The microbiology of the vagina is complex, containing 10⁹ bacterial colony forming units per gram of secretions and potentially dozens of species. It is mainly dominated by members of the genus Lactobacillus, which are capable of H₂O₂-production and lactic acid, thereby maintaining the generally acidic vaginal pH. Age, phase of the menstrual cycle, sexual activity, contraceptive choice, pregnancy, presence of necrotic tissue or foreign bodies, and use of hygienic products or antibiotics can disrupt this ecosystem. A disturbed vaginal microbiota is characterised as vaginitis. Vaginitis, whether infectious or not, poses one of the most common problems in gynaecology, and it is one of the main reasons leading women to seek advice from a physician 1. However, diagnosis and treatment can be elusive, if based on clinical symptoms and the characterisation of vaginal discharge alone 23. This may lead to a lack of relief from symptoms and, because of inadequate treatment, to increased costs for the patient and social securities.

Bacterial vaginosis (BV) and candida vulvovaginitis are responsible for 90% of cases of vaginitis 4. BV is characterised by the substitution of the regular, lactobacilli dominated microbiota by a complex and abundant flora of strictly or facultatively anaerobic bacteria (Gardnerella vaginalis, Bacteroides spp., Prevotella spp., Peptostreptococcus spp., and Mycoplasma hominis) 5678. Recent studies suggested a correlation between the metronidazole-resistant anaerobe Atopobium vaginae and BV 9101112.

Many cases of BV are asymptomatic or present with only malodorous vaginal discharge and no inflammatory complaints. In contrast, CV often correlates with scant discharge and marked inflammatory symptoms (pruritus and soreness) 13 and is characterised by an excessive proliferation of Candida spp. in the vaginal flora 14.

Due to the fact that clinical symptoms associated with vaginitis are various, the clinical diagnosis is often deceptive. As defined by Amsel and colleagues 15 a correct diagnosis of BV should comprise at least three of the following criteria: a thin homogeneous white discharge, the presence of "clue cells" on microscopic exam, a pH > 4.5 of vaginal fluid and fishy odour before or after adding 10% KOH (whiff test). Gram-staining of vaginal secretion is also an accepted additional diagnostic test of BV 16171819. A current CV should be diagnosed by performing saline microscopy in order to look for candidal buds or hyphae. Additionally, the diagnostic panel can comprise culturing, PCR or antibody detection 202122.

The purpose of this study was to evaluate, whether practitioners truly rely on the recommended criteria for diagnosis of BV 1518 and if so, whether the use of these criteria leads to a correct diagnosis according to our definition of BV. In addition, we evaluated the accuracy for the diagnosis of CV. In order to do so, we used a combination of culturing methods, DNA hybridisation and PCR for the differentiation of the various vaginal infectious types.

Of the 220 vaginal samples obtained only 197 were included into the study, as 23 samples did not fulfil all evaluation criteria. No pH was recorded in 12 samples, no admission diagnosis was given in 10 samples, and one sample contained only the swab for DNA-hybridisation.

A comparison of frequencies of detection and mean counts of the major genera comprising the vaginal microbiota in relation to the recorded vaginal pH is shown in Table 1. No significant differences in the frequency of isolation of the various lactobacilli types (H₂O₂-producers and non-producers) could be observed between women with a normal pH and women with an elevated pH, respectively. However, cell counts were decreased by one order of magnitude between the two groups. A lower cell count was recorded for the group with an elevated pH. Interestingly, high rates of H₂O₂-producing lactobacilli (> 10⁸CFU/ml) were significantly (P < 0.05) more often reduced in women who were recorded with an elevated vaginal pH. In several samples positive for H₂O₂-producing lactobacilli, colonies which appeared light blue were seen, but their amount was always less than the number of dark blue colonies.

