Secondary Logo

The Use of Sequential Self-Obtained Vaginal Smears for Detecting Changes in the Vaginal Flora


Sexually Transmitted Diseases: April 1997 - Volume 24 - Issue 4 - p 236–239
Original Article

Background and Objectives: The ability to study daily changes in the vaginal flora may provide insight into the pathogenesis of bacterial vaginosis. Because culture of the vaginal fluid is tedious and expensive, the utility of self-obtained vaginal smears for documenting changes in the flora was evaluated.

Goals: To validate the adequacy of self-collected vaginal fluid Gram stains and use them to monitor vaginal flora.

Study Design: Ten asymptomatic premenopausal women collected daily vaginal smears for 30 days. The smears were Gram stained and interpreted using a standardized scoring system (Nugent criteria). In addition, results from self- and clinician-obtained vaginal smears from 18 women were compared to validate the adequacy of self-obtained smears.

Results: Two women had asymptomatic bacterial vaginosis. One woman, who was postpartum, had intermediate flora that toward the end of the collection period changed toLactobacilluspredominant. The remaining seven women exhibited two patterns. One was Lactobacillus morphotypes only; the second consisted ofLactobacillus-predominant days interspersed with days with moderate to high numbers ofGardnerella/Bacteroidesmorphotypes. There was a significant correlation of the point of change in the flora of this group with menses.

Conclusions: The adequacy of self-collected vaginal fluid Gram's stains was validated. Changes in vaginal flora were demonstrated over a 30-day period by use of this methodology.

*From the Department of Medicine, University of Alabama at Birmingham, and the Jefferson County Department of Health, Birmingham, Alabama

The authors thank Janis Adams, CRNP for her assistance with the clinical portion of the study.

Supported by the Sexually Transmitted Diseases Cooperative Research Center (NIH grant AI-94–16).

Reprint requests: Jane R. Schwebke, MD, 1900 University Boulevard, 229 Tinsley Harrison Tower, Birmingham, AL 35294–0006.

Received for publication June 13, 1996, revised September 24, 1996, and accepted September 26, 1996.

ALTHOUGH BACTERIAL VAGINOSIS (BV) is the most common cause of abnormal vaginal discharge,1 its etiology remains unknown. Various behavioral factors have been associated with the presence of BV, but the actual chain of microbiologic events that leads to a perturbation of the normal vaginal flora to cause BV remains a mystery. Unlike acquired infections, in which pathogenic microorganisms either are or are not present, some of the polymicrobial changes associated with BV suggest an “overgrowth” phenomenon, suggesting in turn that the development of the syndrome may be predicted by following changes in vaginal bacterial flora over time. Unfortunately, however, limited sequential vaginal culture data are available describing changes in the vaginal flora in women with or without BV. Accurate quantitation and identification of specific vaginal bacteria by culture is expensive, time consuming, and difficult. The Gram's stain of vaginal fluid has been previously shown to reflect accurately culture results for describing vaginal flora patterns.2 Thus, we decided to evaluate the utility of vaginal fluid Gram's stains to evaluate the bacterial composition of vaginal fluid on a daily basis among a group of 10 volunteers. If linked to behavioral data, this technique could be used to describe microbiologic changes occurring in the vagina and the relationship of these changes to external factors.

Back to Top | Article Outline


Premenopausal, nonpregnant female volunteers without vaginal symptoms were invited to participate in the study. All participants were asked to provide self-obtained vaginal specimens for 30 days. Participants were instructed to insert a sterile cotton swab into the vaginal vault and immediately to roll the specimen gently onto a clean glass slide. Slides were labeled with a pseudonym and the date of collection. Information regarding menses and oral contraceptive use was collected. Slides were Gram stained and interpreted by the Nugent method,3 which quantitates three bacterial morphotypes (large gram-positive rods consistent with Lactobacillus, small gram-negative or gram-variable rods consistent with Gardnerella and Bacteroides, and curved rods consistent with Mobiluncus). The amount of each morphotype is quantitated from 0 to 4+ and then translated into a grading system of 0 to 10, with scores of 0 to 3 classified as normal, 4 to 6 as intermediate, and 7 to 10 as BV. In this system, the greater the number of Gardnerella/Bacteroides or Mobiluncus present, the higher the score. Conversely, the greater the number of lactobacilli present, the lower the score. Slides were interpreted independently by two microbiologists, and results from a subset of the smears compared for interobserver variation.

