Bacterial vaginosis (BV) results from an imbalance of vaginal flora characterized by a decrease in Lactobacillus species and overgrowth of anaerobes, such as Gardnerella vaginalis and Atopobium vaginae [1-3]. An efficient means to diagnose BV in pregnant women is critical because of its association with adverse perinatal outcomes, such as preterm labor and abortion [1-5]. Most clinicians diagnose BV based on the Nugent score [2]; however, this method is limited in that it requires an experienced microbiologist who is well practiced in BV diagnosis because of the diversity of the vaginal flora [6-8]. In addition, it is unable to detect some BV-associated bacteria, particularly Atopobium vaginae [6-9]. For these reasons, some researchers have sought to develop molecular-based methods to accurately diagnose BV. Menard et al. [6, 7] reported that the quantification of A. vaginae and G. vaginalis could be a more sensitive and specific diagnostic tool, and asserted that A. vaginae quantification could differentiate BV from intermediate flora (IF). In addition, Bretelle et al. [8, 9] suggested that it may predict preterm birth. Molecular analysis of Lactobacillus species could also be used to diagnose BV since its quantitative reduction is an important aspect of BV development. For instance, Jesper, et al. reported that the relative prevalence of Lactobacillus species—including L. crispatus and L. iners—was significantly less in BV patients than of those in the normal flora (NF) group [10]; however, L. iners abundance varies across reports [11-14]. Moreover, the relative presence of Lactobacillus species can differ depending on race, ethnicity, residential area, food, and some physical conditions-such as infection and pregnancy [13-16].

Despite the large number of published quantification methods, an effective tool to diagnose BV in pregnant Korean women has yet to be developed. Therefore, the authors aimed to address this gap by developing a diagnostic system based on the relative abundance of five Lactobacillus species, Gardnerella vaginalis, and Atopobium vaginae in the vaginal microbiota for this population

Pregnant women under antenatal care at Eulji University Hospital were enrolled in the study at 10-14 weeks gestation. After providing written consent, each subject recorded her history of preterm birth, abortion, obstetrics, and the presence or absence of an existing genital infection via a questionnaire. Women who received antibiotics and/or vaginal suppositories within the last two weeks and those with a major medical condition were excluded from the study.

The total patient population consisted of 124 women. Of these, ten were excluded because of medical illness or lack of follow-up, leaving 114 study participants. The study was approved by the Institutional Review Board (IRB) of Eulji University Hospital (IRB #2014-10-002).

Vaginal samples were collected from the post fornix of the vaginal wall with a swab for the qPCR and a cotton swab for the Gram stain. The Gram stain swab was smeared on a slide and dried to fix, while the qPCR swab was stored at -80ºC until analysis.

The Nugent scoring system with Gram staining was used to diagnose BV and intermediate flora [3]. From this, 93 women showed normal flora, nine had intermediate flora, and 12 were diagnosed BV.

Vaginal swabs were thawed at room temperature for 30 minutes and diluted in a 1:9 solution of PBS/saline (pH 7.4). Bacterial DNA was isolated with a specific kit and stored in a -80 ºC freezer.

Vaginal flora qPCR for L. crispatus, L. iners, L. jensenii, L. gasseri, L. vaginalis, G. vaginalis, and A. vaginae was performed using SYBR Green Real Time PCR Master Mix and the primers shown in Table 1. Reactions were performed with a real-time system and the amplification conditions shown in Table 1. Experiments were run in 96-well plates with triplicate samples and normal saline as a negative control. Data was analyzed using the 2^-ΔCt method and then log-transformed to yield final values.

Statistical analysis was performed SPSS, version 18.0. Patient demographics and differences in the presence microorganisms among the three diagnostic groups were analyzed by Kruskal-Wallis (statistical significance, p < 0.05). Mann-Whitney tests were used to make comparisons between groups (statistical significance, p < 0.05/3). Significant factors were used for Receiver Operating Characteristic (ROC) curve analysis to identify a diagnostic threshold for BV. Statistical significance was defined as p < 0.05.

Twelve women were diagnosed as having BV, nine had intermediate flora, and 93 had normal flora. No significant differences in patient demographics including age, parity, abortion history, and delivery date were found among three groups (Table 2). Table 3 shows the quantification of the seven microorganisms analyzed in the study. Interestingly, L. crispatus was more prevalent in NF women than in BV and IF counterparts (p-value < 0.001), whereas less L. iners was found in the NF group compared to BV patients (p-value < 0.001). However, the differences between NF and IF, and between IF and BV were not remarkable. Moreover, no significant differences were observed in the relative abundance of L. gasseri, L. Jensenii, or L. vaginalis. As expected, G. vaginalis and A. vaginae were far more prevalent in BV patients than in NF counterparts; and while IF women exhibited more G. vaginalis than the NF group, that A. vaginae was insignificant.

