American Journal of Obstetrics and Gynecology (2005) 193, 746–51

www.ajog.org

Variation in microbiologic profiles among pregnantwomen with bacterial vaginosis

Leonardo Pereira, MD,a,* Jennifer Culhane, PhD, MPH,b Kelly McCollum, MPH,b

Kathy Agnew, BS,c Paul Nyirjesy, MDb

Division of Maternal-Fetal Medicine, Oregon Health and Science University, Portland, ORa; Departments ofObstetrics and Gynecology, Drexel University, Philadelphia, PAb; and University of Washington, Seattle, WAc

Received for publication September 27, 2004; revised January 26, 2005; accepted January 26, 2005

KEY WORDSBacterial vaginosisMobiluncus

Pregnancy

Objective: The purpose of this study was to determine if clinical findings and sociodemographic

variables among bacterial vaginosis (BV)-positive pregnant women are associated with differentmicrobiologic profiles.Study design: Pregnant women were assessed for BV by Nugent criteria. BVC women were

separated into 6 mutually exclusive microbiologic groups. In unadjusted analyses, we compared(1) sociodemographic and behavioral characteristics, and (2) 3 clinical characteristics amongBVC women with and without Mobiluncus (MC vs M�). Unadjusted data were analyzed using

the chi-square test. Multiple logistic regression was used to assess the likelihood of having clinicalsigns of BV in women with and without Mobiluncus spp while controlling for confounders.Results: A total of 1756 BVC pregnant women were followed. The MC group (n = 702) wassignificantly more likely than the M� group (n = 1054) to be non-Hispanic black (80.9% vs

66.2%; P ! .0001), older than 21 years (61.7% vs 48.7%; P ! .0001), and to have had more than3 lifetime sexual partners (66.4% vs 54.9%; P ! .0001). The MC group was also more likely tohave clue cells on wet mount (63.9% vs 47.2%; P ! .0001) and a positive amine odor after

addition of KOH (57.2% vs 45.0%; P = .001). There was no difference in other demographicvariables or physician diagnosis of abnormal vaginal discharge. In the adjusted analyses for eachclinical outcome, all findings were consistent with the unadjusted analyses.

Conclusion: BVC pregnant women withMobiluncus spp are more likely to have clue cells presenton wet mount, a positive amine odor after KOH preparation, and to be older, non-Hispanicblack, and have had more lifetime sexual partners compared to BVC women without any

Mobiluncus species.� 2005 Elsevier Inc. All rights reserved.

Supported by Center for Disease Control and Prevention/ATPM,

grant nos. TS-0561/TS-0626.

* Reprint requests: Leonardo Pereira, MD, Division of Maternal-

Fetal Medicine, Mail Code L-458, Oregon Health and Science

University, 3181 SW Sam Jackson Park Road, Portland, OR 97239-

3098.

E-mail: pereiral@ohsu.edu

0002-9378/$ - see front matter � 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.ajog.2005.01.069

Bacterial vaginosis (BV) is a condition in which thenormal lactobacillary flora are replaced by an over-growth of Gardnerella vaginalis, Prevotella/Bacteroidesspecies, Mobiluncus species, and other organisms.1-3 Theassociation between BV and preterm birth (PTB) hasbeen reported by several authors.4-8 However, studiesaimed at prevention of PTB by treatment of BV have

Pereira et al 747

yielded differing results.9-10 This discrepancy has led theAmerican College of Obstetricians to take the positionthat ‘‘current data do not support the use of BVscreening as a strategy to identify or prevent pretermbirth.’’11

As our clinical recognition of BV has become stan-dardized by adopting Nugent criteria,12 we can nowappreciate that BV is not the same disease in all women.Microbiologic variation exists among BV-positive indi-viduals and can be characterized by Gram-stain analysis.All womenwith aNugent score of 7 to 10 are diagnosed asBV positive, but among these individuals, there is a widevariety of microflora. The purpose of this study was todetermine if the microbiologic variation among BVpositive individuals could explain variations in clinicalsigns of BVanddifferences in the prevalence ofBV amongdifferent race/ethnic groups.

