Diagnostic accuracy of a duplex real-time reverse transcription quantitative PCR assay for detection of African horse sickness virus

► Diagnostic sensitivity and specificity of the AHSV RT-qPCR assay were 97.8% and 99.9%, respectively. ► Diagnostic sensitivity and specificity of AHSV isolation were 44.2% and 99.99%, respectively. ► Utilisation of Standards for Reporting of Diagnostic Accuracy (STARD). ► Use of a 2-test 2-population Bayesian latent class model to estimate diagnostic accuracy without a reference “gold standard”. Blood samples collected from 503 suspect cases of African horse sickness (AHS) and another 503 from uninfected, unvaccinated South African horses, as well as 98 samples from horses from an AHS free country, were tested with an AHS virus (AHSV) specific duplex real-time reverse transcription quantitative PCR (RT-qPCR) assay and virus isolation (VI). The diagnostic sensitivity and specificity of this AHSV RT-qPCR assay and VI were estimated using a 2-test 2-population Bayesian latent class model which made no assumptions about the true infection status of the tested animals and allowed for the possibility of conditional dependence (correlation) in test results. Median diagnostic sensitivity and specificity of the AHSV RT-qPCR were 97.8% and 99.9%, respectively. Median diagnostic specificity of virus isolation was >99% whereas the estimated diagnostic sensitivity was 44.2%. The AHSV RT-qPCR assay provides for rapid, high-throughput analysis of samples, and is both analytically and diagnostically sensitive and specific. This assay is potentially highly useful for demonstrating freedom or infection of horses with AHSV, thus it is appropriate that its reproducibility be evaluated in other laboratories as a global standard for detection of AHSV.