Predictive value for preterm birth of abnormal vaginal flora, bacterial vaginosis and aerobic vaginitis during the first trimester of pregnancy


GG Donders, Department of Obstetrics and Gynaecology, Heilig Hart Hospital, Kliniekstraat 45, 3300 Tienen, Belgium. Email


Introduction  Abnormal vaginal flora (AVF) before 14 gestational weeks is a risk factor for preterm birth (PTB). The presence of aerobic microorganisms and an inflammatory response in the vagina may also be important risk factors.

Aim  The primary aim of the study was to investigate the differential influences of AVF, full and partial bacterial vaginosis, and aerobic vaginitis in the first trimester on PTB rate. The secondary aim was to elucidate why treatment with metronidazole has not been found to be beneficial in previous studies.

Setting  Unselected women with low-risk pregnancies attending the prenatal unit of the Heilig Hart General Hospital in Tienen, Belgium, were included in the study.

Materials and methods  At the first prenatal visit, 1026 women were invited to undergo sampling of the vaginal fluid for wet mount microscopy and culture, of whom 759 were fully evaluable. Abnormal vaginal flora (AVF; disappearance of lactobacilli), bacterial vaginosis (BV), aerobic vaginitis (AV), increased inflammation (more than ten leucocytes per epithelial cell) and vaginal colonisation with Candida (CV) were scored according to standardised definitions. Partial BV was defined as patchy streaks of BV flora or sporadic clue cells mixed with other flora, and full BV as a granular anaerobic-type flora or more than 20% clue cells. Vaginal fluid was cultured for aerobic bacteria, Mycoplasma hominis and Ureaplasma urealyticum. Outcome was recorded as miscarriage ≤13 weeks + 6 days [early miscarriage (EM), n = 8 (1.1%)], between 14 + 0 and 24 weeks + 6 days [late miscarriage (LM), n = 7 (0.9%)], delivery or miscarriage ≤34 weeks + 6 days n = 29 (3.8%)], ≤36 weeks + 6 days n = 70 (9.2%)]. PTB between 25 + 0 and 36 weeks + 6 days was further divided in severe PTB (SPTB, 25 + 0 to 34 weeks + 6 days) and mild PTB (MPTB, 35 + 0 to 36 weeks + 6 days).

Results  Women without abnormalities of the vaginal flora in the first trimester had a 75% lower risk of delivery before 35 weeks compared with women with AVF [odds ratio (OR) 0.26; 95% confidence interval (CI) 0.12–0.56]. The absence of lactobacilli (AVF) was associated with increased risks of PTB (OR 2.4; 95% CI 1.2–4.8), EPTB (OR 6.2; 95% CI 2.7–14) and miscarriage (OR 4.9; 95% CI 1.4–17). BV was associated with increased risks of PTB (OR 2.4; 95% CI 1.1–4.7), EPTB (OR 5.3; 95% CI 2.1–12.9) and miscarriage (OR 6.6; 95% CI 2.1–20.9) and coccoid AV was associated with increased risks of EPTB (OR 3.2; 95% CI 1.2–9.1) and miscarriage (OR 5.2; 95% CI 1.5–17). In women with BV, partial BV had a detrimental effect on the risk of PTB for all gestational ages, but full BV did not. Preterm deliveries later than 24 weeks+ 6 days were more frequent when M. hominis was present (EPTB OR 13.3; 95% CI 3.2–55).

Discussion  Bacterial vaginosis, AV and AVF are associated with PTB, especially LM and severe PTB between 25 and 35 weeks. The absence of lactobacilli (AVF), partial BV and M. hominis, but not full BV, were associated with an increased risk of preterm delivery after 24 weeks+ 6 days. As metronidazole effectively treats full BV, but is ineffective against other forms of AVF, the present data may help to explain why its use to prevent PTB has not been successful in most studies.


