Aetiology of preterm labour: bacterial vaginosis


Prof S Guaschino, Department of Obstetrics and Gynaecology, IRCCS Burlo Garafolo, Via dell’Istria 65, I-34100 Trieste, Italy. Email:


Bacterial vaginosis (BV) is a common condition characterised by a polymicrobial disorder, with an overgrowth of several anaerobic or facultative bacteria and with a reduction or absence of lactobacillus colonisation. The prevalence of BV ranges from 4 to 64%, depending on the racial, geographic and clinical characteristics of the study population. In asymptomatic women, the prevalence varies from 12 to 25%, and similar percentages are observed in pregnant women. Although BV is associated with several adverse outcomes, such as upper genital tract infections, pelvic inflammatory disease, endometritis, preterm birth and low birthweight, many basic questions regarding the pathogenesis of BV remain unanswered. Mucosal immune system activation may represent a critical determinant of adverse consequences associated with BV. An unequal risk for BV acquisition and\or recurrence could derive from different mucosal immune host abilities and\or capability of invading microbes to produce factors that inactivate the local immune response. BV is associated with a two-fold increased risk of preterm birth, with the greatest risk when BV is present before 16 weeks of gestation (odds ratio = 7.55). This may indicate a critical period during early gestation when BV-related organisms can gain access to the upper genital tract and set the stage for spontaneous preterm labour later in gestation. The results of treatment trials for pregnant women with BV have been heterogeneous, with anywhere from an 80% reduction to a two-fold increase in preterm birth among women who received treatment. For this reason, in current clinical practice significant controversy surrounds determining not only who and when to screen but also who and how to treat. Recent evidence shows that individual genetic backgrounds can affect chemokine production. This is an interesting area for future research and could lead to trials of treatment only for women genetically predisposed to preterm birth.


Bacterial vaginosis (BV) is the most common lower genital tract disorder among women of reproductive age. The condition is not a classical infection caused by a single pathogen, but is rather a complex alteration of the vaginal ecosystem, where the physiologic lactobacilli-dominant flora is replaced by an overgrowth of mixed flora, with a high concentration of anaerobic bacteria, normally present in the vagina in substantially fewer numbers.

Although the normal vaginal milieu is characterised by the presence of a large number of bacterial species, under physiologic conditions H2O2-producing lactobacilli account for 95% of these and act against the proliferation of other microorganisms by maintaining an acidic vaginal pH. In the normal vaginal ecosystem, the anaerobe to aerobe ratio is normally kept between 2:1 and 5:1. In the presence of BV, the quantity and quality of H2O2-producing lactobacilli decrease, the vaginal pH increases, and there is a subsequent shift in the anaerobe to aerobe ratio to between 100:1 and 1000:1. The same bacteria found in the healthy vaginal environment are also found in BV, the difference being not in the quality but in the quantity of the microorganisms present. It is not possible to identify any single species as the cause of BV, although Gardnerella vaginalis, Prevotella spp., Bacteroides spp., Mobiluncus spp. and Mycoplasma spp. are the most commonly found in association.

The reason why this shift in the microbial flora occurs remains unknown. Despite the substantial microbial overgrowth, BV does not seem to be associated with signs of local inflammation. Leucocytes cannot be found in the vaginal fluid of women with BV, and for this reason the term vaginosis has been used rather than vaginitis. Around 50% of women are asymptomatic and when symptoms do occur, they are generally mild, with the distinctive feature being the presence of a malodorous vaginal discharge.

BV is a very common condition. The prevalence observed in the literature ranges between 4 and 60%. The lowest prevalence has been found in private offices (4–17%), while the highest in women attending sexually transmitted disease clinics (64%). In asymptomatic pregnant women, the reported prevalence varies between 6 and 35%, with the highest rate observed in studies from the USA. Epidemiological differences in the populations studied can explain this wide range in the prevalence. Several risk factors for the acquisition of BV have been identified. Black race, smoking, sexual activity, contraceptive practice and the use of vaginal douching have all been found to be associated with increased prevalence.

