Diagnosis of bacterial vaginosis: Comparison of Nugent and novel microscopic method

Background/Aim: Bacterial vaginosis (BV) is common cause of vaginal
 discomfort in women. The aim of this study was comparison of Nugent's scoring system and novel microscopy method, introduced in our laboratory and
 used in BV diagnosis. Methods: 705 pregnant and asymptomatic women between
 24 and 28 weeks of pregnancy participated in this prospective study. Degree
 of agreement between methods was determined by kappa index. Sensitivity,
 specificity, positive and negative predictive value of novel microscopy
 method was compared to Nugent's score as standard. Results: Based on scoring
 system of both methods, Nugent and novel microscopy method, BV was diagnosed
 in 21%, and 25% of women, respectively. Despite the disparities among
 diagnostic criteria, which mainly concerned classification of intermediary
 samples, the degree of agreement between categories, determined by kappa
 index, was satisfactory: Nugent vs novel microscopy method (?=0,68; good
 agreement), and Nugent vs novel microscopy method without intermediary
 results (?=0,83; very good agreement). We also demonstrated that compared to
 Nugent, as golden standard, novel microscopy method had high sensitivity and
 specificity (ranging from 75%-99.3%), and positive and negative predictive
 values (ranging from 88.8%-99.5%). Conclusion: novel microscopy method in
 diagnosis of BV, corresponded well with Nugent's scoring system which allows
 it to be an alternative method in diagnosing of BV. Our method is based on
 relative number of bacterial morphotypes, either rod forms (? 1.5?m,
 lactobacilli) or non-rod forms (< 1.5?m, bacterial vaginosis associated
 bacteria) under 200x magnification, which extends the surface of examined
 preparation, but without prolongation of observer's working time.
 Furthermore, novel microscopy method appeared to be flexible and can be
 reorganized in the way to categorize findings into only two groups: normal
 and BV, which makes it comparable to dichotomous Amsel's clinical criterion.


Introduction
The main constituents of healthy vaginal microbiome are lactobacilli. Protective role of lactobacilli is reflected in their ability to antagonise with other bacteria for adherence to vaginal epithelium as well as to synthesise antimicrobials (hydrogen peroxide, lactic acid, bacteriocines) which suppress growth of pathogenic microbes [1][2][3] . Any decrease in number of lactobacilli can result in disturbance of vaginal micro flora and subsequent development of bacterial vaginosis.
Composition of BV is complex. Molecular analysis has shown that BV is not monobacterial disorder, but can be caused by many microbes such as Gardnerella vaginalis, Prevotella spp, Atopobium spp, Mobiluncus spp, Sneathia sanguinegens 4 . Quite often BV can be asymptomatic, which can make this disorder insidious in regard that it can cause obstetric and gynaecological complication without warning. Some of the consequences of BV can be premature birth, or increased risk to encounter additional infection (Trichomonas vagnalis, Neisseria gonorrhoeae, Chlamydia trachomatis, HSV2 and HIV) [5][6][7] .
This disorder can be diagnosed under various criteria (clinical or microscopic) introduced by Amsel, Nugent, Ison/Hay and Claeys (the first two of four are widely accepted as "golden" standards in BV diagnosing, clinical and microscopic, respectively). Amsel method was mainly based on clinical findings and properties of vaginal discharge. According to Amsel criterion, patient is positive for BV when 3 of 4 criteria are present (vaginal pH above 4.5, "milk-like" white-grayish vaginal discharge, positive whiff test and clue cells on microscopic examination) 8 .
Nugent has been categorized the microscopic findings of Gram-stained vaginal smears by quantification of some of the present morphotypes, Lactobacillus, Gardenerella-Bacteroides, and Mobiluncus into: normal, intermediary and BV 9 . The Ison/Hay system is based on the observation of Gram stains to estimate the qualitative ratios of the observed morphotypes rather than the exact number of bacteria. In order to obtain more precise classification, two additional categories have been introduced to Ison/Hay criteria, group 0without bacteria, and group IVwith large amount of Gram-positive cocci 10 . Further modification by Claeys, using culture and molecular identification of vaginal microbiota, introduced even more categories, subdividing grade I samples in several additional categories: Ia, Ib Iab, I-like, I-PMN, in regard to relative concentration of Gram-positive rods (lactobacilli) and BV-associated bacterial morphotypes 11,12 .
