Bacterial Vaginosis: An Update on Diagnosis and Treatment

Hans Verstraelen; Rita Verhelst


Expert Rev Anti Infect Ther. 2009;7(9):1109-1124. 

In This Article

Diagnosis of BV

Signs & Symptoms

Bacterial vaginosis is confined to an asymptomatic state in at least half of the cases. Symptomatic BV, on the other hand, is most typically accompanied by foul-smelling, profuse vaginal discharge in the absence of any appreciable signs of inflammation. Symptoms of vaginitis are overall, however, rather nonspecific and therefore clinical diagnosis will at best give an indication of the presence of vaginitis, and warrants microscopic investigation to determine an infectious cause. In case of BV, such targeted diagnosis can basically be made according to the so-called clinical criteria or according to what is designated as the microbiological criteria, although both gold-standard approaches actually encompass vaginal fluid microscopy.

Clinical Diagnosis

In 1983, Amsel et al. launched clinical diagnostic criteria[1] for BV, which have proved particularly useful in clinical practice and hence are still in use today. The clinical diagnosis of BV is made if three of the four following signs are present:

  • An adherent and homogenous grayish-white vaginal discharge;

  • A vaginal pH exceeding a value of 4.5;

  • The presence of so-called clue cells – vaginal epithelial cells with such a heavy coating of bacteria that the peripheral borders are obscured – on saline wet mount;

  • A fishy or amine odor after the addition of a 10% potassium hydroxide solution (positive whiff or sniff test).

The Amsel's criteria, among others, however, have been criticized because two of the four criteria, in particular the appearance of the discharge and the appraisal of the odor, are rather subjective and hence may lead to misdiagnosis. By contrast, a pH greater than 4.5 is considered the most sensitive criterion, whereas the presence of clue cells has been considered the single most specific predictor of BV.[2,3] However, even today, diagnosis through Amsel's criteria still remains the best option for in-office testing for BV by the clinician.

Gram-stain-based Diagnosis

As an alternative, Gram-stain-based microbiological diagnosis of BV has been proposed. To perform a Gram stain, vaginal fluid or discharge is collected on a glass slide, allowed to air-dry, stained in the laboratory and examined under an oil immersion for the presence of specific bacteria. This diagnostic method has several advantages, including a permanent record, a high frequency of interpretable results, low cost, and ease of transport and storage.[2] In addition, Gram-stained vaginal smears can be evaluated repeatedly or independently by more than one assessor, thereby increasing diagnostic reliability.

The most widely performed method is the Gram-stain-based scoring system developed by Nugent et al.[3] Basically, the Nugent scoring system accounts for three bacterial cell morphotypes – that is, Lactobacillus morphotypes (large Gram-positive rods), Gardnerella and Bacteroides morphotypes (small Gram-variable or Gram-negative rods) and curved Gram-variable rods (typically Mobiluncus spp.) Based on the abundance of each of these morphotypes per oil immersion field, each of the three morphotypes is then quantitated from 0 to 4+, and the summary score is obtained by adding up the morphotypes-specific scores equates the overall Nugent score:

The criterion for BV is a score of 7 or higher. A score of 4–6 corresponds to so-called intermediate vaginal microflora, and a score of 0–3 is considered to represent normal vaginal microflora.

Overall, the Nugent scoring system for Gram-stained vaginal smears has shown high intracenter and intercenter reliability, as well as high intraobserver and interobserver reproducibility. In addition, validity studies have repeatedly documented a high degree of accuracy. However, practitioners are not usually familiar with performing in-office Gram-stain-based diagnosis, and hence point-of-care testing will usually rely on clinical criteria according to Amsel, possibly followed by laboratory-based Gram-stain confirmation. For research purposes, however a defined preference for Gram-stain-based diagnosis of BV seems to exist.