Material and methods

This study is part of a larger, ongoing clinical prevalencestudy assessing the association between maternal stress,BV, and risk of preterm birth. Pregnant women wererecruited from public health centers in Philadelphia,Pennsylvania, between February 1, 1999 and April 20,2003. Women were recruited for participation in thisstudy at the time of their first prenatal care visit, ata mean gestational age of 14 G 6 weeks. All English orSpanish speaking women with singleton pregnancieswere eligible.

Practitioners collected 2 vaginal smears from allstudy participants, one for Gram-stain and one for wetmount. Practitioners also assessed the following clinicalsigns: the presence of clue cells on wet mount, amineodor after addition of potassium hydroxide (KOH), andvaginal discharge.

Air-dried vaginal smears were assessed for BV accord-ing to Nugent criteria.12 Each Gram stain was evaluatedfor the following morphotypes under oil immersion(100!) magnification: large uniform gram-positive rods(corresponding to Lactobacillus spp), small pleomorphicGram-variable rods (G vaginalis morphotypes), smallGram-negative rods (Prevotella/Bacteroides morpho-types), and curved Gram-variable rods (Mobiluncusmorphotypes). After scanning the slide for representativeareas, each morphotype was quantified from 1C to 4Cwith regard to the average number of morphotypes seenon 5 nonadjacent fields, using the scoring system listed inTables I and II. Each morphotype was given a numberand added to yield a score of 0 to 10 using the followingequation: LactobacillusC G vaginalis/Prevotella/Bacter-oides C Mobiluncus = total score. In this scoring system,Lactobacillus is inversely scored, compared toGvaginalis/Prevotella/Bacteroides and Mobiluncus. A score of 0 to 3was considered normal, a score of 4 to 6 corresponded to

intermediate BV status, and a score of 7 to 10 was definedas positive BV status.

Based on the individual components of this classifi-cation scheme, all smears were categorized into 1 of 14mutually exclusive microbiologic groups. Based on themicrobiologic distribution, it was impossible for 8 of thegroups to achieve a Nugent score above 6. Therefore, allBV-positive women (Nugent score 7-10) were classifiedinto one of 6 mutually exclusive microbiologic groups.Three microbiologic groups contained Mobiluncus spe-cies (MC), and 3 did not (M�).

Gram-stain scoring was performed by a microbiolo-gist (KA) who was blinded to the clinical findings anddemographics of subjects. The investigator (LP) whodesignated individuals into respective microbiologicgroups was similarly blinded. The following outcomeswere compared between BV C, MC women and BV C,M� women: clue cells identified on wet mount, positiveamine odor after addition of KOH, and physiciandiagnosis of vaginal discharge. The same comparisonswere carried out for BV intermediate women.

The statistical significance of all categorical data wasdetermined using the chi-square test. To adjust forpotential confounding variables, and to derive maximumlikelihood estimates of adjusted relative odds with 95%

Table I Gram-stain scoring system

Score No. of morphotypes/100! oil immersion field

0 None1C Less than 1 per field2C 1 to 4 per field3C 5 to 30 per field4C More than 30 per field

Table II Points* awarded by Gram-stain score

Morphotype Points

No Lactobacillus spp. 41C Lactobacillus spp. 32C Lactobacillus spp. 23C Lactobacillus spp. 14C Lactobacillus spp. 0No G vaginalis or Prevotella/Bacteroides spp. 01C G vaginalis or Prevotella/Bacteroides spp. 12C G vaginalis or Prevotella/Bacteroides spp. 23C G vaginalis or Prevotella/Bacteroides spp. 34C G vaginalis or Prevotella/Bacteroides spp. 41 to 2C Mobiluncus spp. 13 to 4C Mobiluncus spp. 2

* Points were totaled using the following equation: (Lactobacil-

lus C G vaginalis or Prevotella/Bacteroides C Mobiluncus = total

score). Total scores: 0 to 3 normal, 4 to 6 intermediate flora, 7 to

10 BV positive.

748 Pereira et al

Figure 1 *No Lactobacillus species were detected in all 6 groups.

