Accuracy of Cervical/Vaginal Cytology in the Diagnosis of Bacterial Vaginosis


Sexually Transmitted Diseases:
Original Article

Background and Objectives:: The cytological pattern of bacterial vaginosis (BV) in the Papanicolaou smear has a controversial history (coccoid vs. Gardnerella type) that has not allowed an efficient use of cervical/vaginal cytology in the diagnosis of this condition. Our study is an attempt to clarify this topic.

Goals:: To evaluate the accuracy of cervical/vaginal cytology in the diagnosis of BV.

Study Design:: 1,896 separate examinations were performed on 1,471 women attending our sexually transmitted diseases center between 1990 and 1993. Amsel's composite clinical criteria were used as a gold standard. The Pap smear was prepared on two slides with the addition of a vaginal specimen.

Results:: In the 166 cases of BV, sensitivity is 88.7%, and specificity is 98.8%. Positive predictive value is 89.8, and negative predictive value is 98.7. Kendall's correlation coefficient is 0.88 (p < 0.001).

Conclusions:: Our findings support the validity of the vaginal Pap smear in diagnosing BV and suggest the screening of such a disease.

Author Information

From the *Department of Pathology and CNR‐Institute of Clinical Physiology, Santa Chiara Hospital, Pisa, Italy.

Reprint requests: Prof. Giuseppe Giacomini, Via dei Pecori 1, I‐50123 Firenze, Italy.

Received for publication May 15, 1997, revised August 21, 1997, and accepted August 29, 1997.

Article Outline

BACTERIAL vaginosis (BV), the most common form of vaginitis, is a complex condition where the ecosystem that guarantees vaginal antibiosis is completely altered.1 BV may create a predisposition for serious sequelae such as postoperative infections in both obstetrics and gynecology, pre‐term delivery, chorioamnionitis, urinary tract infections, endometritis, and pelvic inflammatory disease.2,3 The gold standard for BV diagnosis is considered to be the composite clinical criteria described by Amsel et al.4 At least three of the following four objective signs must be present: (1) a milky homogeneous vaginal discharge; (2) a vaginal pH of more than 4.5; (3) a fishy amine odor in the whiff test; and (4) clue cells on a saline wet mount. However, at times, these clinical signs are not fully evident; at other times the clinicians may require a permanent record of the disease, or perhaps a wet mount examination of the vaginal discharge is not feasible, so confirmation of BV diagnosis requires laboratory tests. There are indeed various choices,5 but Gram‐stained smears6 and Papanicolaou‐stained smears7–9 are generally the most routinely available. In this report, evaluation by Papanicolaou cervical/vaginal cytology was compared with evaluation by composite clinical criteria used to diagnose BV in 1,896 examinations performed on 1,471 women attending our sexually transmitted diseases (STD) center.

Back to Top | Article Outline

Materials and Methods

Study Population

Women attending the Sexually Transmitted Diseases Center, which is affiliated with the Diagnostic Cytology Section, Department of Pathology, Santa Chiara Hospital, Pisa, during 1990 to 1993 were considered in this study. Women, regardless of their age, showing vaginal atrophy and those treated with vaginal, oral, or parenteral antibiotics in the previous 15 days were excluded from this research. Women who had intercourse in the previous 3 days or with menses or spotting were asked to return later for examination. A total of 1,896 examinations were performed on 1,471 patients.

Back to Top | Article Outline
Clinical and Laboratory Evaluation

All patients were interviewed and examined by the same doctor, and results were recorded. An unlubricated speculum was inserted to display the vagina and the cervix under the colposcopy light. Firstly, the composite clinical criteria were investigated. Where three of the four clinical signs (discharge, pH, odor, and wet mount) were positive, a diagnosis of BV was made.

Discharge was examined for volume, color, consistency, and frothiness. A thin, homogeneous, and gray discharge suggested BV. A little fluid was taken with a spatula from the lateral wall of the vagina, and pH was measured by Special Indicator paper pH 4.0–7.0 (Macherey‐Nagel +Co, 516 Düren, Germany). Vaginal pH higher than 4.5 was considered abnormal. The wall was scraped again with a spatula, and the fluid was smeared on two glass slides. One drop of 10% potassium hydroxide was placed on one of two slides and immediately evaluated by the whiff test. Then a cover slip was placed over the same slide, which later was examined by phase contrast microscopy for Candida organisms. One drop of normal saline was placed on the second slide, and a cover slip was put over it: 20 fields of this last smear were examined by phase contrast. Where clue cells made up 20% or more of the total cells, the examination was considered wet mount positive for BV diagnosis. For cervical/vaginal cytology we used the two‐slide Pap test, which is a vaginal‐cervical‐endocervical (VCE) smear10 applied to two slides instead of one. The vaginal specimen is smeared crosswise at the extreme end of the first slide to enable examination side by side with the endocervical brushing placed on the rest of the first slide. The second slide is completely reserved for the ectocervical scraping specimen. The Pap smears were examined by a senior cytotechnician who had no knowledge of the clinical findings. The first 30 cases were also examined blind by two other cytotechnologists who reported BV.