In 45 cases an elevated pH was accompanied by a decrease in the frequency of H₂O₂-lactobacilli as well as by the presence of anaerobes, and was accordingly classified as bacterial vaginitis (BV). Candida spp. were detected in 46 cases and mixed infections in one case, whereas in 105 cases no sign of an infectious vaginitis could be identified. The mean counts of total aerobes, anaerobes, Staphylococcus agalactiae, Streptococcus spp., Bacteroides spp., and bifidobacteria were similar in women with a normal pH and women with an elevated pH. However, the amount of Bacteroidesspp. was elevated by one order of magnitude in women with an elevated pH. No other anaerobic bacteria were identified. Furthermore, no significant differences in the presence of Gardnerella vaginalis, Candida spp. and Trichomonas vaginalis were detected. However, there was a significant correlation between Atopobium vaginae and an elevated pH. Interestingly, A. vaginae was always detected in combination with G. vaginalis. Within the study group no correlations between pre- and post-menopausal women, vaginal complaints and the microbiota could be drawn.

The results of the 197 questionnaires are summarised in Tables 2 and 3. Itching and burning were the most common symptoms reported, followed by discharge, redness, and odour (Table 2). None of the reported symptoms could be associated with the presence of a specific organism (P > 0.05; data not shown). The diagnosis of the participating physicians was BV in 80 cases, candida vulvovaginitis (CV) in 109 cases, and mixed infection in 8 cases (Table 3). Amsel-criteria were used in 54 out of 80 cases for the diagnosis of BV, microscopy in 60 out of 109 cases for the diagnosis of candida vulvovaginitis, and both were used in four out of eight cases for mixed infections (Table 3). Both, Amsel criteria and microscopy, were predominantly but not exclusively used by gynaecologists (data not shown). Not a single physician based his diagnosis on Nugent-score 18.

The diagnosis of BV by physicians agreed only fairly with our diagnosis (κ = 0.29; 95% CI 0.16, 0.42). There was an even poorer agreement for the diagnosis of candidiasis (κ = -0.01; 95% CI -0.12, 0.10). Regardless of the implementation of the various recommended criteria for evaluation of vaginitis 2128, the error of judgement for BV, CV or a mixed infection was high (Table 3). The level of misdiagnosis exceeded 70%. The misjudgement of a mixed infection was highest (87%), followed by CV (77%) and a manifested BV (>61%). The use of the Amsel criteria for the detection of BV or microscopy for the detection of Candida spp. did not enhance the diagnostic correctness (Table 3). Compared to the detection of a pathological vaginal colonisation as reference standard, which was based on our microbial verification of the pathologic agent or altered vaginal microbiota, the diagnostic classification by physicians was of relatively minor accuracy. Sensitivity and specificity (95% confidence interval), respectively, were 68.9% (55–82) and 67.8% (60–75) for the diagnosis BV, whereas only 54.3% (40–69) and 44.4% (36–52) for the diagnosis of a CV. Positive and negative predictive values as well as likelihood ratios (LR) of a positive and negative result were 38.8% (28–50), 88% (80–93), 2.1 (1.6–2.9) and 0.46 (0.29–0.71), respectively, for the diagnosis BV. The corresponding values for the diagnosis of a CV were 22.9% (15–31), 76.1 (65–84), 0.97 (0.72–1.31) and 1.03 (0.72–1.47), respectively. From these values the diagnostic odds ratios were calculated for the diagnosis of BV and Candidiasis with values of 4.66 (95% CI: 0.51, 2.16) and 0.95 (95% CI: 0.49, 1.84), respectively, indicating again that the diagnosis of a CV by physicians does not discriminate between diseased and not diseased.

Studies showed that bacterial vaginosis (BV) resulted from an imbalance of the normal vaginal flora, with an overgrowth of gram-negative anaerobes and a reduction in lactobacilli, especially those that produced H₂O₂ 29303132. Candida vulvovaginitis (CV) was primarily associated with Candida albicans, but other species are emerging 3334. The standard diagnosis for BV should enclose the Amsel-criteria 28. For the identification of fungal infections, culturing or microscopy was recommended 21.