Diagnosis of BV

One factor that may influence the detection of BV is the method used for the diagnosis. Since BV is not an infection caused by a single pathogen, but rather an imbalance in the proportions of the normal vaginal microflora, the microbiological culture, classically considered as the gold standard in the diagnosis of any infection, cannot reliably predict the presence of BV. Consequently, different diagnostic techniques have been developed to test for the condition. The two most widely used are clinical criteria and Gram staining of vaginal secretions.

The clinical diagnosis of BV can be made in the presence of three of the following four signs first described by Amsel in 19831:

  • • the presence of an adherent and homogeneous vaginal discharge
  • • A vaginal pH >4.5
  • • detection of clue cells (vaginal epithelial cells so covered by bacteria as to render the borders indistinct) on saline wet mount
  • • an amine odour (positive ‘whiff test’) after addition of the amine potassium hydroxide (10%) to the vaginal secretions.

These criteria have some limitations, being subjective, not reproducible and somewhat unpleasant. As a consequence, the Gram staining of an air-dried smear of vaginal secretions has become the most widely used diagnostic method. Two different systems have been introduced to interpret and score the Gram stain; the first proposed by Spiegel in 19832 is less commonly used, while the second and most widely applied is the score developed by Nugent in 1991.3 In the Nugent’s score system, each microbial morphotype is quantified from 1+ to 4+ based on the number of morphotypes per oil immersion field, and a corresponding total score is assigned. The scoring system allows for gradations in severity: a diagnosis of BV can be made when the score is 7 or more, a score of 4 to 6 is considered intermediate, while 0 to 3 is normal. The Gram stain method has been shown to be highly reliable and reproducible in diagnosing BV, with a high inter-observer agreement in the examination of the specimens. In addition, a good correlation between Gram stain scores and clinical signs of BV has been shown.

Preterm birth

BV has long been regarded as an inconvenient but not a pathological problem. Increasingly, BV has attracted interest because of an observed association between its presence and a number of significant obstetric and gynaecological complications, such as preterm birth, preterm prelabour rupture of membranes (PPROM), chorioamnionitis, postpartum endometritis, postsurgical infections and pelvic inflammatory disease.

The association between BV and preterm birth has been the subject of a large number of case–control and cohort prospective studies in the last 20 years, the majority of which have shown a significant association. In 2003, a meta-analysis of studies involving 20 232 women showed that BV doubles the risk of preterm delivery at less than 37 weeks of gestational age (odds ratio [OR] 2.19; 95% CI 1.54–3.12).4 The studies included in the meta-analysis show considerable differences in the methods of screening, in the population evaluated (low-risk, high-risk, mixed-low and high-risk, single or twin pregnancies, women with symptoms of spontaneous preterm labour) and in the gestational age at screening, which may partly explain the heterogeneity in the statistical analysis. A subgroup analysis of the included studies suggests that the risk of preterm birth is further increased when the diagnosis of BV is made early in pregnancy. The rate of preterm birth calculated in the meta-analysis is seven-fold higher if BV is diagnosed before 16 weeks of gestation and more than four-fold in women with BV diagnosed before 20 weeks. These findings have not been confirmed in other observational studies. In a recent analysis of 12 937 women screened for BV at gestational ages between 8 and 22 weeks as part of a multicentre clinical trial from the USA, BV was found to be significantly associated with preterm birth, with an OR between 1.1 and 1.6.5 This OR did not vary significantly by gestational age at diagnosis. It is possible that variations in the population characteristics account for the different observations coming from the USA when compared with Europe.

Only a small number of women with multiple pregnancies have been evaluated, showing no association between the presence of BV and preterm birth. Another notable finding of the meta-analysis is the absence of a significant association of BV with early preterm birth at less than 34 or less than 32 weeks of gestation. It should be noted that few studies included in the meta-analysis used these outcomes, and none of the studies that used such early screening was considered in the analysis.4

Despite the voluminous literature evaluating the association of BV with spontaneous preterm labour and preterm birth and the demonstration of the presence of a significant association, many doubts remain about the optimal time for diagnosis, the role of BV in cases of very early preterm birth, its association with the most serious neonatal sequelae, and the different strength of the association between BV and preterm birth observed in different populations. What is surprising about BV is why such a prevalent condition, caused by relatively low virulence microorganisms, which is often asymptomatic and without severe inflammatory symptoms, can be responsible for serious adverse sequelae.

