The importance of correct diagnosis of bacterial vaginosis and of more detailed characterization of the vaginal microflora
Although not causing a vaginal inflammatory response, bacterial vaginosis is considered to be the most common cause of vaginitis in pregnant and non-pregnant women and prevalences between 4.9% and 36.0% have been reported from European and American studies . Several studies suggest the possibility that bacterial vaginosis increases the risk of acquiring HIV [24, 25] and that the bacterial flora associated with bacterial vaginosis increases genital-tract HIV shedding . A recent meta-analysis by Leitich et al.  established an odds ratio of 8 for preterm birth in association with bacterial vaginosis during early pregnancy. Spontaneous preterm birth occurs in 7–11% of pregnancies but accounts for three quarters of perinatal morbidity and mortality and for half of long term neurological impairment in children [28, 29].
Bacterial vaginosis is characterized by the replacement of the normal vaginal microflora of lactobacilli by Gardnerella vaginalis and anaerobic organisms. Recently, different groups showed that the strict anaerobe Atopobium vaginae is another organism that is strongly associated with bacterial vaginosis [12, 13, 16, 17]. The association between A. vaginae and bacterial vaginosis might help explain why some women suffer from recurrent bacterial vaginosis. For example, a recent study pointed to great in vitro efficacy of metronidazole, since this antibiotic inhibited growth of 99% of the vaginal isolates from bacterial vaginosis samples , but most likely overlooked the fastidious metronidazole resistant A. vaginae, shown in this study to be present in 86.4% of bacterial vaginosis samples when detected with species specific PCR.
Given the possibility that certain not yet characterized subgroups within the presumably heterogenic clinical entity of women with bacterial vaginosis could identify a group at higher risk for preterm birth than women with bacterial vaginosis as a whole and that adequate treatment of women from this higher risk group may allow for more targeted preterm birth prevention, better understanding of the composition and dynamics of the vaginal microflora and accurate diagnosis of bacterial vaginosis are warranted. Also, our data indicate that refined characterization of vaginal microflora may be necessary for more accurate interpretation of the results of clinical studies. For example, thus far Atopobium vaginae has been overlooked in clinical studies and furthermore, the fact that different Lactobacillus species may confer different strengths of colonisation resistance [10, 11] has not been taken into account, partly because most laboratories lack the access to rapid and accurate methods for the identification of lactobacilli to the species level. In other words, several studies concerning the relation between the status of the vaginal microflora and different gynecologic and obstetric diseases and their treatments thus far may have reached biased conclusions due to insufficiently precise characterization of the microflora.
Criteria for microbiological categorization of vaginal microflora status
Spiegel et al.  defined a scoring system based on some of the bacterial cell types that can be seen in Gram stained smears of vaginal secretion. This was later refined by Nugent et al. , who provided a scoring system that evaluates the changes in vaginal microflora, from the normal condition to bacterial vaginosis status, as a continuum. Although the Nugent criteria have gained wide acceptance for the evaluation of the condition of the vaginal microflora [2, 32], further refinement is warranted for several reasons. First, no definite criteria have been described to distinguish the Lactobacillus cell types from the Gardnerella and Bacteroides-Prevotella cell types. In practice and in our experience, 'morphotypes' are often difficult to assign to one of these groups. Also, some genera and species that are clearly associated with bacterial vaginosis, like Peptostreptococcus spp.  and A. vaginae [12, 17, 13] are not included in the Nugent score. Furthermore, Forsum et al.  found major discrepancies in scoring when the lactobacillary cell types were few in number and Larsson et al.  reported several problems in the interpretation of smears. For example, using the Nugent criteria, the presence of different Lactobacillus cell types in smears from patients with bacterial vaginosis can lead to assignation to grade II, whereas patients without bacterial vaginosis but with smears with more than 300–500 pleomorphic Lactobacillus cells may be regarded as containing G. vaginalis, also because some of these cells are very small. Additionally, the Nugent scoring system conflates women with potentially very different vaginal microflora in a single category .
In this study, the clinical microbiologist (GC) could not grade some of the smears due to the presence of cell types not scored in the system developed by Nugent  and classified these samples as grade I-like. Further detailed observation lead to the splitting up of grade I samples into subcategories designated grade Ia, grade Ib and grade Iab. After blind grading of the vaginal smears into grades Ia, Ib, Iab, I-like, II, III and IV, this classification was compared with the culture results and with species specific PCRs.
Grade Ia and Iab: Agreement with the presence of L. crispatus
From this comparison it became obvious that it is possible to recognize the presence of L. crispatus by means of Gram stain, since this species was cultured in 81.9% of the grade Ia samples and 76.7% of the grade Iab samples. Nevertheless, L. crispatus was not cultured from 21 of the 162 grade Ia samples. This may be explained by the fact that L. crispatus is not as easily cultured as other lactobacilli. Indeed, L. crispatus colonies were quite often observed as satellites of other bacteria and in some cases no growth at all was observed in samples with numerous L. crispatus-like lactobacilli on Gram stain. Using non culture dependent t-RFLP-analysis (data not presented) the Ia samples negative for L. crispatus culture were tested and 16 were positive for L. crispatus, bringing the agreement between Gram stain grading as grade Ia and the presence of L. crispatus to 96.9%. Similarly, when taking into account t-RFLP-analysis results, the agreement between categorization as grade Iab and t-RFLP-analysis positive for L. crispatus was 92.9% whereas L. crispatus was detected by t-RFLP-analysis only in 27.3%, 20.0%, 22.5% and 0% for grades Ib, I-like, II and III, respectively. These results indicate that - for a trained microbiologist - it is possible to recognize L. crispatus bacteria upon cell morphology, a finding that is of importance since this species is clearly associated with healthy microflora, and possibly better ensures stable healthy microflora than other lactobacilli . Samples were scored as grade Ib when no L. crispatus cell types were observed, but other Lactobacillus cell types were predominant. The agreement with culture results was high: only 13.3% contained L. crispatus upon culture, whereas L. gasseri, L. iners, and L. jensenii were present in respectively 32.0, 39.8, and 24.3% of the grade Ib samples. These were clearly grade I samples since bacterial vaginosis-associated organisms were mostly absent.