Four microorganisms—L. crisp, L. iners, G vaginalis, and A. vaginae—showed significant differences among the three groups. These four factors were analyzed by ROC curve analysis to determine a threshold for BV diagnosis. In addition, L. crispatus was less prevalent in BV than in NF patients, while the other species were increased in BV when compared to NF. Therefore, ROC curve was analyzed with two curves (Figures 1 and 2).

Figure 1.

— Receiver operating characteristic (ROC) curve of L. crispatus for bacterial vaginosis diagnosis. Area under curve (AUC), 0.742.

Figure 2.

— ROC curve of L. iners, G. vaginalis, and A. vaginae for bacterial vaginosis diagnosis. AUC for G, vaginalis, A. vaginae, and L. iners were 0.934, 0.918, and 0.763, respectively. G. vaginalis: sensitivity, 0.917; specificity, 0.902; cut-off value, 3.724. A. vaginae: sensitivity, 0.833; specificity, 0.833; cut-off value, 3.442.

The area under curve values were 0.934, 0.918, 0.763, and 0.742 for G. vaginalis, A. vaginae, L. iners, and L. crispatus, respectively (Figures 1 and2). ROC thresholds (Log 2^-ΔCt) for G. vaginalis and A. vaginae were 3.724 and 3.442, respectively (G. vaginalis: sensitivity, 0.917; specificity, 0.902; A. vaginae: sensitivity, 0.833; specificity, 0.833).

The G. vaginalis and A. vaginae binaries for BV diagnosis were set based on the ROC thresholds and analyzed by Chi-square testing (Table 4). Notably, both measures were statistically accurate diagnostic modalities with G. vaginalis and A. vaginae predicting BV in 11/12 (91.7%, p < 0.001) and 10/12 (83.3%, p < 0.001) of cases, respectively.

BV is a critical disease to diagnose in pregnant women because of its association with preterm birth [2]. However, the current methods for BV diagnosis are limited and could be improved by the development of a molecular test [5-12]. For instance, the quantification of A. vaginae in the vaginal microflora is reported to predict preterm birth [6-9]. Since the vaginal microbiome varies based on race, ethnicity, living area, or medical conditions, such as infection or pregnancy [13-15]; therefore, the present authors attempted to analyze the bacterial distribution in pregnant Korean women to identify useful markers to diagnose BV.

In this study, the authors showed that the quantification of L. crispatus, L. iners, G. vaginalis, and A. vaginae were significantly different in the BV group versus the NF group. A ROC curve was used to generate a threshold for BV diagnosis, which was set at 0.918 and 0.934 for G. vaginalis and A. vaginae, respectively. Notably, these values led to a detection rate of 91.7%, and 83.3% in the present patient population. Menard et al. [6, 7] asserted that the quantification of A. vaginae and G. vaginalis could serve as more sensitive and specific diagnostic tool for BV diagnosis in France, and also asserted that the prevalence of A. vaginae could differentiate BV in IF [7]. However, despite several published quantification methods, an effective tool to diagnose BV in pregnant Korean women has yet to be developed. This is particularly important as vaginal microbiota composition can vary according to ethnicity, residence, and other conditions, such as pregnancy and vaginal infection [13-15]. Therefore, the present authors aimed to develop a diagnostic system based on the relative abundance of five Lactobacillus species, Gardnerella vaginalis, and Atopobium vaginae in the vaginal microbiota of this population.

The present results show that the levels of L. crispatus and L. iners were relatively unchanged when compared to those of A. vaginae and G. vaginalis; however, the significance of L. iners in BV is controversial. Jesper et al. [10] previously reported that L. iners is much more prevalent in NF women than BV counterparts, yet the present data showed the opposite (AUC = 0.76). Moreover, L. crispatus prevalence was decreased in BV patients (AUC = 0.74) and has been reported to have stronger bacterial inhibitory activity than L. iners [17]. However, the present authors have known that it had less diagnostic value than two anaerobes in spite of its significant decrease in BV; nevertheless, these results should be verified with additional studies.

In conclusion, this study demonstrated that G. vaginalis and A. vaginae quantification in vaginal secretions could be a useful measure to diagnose BV in pregnant Korean women and may facilitate the development of a more accurate microbiome-based molecular diagnostic for BV in the future.