It is possible that some of these contradictions can be explained by shifting from the concept of BV as a disease directly causing severe complications to the more intriguing idea of BV as a ‘microbial/mucosal immunity disorder’. Only limited proportions of women with BV develop adverse pregnancy outcomes. In one study, only 6.3% of women had a low birthweight (LBW) preterm birth and 3.4% had an LBW full-term infant.6 BV must therefore be considered a heterogeneous condition, where the most serious complications involve only a limited subgroup of women predisposed to the progression of the disease. The factors causing this variable expression of BV are still largely unknown. It is possible that variations in the degree of virulence of the microorganisms involved and, more likely, a difference in the immune response mounted by the host against the infection can determine the severity of the condition. Our group has recently postulated the presence of a subgroup of women with BV, who show a higher risk of preterm birth and LBW, characterised by the presence of an impaired immune response against the infection.7,8 These women show a low or negative immunoglobulin (Ig) A response against the haemolytic G. vaginalis toxin associated with a high level of sialidase and/or prolidase activity in vaginal secretions. Both sialidases and prolidases are enzymes capable of impairing the host immune defence by inactivating mucosal factors involved in innate and acquired immune mechanisms. The innate immune response is the first-line mucosal immune defence, and it is considered critically important to clear the mucosa of pathogens. Neutrophil recruitment and activation are the main innate response, and this is regulated by the presence of locally secreted chemokines.9 Innate immunity is also a necessary precursor to an adaptive immune response. In 2002, it has been shown by our group that elevated levels of interleukin-1 beta (IL-1β), a crucial proinflammatory chemokine, can be found in 30% of BV-positive women, and these women show a high anti-Gvh IgA response and a high number of leucocytes.10 The IL-1β rise in these women should therefore activate vaginal innate and adaptive immunity against microbes and is likely to be beneficial to the host by promoting an array of cascade responses devoted to fighting the damaging effect of abnormal vaginal colonisation. It is possible that women exposed to anaerobic mucosal invasion who do not mount a strong innate and adaptive immune response may be more prone to poor outcomes. Further studies are needed to assess this issue, as IL-1β is a highly inflammatory cytokine and the margin between beneficial and adverse effects of IL-1β is considered to be exceedingly narrow. Recently published data are in contrast with our hypothesis. Genc et al. found that an imbalance in the IL-1β and IL-1 receptor antagonist response in the cervicovaginal secretions of women with vaginal colonisation by anaerobic Gram-negative rods or G. vaginalis was associated with spontaneous preterm labour and preterm birth.11 They concluded that an elevated IL-1β concentration not balanced by the presence of a similarly high level of IL-1 receptor antagonist, which is normally produced to counteract the proinflammatory effect of IL-1, can identify women at higher risk of preterm birth. In these women, the immune response is suboptimal, with an excessive inflammatory response mounted against the infection. In contrast, Simhan et al. reported that a decreased rather than increased lower genital tract concentration of several cytokines including IL-1β, is associated with subsequent chorioamnionitis.12

It is likely that all these contradictory findings can be explained by the idea of a subtle balance governing the ‘healthy’ immune response. We must always consider that the immune response is complex and involves the interplay of many different factors. It may be postulated that both an imbalance towards an exaggerated immune activation and a similar imbalance towards a depressed response against infection can predispose to the adverse obstetric outcomes observed in women with BV. As suggested by Romero et al. in 2004, ‘individuals differ in their ability to mount an inflammatory response’,13 but what influences ability of the individual to mount a response is largely unknown. It would appear that, together with environmental factors, the individual genetic background could play a major role. It has been shown by Macones et al. in 2004 that an interaction between genetic susceptibility and environmental factors (BV) is associated with an increased risk of spontaneous preterm labour and preterm birth.14 In this case–control study, the presence of BV was associated with an increased risk of preterm birth (OR 3.3), and women carrying the tumour necrosis factor-alpha (TNF-α) allele 2 (which leads to the overproduction of TNF) also had an increased incidence of preterm birth (OR 2.7). When both the conditions (BV and presence of TNF-α allele 2) are present in the same woman, the OR for preterm birth increased considerably to 6.1, suggesting that a gene–environment interaction can confer a higher risk. It is likely that other gene–environment interactions, involving both maternal and fetal genetic background, will be found in the future. These interactions could play a major role in determining the susceptibility to ascending infection, to the microbial invasion of the amniotic fluid and the fetus. They could also be a key in understanding the mechanism of damage to the neonate, since this damage seems to be caused mainly by the effects of a deregulated inflammatory response directed against the fetus.