For decades, the important objective of reducing the preterm birth (PTB) rate has presented a challenge. Socio-economic variables, maternal smoking, genital infections and short cervix are among the recognised risk factors for PTB that can be addressed, at least theoretically.1

Many studies suggest that the presence of AVF early in pregnancy increases the risk of PTB, premature rupture of the membranes and low birth weight, as does the presence of anaerobic overgrowth, for example in bacterial vaginosis (BV), and aerobic overgrowth, for example in aerobic vaginitis (AV) or trichomoniasis.2,3 BV can be treated, with similar clinical cure rates, with either metronidazole or clindamycin, which are both efficient antibiotics for anaerobic infections. However, clindamycin has a broader spectrum than metronidazole and has in vitro activity against anaerobes as well as aerobic organisms such as streptococci and Staphylococcus aureus. During pregnancy, treatment of BV/AVF with metronidazole has been found to be unsuccessful, but some randomised studies have shown a clear benefit of treatment with clindamycin over placebo in reducing the rates of PTB and preterm rupture of membranes.

The Preterm Risk Early in Pregnancy (PREP) study was designed to detect risk factors for preterm delivery at prenatal visits before the 16th week of pregnancy. In this report, we describe the influences of different subtypes of AVF on pregnancy outcome.

Materials and methods

Participants and procedures

During the period June 2000 to December 2001, 1026 unselected pregnant women presenting for their first prenatal visit at the Heilig Hart General Hospital in Tienen, Belgium, were asked to participate in a surveillance study to assess the importance of first trimester screening in the prevention of preterm delivery. The study was reviewed and approved by the ethical committee of the Heilig Hart General Hospital and prior written informed consent was obtained from all patients.

All women underwent a vaginal ultrasound examination at the first prenatal consultation to confirm the gestational age of the pregnancy, and this was corrected, if the menstrual and the ultrasound dates differed by more than 1 week. A vaginal smear was taken to assess vaginal microflora by phase contrast microscopy. Two or three electronic images for each slide were stored for later review, if necessary. Vaginal swabs were taken for detection of aerobic bacterial overgrowth, Candida colonisation and significant Mycoplasma hominis and Ureaplasma urealyticum colonisation. Urinary cultures were performed for the detection of significant bacterial colonisation.

For this study, we selected only women who were having a singleton pregnancy, who attended the hospital for their first antenatal visit at between 9 and 16 weeks of pregnancy, for whom the gestational age of the foetus was confirmed by ultrasound before 16 weeks, who had complete data available regarding intake wet mount microscopy and M. hominis cultures and who had confirmed outcome data available. Obstetric data on 801 women (84%) were thus collected. One box of 42 fresh slides of vaginal fluid was accidentally discarded in the laboratory before the slides could be read. The final number of fully evaluable women was therefore 759.

The obstetric outcome was assessed using the mean birth weight (low birth weight was defined as <2500 g) and gestational age at delivery. Outcome was recorded as miscarriage ≤13 weeks + 6 days [early miscarriage (EM), miscarriage or delivery between 14 + 0 and 24 weeks + 6 days [late miscarriage (LM), delivery or miscarriage ≤34 + 6 and ≤36 weeks + 6 days. Delivery between 25 + 0 and 36 weeks + 6 days was further divided into severe PTB (SPTB, 25 + 0 to 34 weeks+ 6 days) and mild PTB (MPTB, 35 to 36 weeks +6 days). Patients delivering newborns with life-threatening or major morphological abnormalities or twins were excluded from the study.

Microscopic assessment of the vaginal microflora

After insertion of a water-lubricated speculum, a smear was taken from the upper lateral vaginal wall with a polystyrene Ayre spatula and spread on a glass slide. In the central laboratory, a droplet of saline was added and microscopy was performed at ×400 magnification with a phase contrast microscope (Leitz Biomed, Wetzlar, Germany) and stored in a digital archive.