Although widely used, all of these methods mentioned above, had certain insufficiencies. Nugent method categorizes the smears by quantification of bacterial morphotypes, which demands noticeable time and skill of an observer (experienced microbiologist) 9 . Additionally, the Nugentscoring system includes only three bacterial morphotypes and therefore it may not match the heterogeneity and complexity of the vaginal microflora. Albeit that Ison/Hay and Claeys had overcome some deficiencies of Nugent method by introducing qualitative assessment of vaginal smears, their method is still based on observation of small slide area (under 1000x magnification). Observing 5-20 fields of view under the 1000x magnification the actual scanned surface makes only a tiny fraction of the slide surface thus being a source of sampling error 13,14 .
In regard to overcome some insufficiencies of previously mentioned criteria: time consuming, a complicated numerical summing with narrow intervals, a need for experienced personnel, a demand for standardizing surface of the microscopic field of view, and evaluation of only three bacterial morphotypes, we established a novel method of microscopic examination of Gramstained vaginal smears based on qualitative examination of preparations under 200x magnification 15 . The categorization system of our method refers to six groups: three normal and three BV, which can make easier comparison of microscopic method and dichotomous clinical assessment of samples such as Amsel"s method. To test its value, we compared our method to already established Nugent"s method.

Study population and design
This prospective study comprised of 705 pregnant and asymptomatic women between 24 and 28 weeks of pregnancy, seen during a regularly planned appointments in Military Medical Academy hospital from 2012 to 2014. Patients younger than 18 and older than 40 years, patients with multiple pregnancies, anomalies of the uterus, cervical conization, or patients with previous preterm delivery were excluded from this study. Patients who were under any kind of therapy two weeks before examination, as well as patients who had sexual intercourse a week before appointment didn't enrol the study also. The institutional Ethical Board approved the study protocol and all study subjects agreed to participate through a written informed consent.

Sampling and data collection
The specimens were prepared under standard ethical and laboratory protocols. After clinical examination, vaginal samples were collected by inserting sterile dacron-tipped swab into vagina.
The swab was rolled round through 360 degrees against the vaginal wall at the mid portion of the vault and carefully withdrawn to prevent contamination. Swabs were then smeared on a plain glass slide and air-dried at room temperature. The slides were Gram stained and categorized according to Nugent (viewed under immersion, 1000x magnification), and novel method of microscopic examination (viewed under immersion, 200x magnification) which will be further denoted here as the criterion of Dane (15).

Analysis of data
Nugent scoring system implies categorization of Gram-stained smears into three groups in regard to morphotypes of bacteria under microscope 1000x magnification. Morphotypes are scored by their presence/absence as the average number seen per oil immersion field (5-20 fields) (9). For example, if more than 30 lactobacilli are recognized in the visual field, the score is 0; if no lactobacilli are detected, the score will be 4 points. If Gardnerella-like bacteria are absent, the score is 0; if more than 30 are observed, the score will be 4. The presence of other microorganisms, such as Mobiluncus, can add additional 2 points. According to final score, all findings are designated as: I-normal (0-3), II-intermediate (4)(5)(6) and III-bacterial vaginosis (7-10). Dane"s scoring system is based on the examination of Gram-stained vaginal smears under 200x magnification and their categorization depending on presence of either rod forms (RFs) or non-rod forms (NRFs) (15). The shortest length still observable as a rod at the 200x magnification is 1.5μm. Based on this fact, under 200x magnification, there is no obstacles to recognize predominance of either RFs (˃ 1.5μm, lactobacilli) or NRFs (< 1.5μm, bacterial vaginosis associated bacteria). Number of RFs and NRFs was estimated semi-quantitatively like this: numerous bacteria covering the most of slide surface between, around and over epithelial cells were labeled as "full"; bacterial forms rare or absent between, but found mostly around and on epithelial cells were designated as "mid"; and absence of bacterial forms with only rare elements seen around and on epithelial cells were termed as "empty". According to predominance of either RFs or NRFs, each of these three categories was additionally subdivided into normal (N) and bacterial vaginosis (BV) subgroup, respectively. In this way all slides were categorized into 6 groups, three normal: normal full -NF, normal mid -NM and normal null-NN and three bacterial vaginosis varieties: BV full -BVF, BV mid -BVM, and BV null -BVN.