Finally, colposcopy and standard pelvic examination were performed.

The cytological criteria for BV diagnosis have been previously mentioned.7 The most important is the presence of bacterial "sand" or "sludge", i.e., a background covering of tiny cocco‐rods broken here and there by open spaces around the cells (Figures 1 and 2). This background is clearly visible at even low magnification. This pattern and the old "coccoid bacterial type"11,12 coincide perfectly: bacterial vaginosis and coccoid bacteria are the same in vaginal cytology. This pattern is easily identified in the vaginal smear and a little less in the ectocervical scraping smear. Bacterial flora can be analyzed under powerful magnification, and the absence of lactobacilli and the eventual presence of commas and other forms of filaments and cocci are sought. Other signs are enhanced cellular desquamation and the occasional presence of anucleated squamous cells. Anucleated squamous cells are not always caused by hyperkeratosis as has been asserted13 but also to BV because of the toxic effect of the amines. Sometimes clue cells may be detected in the stained smear, but, as a rule, they are more frequent in the wet mount according to Gardner.14

Back to Top | Article Outline
Statistical Analyses

Sensitivity, specificity, positive and negative predictive values, and relative 95% confidence intervais (CI) of the Pap test were calculated in comparison to composite clinical criteria. Kendall's tau coefficient was used to correlate the Pap test to the reference diagnosis. The K coefficient and its standard error (SE(K)) were computed to evaluate agreement, corrected for chance, between the Pap test and each one of the four clinical signs.

Back to Top | Article Outline


Of the total 1,896 examinations performed, 318 were excluded because composite clinical criteria were incomplete (9 had cytological diagnosis for BV). Of the remaining 1,578 valid examinations, 166 were diagnosed as BV. Table 1 compares diagnosis of BV using composite clinical criteria and cytological diagnosis of BV. Cytological diagnosis of BV had a specificity of 98.8 (CI 98.0–99.3), a sensitivity of 88.7 (CI 82.7–92.9), a positive predictive value of 89.8 (C.I. 83.9–93.7), and a negative predictive value of 98.7 (CI 97.9–99.2). Kendall's correlation coefficient was 0.88 (p < 0.001). Comparison of the Pap test and single clinical criteria is shown in Table 2. Best agreement is seen between the Pap test and the wet mount diagnosis, whereas there is poor agreement between pH and the Pap test. There is a high number of abnormal pH values in non‐BV patients, but most of these cases had pH value of 4.6 (556 over 869, about 64%). In fact pH 4.6 is a borderline value, and Eschenbach et al2 considered pH 4.7 as the cutoff point for BV diagnosis. The cases of more abnormal pH are because of candidiasis, trichomoniasis, contaminations from endocervical secretions in women with large eversion, or nonreported douching.

Back to Top | Article Outline


The normal vaginal flora is characterized by hydrogenperoxide producing Lactobacillus. In bacterial vaginosis, lactobacilli are replaced by a massive overgrowth of catalase negative organisms, such as Gardnerella vaginalis, Bacteroides spp, Peptostreptococcus spp, Mobiluncus spp and Mycoplasma hominis.5 It is believed that Gardnerella and anaerobes generate amino acids and decarboxilases that produce volatile malodorous amines from amino acids in the presence of elevated pH. This fishy odor is stronger with the addition of a drop of 10% potassium hydroxide.15 The presence of amines may also increase vaginal epithelial shedding4,16 and inhibits leukocyte chemotaxis.

The most popular laboratory support for BV diagnosis is the Gram‐stained smear.6 However, our results show that also cytological examination is highly specific (98.8%), similar to that reported by Platz‐Christensen et al in 1989.8 In our study, the clinical criteria were standardized by using one observer.