Some concern remains, however, over the performance of the Nugent scoring system. First, it has been acknowledged that Nugent's criteria are widely applied in the absence of standardized pre-analytical and analytical conditions, which may impinge on Gram-stain diagnosis. Forsum et al. emphasized the need for quality specifications in this respect,[4] as different sampling devices and procedures, different ways of spreading the vaginal specimen on the glass slide leading to differences in homogeneity of the sample and in the thickness of the smear, different fixation methods and time, and differences in the area of the high-power oil immersion field at magnification ×1000 all may affect Gram-stain interpretation.[4–6] Second, Gram-stain interpretation with regard to Nugent's criteria per se is also a matter of concern, since no definite criteria have been proposed to distinguish between the three basic morphotypes handled in the Nugent scoring system. Albeit in two international workshops the interobserver reliability in scoring Gram-stained vaginal smears was generally good,[4,5] specific problems occurred; in particular there seems to be disagreement between researchers as to which morphotypes should be considered Gram-positive rods and hence which morphotypes are scored as lactobacilli, the differentiation between cocci and small rods varies among investigators and, importantly, small bacteria morphotypes such as Gardnerella (and Prevotella) may vary in size from round to more elongated – as they may vary in Gram-staining aspect – which may lead to confusion in their categorization.[4,5]

Commercial Point-of-care Tests for the Diagnosis of BV

Several rapid point-of-care diagnostic tests for BV have been developed and commercialized, although none of these tests are being widely used. Here we provide a brief, nonexhaustive overview of tests that may be of use in clinical practice or in epidemiological settings.

Self-test pH Glove A glove with an integrated pH indicator paper was developed in Germany in the early 1990s by which women can monitor their vaginal pH by inserting one finger into the vagina. The glove has been broadly applied in population-based screening programs in Germany for the prevention of preterm birth, by which women were instructed to consult their physician if the pH was 4.7 or more.[7] As mentioned earlier, an increased vaginal pH is a very sensitive, although nonspecific, indicator of BV, and hence still warrants further assessment as outlined previously.

Testing for the Presence of Trimethylamine (Electronic Sensor Array or Electronic Nose) The volatile organic amino acids responsible for the characteristic odor in BV have been used as a target for BV diagnosis. This was initially accomplished through gas–liquid chromatography[8] – a laborious procedure. This knowledge was then translated to the development of an electronic sensor array by which vaginal fluid is passed over an application-specific array of conducting polymer sensors, each of which has specific interactions with different volatile organic species based upon their size, shape and functional group.[9] In a large diagnostic study, Hay et al. obtained a sensitivity of 81.45% and specificity of 76.1% with the electronic nose compared with Amsel's criteria, and a sensitivity and specificity of 82.9 and 77.3% compared with Gram-stain diagnosis.[10]

Testing for the Presence of Trimethylamine in Combination with Vaginal pH Assessment This point-of-care test, the FemExam® (CooperSurgical, Inc, CT, USA) test card, is based on determining pH and trimethylamine levels in vaginal fluid for the diagnosis of BV; however, the test did not compare favorably with Amsel's criteria or with Nugent criteria in published studies.[11–13]

Testing for Sialidase Activity The BVBlue® (Gryphus Diagnostics, AL, USA) system is a chromogenic diagnostic test based on the presence of elevated sialidase enzyme activity in vaginal fluid samples. This point-of-care test has consistently shown good sensitivity, specificity, and positive and negative predictive values when weighted against both Amsel's criteria and Nugent criteria.[14–16]

Testing for Proline Aminopeptidase Activity Promising results in detecting proline aminopeptidase activity (Pip Activity TestCard™, Quidel Corp., CA, USA and CooperSurgical CT, USA) of anaerobes, especially Gardnerella vaginalis, in vaginal discharge for the diagnosis of BV were reported by Schoonmaker et al. in 1991.[17] Two subsequent studies obtained a high diagnostic accuracy with the test for proline aminopeptidase activity.[18,19]