To validate the adequacy of self-obtained vaginal swabs, 18 women attending the Jefferson County Department of Health Sexually Transmitted Diseases Clinic were asked voluntarily to provide a self-obtained vaginal swab before their examination, which was handed to the clinician who prepared the slide. During the examination, a second swab specimen for Gram's stain was collected by the clinician, and the results of these two smears compared. This study was approved by the Institutional Review Board at the University of Alabama at Birmingham.

Back to Top | Article Outline

Statistical Methods

The Spearman product moment correlation was calculated to compare the Nugent scores between the patient- and clinician-obtained specimens. The correlation coefficient provides a measure of the degree of agreement between the scores, with values close to 1 indicating a strong agreement. The associated P-value for the correlation coefficient tests whether this agreement is significantly different from zero. Likewise, correlation coefficients were calculated for the Nugent score as well as the individual scores for each morphotype to determine the agreement between the two microbiologists who evaluated the slides.

The nonparametric change-point test was used to detect whether there was a change in the observed distribution of Nugent scores over the 30-day period. This test was done for each of the five women with variable flora and for each of the two women with consistent flora. The z-score was calculated, and the associated P-value tests whether there was no change in the Nugent score over the 30-day period. In addition, an estimate of the point/day at which there is a change in the pattern over the 30-day period is provided in this analysis.

Back to Top | Article Outline


A subset of 25 slides was interpreted independently by the two microbiologists. There was a statistically significant correlation in Nugent scores between the two readers, indicating excellent agreement in the readings (r = 0.83, P < 0.001). Similarly, strong agreement was evident in the correlation of scores for each of the individual morphotypes—Lactobacillus (r = 0.78), Gardnerella/Bacteroides (r = 0.86), and Mobiluncus (r = 0.83).

Evaluation of the self- and clinician-obtained pairs of smears found the self-obtained smears to be as adequate for interpretation as those obtained by the nurse. Of 18 pairs, 9 had identical scores. Of the remaining pairs, seven had scores within 1 point of each other, whereas two had scores which differed by 2 points (Spearman's correlation coefficient r = 0.74, P < 0.001). Eleven of the pairs of smears obtained from this portion of the study were graded as BV, and five as intermediate. For two of the patients, one slide in the pair had an intermediate score, whereas the other slide had a score consistent with BV.

Slides for the sequential portion of the study were obtained from 10 women. Seven of the women had a Lactobacillus-predominant flora, whereas two of the women had flora consistent with BV. The remaining volunteer, who was postpartum, had an intermediate pattern of vaginal flora consisting of cocci only, which subsequently became Lactobacillus predominant during the study period presumably because of resumption of the menstrual cycle.

Among the seven women with Lactobacillus-predominant flora, two patterns were observed. The first pattern consisted of Lactobacillus morphotypes only and was seen in two women. The second pattern, found in the remaining five women, consisted of days in which Lactobacillus was the sole morphotype, interspersed with days in which Gardnerella and Bacteroides morphotypes appeared along with normal or reduced numbers of lactobacilli. In addition, gram-negative diplococci frequently accompanied these changes. Although these changes resulted in Nugent scores as high as 7 to 8 in three of the women, the longest such a change persisted was 2 days.