Bacterial vaginosis

Bacterial vaginosis is an ecological disorder that ensues when normal lactobacilli are replaced by large numbers of Gardnerella vaginalis, Prevotella sp., Bacteroides sp., Mobiluncus sp. and M hominis.4 Patients have no symptoms or present with a combination of a grey, watery vaginal discharge, fishy odour, vaginal pH above 4.5 and typical findings during fresh wet mount examination of vaginal fluid (e.g. ‘clue cells’). As amine odour and increased discharge only have sensitivities of 34% and 56%, respectively,5 for the diagnosis of BV, we used centralised microscopy preformed on rehydrated fresh wet mounts by a trained team to analyse the vaginal microflora. ‘Full BV’ was defined as a predominant granular microflora with uncountable bacteria all over the slide, and more than 20% of epithelial cells covered with bacteria (clue cells) (Figure 1). Mixed areas with streaks of BV-like flora or sporadic clue cells combined with other types of microflora (normal-appearing microflora, flora with small bacilli or aerobic coccoid flora) were classified as ‘partial BV’, as described elsewhere.6 These BV streaks are characterised by small, uncountable bacteria that overlie one another so that they cannot be distinguished or counted individually, and appear to be identical to the streaks seen in full BV. In partial BV, however, these streaks occur on the same slides as AV flora or normal flora.

Figure 1.

 (A) Partial bacterial vaginosis (BV), defined as streaks of BV flora in a smear that also shows areas of normal or different flora. (B) Full BV, defined as a complete absence of lactobacilli, typical granular microflora with uncountable numbers of bacteria, and clue cells. (C) Coccoid AV flora. (D) Mainly staphoid aerobic vaginitis (AV) flora (a Candida blastospore is also visible). Similarities between the flora in the upper left corner of (A) (showing partial BV) and the flora in (C) and (D) are striking.

Aerobic vaginitis

Aerobic vaginitis corresponds to another type of disturbed microflora, in which the lactobacilli are replaced by aerobic facultative pathogens (intestinal microorganisms, such as Escherichia coli, enterococci, Staphylococcus spp. and group B streptococci), vaginal leucocytosis and parabasal cells. Sexually transmitted infections with organisms, such as Chlamydia trachomatis, Neisseria gonorrhoeae and Trichomonas vaginalis have to be excluded.7 The clinical picture of severe AV often includes a red, inflamed vaginal mucosa, a yellowish sticky discharge, a high pH above 6 and an odour that is unpleasant but not like fishy odour.8 Such a severe form of AV is thought to occur only rarely in pregnancy, but less severe forms may be more frequent. Because of the infrequent occurrence of severe AV during pregnancy, coccoid microflora was used as a substitute criterion for AV. Coccoid microflora is defined as microflora containing easily recognisable cocci on microscopic examination: separate cocci, or cocci bunched together in little collections, or cocci regularly arranged in chains. Coccoid flora can consist of round cocci, ovaloid cocci or cocci-bacillary morphotypes, but aerobic cocci are always thicker and more pronounced than anaerobic bacteria, and, unlike anaerobic, BV-associated flora, can be distinguished and counted individually.

Lactobacillary grades (LBGs)

Lactobacillary grades are modifications of Schröder’s classification.9 LBG I corresponds to normal microflora with a predominant presence of Lactobacillus morphotypes. LBG II corresponds to intermediate, mixed flora, LBG IIa being near-normal with lactobacilli outnumbering other microorganisms, and LBG IIb having other microorganisms outnumbering lactobacillary morphotypes. LBG III corresponds to completely disrupted flora in which only bacteria other than Lactobacillus morphotypes are present.


The lowest leucocyte score of 0 corresponds to fewer than ten leucocytes per high-power field (HPF; ×400 magnification). For more than ten leucocytes per HPF, a score of 1 corresponds to fewer than ten leucocytes per epithelial cell and a score of 2 to more than ten leucocytes per epithelial cell. This system corresponds well with severity of symptoms in nonpregnant patients8 and pregnant patients.10

Laboratory tests for the assessment of the vaginal microflora

Aerobic cultures

For cultures, samples were allowed to grow on a sheep blood chocolate agar at 37°C for 48 hours and semiquantitative growth 2+ or more was taken to be significant. Urinary culture was significant, if >100 000 colony-forming units (CFU)/ml were cultivated, in which case treatment with furadantoin was advised unless resistance studies suggested otherwise.