CI, multivariate logistic regression analyses were per-formed. Potential confounders were identified a priori byreviewing the literature and based on theoretical consid-erations. The following confounders were included in allof the multivariate models: race, maternal age, age atsexual debut, number of lifetime sexual partners, recentsmoking, douching, parity, marital status, education,maternal annual income, and foreign-born status. TypeI error was set at 0.05 (two-sided). All analyses wereconducted using STATA 8.0 (College Station, Tex).

Results

A total of 4361 women met inclusion criteria and wereeligible to participate in this study. Of these, 3881(89.0%) consented to participate, and 3765 (86.3%)completed the study. There were 1756 (46.6%) BVCwomen, and 553 (14.7%) BV-intermediate women. Thelargest subgroup of BVC women (57.8%) demonstrated

G vaginalis and Prevotella/Bacteroides on Gram stain,and an absence of Lactobacillus and Mobiluncus species.The second largest subgroup (39.7%) had the sameprofile, except that Mobiluncus species were present onGram stain. The other 4 possible classifications wererarely identified, accounting for just 2.5% of subjects.Because of this distribution, women with Mobiluncusspp (40.0% combined 3 groups, MC) were comparedto women without Mobiluncus spp (60.0% combined3 groups, M�) (Figure 1).

In unadjusted analyses, BVC women with Mobilun-cus were significantly more likely to be non-Hispanicblack (80.9% vs 66.2%; c(df)

2 = 45.2(1); P ! .0001), tobe above the median age of 21 (61.7% vs 48.7%;c(df)

2 = 28.7(1); P ! .0001), and to have had morethan 3 lifetime sexual partners (66.4% vs 54.9%;c(df)

2 = 22.9(1); P ! .0001) compared with BVCwomen without Mobiluncus. However, the presence ofMobiluncus spp was not associated with age at sexual debut,recent smoking, or douching (Figure 2). Furthermore,

Pereira et al 749

Figure 2 BV, bacterial vaginosis; MC, Mobiluncus present; M�, Mobiluncus absent; yo, years of age.

in adjusted analyses, women with Mobiluncus spp weremore likely to have clue cells on wet mount (63.9% vs47.2%; c(df)

2 = 23.2(1); P ! .0001) and a positive amineodor after addition of potassium hydroxide (57.2% vs45.0%; c(df)

2 = 11.6(1); P = .002). However, there wasno difference in the physician finding of an abnormalvaginal discharge (64.6% vs 61.5%; c(df)

2 = 1.18(1);P = .28). In the adjusted analysis for each clinicaloutcome, all findings were consistent with those fromthe unadjusted analyses. MC women were 1.7 timesmore likely to have clue cells present on wet mount(95%CI: 1.3–2.3), and 1.5 times more likely to haveamine odor after the addition of KOH to the wet mountcompared with M� women. In addition, there was nodifference between the 2 groups with regard to physicianfinding of abnormal vaginal discharge (OR 1.1; 95% CI0.8–1.4) (Table III).

When we conducted this same analysis in BV-interme-diate women, all of the demographic distributions andclinical outcomes followed identical trends when compar-isons were made between MC (n = 35) and M�(n = 518) BV-intermediate women (data not shown).However, none of the findings were statistically signifi-cant.

This study had sufficient power (O80%) to detecta minimum odds ratio of 1.5 for the association betweenthe presence of Mobiluncus spp and each of the 3 clinical

outcomes. Nonetheless, we did not find an associationbetween the presence of Mobiluncus and physiciandiagnosis of abnormal vaginal discharge (Table III),suggesting that this null finding is not likely to be theresult of a type II error.

