Assessing a diagnosis tool for bacterial vaginosis

Diagnosis of bacterial vaginosis (BV) in resource-poor settings relies on semiquantitative microscopy algorithm such as the Nugent score (NS). We evaluated a quantitative real-time PCR (qPCR) assay to detect and quantify individual BV-associated bacterial communities. Vaginal swabs from 247 South African women attending an STI clinic were evaluated for BV using NS. We used qPCR to analyze DNA from vaginal swabs for eight BV-associated bacteria, Gardnerella vaginalis (GV), Prevotella bivia (PB), BV-associated bacteria 2 (BVAB2), Megasphaera-1 (M-1), Atopobium vaginae (AV), Lactobacillus crispatus (LC), Lactobacillus jensenii (LJ), and Lactobacillus iners (LI). Sensitivities and specificities were generated for each qPCR assay. Using a ROC analysis, cutoffs were calculated for each bacterial species. A logistic regression model was used to determine the strongest predictors of BV status. Nugent scores indicated 35.6% of patients harbor BV-associated flora (NS 7-10). AV, GV, GAMB (GV + AV + M-1 + BVAB2), and LC + LJ showed the highest AUC, sensitivities, and specificities (listed respectively): AV (0.96; 96%; 93%), GV (0.88; 78%; 79%), GAMB (0.9; 87%; 82%), and LC + LJ (0.84; 82%; 72%) (all p