DNA Probe for G. vaginalis rRNA The Affirm™ VP III (BD Diagnostic Systems, NJ, USA) G. vaginalis DNA hybridization assay is a DNA hybridization test that is positive only for concentrations of G. vaginalis in excess of 2 × 105 bacterial cells per ml of vaginal fluid[20] and should therefore be positive most often in women with BV and rarely in women with normal vaginal microflora. The Affirm system can also detect the presence of Candida spp. and Trichomonas vaginalis in the same specimen, making it a quite attractive tool for evaluating women with vaginal discharge. The procedure requires between 30 and 45 min to complete. The test can be performed in an office setting but this is not very efficient and it is usually better done in a laboratory setting. Nonetheless, the rapid turnaround time permits return of results to clinicians within 24 h. Promising results have been reported with this approach in two populations, one in pregnant women and one in otherwise healthy women of childbearing age.[12,20] Both studies concluded that the probe can be used as a supplement to the Amsel and Nugent methods.

Molecular Diagnosis of BV

Following the recent surge of molecular analysis-based studies of the vaginal microflora, several prospects for the molecular diagnosis of BV have emerged. This approach might overcome at least some of the aforementioned problems with regard to the reliability and reproducibility of Gram-stain diagnosis.

Until the first study in 2002 that used broad-range PCR to characterize the vaginal microflora,[21] essentially all our knowledge of the vaginal microflora has been obtained from isolating organisms by culture and subsequently identifying them by phenotypic means. This approach has long been the mainstay for studies on the human vaginal ecosystem. However, cultivation of microbes as a means to characterize microbial communities in a natural ecosystem has major shortcomings, as it is recognized that many microbes in different ecosystems cannot be cultivated using standard culture techniques.[22]

Since the beginning of this decade, several research groups have applied PCR-based culture-independent methods to study the bacterial microflora of the human vagina and showed – by means of cloning of the 16S rRNA gene or the chaperonin-60 gene, by species-specific PCR (sPCR), by denaturing gradient gel electrophoresis of the 16S rRNA gene, by terminal restriction fragment length polymorphism of the 16S rRNA gene or by FISH – that previously unrecognized difficult-to-culture organisms are part of the vaginal microflora.[23,24] These broad-range bacterial PCR studies identified key organisms related to BV and opened the door for the detection of these bacteria by either conventional sPCR or quantitative real-time (QRT)-PCR.

In the past, several sPCR assays have been developed for sensitive detection of vaginal bacteria that are either characteristic for a normal microflora or are BV related. Several assays for the detection of G. vaginalis, targeting, respectively, the 16S–23S rRNA spacer region[25–27] or the 16S rRNA gene[28] were designed. In addition, the fastidious anaerobic Mobiluncus spp.[25,29] and Mycoplasma spp.[27] were interesting species for specific detection, especially the latter, as this species is not detected by Gram staining. Recently, several research groups applied sPCR for Atopobium vaginae, since this species was found to be more specific for the detection of BV than G. vaginalis.[30–33]