Eschenbach17 has reviewed the history of BV, dividing it into phases. We have done something similar for cervical/vaginal cytology of BV. The first phase began in 1953 when Wied and Christiansen18 accurately described a pattern they called "Streptococcen‐Typ." They observed abundant background film of coccobacillary flora, broken around the cells, and described the total absence of lactobacilli. A year later Wied renamed this pattern "kokkentyp."19 It was a distinctive, unmistakable pattern that in American literature was called "coccoid bacteria." Papanicolaou,11 in 1958, agreed with the name coccoid bacteria; it was reported and illustrated in books published at that time as one of the patterns most characteristic of cervical/vaginal cytology.12 Later, one year after the basic paper by Gardner and Duke,20 Pundel noted that the "streptococcen" smear of Wied and Christiansen was in fact mostly Haemophilus vaginalis.21 The same cytological pattern was seen in women harboring heavy infection with Mycoplasma hominis22 and was known for some time as "dirty smear."

In the second phase Wied and Bibbo23 developed a microbiologic classification for cell samples that made a distinction between Haemophilus vaginalis (presence of clue cells) and coccoid bacteria. The classification of Wied and Bibbo also comprised "mixed bacteria," where a combination of lactobacilli and coccoid bacteria was found. This recalls Schroder's 1921 theory24 of an intermediate stage between normal flora and that characteristic of what we now call BV.

In the present phase, the third, some authors7,25 have suggested grouping coccoid bacteria and Gardnerella vaginalis together because the massive overgrowth of coccobacillary flora and the presence of clue cells are two aspects of the same condition, now called BV. The Bethesda System (TBS) for reporting cervical/vaginal cytological diagnoses developed in United States in 198826 adopted this suggestion using the single expression "microorganisms morphologically consistent with Gardnerella species." Giacomini and Schnadig27 objected that BV is a polymicrobial syndrome caused not only by Gardnerella. They proposed naming the cytological pattern "microorganisms morphologically consistent with the bacteria of bacterial vaginosis." The revised 1991 TBS made a further change, replacing its previous expression with "predominance of coccobacilli consistent with shift in vaginal flora."28 This suggested that the condition represents a continuum of changes, with a lactobacilli domination at one end and BV at the other, so that seemingly only the composite clinical criteria or examination of Gram‐stained smears could give reliable diagnosis of BV.

Recently Schwebke et al considered the use of Gram's stain as the criterion standard for the diagnosis of BV.29 Nevertheless we believe that the cytological pattern of BV is typical and unmistakable; it was one of the earliest described in vaginal cytology and as a rule is associated with positive clinical signs. According to our experience, the intermedial stage in vaginal flora does not indicate BV because massive overgrowth of coccobacilli is lacking and clinical signs are negative.

The added advantage of using the Pap test to diagnose BV is that other cervical/vaginal pathologies can be identified at the same time. Moreover, Pap smears are the most widely used tests for women and the most frequent microscopic examinations of vaginal discharge. Full recognition of the cytological patterns connected with BV would also improve evaluation of Pap smears by cytotechnicians.

The Pap smear reading for BV in pregnancy could identify women in which treatment is considered. It soon will be standard care to screen for and treat BV in all women at high risk for preterm delivery.30 BV is related to preterm low birth weight.3 The Pap smear is usually performed in pregnancy, and if performed, would mean that pregnant women would not need rescreening for BV by another method. This would save time and expense. In the very likely possibility that women at low risk for preterm delivery are found to benefit from treatment of BV, all pregnant women would need to be screened for BV and the issue of cost‐effective screening then becomes enormous.