The colonisation resistance conferred differs between Lactobacillus species
Overall the frequency of isolation of all Lactobacillus species together was comparable for the different grades in our population, since lactobacilli were cultured from 96.9% of grade Ia, 94.5% of grade Ib, 96.7% of grade Iab, 78.9% of grade I-like, 93.5% of grade II, 59.1% of grade III and 62.5% of grade IV samples. This is in agreement with previous reports [32, 34]. However, we observed a clear difference with regard to the Lactobacillus species frequency distribution for the different grades. While L. crispatus, known as a strongly H2O2-producing species [7, 8], was cultured from 87.0% of grade Ia specimens, it was absent in grade III specimens and only present in 2.8% of grade I-like specimens. In contrast, L. iners, reported as a weakly H2O2-producing species [7, 8], was present in only 3.7% of Ia specimens but in 39.8% of grade Ib and 31.8% of grade III specimens. Whether it is the hydrogen peroxide production by L. crispatus that confers colonisation resistance remains a matter of debate, since a correlation between the presence of hydrogen peroxide production and the type of vaginal microflora was found by some , though not by others . Possibly other species specific characteristics, present in L. crispatus, but absent in species like L. gasseri and L. iners, confer colonisation resistance. It has also been hypothesized that the onset of perturbation leading to bacterial vaginosis may be due to competition between Lactobacillus species , a situation possibly reflected by grade Iab specimens.
Grade I-likes: a separate category of vaginal microflora status
A number of samples were initially difficult to score because the predominant cell types could not be categorized as Lactobacillus, Gardnerella, Bacteroides-Prevotella or Mobiluncus cell types. These samples were considered as belonging to a separate category because of the presence of Gram positive rods, either quite small and short or otherwise irregularly shaped with clubbing, curved edges and irregular staining and often arranged like Chinese letters ('diphtheroid cell types'). Since it is likely that most microbiologists would score this cell type as 'Lactobacillus-like' and that therefore it would be scored in most cases as grade I, we designated it as 'grade I-like'. Culture and species specific PCR confirmed that indeed these samples represent a separate kind of vaginal microflora. This is reflected by the increased species diversity of 0.83, which is much higher than that for grades Ia, Ib and Iab (0.15–0.30) and which is comparable to that of grade II (0.76), but even more so by the virtual absence of L. crispatus (cultured from only one of 36 samples) as well as of G. vaginalis and A. vaginae (cultured from respectively 1 and 0 samples) and the presence of Bifidobacterium spp. in 19 of 36 samples, a much higher prevalence than in samples from all other grades. This was confirmed by cloning of two grade I-like samples, which contained only L. delbrueckii, L. gasseri and B. breve.
Rosenstein et al.  mentioned a category of vaginal smears with aberrant morphology, which they designated as grade I revertants. At first sight, their category shows resemblance with the category we describe here as I-like, because of low numbers of G. vaginalis and increased numbers of bifidobacteria, but on the other hand they reported even more bifidobacteria in their grade II and grade III samples and they designated this category as grade I revertants because the vaginal microflora of all 41 women with such smears reverted to grade I, which was not the case in our study (data to be presented elsewhere).
Importantly, since Gram stain based categorizing can result in the interpretation of grade I-like samples as genuine grade I samples (whereof their designation), this class of samples may jeopardize - and probably has done so in the past - the interpretation of the results of clinical studies.
Grade II: a microbiologically intermediate stage between healthy microflora and bacterial vaginosis
Our results confirm that grade II samples represent a microbiologically clearly intermediate status between grade I and III. L. crispatus is still present in 10.9% of the samples (compared to 59.0% of grade I and 0% of grade III samples), whereas the number of samples with L. gasseri (54.3%) is increased compared to grade I (21.3%) and grade III (9.1%). Species diversity of grade II is intermediate between that for grade I and grade III and species typically associated with bacterial vaginosis, like A. neuii, A. christensenii, A. vaginae, B. ureolyticus, F. magna, G. vaginalis, Peptoniphilus sp. and V. cambriense, are present, but again in a lower number of samples than in grade III specimens.
Grade III: Characterization of bacterial vaginosis -related organisms
The following species are generally considered as bacterial vaginosis related anaerobe organisms: Anaerococcus (Peptostreptococcus) tetradius, A. (Peptostreptococcus) vaginalis, Atopobium vaginae, Bacteroides ureolyticus, Finegoldia (Peptostreptococcus) magna, G. vaginalis, Gemella (Peptostreptococcus) morbillorum, Mobiluncus curtisii, Mycoplasma hominis, Peptoniphilus sp., Peptostreptococcus sp., Prevotella bivia, Prevotella ruminicola and Prevotella sp. [37, 38]. Using tDNA-PCR we were able to identify 87.8% of the cultured isolates to the species level and found our results to be largely in agreement, but in addition we cultured Actinomyces neuii, Aerococcus christensenii, Dialister sp. and Varibaculum cambriense, whereas Mobiluncus spp., Mycoplasma hominis and Ureaplasma urealyticum were not or only sporadically cultured from grade II and grade III specimens. The absence of the latter species in our study can be explained by the fact that we did not use the specific culture methods for these fastidious organisms.
In this study we confirmed the strong association, as established previously [12, 13, 17], between A. vaginae and bacterial vaginosis.