In all five of the women with this the variable pattern, the days in which the flora varied were concentrated around the time of menses (Figure 1), although there were sporadic marked variations that also occurred throughout the cycle of two subjects (Figure 2). For one of the women in the group with variable flora, the estimated change point over the 30-day period was on day 7, right at the end of this woman's menses. The corresponding test of significance resulted in a z-score of 3.68 (P < 0.05), indicating a statistically significant change in the pattern of bacterial morphotypes beginning on day 7 of the 30-day period. Three additional women in this group had similar statistically significant change points around the time of menses. One of these had two menses during the collection period and had change points evident with each of her cycles. The fifth woman in this group also had an estimated change in the pattern of her flora at the beginning of menses; however, the observed change did not achieve statistical significance. No change in the pattern of the Nugent score was observed for one of the women with consistent flora, whereas the pattern of the other woman in this group showed minimal variation, but this did not achieve statistical significance (z-score = 0.53, P > 0.05). Two women, one with each of these two patterns, repeated the collection of daily smears approximately 6 months after the initial collection. The vaginal flora patterns for these women matched those of their initial collections. The only women with curved rod morphotypes (Mobiluncus) were the volunteers with BV by Nugent criteria.

Fig. 1

Fig. 1

Fig. 2

Fig. 2

No information was available on douching or sexual activity among this group of women; however, the presence of spermatozoa on the slides did not appear to correlate with any changes in the vaginal flora pattern. Five women were using oral contraceptives, including both women in the Lactobacillus-only group and three women in the variable group.

Back to Top | Article Outline


Detailed microbiologic data on the composition of the normal vagina as well as in women with BV are available only in isolated, cross-sectional fashion.4–6 The actual mechanisms that control the composition of the vaginal flora are not well understood, with the exception of the effects of estrogen, which appears to promote a Lactobacillus flora.7,8 The factors leading to the development of BV remain unclear. Sequential vaginal cultures have been examined previously in a limited manner. Bartlett et al.9 analyzed vaginal cultures obtained at 3- to 5-day intervals during the menstrual cycle from five volunteers. They found a decrease in the concentrations of facultative gram-positive cocci and bacilli in all five women in the premenstrual week, whereas the concentration of anaerobic organisms remained constant. Sautter and Brown10 did a similar study with seven volunteers. Among their volunteers, lactobacilli or Gardnerella vaginalis were always the predominant organisms, and when lactobacilli were present, they were noted to be present throughout the cycle at a higher concentration then any other organism. All subjects had G. vaginalis recovered from their cultures at some time during the study. Also noted was a cyclic variation of Bacteroides spp., with most isolates present during the first half of the menstrual cycle. Others have also demonstrated increased variance of the vaginal flora by use of cultures before or during the time of menses.11,12

The vaginal fluid Gram's stain as interpreted by the Nugent method has been shown to be highly reproducible and to correlate well with the results of microbiologic culture.2,3,13 The reported use of this technique in a serial prospective manner is limited to a study of the vaginal flora patterns of pregnant women in which vaginal smears were obtained at 23 to 26 weeks' gestation and then again at 31 to 36 weeks.2 Of the women with normal vaginal smears at enrollment, 81% remained normal at follow-up, whereas 12% became intermediate and 7% developed BV. A similar shift in patterns was found for women with BV at enrollment. Women with intermediate flora had the greatest microbiologic shifts at follow-up, with only one third maintaining the intermediate pattern. These data suggest that the vaginal flora, at least in a subset of pregnant women, is dynamic.

The analysis of daily self-obtained vaginal fluid smears has been previously described,14,15 and our comparison of self-obtained versus clinician-obtained smears presented in this study further validates the adequacy of self-obtained specimens. The results from the sequential portion of our study demonstrate that this technique is able to detect day-to-day changes in the vaginal flora. Although our numbers are small, there appear to be two groups of “normal” patterns, one of which is homogeneous throughout the menstrual cycle, whereas the other exhibits variation consisting of the appearance of Gardnerella and Bacteroides morphotypes. Further, our results concerning the timing of changes are consistent with the patterns that have been described previously with the use of sequential cultures.9–12 The changes we observed were transient and did not lead to the development of BV within our period of observation.