Back to Top | Article Outline


1. Mårdh P-A. The vaginal ecosystem. Am J Obstet Gynecol 1991;165:1163-1168.
2. Eschenbach DA, Hillier S, Critchlow C, et al. Diagnosis and clinical manifestations of bacterial vaginosis. Am J Obstet Gynecol 1988; 158:819-828.
3. Hillier SL, Nugent RP, Eschenbach DA, et al. Association between bacterial vaginosis and preterm delivery of a low-birth-weight infant. N Engl J Med 1995; 333:1737-1742.
4. Amsel R, Totten PA, Spiegel CA, et al. Nonspecific vaginitis. Diagnostic criteria and microbial and epidemiologic associations. Am J Med 1983; 74:14-22.
5. Hillier S, Holmes KK. Bacterial vaginosis. In: Holmes KK, Mårdh P-A, Sparling PF, et al., eds. Sexually Transmitted Diseases. New York: McGraw Hill, 1990:547-559.
6. Spiegel CA, Amsel R, Holmes KK. Diagnosis of bacterial vaginosis by direct Gram stain of vaginal fluid. J Clin Microbiol 1983; 18:170-177.
7. Giacomini G, Reali D, Vita D, Barahmandpour S, D'Acunto A. The diagnostic cytology of nonspecific vaginitis. Diagn Cytopathol 1987; 3:198-205.
8. Platz-Christensen J-J, Larsson P-G, Sundström E, Bondeson L. Detection of bacterial vaginosis in Papanicolaou smears. Am J Obstet Gynecol 1989; 160:132-133.
9. Schnadig VJ, Davie KD, Shafer SK, et al. The cytologist and bacterioses of the vaginal-ectocervical area. Clues, commas and confusion. Acta Cytol 1989; 33:287-297.
10. Wied GL, Bahr GF. Vaginal, cervical and endocervical cytologic smears on a single slide. Obstet Gynecol 1959; 14:362-367.
11. Ayre JE, Berger J, Graham RM, et al. Opinion poll on cytological terminology. Acta Cytol 1958; 2:112-114.
12. Nieburg HE. Diagnostic Cell Pathology in Tissue and Smears. New York: Grune & Stratton, 1967:88.
13. Kern SB. Significance of anucleated squames in Papanicolaou-stained cervicovaginal smears. Acta Cytol 1991; 35:89-93.
14. Gardner HL. Hemophilus vaginalis vaginitis (Corynebacterium vaginale vaginitis). In: Gardner HL, Kaufman RH, eds. Benign Diseases of the Vulva and Vagina. Boston: Hall Medical Publishers, 1981:273-299.
15. Pheifer TA, Forsyth PS, Durfee MA, Pollock HM, Holmes KK. Nonspecific vaginitis. Role of Haemophilus vaginalis and treatment with metronidazole. N Engl J Med 1978; 298:1429-1434.
16. Chen KCS, Forsyth PS, Buchanan TM, Holmes KK. Amine content of vaginal fluid from untreated and treated patients with nonspecific vaginitis. J Clin Invest 1979; 63:828-839.
17. Eschenbach DA. History and review of bacterial vaginosis. Am J Obstet Gynecol 1993; 169:441-445.
18. Wied GL, Christiansen W. Bedeutung und einfluss der bakterienflora im zytologischen vaginalausstrich. Zbl f Bakt Parasitenk 1953; 160:413-424.
19. Wied GL. Zytologie der gravidität und der menopause. Bakteriell bedingte veränderungen im zytologischen ausstrich. Runge H, ed. Gynäkologische Zytologie. Dresden und Leipzig: Verlag Von Theodor Steinkopff, 1954:24-43.
20. Gardner HL, Dukes CD. Haemophilus vaginalis vaginitis. A newly defined specific infection previously classified "nonspecific" vaginitis. Am J Obstet Gynecol 1955; 69:962-976.
21. Pundel JP. Précis de Colpocytologie Hormonale. Paris: Masson, 1996:164-168.
22. Jones DM, Davson J. Mycoplasma hominis in Ayre's smears. Nature 1967; 213:828-829.
23. Wied GL, Bibbo M. Question VI: Microbiologic classification of the cellular sample. In: Zuspan FP (Coordinator), Bibbo M, Gardner HL, et al. Management of Patients With Vaginal Infection: An Invitational Symposium. J Reprod Med 1972; 9:1-16.
24. Schroder R. Zur pathogenese und klinik des vaginalem fluors. Zentralb Gynäkol 1921; 38:1350-1361.
25. Giacomini G, Paavonen J, Rilke F. Microbiologic classification of cervicovaginal flora in Papanicolaou smears. Acta Cytol 1989; 33:277-278.
26. National Cancer Institute Workshop. The 1988 Bethesda system for reporting cervical/vaginal cytological diagnoses. JAMA 1989; 262:931-934.
27. Giacomini G, Schnadig VJ. The cervical Papanicolaou smear: Bacterial infection and the Bethesda system. Acta Cytol 1992; 36:109-110.
28. National Cancer Institute Committees. The Bethesda system for reporting cervical/vaginal cytologic diagnoses. Acta Cytol 1993; 37:115-124.
29. Schwebke JR, Hillier SL, Sobel JD, McGregor JA, Sweet RL. Validity of the vaginal Gram stain for the diagnosis of bacterial vaginosis. Obstet Gynecol 1996; 88:573-576.
30. Hauth JC, Goldenberg RL, Andrews WW, DuBard MB, Copper RL. Reduced incidence of preterm delivery with metronidazole and erytromycin in women with bacterial vaginosis. N Engl J Med 1995; 333:1732-1736.
© Copyright 1998 American Sexually Transmitted Diseases Association