Comment

In the past 20 years of obstetric research, few topics havebeen as widely studied as BV. A MEDLINE searchusing the keywords ‘‘bacterial vaginosis,’’ and ‘‘preg-nancy’’ produced more than 500 corresponding articlessince 1984. In fact, the association between BV andpreterm birth (PTB) has been so consistently reportedthat there is little doubt in today’s medical communitythat a true relationship exists between them.4-8 Despitethis body of evidence, however, an understanding of theexact nature of the relationship between BV and PTBremains elusive.

Strong association does not equal causation, andperhaps this explains, at least in part, why treatmenttrials aimed at eradication of BV to decrease the rate ofPTB have yielded differing results.9-10 Another possibil-ity is that although the relationship with BV and PTBmay in fact be causative, it may be causative only forcertain individuals. This may be due to differences inhost response to BV, confounding risk factors for PTB,

750 Pereira et al

Table III Multivariate logistic regression of the association between the presence of Mobiluncus species in BV-positive women and3 clinical signs*

Dependent variableMobiluncus speciespresenty OR (95% CI)

Mobiluncus speciespresent (%)

Mobiluncus speciesabsent (%) P value

Clue cells on wet mount 1.7 (1.3, 2.3) 63.9 47.2 ! .00001Amine odor after KOH addition 1.5 (1.1, 2.0) 57.2 45.0 .002Physician diagnosis of abnormal vaginal discharge 1.1 (0.8, 1.4) 64.6 61.5 .28

* All models adjusted for the following variables: race, maternal age, number of lifetime sexual partners, recent smoking, douching, parity, marital

status, education, maternal annual income, and foreign-born status.y Mobiluncus species negative women are the reference group for all models.

or variations in the profile of pathogens constituting BVamong different women.

The importance of the Nugent scoring system inproviding standardized diagnostic criteria for BV cannotbe overstated. However, the diagnosis of BV by Nugentscore alone is limited because it clusters women withdifferent vaginal flora into one group of BV-positiveindividuals. The limitations of Nugent criteria havealready been recognized by some authors who havereported subsets of women with intermediate BV scoreswho appear to be at similar risk of preterm birth as theirBV-positive counterparts.13-15 By further refining ourdiagnostic criteria for BV, we may be able to betteridentify which individuals are at highest risk for pretermbirth.

We were able to identify 6 different morphotypesamong women with a Nugent score of 7 to 10 by Gram-stain analysis. We observed that BV-positive pregnantwomen with any Mobiluncus spp were significantly morelikely to be non-Hispanic black, above age 21, to havehad more than 3 lifetime sexual partners, to have cluecells on wet mount, and to have a positive amine odorafter KOH preparation compared to those withoutMobiluncus spp (Figure 2 and Table III). The associa-tion between Mobiluncus spp and a positive amine odorafter KOH preparation (‘‘whiff test’’) has been pre-viously reported,16 and our findings in this studyconfirmed the hypothesis that variations in the micro-biologic profile of BV are associated with differentdemographic variables and clinical signs.

An important limitation of our study is the relianceon Gram-stain analysis to differentiate the different BVmorphotypes. Ideally, this should be done on the basisof culture results because Gram-stain interpretation,specifically differentiating Gram-variable, pleiomorphicbacillus (G vaginalis) from more uniform Gram-negativebacillus (Prevotella/Bacteroides spp) can be difficult andintroduces a subjective element.12 However, since ourresults focus primarily on the presence or absence ofMobiluncus spp, which can be identified with excellentreliability on Gram stain,12 this should not affect thevalidity of our observations. Furthermore, our observa-tion that a significantly higher proportion of BVC blackwomen haveMobiluncus spp compared with BVC white

women is consistent with the findings of previousauthors.

