The misdiagnosis of intramammary infections (IMI) with Streptococcus agalactiae (S. agalactiae) could lead farmers to treat or cull animals unnecessarily. The objective of this field study was to estimate the sensitivity (Se) and specificity (Sp) of real-time PCR at different cut-offs for cycle threshold (Ct) values against bacterial culture (BC) for diagnosis of S. agalactiae IMI using latent class analysis to avoid the assumption of a perfect reference test. A total of 614 dairy cows were randomly selected from 6 herds with bulk tank PCR Ct value <= 39 for S. agalactiae and S. aureus. At milk recording, 2456 quarter milk samples were taken aseptically for BC and the routinely taken cow level milk samples were analyzed by PCR. Results showed that 53 cows (8.6%) were positive for S. agalactiae IMI by BC. Sensitivity of PCR at cut-offs; <= 39, <= 37, <= 34, and <= 32, was 96.2%, 91.9%, 87.2% and 73.9%, while Se of BC was 25.7%, 29.9%, 59.9% and 72.1%. Specificity of PCR at cut-offs; <= 39, <= 37, <= 34, and <= 32, was 96.8%, 96.9%, 96.7%, and 97.22%, while Sp of BC was 99.7%, 99.5%, 99.2%, and 98.9%. The estimated prevalence of S. agalactiae IMI by PCR was higher than the apparent prevalence at the tested cut-offs, indicating under estimation of S. agalactiae IMI in the examined dairy cows. In conclusion, Se of PCR is always higher than Se of BC at all tested cut-offs. The lower cut-off, the more comparable becomes Se of PCR and Se of BC. The changes in Se in both PCR and BC at different Ct-value cut-offs may indicate a change in the definition of the latent infection. The similar Se of both tests at cut-off <= 32 may indicate high concentrations of S. agalactiae viable cells, representing a cow truly/heavily infected with S. agalactiae and thus easier to detect with BC. At cut-off <= 39 the latent definition of infection may reflect a more general condition of cows being positive for S. agalactiae. Our findings indicate that PCR Ct-value cut-offs should be chosen according to the underlying latent infection definition of interest. Latent class analysis proposes a useful alternative to classic test evaluation of diagnostic tests used for detection of S. agalactiae IMI in milk. (C) 2012 Elsevier B.V. All rights reserved.