Bayesian estimation of the seroprevalence of brucellosis in humans and livestock in Kyrgyzstan.

Kyrgyzstan reported 77.5 new cases of human brucellosis per 100,000 inhabitants in 2007, which is one of the highest incidences in the world. However, because this number is based on official records, it is very likely that the incidence is underreported. The diagnostic tests most commonly used in Kyrgyzstan are the Rose Bengal test in ruminants and the Huddleson test in humans. The sensitivity and specificity of these tests have never been evaluated under field conditions in Kyrgyzstan, where the strains circulating in livestock and humans are unknown. Therefore, a representative national cross-sectional serological study was undertaken in humans, cattle, sheep and goats to assess the true seroprevalence and to compare different serologicaltests. In the year of study (2006), few animals were vaccinated against brucellosis in Kyrgyzstan. A total of 5,229 livestock sera and 1,777 human sera from three administrative regions were collected during spring 2006 and submitted to a range of serological tests. The true seroprevalence of brucellosis, estimated using Bayesian methodology, was 7% (95% credibility interval 4%-9%) in humans, 3% (1%-5%) in cattle, 12% (7%-23%) in sheep and 15% (7%-30%) in goats. The Rose Bengal test was confirmed as a useful screening test in livestock and humans, although its sensitivity was lower than that of other tests. The estimates of specificity of all tests were significantly higher than those for sensitivity. The high seroprevalence of brucellosis in humans, cattle and small ruminants in Kyrgyzstan was confirmed. Bayesian statistical approaches were demonstrated to be useful for simultaneously deriving test characteristics and true prevalence estimates in the absence of a gold standard.