Royce et al previously reported significant race/ethnicvariation in bacterial morphology among BV-positivepregnant women, with black women 9.26 times as likelyto have Mobiluncus compared with their white counter-parts.17 In a nonpregnant population of women seekingtreatment at an STD clinic, Hillier et al also reportedrace/ethnic differences in the presence of Mobiluncus sppidentified by Gram stain, with black women 2.5 times aslikely to be MC compared with white patients.18 Whyblack women exhibit more colonization of Mobiluncus isnot known. To date, sexual transmission does not seemto be the primary method of acquisition of Mobiluncusspp, and a gastrointestinal reservoir has been sug-gested.19 It is possible that race/ethnic differences inhost factors that modulate bacterial binding to vaginaland gastrointestinal cells may account for the observedrace/ethnic differences in Mobiluncus carriage rates. Forexample, Lewis antigens are thought to inhibit thebinding of infectious organisms to human cells. Lewisantigen deficiency has been associated with increasedrate of urinary tract infections in women,20 recurrentvaginal candidiasis,21,22 and Helicobacter pylori infec-tions.23 Some data suggest that blacks are more likely tobe Lewis antigen-negative, and differences in Lewisantigen status or a similar host factor could potentiallyexplain the race/ethnic variations in Mobiluncus coloni-zation rates.24

In conclusion, different BV morphotypes exist amongindividuals. Demographic variations (age, race/ethnic-ity, etc) are observed when comparing BVC women byMobiluncus spp status. Microscopy also demonstratesthat Mobiluncus spp is associated with cellular variation(more clue cells) and diagnostic variation (a positiveamine odor after KOH preparation). However, thisdistribution of BV morphotypes does not seem toexplain why some BV-positive individuals are diagnosedwith abnormal vaginal discharge while others are not.Differential expression of Lewis antigens may providea genetic mechanism for the increased prevalence ofMobiluncus among BV-positive non-Hispanic blacksand warrants further study. In the future, we plan toinvestigate whether the BV morphotypes characterized

Pereira et al 751

in this study are associated with different vaginal levelsof hydrolytic enzymes and proinflammatory cytokinesand, ultimately, with different risks of preterm birth. Asour understanding of BV evolves, it is important torecognize that microbiologic and clinical variation existsamong BV-positive women, and our thinking of BV-positive women as a homogeneous population harbor-ing the same disease must evolve as well.

References

1. Eschenbach DA, Gravett MG, Chen KC, Hoyme UB, Holmes

KK. Bacterial vaginosis during pregnancy. An association with

prematurity and postpartum complications. Scand J Nephrol

(Suppl) 1984;86:213-22.

2. Hillier SL, Krohn MA, Rabe LK, Klebanoff SJ, Eschenbach DA.

The normal vaginal flora, H2O2-producing lactobacilli, and

bacterial vaginosis in pregnant women. Clin Infect Dis 1993;

16(Suppl 4):S273-81.

3. Holst E, Goffeng AR, Andersch B. Bacterial vaginosis and vaginal

microorganisms in idiopathic premature labor and association

with pregnancy outcome. J Clin Microbiol 1994;32:176-86.

4. Hillier SL, Nugent RP, Eschenbach DA, Krohn MA, Gibbs RS,

Martin DH, et al. Association between bacterial vaginosis and

preterm delivery of a low-birth-weight infant. The Vaginal

Infections and Prematurity Study Group. N Engl J Med 1995;

333:1737-42.

5. Gravett MG, Hummel D, Eschenbach DA, Holmes KK. Preterm

labor associated with subclinical amniotic infection and with

bacterial vaginosis. Obstet Gynecol 1986;67:229-37.

6. Goldenberg RL, Hauth JC, Andrews WW. Intrauterine infection

and preterm delivery. N Eng J Med 2000;342:1500-7.

7. McGregor JA, French JI, Seo K. Premature rupture of membranes

and bacterial vaginosis. Am J Obstet Gynecol 1993;169:463-6.

8. Gibbs RS. Chorioamnionitis and bacterial vaginosis. Am J Obstet

Gynecol 1993;169:460-2.