A vaginal swab was tested for mycoplasmata (M. hominis and U. urealyticum) using the Mycoscreen® test. A Dacron swab was first rotated in the cervix and then soaked in the vaginal fluid, and transported to the laboratory in A3 transport medium. The swab was incubated for 48 hours in lyophilised urea–arginine broth at 37°C. Positive (phenol red) samples were cultured for 48 hours on A7 agar at 37°C to identify and quantify the mycoplasmata, and Mycofast® was used for the antibiogram. Women with cervicitis caused by N. gonorrhoeae (as determined on chocolate blood agar) or C. trachomatis (as determined using Chlamydiazym®, Abbott, Chicago, IL, USA) were treated and excluded from further analysis. The results of the mycoplasma culture and vaginal microscopy were not communicated to the investigator and treatment for mycoplasmata was not given.

Statistical methods

Normally distributed continuous variables were compared using Student’s t-test, while a nonparametric Welch test was used for data that were not distributed normally. Stochastic variables were compared with chi-square or Fisher’s exact test as appropriate. P < 0.05 was considered significant.


The mean age of the women participating in this study was 29.0 ± 4.4 years, 95% were Caucasian, and 649 (61%) were nulliparous, 297 (28%) primiparous and 110 (10%) multiparous (two or more children) (Table 1). The 983 singleton newborns with a birth weight above 500 g and a gestational age at birth of at least 24 weeks + 6 days had a mean birth weight of 3239 ± 502 g and a mean gestational age of 38.5 ± 1.9 weeks; of these, 58 neonates (5.9%) had a birth weight of <2500 g and 70 (9.2%) were miscarried or delivered at or <36 weeks + 6 days gestational weeks. Of these, 15 were miscarried, or born at or before 24 weeks + 6 days (M; 2%), 14 were delivered between 25 + 0 and 34 weeks + 6 days [severe preterm births (SPTB); 1.8%], and 41 were born in the gestational age range 35 + 0 to 36 weeks +6 days [mild preterm births (MPTB); 5.4%]. Minor morphological malformations were found in 9% of neonates but none was associated with a change in mean birth weight or gestational age. At the inclusion visit, 8.4% of the women had bacterial vaginosis.

Table 1.   Demographic characteristics and general pregnancy outcome data of the study patients
  1. SD, standard deviation; *percentage of subgroup of multipara; **only singletons, >24 weeks + 6 days and >500 g.

Age, mean ± SD29.0 ± 4.4 yr
Birth weight, mean ± SD**3239 ± 502 g
Weeks at delivery, mean ± SD**38.5 ± 1.9 wk
Birth weight, mean ± SD**3239 ± 502 g
Preterm delivery rates, n (%)**
Miscarriage (<25 wk)15 (2.0)
Severe preterm delivery (25 + 0 to 34 wk + 6 days )14 (1.8)
Mild preterm delivery (35 + 0 to 36 wk + 6 days)41 (5.4)
Term delivery (more than 37 full ended weeks)689 (90.8)
Parity, n (%)
P = 0649 (61.5)
P = 1297 (28.1)
P = ≥2110 (10.4)
Previous preterm birth*, n (%)
All (parous women only)47/407 (15.3)
More than one (parous women only)10/407 (2.4)
Smoking habits, n (%)
Current smoker120/761 (15.8)
Previous smoker129/761 (17)
Treatment with antibiotics prior to pregnancy, n (%)
Cervix infection/PID72/698 (10.3)
Herpes7/690 (1.0)
BV (treated)23/690 (3.3)
Candida (treated)270/690 (39)
UTI318/757 (42)
Bleeding during pregnancy153/801 (19)
Work outside during pregnancy, n (%)
Parttime170/758 (22.4)
Fulltime482/758 (63.6)

Flora was defined as normal (n = 611; 81.4%) when none of the following was present: LBG III, bacterial vaginosis, coccoid AV flora, increased vaginal leucocytosis or M. hominis. Compared with women with AVF (LBG III), women with normal flora had lower rates of deliveries before 25 weeks (1.3% versus 5.8% for women with normal flora versus AVF, respectively; P = 0.02) and between 25 + 0 and 34 weeks + 6 days (1.3% versus 8.5%; P = 0.001) (Table 2). BV was found in 8.4%, of whom 30 (4.0%) had full BV and 34 (4.4%) partial BV, coccoid microfora in 8.3% and increased numbers of leucocytes in 3%. Specific cultures revealed that 1.8% of the women were infected with M. hominis, and 28.9% with U. urealyticum. As expected, LBG III overlapped with BV: 16 women (52%) with partial BV and 27 (90%) with full BV also had LBG III, while 30% of women with coccoid flora also had LBG III. Twenty per cent of women with partial BV also had coccoid flora, while only 6.5% of women with full BV had this condition. Mycoplasma hominis was associated with BV (17% of women with full BV, 6.5% of those with partial BV and 0.5% of those without BV had M. hominis infection).