The factors responsible for the changes in the flora are unknown. Because we did not collect data on behavioral practices, it is not possible to speculate concerning the role of douching or sexual intercourse. The occurrence of most of the microbiologic changes around the time of menses concurs with the findings of others that estrogen may exert important influences on the vaginal flora.7,8 However, not all of our subjects exhibited cyclic changes. A dramatic example of the probable influence of estrogen was noted in our subject whose postpartum state resolved during the study. Although BV developed in none of the women exhibiting these cyclic changes, it is interesting to speculate that this time of variation or instability of the vaginal flora may represent a “critical period” during which additional exogenous factors could tip the balance in favor of BV.

In summary, sequential self-obtained vaginal fluid Gram's stains are an accurate and efficient means of detecting changes in the vaginal flora. The use of this technique along with carefully collected behavioral data could provide valuable information concerning the pathogenesis of BV.

Back to Top | Article Outline


1. Rein MF, Holmes KK. “Non-specific vaginitis,” vulvovaginal candidiasis, and trichomoniasis: Clinical features, diagnosis and management. Curr Clin Top Infect Dis 1983; 4:281–315.
2. Hillier SL, Krohn MJ, Nugent RP, Gibbs RS. Characteristics of three vaginal flora patterns assessed by Gram stain among pregnant women. Am J Obstet Gynecol 1992; 166:938–944.
3. Nugent RP, Krohn MJ, Hillier SL. Reliability of diagnosing bacterial vaginosis is improved by a standardized method of Gram stain interpretation. J Clin Microbiol 1991; 29:297–301.
4. Paavonen J. Physiology and ecology of the vagina. Scand J Infect Dis 1983; S40:31–35.
5. Spiegel CA, Amsel R, Eschenbach D, Schoenknecht F, Holmes KK. Anaerobic bacteria in non-specific vaginitis. N Engl J Med 1980; 303:601–607.
6. Thomason JL, Gelbart SM, Scaglione NJ. Bacterial vaginosis: Current review with indication for asymptomatic therapy. Am J Obstet Gynecol 1991; 165:1210–1217.
7. Spiegel CA. Bacterial vaginosis. Clin Microbiol Rev 1991; 4:485–502.
8. Redondo-Lopez V, Cook RL, Sobel JD. Emerging role of lactobacilli in the control and maintenance of the vaginal bacterial microflora. Rev Infect Dis 1990; 12:856–872.
9. Bartlett JG, Onderdonk AB, Drude E, et al. Quantitative bacteriology of the vaginal flora. J Infect Dis 1977; 136:271–277.
10. Sautter RL, Brown WJ. Sequential vaginal cultures from normal young women. J Clin Microbiol 1980; 11:479–484.
11. Johnson SR, Petzold CR, Galask RP. Qualitative and quantitative changes of the vaginal microbial flora during the menstrual cycle. Am J Reprod Immunol Microbiol 1985; 9:1–5.
12. Brown WJ. Variations in the vaginal bacterial flora. Ann Intern Med 1982; 96:931–934.
13. Joesoef MR, Hillier SL, Josodiwondo S, Linnan M. Reproducibility of a scoring system for Gram stain diagnosis of bacterial vaginosis. J Clin Microbiol 1991; 29:1730–1731.
14. Mengel MB, Berg AO, Weaver CH, et al. The effectiveness of single-dose metronidazole therapy for patients and their partners with bacterial vaginosis. J Fam Pract 1989; 28:163–171.
15. Gwyther RE, Addison LA, Spottswood S, Bentz EJ, Evens S, Abrantes A. An innovative method for specimen auto-collection in the diagnosis of vaginitis. J Fam Pract 1986; 23:487–488.
© Copyright 1997 American Sexually Transmitted Diseases Association