9. Carey JC, Klebanoff MA, Hauth JC, Hillier SL, Thom EA, Ernest

JM, et al. Metronidazole to prevent preterm delivery in pregnant

women with asymptomatic bacterial vaginosis. N Engl J Med

2000;342:534-40.

10. McDonaldHM,O’Loughlin JA, VigneswaranR, Jolley PT,Harvey

JA,BofA, et al. Impact ofmetronidazole therapyonpretermbirth in

women with bacterial vaginosis flora (Gardnerella Vaginalis):

a randomized, placebo-controlled trial. BJOG 1997;104:1391-7.

11. American College of Obstetrics and Gynecologists. Assesment of

risk factors for preterm birth. ACOG Practice Bulletin no. 31.

Obstet Gynecol 2001;98:709-16.

12. Nugent RP, Krohn MA, Hillier SL. Reliability of diagnosing

bacterial vaginosis is improved by a standardized method of gram

stain interpretation. J Clin Microbiol 1991;29:297-301.

13. Ugwumadu A, Manyonda I, Reid F, Hay P. Effect of early oral

clindamycin on late miscarriage and preterm delivery in asymp-

tomatic women with abnormal vaginal flora and bacterial vagi-

nosis: a randomised controlled trial. Lancet 2003;361:983-8.

14. Goffinet F, Maillard F, Mihoubi N, Kayem G, Papiernik E,

Cabrol D, et al. Bacterial vaginosis: prevalence and predictive

value for premature delivery and neonatal infection in women with

preterm labour and intact membranes. Eur J Obstet Gynecol

Repro Biol 2003;108:146-51.

15. Cauci S, Hitti J, Noonan C, Agnew K, Quadrifoglio F, Hillier SL,

et al. Vaginal hydrolytic enzymes, immunoglobulin A against

Gardnerella vaginalis toxin, and risk of early preterm birth among

women in preterm labor with bacterial vaginosis or intermediate

flora. Am J Obstet Gynecol 2002;187:877-81.

16. Hallen A, Pahlson C, Forsum U. Bacterial vaginosis in women

attending STD clinic: diagnostic criteria and prevalence of

Mobiluncus spp. Gentiourin Med 1987;63:386-9.

17. Royce RA, Jackson TP, Thorp JM Jr, Hillier SL, Rabe LK,

Pastore LM, et al. Race/ethnicity, vaginal flora patterns, and pH

during pregnancy. Sex Transm Dis 1999;26:96-102.

18. Hillier SL, Critchlow CW, Stevens CE, Roberts MC, Wolner-

Hanssen P, Eschenbach DA, et al. Microbiological, epidemiolog-

ical and clinical correlates of vaginal colonisation by Mobiluncus

species. Genitourin Med 1991;67:26-31.

19. Holst E. Reservoir of four organisms associated with bacterial

vaginosis suggests lack of sexual transmission. J Clin Microbiol

1990;28:2035-9.

20. Sheinfeld J, Schaeffer AJ, Cordon-Cardo C, Rogatko A, Fair WR.

The association of the Lewis antigen blood group phenotype with

recurrent urinary tract infections in women. N Engl J Med

1989;320:773-7.

21. Hilton E, Chandrasekaran V, Rindos P, Isenberg HD. Association

of recurrent candidal vaginitis with inheritance of Lewis blood

group antigens. J Infect Dis 1995;172:1616-9.

22. Chaim W, Foxman B, Sobel JD. Association of recurrent vaginal

candidiasis and secretory ABO and Lewis phenotype. J Infect Dis

1997;176:828-30.

23. Rothenbacher D, Weyermann M, Bode G, Kulaksiz M, Stahi B,

Brenner H. Role of Lewis A and Lewis B blood group antigens in

Helicobacter pylori infection. Helicobacter 2004;9:324-9.

24. Molthan L. Lewis phenotypes of American Caucasians, American

Negroes and their children. Vox Sang 1980;39:327-30.