However, error in diagnosis was rather high within the group of surveyed physicians, no matter whether they used the recommended techniques or not. The establishment of a wrong diagnosis was highest for mixed infections followed by CV and BV. This fact may be explained by an inadequate use of microscopy, as microscopy is the key technique in recognition of a vaginal infection. Microscopy without specific staining generally leaves room for interpretation. In example, none of the participating physicians detected A. vaginae under the microscope or interpreted the presence of small elliptical cocci as conspicuous. When asked in retrospect, none of the participating physicians was aware of the existence of this species. Taken into account that A. vaginae is resistant to metronidazol, a common antibiotic given for the treatment of BV, a proof of absence of this organism should be mandatory at least in women with recurring vaginal complaints.

According to its frequent citing in scientific publications the Nugent-score is well established for the diagnosis of BV 7173536. However, the herein presented results suggest that it is not commonly used on a day to day basis within the practicing community of physicians.

In recent years, the capability of lactobacilli to protect against vaginal infections by altering vaginal pH through the production of metabolic end-products including lactic acid, and by their ability to produce hydrogen peroxide (H₂O₂) has been widely recognised 3132404142. If those lactobacilli, which predominantly produce H₂O₂ are missing or underrepresented, serious problems including BV 4243, recurrent urinary tract infections 32 or even an increased risk of pre-term delivery are likely to occur 3044. Interestingly, in over 50% of cases within our study, no association of complaints with BV or candida vulvovaginitis could be drawn. However, within this group of patients, low cell counts of H₂O₂-producing lactobacilli were observed. Lactobacilli, especially those capable of H₂O₂-production, are responsible for the maintenance of the normal vaginal ecosystem 293145. Therefore, in cases where lactobacilli cell counts are low it may be favourable to supplement the vaginal flora with lactobacilli to help restore the usual conditions.

Concerning the cases of misdiagnosed fungal infections, which occurred more frequently than cases of misdiagnosed BV, our results are in accordance with the literature 222. Because microscopic evaluation requires special diagnostic skills not available to all practitioners, misjudgement is common, therapy is frequently empirical and the relapse rate is high 20. In our study, the recognition or the exclusion of the diagnosis candidiasis by surveyed physicians was nearly as likely in patients with the disease as in patients without the disease. Physicians diagnosed 54.3% of all candida infections correctly, but had a false positive CV diagnosis for 55.6% of all not infected cases. This is also demonstrated by the LR values close to 1, with which tests cannot help a clinician to rule in or rule out the target disease. Therefore, the diagnosis of a CV by physicians was not better than those expected by chance as judged by the positive and negative likelihood ratios. In one recent study, the sensitivity of microscopy for diagnosis of candida vulvovaginitis was only 22% [46]. Therefore, a more reliable test system, like the herein implemented hybridisation system (Affirm VPIII), which detects several Candida spp. (C. albicans, C. glabrata, C. kefyr, C. krusei, C. parapsilosis and C. tropicalis) is necessary.

Our results are in accordance with a recent study, which concluded that the clinical diagnosis of vaginal infections (history and physical examination, vaginal pH, wet mount microscopy, whiff test) is inadequate and should be confirmed by microbiological testing [47]. This was also reported by Anderson et al., who evaluated the role of clinical examinations in published articles from 1966 to 2003 [48]. The poor performance of individual symptoms and signs left one third of women with vaginal complaints without an adequate diagnosis. In general, laboratory tests should be more emphasized in the diagnosis of vaginal complaints.

It becomes evident from this study that the misjudgement of vaginal complaints is rather high and that the so called "gold-standards" for the diagnosis of vaginal infections are not widely and/or correctly used. Ultimately, a correct diagnosis is not yielded; it is rather throwing a dice for the diagnosis.

Herewith, the authors declare that they obtain, except for DT, funding by the Institute for Microecology.

AS, KR and VR planned and carried out the design of the study. AS and KR participated in laboratory studies and interpretation of the data. AS wrote the first draft of the manuscript. DT carried out the statistics. All authors had intellectual contribution, and all read and approved of the final manuscript.