Table 2.   Pregnancy outcome for all women with singleton pregnancies, complete information regarding intake microscopy, sonographic confirmation of gestational age, and complete information regarding pregnancy outcome data
 n<14 wk (EM)14 + 0 to 24 wk + 6 days (LM)25 + 0 to 34 wk + 6 days (SPTB) 35 + 0 to 36 wk + 6 days (MPTB) >36 wk +  6 days (term) <37 wk (PTB)<35 wk (EPTB)<25 wk (M)
  1. The subcategories of abnormal vaginal flora (AVF), namely full and partial bacterial vaginosis (BV), aerobic vaginitis (AV) and Mycoplasma infection, can occur simultaneously and hence overlap. M: miscarriage (<25 weeks), EM: early miscarriage (<14 weeks), LM: late miscarriage (14 + 0 to 24 weeks + 6 days), PTB: preterm birth (<37 weeks), EPTB: early preterm birth (<35 weeks), further divided in SBTB: severe PTB (25 + 0 to 34 weeks + 6 days ) and MPTB: mild PTB (35 + 0 to 36 + 6 weeks). Term pregnancies are defined as 37 weeks or more at delivery. *OR 7.4 (2.5–22.0) P = 0.0012; **OR 10.4 (1.8–59.3), P = 0.03; ***OR 5.2 (1.1–25.7) P = 0.08; ****OR 18 (3.2–97.0) P = 0.011; *****OR 6.3 (1.6–24.6) P = 0.024.

Total number of women7598 (1.1%)7 (0.9%)14 (1.8%)41 (5.4%)689 (90.8%)70 (9.2%)29 (3.8%)15 (2%)
Normal6114 (0.7%)4 (0.7%)8 (1.3%)34 (5.6%)561 (91.8%)
Abnormal vaginal flora (LBG III)712 (2.7%)2 (2.7%)6 (8.5%)*4 (5.6%)57 (80.3%)2.4 (1.2–4.8)
P = 0.022
6.2 (2.7–14.4)
P < 0.0001
4.9 (1.4–16.9)
P = 0.022
Bacterial vaginosis (BV)642 (3.3%)3 (4.6%)3 (4.6%)3 (4.6%)53 (82.8%)2.43 (1.1–4.7)
P = 0.035
5.3 (2.1–12.9)
P = 0.0009
6.6 (2.1–20.9)
P = 0.0041
Partial BV341 (2.9%)2 (5.9%)**2 (5.9%)***2 (5.9%)27 (79.4%)2.4 (1.2–7.1)
P = 0.022
7.2 (2.4–21.0)
P = 0.0022
3.3 (1.3–8.0)
P = 0.014
Full BV30(3.3%)*1 (3.3%)1 (3.3%)1 (3.3%)26 (86.7%)nsnsns
Coccoid aerobic flora63 4 (6.4%)1 (1.7%)3 (4.8%)55 (87.3%)ns3.2 (1.4–9.1)
P = 0.038
5.2 (1.5–17.7)
P = 0.019
Increased number of leucocytes23 001 (4.4%)22 (95.6%)nsnsns
Positive culture of Mycoplasma hominis141 (7%)2 (14%)****3 (21%)*****8 (57%)8.5 (2.8–25.5)
P = 0.0006
13.3 (3.2–55)
P = 0.0039