For the purpose of the study, the Nugent"s score was taken to be the gold standard.
With aim to compare our results with Nugent as "golden" standard, we grouped our findings

Statistical analysis
Complete statistical analysis was conducted with commercially available statistical software SPSS v17.0. Variables were presented as frequencies of individual parameters (categories), and statistical significance of differences was evaluated using Chi-square test. Degree of agreement between categories (scale of measurement) was determined by kappa index. Sensitivity and specificity were calculated in an ordinary manner. Statistically significant difference was evaluated on minimal level p<0,05.

Results
On the basis of NMM and Nugent's scoring, bacterial vaginosis was diagnosed in 25% and 21% of women, respectively (figure 1). Normal finding was observed in 75% of women by NMM, and 63% by Nugent, while 16% of patients were classified as intermediate under Nugent. According to chi-square test, association has been found between NMM and Nugent's categorization (χ 2 =669,800; df=10; p<0.001). When we observed normal finding group, the best association has been found between groups with intermediary result by Nugent and NF (normal full by NMM, 96.0%), and between intermediary group by Nugent and NM (normal mid by NMM, 80%) ( Figure 2). Group with intermediary result by Nugent, was in significant association with BVN (bacterial vaginosis normal, 57%) and NN (normal null, 44%). BV group has shown the best association with BVF (bacterial vaginosis full, 99%) and BVM (bacterial vaginosis mid, 78%). It can be observed, that around half of patients from Nugent's intermediary group, was grouped as NMM's NULL groups (hipocellular: NN+BVN).
In the figure 3 it can be seen that intermediary group under NMM was formed by adding NN (normal null) to BVN (bacterial vaginosis null) considering that the majority of Nugent"s intermediary patients were contained within these groups (figure 2). It was shown that the best association (χ 2 =634,442; df=4; p<0.001) was found between patients with normal finding (91%) and those with bacterial vaginosis (89%) (Figure 3). On the other hand, the weakest association was observed in patients with intermediary result (49%).
When we observe results presented in the figure 4, the best association was found in groups with normal finding (82%) and groups categorized as bacterial vaginosis (74%) (χ 2 =437,40; df=2; p<0.001). 22% of Nugent's intermediary group was categorized as BV according to NMM, while 14% of Nugent's intermediar finding was clasified as NMM's normal group.
In Sensitivity, specificity, positive and negative predictive value of NMM, compared to Nugent"s score as standard, is given in table 2. Intermediary score, grade II, was considered either as positive, negative or excluded. In case when the intermediary score was considered negative/normal (table 2 a) the sensitivity and specificity of Dane"s criterion was lower, but still high, with high positive and negative predictive value. When IMD and BV samples were analysed as one group ( Thus the sensitivity and specificity of NMM were very high, 98.2% and 99.3%, respectively, as well as positive and negative predictive values (97,4% and 99,5%, respectively).