Although these PCR assays proved to be highly sensitive for BV detection, none of these individual assays gained a footing in the diagnosis of BV. Attempts were made to combine several sPCRs in order to obtain more predictive assays. Obata-Yasuoka et al. proposed, as a diagnostic test for BV, a multiplex PCR assay using primers specific to 16S rRNA genes of Mobiluncus mulieris and Mobiluncus curtisii, the nanH gene of Bacteroides fragilis, and an internal spacer region of the rDNA of G. vaginalis.[25] The diagnostic sensitivity, specificity, positive predictive value and negative predictive value of multiplex PCR in comparison with Gram-stain examination were 78.4, 95.6, 82.9 and 94.2%, respectively. Verhelst et al. used a combination of the sPCRs for G. vaginalis and A. vaginae for diagnosing BV.[34] The simultaneous presence of A. vaginae and G. vaginalis in a vaginal swab specimen as detected by species-specific PCR had an accuracy of 90% (95% CI: 86–92%), a sensitivity of 82% (95% CI: 59–94%) and a specificity of 90% (95% CI: 87–92%) in assessing BV. This preliminary predictive model further showed a negative predictive value as high as 99% (95% CI: 98–100%); but a positive predictive value of merely 26% (95% CI: 17–39%) in assessing BV, the latter resulting from the poor discriminative value of qualitative differentiation between normal and BV microflora, considering the common presence of G. vaginalis in low numbers with normal microflora. The most thorough sPCR approach was reported by Fredricks et al., who targeted 17 different vaginal bacteria that were previously found to be either highly specific for BV or novel.[32,35,36] The study of 264 vaginal samples obtained from 81 subjects with BV and 183 subjects without BV in two clinics in Seattle (WA, USA), revealed that while Lactobacillus crispatus was inversely associated with BV, three novel bacteria from the Clostridiales order (BVAB 1–3), as well as Atopobium, an Eggerthella-like bacterium, Sneathia/Leptotrichia, Megasphaera types 1 and 2, and a bacterium from the TM7 division were highly specific for BV. Moreover, detecting the combination of one of the Clostridiales bacteria (BVAB2) or Megasphaera type 1 produced the highest sensitivity and specificity for sPCR diagnosis of BV so far reported (sensitivity 99% and specificity 89%).

Although conventional sPCR merely allows assessing the presence or absence of a particular target, it revealed the presence of noncultivable species, some of which allowed the diagnosis of BV with much higher specificity than G. vaginalis detection. Of note is that sPCR studies reported the traditional BV-associated species Mobiluncus[25,29] and Mycoplasma[27,37,38] at approximately the same incidence as culture-based studies, while broad-range bacterial PCR studies failed to detect these species. Most importantly, sPCR studies that applied combinations of key organisms suggested that PCR-based amplification might potentially be used for the molecular diagnosis of BV and resulted in the application of QRT-PCR, allowing us to also take into account the inoculum of bacteria.

The first efforts to investigate the applicability of QRT-PCR for the diagnosis of BV came from Zarrifard et al. who determined the feasibility of using this technique to detect and quantify Lactobacillus spp., G. vaginalis and Mycoplasma hominis in the genital tract of 21 women using stored vaginal samples.[27] The results show that samples from women with BV who were clinically diagnosed had significantly higher numbers of G. vaginalis, but significantly lower numbers of lactobacilli. Moreover, there was a noticeable pattern where low numbers of lactobacilli were found in samples with high numbers of G. vaginalis and, conversely, low numbers of G. vaginalis organisms were seen in samples that had high numbers of lactobacilli. In a following study,[39] this research group compared Nugent score with Amsel criteria and quantitative bacterial PCR for diagnosing BV in 203 cervicovaginal lavage samples from women with Nugent scores of 7–10 (BV group) and 203 samples from women with BV Nugent scores of 0–3 (no-BV group). Although there was significant overlap in the log10 Lactobacillus counts between the two groups, their data demonstrate that quantitative bacterial PCR for G. vaginalis, M. hominis and lactobacilli significantly correlates with the Nugent Gram-stain method to diagnose BV. In addition, they were able to identify cut-off points for G. vaginalis and M. hominis that differentiated the BV group from the no-BV group. Utilizing all three log10 bacterial counts (G. vaginalis, M. hominis and lactobacilli), the sensitivity and specificity of the PCR assay were 83 and 78%, respectively, in comparison with Nugent score, while the sensitivity and specificity of the Amsel criteria compared with the Nugent score were 37 and 99%, respectively.

