective of the study was to estimate the accuracy of three enzyme-linked immune-sorbent assays (ELISA) for BPAGs to diagnose pregnancy in dairy cattle under South African field conditions compared to transrectal ultrasound (TRUS) as a gold standard. A multi-centre prospective study was conducted at five sites in South Africa utilizing the assistance of experienced veterinary practitioners. A total of 1036 plasma and milk samples from dairy cows were analysed. Transrectal ultrasonography was performed immediately after sample collection. The milk and blood samples were analysed with commercial ELISA kits. A total of 532 and 395 cows were found to be pregnant and open, respectively based on ultrasound. The pregnancy rate was 51.4% (532/1036) based on TRUS. Results for sensitivity of ELISA analyses of serum, milk and serum visual ELISA was 99.8%, 99.4%, and 99.6%, respectively. Specificity for the serum, milk and serum visual ELISA was 57.3%, 55.3% and 55.0%, respectively. The results for factors affecting sensitivity and specificity of all the BPAG ELISA analyses (ELISAs) showed that days after AI, breed, DIM and lactation number did not have an effect on sensitivity of any of the BPAG ELISAs. Days after AI and breed influenced the specificity of all the BPAG ELISAs. There was a negative correlation between days after AI and specificity of all the BPAG ELISAs. Specificity decreased gradually as days after AI increased from day 28 to 35. Additionally, pregnancy loss influenced the specificity of the three BPAG ELISAs from 28 to 35 days after AI.
In conclusion, the BPAG ELISAs were sensitive tests for detecting pregnancy relative to TRUS B from days 28-35 and 42-49 d after breeding but poorly specific in identifying non-pregnant cows. The specificity estimates were under-estimated due to the incidence of pregnancy loss during early gestation.