Discussion
The human vaginal microbiome is very important for health of women. It can be changed by hormonal status (it is not the same before puberty, during reproductive period or among menopausal women), certain sexual behaviour and it varies according to ethnical affiliation. Nevertheless, although BV may appear at any age, yet it is the most frequent in reproductive period. The most striking event in shifting of healthy vaginal environment towards BV is replace of dominant lactobacilli by mixture of mainly anaerobic bacteria such as: Gardnerella vaginalis, Atopobium vaginae and Prevotella spp [16][17][18] . Adequate diagnosis of BV is demanding and choice of right method for its diagnosis requires review of hardly explicable results such as intermediary results 19 . Although there are many criteria and scoring systems which are mutually comparable, it is not convincing that they will always classify the same category of patients. As is well known, demonstration of infectious agent existence is often a basic criterion in diagnosing of the infective disease. This is not the case with BV, since the real cause of the disorder is not yet defined. Thus the patient must meet clinical or laboratory criteria which do not consider the presence or quantity of a specific bacterium. It is important to keep this in mind when comparing different diagnostic methods. BV does not evolve from a commonly defined bacterial infection caused by one agent, but can rather be compared to consequences caused by anaerobic mixed flora in other parts of the organism. Diagnosis based on diagnostic criteria is actually the weighting of criteria to provide the best possible agreement between the criteria and the presence of BV. It is important that the examiner, whether a clinician or laboratory technician, is well trained and able to evaluate the clinical adequacy of the diverse methods available for BV diagnosis.
Clinically, BV is usually diagnosed by physical examination, pH of vaginal discharge, Whiff test and presence of clue cells which represents diagnostic system proposed by Amsel in the early 1980s 8 . The Nugent"s criterion is the method mostly used for diagnosing BV and it is considered to be golden standard among microscopic methods 13 . However, its score intervals are very narrow, differing in only a few bacteria, and the observed number of bacterial morphotypes may vary depending on the examiner. The homogeneity and thickness of the specimen may be influenced by the way of spreading the sample on the glass slide 14 . To avoid demanding counting of bacterial morphotypes, qualitative microscopic examination was introduced by Ison/Hay and Claeys 10,11 . These methods give advantage in saving observer"s time and more precise differentiation of lactobacillus morphotypes, but in the other hand they examine small microscopy field which can influence the results (because of unequally scattered smears over the slides).
In our institutions, clinical examinations, as well as, microscopy are in routine use in diagnosis of BV, but often there is neither sufficient time nor expertise available to practice the quantitative scoring systems. Therefore, the main goal of our study was to validate simpler grading schemes for microscopic diagnosis of BV, previously described by Dane et al (novel microscopy method) 15  However, regardless of the microscopic method used for diagnosis of BV, for an accurate diagnosis of the disease it is necessary to evaluate the clinical aspects and clinical adequacy of diverse methods available. Besides various methods currently used, clinicians still have difficulties to make decision about patients with BV which should be treated. What make this decision even more difficult are discrepancies in classification of intermediary findings. I has been shown that composition of intermediary flora is divided among lactobacilli and bacteria associated with BV which isthe main reason why intermediary "phase" is considered "transient phase" between healthy vaginal microbiome and BV 20,21 .
From our study, we could indirectly asume that the most of patients with Nugent intermediary finding actually belonged to the group with low number of bacterial forms.
Possible explanation for these "illogical" results are narrow intervals in Nugent categorization criterion. For example, in original Nugent criterion counting is performed on 5-10 visual fields under the magnification x1000, notified as interval on ordinary scale (in range from 0-1.000.000 bacteria per visual field). Evaluation of bacterial number in intervals is carried out assuming that number of bacteria from 1-30, counted on part of visual field, can be used for aproximate bacterial number estimation on entire visual field.