Absence of lactobacilli (LBG III) at the first visit was associated with increased risks of PTB [birth at <37 weeks; odds ratio (OR) 2.4; 95% confidence interval (CI) 1.24.8], EPTB (birth at < 35 weeks; OR 6.2; 95% CI 2.714.4) and miscarriage (M) (at < 25 weeks; OR 4.9; 95% CI 1.416.9) (Table 2). Bacterial vaginosis at the 10–14 weeks visit was associated with PTB (OR 2.4; 95% CI 1.1–4.7), EPTB (OR 5.3; 95% CI 2.1–12.9) and M (OR 6.6; 95% CI 2.1–20.9) and coccoid AV flora with EPTB (OR 3.2; 95% CI 1.4–9.1) and M (OR 5.2; 95% CI 1.5–17.7). Of the women with BV, roughly half had ‘full BV’ and half had a mixed flora of BV with other types of microflora (‘partial BV’). As a predictor of preterm delivery, ‘full BV’ was not a significant marker, but ‘partial BV’ was a predictor of delivery before 25 weeks (OR 3.3; 95% CI 1.3–8.0), before 35 weeks (EPTB, OR 7.2; 95% CI 2.421.1) and before 37 weeks (PTB, OR 3.3; 95% CI 1.3–8.0).

Only two subtypes of abnormal flora were associated with the important subgroup of women giving birth between 25 + 0 and 34 weeks + 6 days (SPTB). Patients presenting with an absence of lactobacilli (AVF) had a 7.4-fold increased risk (OR 7.4; 95% CI 2.5–22.0) and patients with partial BV had a 5.2-fold increased risk of SPTB (OR 5.2; 95% CI 1.1–25.7).

The presence of vaginal M. hominis was strongly associated with lower birth weight (2715 ± 768 g versus 3248 ± 515 g; P = 0.028) and lower gestational age (37.2 ± 1.7 versus 38.8 ± 1.8 weeks; P < 0.0001) at delivery (Table 2). In M. hominis-positive patients, SPTB (25 + 0 to 34 weeks +6 days) and mild PTB (34 + 6 to 36 weeks + 6 days) were, respectively, 13 and 8.5 times more frequent (P < 0.005). Nearly half of the patients harbouring M. hominis delivered preterm, which, despite low numbers, is significantly different from the number of women with normal flora delivering preterm (43% versus 8.1%, respectively; OR 8.1; 95% CI 2.825.5; P = 0.0006). Vaginal colonisation with U. urealyticum, group B streptococci (GBS) or Candida albicans was not associated with PTB or low birth weight (data not shown).


The observed PTB rate of 7% in this study is similar to the expected rate in Flanders during the same period (7.4%; singletons 6.3%).11 Numerous studies have consistently shown that bacterial vaginosis and AVF at 12–24 weeks are associated with increased risks of PTB, preterm rupture of the membranes, chorioamnionitis, fetal infection and cerebral palsy.12–22 However, studies applying strict criteria for full bacterial vaginosis (a Nugent score above 7, overall granular flora, >20% clue cells and LBG III) did not find a significant correlation between BV and the risk of PTB.14,18,23

Helen McDonald pointed out in the early 1990s that women with an increased risk for PTB have two types of AVF, one consisting of predominantly BV flora, and the other of aerobic microorganisms, such as Klebsiella and Escherichia coli.24 Also, in case–control studies of women in labour investigating the difference between typical bacterial vaginosis with overgrowth of Gardnerella vaginalis/anaerobes and the presence of other abnormal flora types, an association between BV in labour and subsequent PTB was consistently lacking, while the association of PTB with the presence in labour of Klebsiella sp., E. coli, staphylococci and streptococci,2,24 with or without signs of inflammation or neutrophils in the vagina,25–28 was rather strong and consistent. It should therefore come as no surprise that treatment with oral or vaginal metronidazole, which efficiently kills anaerobes, but not aerobic bacteria, reduces BV in pregnancy, but fails to decrease the risk of PPROM or PTB in randomised, placebo-controlled series.29–31 Carey et al.2 re-analysed their data after they failed to find a protective effect of metronidazole in a large population-based placebo-controlled treatment trial. They found that, at delivery, Klebsiella and E. coli were more often present in the flora of women delivering preterm than in the flora of other women.