It is clear that the assessment of the vaginal ecosystem in health and disease cannot be based only on the evaluation of the microbial flora, but must also integrate information on the host immune response and, possibly, on the host genetic background to identify those individuals at high risk of serious adverse sequelae.

BV is common, even in low-risk populations of women, and as it is amenable to treatment. Identification during pregnancy and treatment may present a rare opportunity to reduce the rate of preterm birth and the consequent risk of neonatal mortality and morbidity. Such interventions may also reduce other adverse perinatal outcomes such as puerpural infection. The question of why BV is associated with preterm birth in some women but not in others remains unanswered, and the exact mechanism by which the organisms associated with BV may effect the initiation of spontaneous preterm labour remains unclear. Hence, the results of randomised controlled trials of treatment are needed to provide more direct evidence of the role of BV in preterm birth.


Our data suggest that treating asymptomatic BV with topical clindamycin neither markedly prolongs pregnancy nor increases birthweight,15 although the results of this trial are limited by the small number of women enrolled. Whether treatment of BV in pregnancy reduces the frequency of preterm birth remains confused. Twelve studies have investigated the potential to reduce the incidence of spontaneous preterm labour and preterm birth through treating women with BV (Table 1). The results are contradictory. Four studies indicate a reduced risk of spontaneous preterm labour and preterm birth with treatment, while others do not, and one study indicated an increased risk. Ugwumadu et al. showed that the treatment of asymptomatic intermediate abnormal vaginal flora and BV in a general obstetric population reduces the occurrence of late miscarriage and spontaneous preterm labour and preterm birth.25 These results are important since no previous studies have assessed the role of antibiotic treatment in the reduction of risk of late miscarriage in women with abnormal vaginal flora. In contrast to these results, Carey et al. found no benefit of metronidazole on the incidence of preterm birth (<week 32) in a in a low-risk population.21 One possible explanation for these conflicting results is that therapy was introduced relatively late in pregnancy (mean gestational age week 20).

Table 1.  Meta-analysis of the relative risks of preterm delivery following treatment for bacterial vaginosis
AuthorInclusion weeknTreatmentPreterm delivery in treatment groupPreterm delivery in placebo groupOR (95% CI)Reduction in preterm deliveries (%)
  1. ns, not significant.

Morales et al.1613–2080Metronidazole orally8/4416/360.28 (0.10–0.76)60
McGregor et al.176–27129Clindamycin vaginal9/605/692.26 (0.71–7.16)ns
Hauth et al.1822–24258Metronidazole + erythromycin orally54/17242/860.48 (0.28–0.82)40
Joesoef et al.1914–26681Clindamycin vaginal51/3509/3411.13 (0.74–1.74)ns
McDonald et al.2016–26480Metronidazole orally11/24215/2380.71 (0.32–1.57)ns
Carey et al.2116–231919Metronidazole orally116/953121/9660.97 (0.74–1.27)ns
Kekki et al. 2210–17375Clindamycin vaginal9/1877/1881.31 (0.48–3.59)ns
Kurkinen-Raty et al. 2312101Clindamycin vaginal1/510/501.13 (0.74–1.74)ns
Lamont et al.2413–20409Clindamycin vaginal8/20819/2010.38 (0.16–0.90)50
Guaschino et al.1514–25112Clindamycin vaginal6/498/510.75 (0.24–2.34)ns
Ugwumadu et al.2512–22486Clindamycin vaginal13/24438/2410.30 (0.16–0.58)75
Kiss et al.2615–204229Clindamycin vaginal + clotrimazole48/205888/20970.55 (0.38–0.78)50
All studies—pooled 9506 334/4942368/45640.83 (0.71–0.96)20

Currently, treatment of asymptomatic BV in pregnancy is controversial, with the possible exception of women with a history of preterm birth. Women symptomatic for BV were generally excluded from many trials because they are treated, and these women may be at high risk of adverse pregnancy outcome.