What was illogical in Nugent categorization is that patients with 4 or less bacterial forms were assigned as 0, 1 or 2 points, while patients with bacterial number above 4, and above 30 were assigned with 3 and 4 points. Therefore, 0 points will get only patient with finding of 0 bacterial number on 5 observed visual fields. Accordingly, if we imagine finding which is under Ison/Hay "clean" and if we didn"t find any of lactobacillus on 5 visual fields, patient would receive 4 points. If we observed, on latter patient case, G. vaginalis also, with average number of ˃4 per field, this finding should be assigned with 3 additional points (total 7 points) and categorized as BV. There is possibility in re-observation of aforementioned Gram stained smear, to count bacteria slightly different with average number of exactly 4 bacterial forms per visual field. In this case, patient will be assigned with 2 points instead 3 points, and categorized as intermediary group by Nugent. Taking all previously mentioned in consideration, we can conclude that cases with small bacterial number could be tricky for observer which is furthermore complicated by narrow diagnostic criteria. Given that low number of bacteria does not mean absence of disorder (inflammation and potential risk of miscarriage), we should pay more attention to these groups: "clean" under Ison/Hay, intermediary by Nugent, and NN and BVN under novel microscopy method. The reasons could be various, but we will try, on practical examples, to discuss some of them.
In the repeated observation of preparations, probability of analyzing the same 10 visual fields almost does not exist. As follows, during first examination, we can see small bacterial number and categorize patient as intermediary result. While we observe the same sample again, there is a possibility to assign preparation with 7 points (BV), which means: the less cellularity, the greater chance to misinterpret finding through repeated observation.
If we involve additional observers, likelihood of different preparation "reading" can become even higher. It is important to stress another yet observed rule from our study: homogeneity of preparation is proportional to cellularity. For example, according to novel microscopy method, the highest homogeneity is noted in patients with BVF and NF.
According to our investigation, probability to find areas BVN in BVF preparations was very low (under 5%), while the possibility to find BVN areas in BVM preparations was slightly higher (10-15%). However, in our opinion, Nugent has two crutial adventages in relation to other diagnostic criteria: first, as we said before, it is well established and widely used criterion because its simplicity (golden standard); and second, Nugent has intermediary group (compared to Amsel). We cannot diminish significance of intermediary group without an explanation.
In accordance with presented findings from our study, we have clearly shown that tested methods are reliable in diagnosing extreme categories, either BV negative or positive, but problem arises in the classification of intermediary group, the most difficult to interpret. As we have shown, the difference in percentage of IMD patients is significant (16% according to Nugent). Besides the fact that the largest number of disagreements was observed in IMD samples, kappa value was not low because differences occurred between successive categories, but not between extreme categories. Therefore, excluding IMD patients from our analysis we found nearly perfect agreement between tested criteria, with very high sensitivity, specificity, positive and negative predictive value. By shifting criteria, in order to conjoin IMD patients to normal or BV, both sensitivity and specificity have been decreased, confirming that IMD samples truly represent an intermediary status between normal and BV.
These findings suggest that Nugent"s method seems to recognize higher number of positive cases compared to other methods for scoring Gram stained samples. Also, other studies indicated that, compared to Amsel"s method, Nugent"s criterion can overestimate the real prevalence of BV and may even interpret healthy individuals to be diseased 22-24 . Nevertheless, measuring agreement between two sets of criteria usually take one set as working definition of the disease, but it is unable to determine superiority between them because of basic differences between these two methods. What further complicate problem of BV diagnosis is that Amsel"s method, mainly based on clinical findings, is dichotomous, having only two categories, whereas microscopic methods allow assessment of variation in vaginal microflora as a continuum and have three or more categories. Therefore, it is of great importance to provide the best possible agreement between clinical and/or laboratory criteria and the presence of BV. In line with that, to improve clinical adequacy of BV diagnosis revised set of criteria that combines clinical and microbiological parameters is needed.
In conclusion, novel microscopy method scoring system seems to constitute a good classification method, as it allows the microscopist to formulate an impression based on the relative numbers of RFs and NRFs morphotypes while the influence of surface area and bacterial density are lessened. Furthermore, novel microscopy method is flexible and can be reorganized in the way to categorize findings into only two groups: normal and BV, the fact that may have important clinical implications.     Table 2. Sensitivity, specificity, positive and negative predictive value of NMM (novel microscopy method), compared to Nugent"s score as standard, is given in table 2, in next order: a. when IMD group was added to normal Nugent results, b. when IMD group was added to BV group by Nugent, and c. when Nugent"s IMD group was excluded from data analysis.