In a randomised study comparing repeated courses of metronidazole versus vitamin C, patients with BV receiving metronidazole were found to have a lower gestational age at delivery than women without BV or those with BV treated with vitamin C only, and this difference was even more pronounced in multiparous women with previous preterm delivery (43% PTB versus, respectively, 29% and 24%).32 In another large randomised study in which metronidazole was used to treat Trichomonas infection during pregnancy, a significant increase in the PTB rate was also found in metronidazole-treated women versus those given placebo.33

Not unexpectedly, recent meta-analyses independently concluded that prophylactic treatment of BV during pregnancy is not adequate to reduce infection-related PTB rates in low-risk women.34–37 A reduction in the rate of PTB was demonstrated only in studies in which women were not only treated early in pregnancy but also given broader spectrum antibiotics such as clindamycin or metronidazole in combination with erythromycin.34,38–42 The impact of BV on pregnancy outcome has been addressed in depth in previous studies. However, as ‘intermediate flora’ (score 4–6 in the Nugent system on Gram-stained specimens) does not correspond closely to ‘intermediate BV’, but rather to another less well-defined disturbance of the flora, it is not clear how these data should be interpreted.6,43–45 In a recent update of the Cochrane database, Helen McDonald et al.34 concluded that screening and treatment for BV or intermediate flora are not useful to prevent complications of pregnancy when carried out after 20 gestational weeks. Treatment with vaginal or oral metronidazole or clindamycin at <20 weeks effectively decreased the rate of PTB at <37 weeks (five studies), although this effect was not observed in the smaller databases of two studies reporting data for delivery before the 32nd week of gestation. The lack of an effect of treatment in lowering the PTB rate may be a result of the inclusion of some women with intermediate flora, which may not respond to metronidazole. However, only 20% of all women included in the five studies on women at <20 weeks of gestation had intermediate flora, suggesting that other misclassifications, such as partial BV or AV, may also have been involved. Similar findings were reported in observational studies by Hay et al.46 and ourselves,47 suggesting that intermediate flora is more dangerous in terms of the risk of PTB, in particular for mid-term pregnancy losses, than full BV.48 The short-term association of BV with miscarriages before 16 weeks, however, was confirmed in some studies.49–51

We suggest that the diagnosis of AVF should be refined and a distinction made between abnormal anaerobic-type (granular) flora (BV flora) and aerobic microflora (short bacilli or cocci; AV flora), and that the inflammatory response in the vaginal fluid should be reappraised, as aerobic microflora and the finding of increased vaginal leucocytosis correlate with greater concentrations of pro-inflammatory cytokines present in the vagina,8,38,43 and with enzymatic activity leading to preterm contractions and intrauterine infection.20,28,41,44,45 An increasing number of studies show an higher risk of preterm rupture of the membranes and PTB when an the number of leucocytes were found to be elevated in the vaginal fluid, especially when counted in proportion to the epithelial cells.26,27,52 In the present study, the sole criterion of increased vaginal leucocytes was not associated with preterm delivery.

To elucidate the roles of full or partial BV,6 AV8 and intermediate flora46 in the causation of preterm labour, we studied the impacts of these specific subcategories in detail in the present study. We confirmed the previous finding that AVF (LBG III) in the first trimester of pregnancy is most consistently associated with preterm delivery, early preterm delivery and mid-trimester pregnancy losses. Furthermore, after exclusion of pregnancy losses at nonviable gestational ages of <25 weeks, AVF remained associated with a 10-fold higher risk of the most critical PTB, those between 25 + 0 and 34 weeks + 6 days. In general, not only the presence of BV flora but also of microflora suggestive for aerobic bacteria (AV flora), was associated with mid-trimester pregnancy losses and early preterm delivery at <35 weeks. Unlike full BV (no lactobacilli, overall granular flora with >20% clue cells) that was related to EM, partial BV (patches of BV but also other flora present) was related to LM, to early preterm delivery and preterm delivery. Therefore, like AV, partial BV may have a more profound effect on the risk of PTB. However, caution in drawing firm conclusions is warranted because of the low numbers in this study, and confirmation of these findings must be awaited.

The exact composition of partial BV is not known, but in general the microflora seen in combination with the anaerobic granular plaques of BV is often of the coccoid and staphoid AV flora type, suggesting that AV and BV can co-exist in these women (see also Figure 1). If partial BV represents a mixed flora of AV and BV, it may be speculated that the aerobic component of the microflora may be the main risk factor during pregnancy.

