Classification and Regression Trees for Bacterial Vaginosis Diagnosis in Pregnant Women Based on High-Throughput Quantitative PCR

Abstract

Bacterial vaginosis (BV) diagnosis in pregnancy is based on the Nugent score, which consists of semiquantitation of bacterial morphotypes. Limited data exist concerning molecular-based diagnosis in asymptomatic pregnant women. Using high-throughput quantitative PCR, 34 microorganisms were screened in asymptomatic pregnant women and compared with the Nugent score. Three-hundred and four vaginal samples had a Nugent score <7 (69.9%) and 131, a Nugent score ≥7 (30.1%), consistent with BV. More pregnant women with BV share Atopobiumvaginae, bacterial vaginosis associated bacteria-2, Gardnerella spp., Mobiluncus curtisii, Mo. mulieris, Mycoplasma hominis, Ureaplasma urealyticum, Prevotella bivia, Megasphaera 1, and Megasphaera 2 in their vaginal sample. Fewer pregnant women with BV share Lactobacillus crispatus, L. gasseri, L. jensenii, and Enterococcus faecalis in their vaginal sample (P<0.001). Classification and regression tree analysis was performed to determine which combinations of detected bacteria optimally diagnose BV in this population. A set of only four bacteria of 34 microorganisms (A. vaginae, Gardnerella spp., L. crispatus, and P. bivia) was the best combination to identify BV in a cohort of asymptomatic pregnant women, with a sensitivity of 77.1%, and specificity of 97.0% compared with the Nugent score. The quantitative PCR in the present study responds to the limits of the Nugent score by implementing an easily reproducible quantitative assay to assess the absence of BV in pregnancy.