Several research groups also applied an A. vaginae QRT-PCR assay to study the role of this species in BV.[40–43] Bradshaw et al. and Ferris et al. studied the presence of A. vaginae before and after treatment with metronidazole.[40,41] Bradshaw found higher recurrence rates in women in whom both A. vaginae and G. vaginalis were detected pretreatment, and Ferris' QRT-PCR assay indicated that high pretreatment A. vaginae concentrations are predictive of adverse treatment outcomes for BV patients. Tabrizi et al. studied the occurrence of G. vaginalis and A. vaginae in 44 women who were virgins and found that 45% had G. vaginalis and 7% had A. vaginae.[42] De Backer et al. confirmed by QRT-PCR that the presence of A. vaginae seems to be a diagnostically more valuable marker for BV than the presence of G. vaginalis.[44] Zozaya-Hinchliffe et al. assessed the prevalence and abundance of uncultivated Megasphaera-like bacteria in the vaginal niche using QRT-PCR.[45] Megasphaera type 1 concentrations were higher in subjects with BV (up to five orders of magnitude) than subjects without BV, and this bacterium was significantly associated with BV (p = 0.0072), as was Megasphaera type 2 (p = 0.0366).

Recently, a few research groups applied broader sets of QRT-PCR assays for the diagnosis of BV. Menard et al. used a series of species-specific primers for QRT-PCR targeting Lactobacillus species, G. vaginalis, M curtisii, Mobiluncus mulieris, Ureaplasma urealyticum, A. vaginae, Candida albicans and M. hominis.[46] They were able to document that the presence of A. vaginae at a level of 108 copies/ml or more and of G. vaginalis at a level of 109 copies/ml or more was highly predictive for the diagnosis of BV, with a sensitivity of 95%, a specificity of 99%, a negative predictive value of 99% and a positive predictive value of 95%.[46] Interestingly, Menard et al. also showed that, according to this criterion, 57% of the samples rated as intermediate microflora on Gram stain actually involved BV.[46] Fredricks et al. used eight QRT-PCR assays targeting both easily cultivated vaginal bacteria (G. vaginalis and L. crispatus) and fastidious bacteria (BVAB1, BVAB2, BVAB3, Leptotrichia/Sneathia, Atopobium and Megasphaera-like species) to determine how concentrations of vaginal bacteria change in women with BV by comparing women who were cured with women with persistent BV 1 month following vaginal metronidazole treatment.[47] Successful antibiotic therapy resulted in 3- to 4-log reductions in median bacterial loads of BVAB1, BVAB2 and BVAB3, a Megasphaera-like bacterium, Atopobium species, Leptotrichia/Sneathia species and G. vaginalis. By contrast, median post-treatment bacterial levels did not change significantly in subjects with persistent BV except for a decline in BVAB3 levels. Fredricks et al. therefore concluded that the presence or absence of BV is reflected by vaginal concentrations of BV-associated bacteria such as BVAB1, BVAB2, Leptotrichia/Sneathia species, Atopobium species, G. vaginalis and a Megasphaera-like bacterium, and hence that these bacteria may be suitable markers of disease and treatment response.[47] In conclusion, there are now some prospects for the molecular diagnosis of BV. The study by Menard et al. that assessed the utility of quantitative loads of G. vaginalis and A. vaginae as a diagnostic tool for BV clearly seems to provide a sound basis, thereby possibly overcoming the aforementioned problems with regard to the reliability and reproducibility with Gram-stain diagnosis.[46] Indeed, it might be hypothesized that BV is characterized by a pathological core of G. vaginalis and A. vaginae in a biofilm configuration with a unique virulence profile regardless of the variability of associated species.[48] However, as recently reiterated by Kalra et al., distinct BV profiles consisting of differing species might predispose to different health outcomes.[49] As a consequence, the broader set of QRT-PCR assays applied by Menard et al. or by Fredricks et al. will be required for subtype-specific BV profiles.[46,47] In addition, the molecular diagnosis of the normal, lactobacilli-dominated vaginal microflora may be particularly interesting in the identification of women at risk of developing BV and associated adverse health outcomes.[44,50,51] These ambiguities can only be resolved by large, prospective cohort studies in which microbiological profiles obtained through molecular techniques are correlated with known reproductive and infectious health outcomes.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.