Characterization of the vaginal microflora in health and disease.

BACKGROUND Bacterial vaginosis (BV) is an imbalance of the vaginal bacterial microbiota and its aetiology is still unknown. Our aims were to investigate the diagnostic potential of species/genus specific quantitative PCR (qPCR) for bacteria present in swabs and first-void urine (FVU) samples using Nugent's and Claeys' criteria and 454 sequencing of the vaginal microbiome as reference. METHODS Self-collected swabs, vaginal smears and FVU were obtained from 177 women from Greenland (Study I and III) and physician-collected vaginal swabs and smears were obtained from 163 Swedish women (Study II). BV was diagnosed by Nugent's criteria in Study I and III and by Amsel's criteria in Study II. The vaginal swabs and FVU samples were analysed by qPCR for selected vaginal bacteria in all three studies and for four sexually transmitted infections (STIs) in Study I. RESULTS Study I: STIs were common in women from Greenland and BV was found in 45% of these women but was not associated with individual STIs. In multivariate logistic analysis, Atopobium vaginae and Prevotella spp. were both independently associated with BV in swabs. BV could be subdivided into clusters dominated by a single or a few species together. Seven vaginal bacteria (A. vaginae, Prevotella spp. Gardnerella vaginalis, Bacterial vaginosis associated bacterium (BVAB) 2, Eggerthella-like bacterium, Leptotrichia amnionii and Megasphaera type 1) had areas under the receiver operating characteristic (ROC) curve > 85% in swabs, suggesting that they were good predictors of BV according to Nugent. Study II: For the majority of species/genera, the kappa values indicated fair to good agreement when their presence was determined by 454 pyrosequencing versus real-time PCR. The same seven vaginal bacteria as found in Study I, had areas under the ROC-curve > 85% in swabs from Swedish women, demonstrating a good diagnostic accuracy for BV according to Amsel. Study III: In a multivariate model, Megasphaera type 1 and Prevotella spp. remained significantly associated with BV in FVU samples. A linear regression analysis showed good agreement between bacterial load from swabs and FVU, but Prevotella spp. could be detected in high numbers in a few FVU samples without being present in swabs. After applying ROC curve analysis, the same seven vaginal bacteria as previously mentioned showed good prediction for BV according to Nugent in FVU. BV could be detected with comparable sensitivity in FVU and vaginal swabs. CONCLUSION BV can be diagnosed by molecular methods performed either on swabs or urine but it is important to apply thresholds in order to improve the accuracy of the diagnosis. Furthers it was possible to identify clusters of BV dominated by single or paired bacteria, and these clusters could classify BV into subgroups, providing a more detailed understanding of the condition. Seven vaginal bacteria were highly accurate for BV diagnosis both in swabs and FVU. Finally a good agreement between Nugent and Claeys was found.