Four meta-analyses of trials involving the screening and treating of BV in pregnancy have been published in the past 4 years.4,27–29 All have showed no reduction in preterm birth with treatment for BV. In the continuing Cochrane Review of antibiotics for treating BV in pregnancy, an analysis of 13 high-quality trials involving 5300 women found that antibiotic treatment was effective in eradicating BV in pregnancy (OR 0.21, 95% CI 0.19–0.24), but did not significantly reduce the risk of preterm birth, PPROM or subsequent preterm birth in women with a previous preterm birth.27 Only in women with a previous preterm birth was the use of antibiotics associated with a decreased risk of PPROM and LBW.

It should be remembered that meta-analyses are liable to numerous biases despite quality control measures, and their results should be interpreted with caution. Varma and Gupta in 2006 undertook a repeat meta-analysis of screening and treating BV in pregnancy, in which they included recently published trials and applied strict criteria on data extraction.30 The results of this analysis suggested a reduction in the incidence of preterm birth following screening and treatment for BV in low-risk women but not in high-risk women, an apparent contradiction as one would normally expect treatment to exert greater risk reduction in the higher risk group. These results could be explained by several factors including the differences in antibiotic sensitivity between high- and low-risk groups, and the hypothesis that preterm birth in high- and low-risk pregnant women are different entities and not linear extremes of the same syndrome. In addition, statistical investigations, such as subgroup analyses or meta-regression, used to correct for heterogeneity, may also be prone to erroneous interpretation and may be unreliable unless large number of studies are available.

One important consideration emerging in recent years is the possibility that the conflicting results of these studies could be related to antimicrobial resistance associated with treatment of BV. The recommendations of the Centers for Disease Control and Prevention for the treatment of BV suggest the use of either metronidazole or clindamycin (oral and intravaginal preparations).31 Despite clinical equivalence in efficacy between intravaginal metronidazole and clindamycin, the two agents produce different antibiotic resistance profiles among anaerobic bacteria present in the vagina after treatment for BV.32 Intravaginal clindamycin use was associated with significant and sustained evidence for anaerobic resistance after treatment, whereas intravaginal metronidazole use was not associated with significant antibiotic resistance yet four studies in low-risk women have shown a statistically significant reduction in the incidence of preterm birth when BV was treated with clindamycin in early pregnancy.24–26,33 Further studies are needed to evaluate the full implications of an increase in the reservoir of resistant bacteria in the vagina.


In this paper, we have outlined the conflicting results from several clinical trials of BV in pregnancy, and considered a number of possible reasons for these discrepancies (gestational age at diagnosis/enrolment, history of preterm birth, choice of antimicrobial agent etc.). The most likely scenario is that only a subgroup of women is truly at risk for infection-associated preterm birth. Infectious diseases and their consequences result not only from the ill-effect of microbial invasion but also the nature of host response. Romero et al. proposed that the maternal immune response is crucial in the different outcome of pregnancy.13 Immunologically hyporesponsive women may not be able to control the microbial burden, and this in turn may predispose to ascending intrauterine infection and clinical chorioamnionitis. In contrast, hyperresponders would develop an excessive local inflammatory response, clinical symptoms of vaginitis, and be at risk for preterm birth if microorganisms gain access to the decidua. The ‘optimal’ host is capable of mounting a measured and proportionate inflammatory response, which can deal with changes in the vaginal ecosystem without paying the price of adverse pregnancy outcome. The genetic profile of these women, the gene/environment interaction and the relationship between changes in the vaginal ecosystem/vaginal inflammation and host cells are interesting areas for future research.