Aerobic vaginitis is a condition in which disturbed microflora with an absence of lactobacilli is not overwhelmed by anaerobic bacteria, as in typical BV, but rather contains a significant number of aerobic facultative pathogenic flora from the bowel. In moderate and severe cases, an elevated host immune reaction can be demonstrated by an increased number of leucocytes, which sometimes have a ‘toxic’ appearance.8 As this condition produces large amounts of pro-inflammatory cytokines, it was suggested that it can cause preterm labour by initiating the prostaglandin cascade.53 In the present study, the association between AV and increased PTB rate, particularly EPTB before 35 weeks, was confirmed, as was the association of partial BV with PTB, suggesting a mixed AV/BV flora. In full BV, we and others found unexpectedly low levels of interleukin IL-8 and a striking absence of leucocytes, in spite of the presence of moderate amounts of IL-1β.8,54 It can therefore be postulated that BV is a condition in which there is relative immune suppression in response to bacterial overgrowth, whereas in AV there is rather a sepsis-like local overreaction of the immune response.6 In Latvia, Receberga et al. showed a strong relation between the presence of overgrowth of AV-associated bacteria in vaginal cultures at the first prenatal visit and a risk of inflammation of the umbilical cord (funisitis) at birth, while BV-associated bacteria were correlated with chorioamnionitis but not with funisitis.55 All of these findings stress the importance of recognising AV flora early in pregnancy.

A remarkable finding was the association of M. hominis with preterm delivery, although numbers were small. In the literature, the presence of M. hominis has generally been related to an increased risk of miscarriage,51,56 and preterm delivery if found in combination with bacterial vaginosis.57,58 Although the paediatric literature often also suggests maternal U. urealyticum to be responsible for complications in preterm babies, especially if present in the amniotic fluid or when neonates are of extreme low birth weight <1500 g,59,60 we could not confirm a significant role of this organism in the causation of PTB. This is in accordance with the findings of Paul et al.61, who also found no role for U. urealyticum infection as a possible cause of preterm labour.

We recognise that there are weaknesses in the present study. As a consequence of the splitting of abnormal flora into categories, the numbers are rather too small to draw firm conclusions, but as they showed significant differences, further investigation is warranted. Furthermore, some data could not be recovered, resulting in a loss of power. The use of wet mounts instead of Gram stains to diagnose abnormal flora types may be seen as a disadvantage by some, but as a necessary prerequisite by investigators with experience and expertise in its broad diagnostic abilities. The value of visual recognition of the different vaginal flora types on wet mounts is supported by the detection of a great number of aerobic and anaerobic bacteria with modern electrophoresis techniques and real-time polymerase chain reaction (PCR).62 Its prospective design, the absence of significant information bias and the detailed analysis of the vaginal flora, however, are strong advantages of this study. Larger studies of this type would be welcomed, and it is to be hoped that such studies might unravel some of the poorly understood relationships between AVF and complications of pregnancy, and allow trials of appropriate treatments for the prevention or PTB to be carried out.

We conclude that different patterns of AVF can be recognised at the beginning of pregnancy, and these require further study in order to elucidate the pathogenesis of infection-related preterm delivery and the (in)efficacy of prophylactic antibiotic treatment regimens to prevent PTB. Specific markers of AVF, such as the absence of lactobacillary morphotypes on fresh microscopic smears, signs of AV or partial BV, and the presence of M. hominis, seem to be better predictors of infection-related PTB than formerly used criteria such as full bacterial vaginosis. As full BV was not related to PTB in this small study and metronidazole treatment was found to increase preterm delivery rates in others, the option to ‘screen and treat’ for asymptomatic BV with metronidazole needs to be revised after more detailed study of the different subtypes of vaginal microflora.

Disclosures of interest


Contribution to authorship

G. Donders: design, coordination, investigation, primary statistics, technical diagnosis (microscopy) and writing; G. Bellen: data management, writing; I. Riphagen: investigation; T. Van den Bosch: investigation; R. Reybrouck: laboratory tests; S. Van Lierde: follow-up of newborns.

Ethics approval

EC Committee of Heilig Hart General Hospital, Tienen, 5 January 2000 (no reference number at that time).


Funding was received from Femicare, an official nonprofit organisation supporting and organising clinical research for women.