Relative performance of three methods for diagnosing bacterial vaginosis during pregnancy.

Objective: This study measures the relative performance of three methods for diagnosing bacterial vaginosis (BV) during pregnancy and assesses the implications of measurement for clinical practice and surveillance.

Methods: A sample (n = 1,780) of English or Spanish speaking women, with a singleton intrauterine pregnancy and receiving prenatal care at a consortium of public health centers in Philadelphia were consecutively enrolled. Gram stain, clinician's diagnosis, and a commercial test were the three diagnostic methods used to assess BV. Sensitivity, specificity, and the positive and negative predictive values of clinical diagnosis and the commercial test were assessed using the gram stain/Nugent score as a gold standard.

Results: The prevalence of BV, measured on the same population, differed considerably depending on the diagnostic test used. The measured prevalences were 55% (Gram stain), 28.5% (clinician's diagnosis), and 12.6% (commercial test). The prevalence of BV (diagnosed by gram stain) was twice as high among African American women compared to White women. Only 69% BV-positive high-risk women were treated for BV.

Conclusions: Inaccurate diagnosis of BV leads to missed cases. The identification of true cases is critical for assigning treatment and for assessing treatment effectiveness. Clinician's routine diagnosis fell short of recommended procedures and performed poorly compared to gold standard in case ascertainment. This inability to ascertain cases may have an impact on our ability to prevent preterm birth.