Abstract
Summary
The cause of bacterial vaginosis remains controversial. The two leading hypotheses are that Gardnerella vaginalis is the specific aetiologic agent versus the polymicrobial hypothesis that G. vaginalis acts in concert with other bacteria, principally anaerobes, to produce the disease. Here we reassess the prevailing polymicrobial hypothesis, finding it conceptually problematical. No host initiating factors as postulated have been identified, nor is there conclusive evidence that G. vaginalis lacks diagnostic specificity for the disease. The polymicrobial hypothesis, presupposing a unique disruption of vaginal microecology, is inconsistent with the epidemiological profile of the syndrome, which is that of a sexually transmitted disease. The epidemiological and clinical similarities between bacterial vaginosis and trichomoniasis suggest a similar pathogenetic process; i.e., primary causation by a specific agent, with secondary anaerobic activation and resultant amine production.
Introduction
Bacterial vaginosis (BV) is the most common vaginal infection worldwide and has been linked to serious public health consequences including preterm birth, postoperative infections and acquisition of human immunodeficiency virus (HIV) by women.1,2 Despite extensive study of the altered vaginal bacterial environment associated with BV, the microbiological cause remains controversial. Until the pathogenesis is clarified it will not be possible to improve the diagnosis, treatment and prevention of this disease. The two leading hypotheses of causation are that Gardnerella vaginalis (originally designated Haemophilus vaginalis) is the specific and sexually transmitted aetiological agent3,4 versus the polymicrobial aetiological hypothesis that G. vaginalis acts in concert with other bacteria, principally anaerobes, to produce the disease. 5 The latter hypothesis has held sway for more than two decades. However, it is imprecise in its formulation, leaving a number of questions unanswered. In this article, we assess the polymicrobial aetiological hypothesis in the light of current knowledge.
Overview of the Polymicrobial Hypothesis
Numerous studies have confirmed that G. vaginalis can be cultured from the vaginal fluid of virtually all women having a diagnosis of BV by standardized criteria. Other associated bacteria, recovered at variable rates, are the genital mycoplasmas and various anaerobic organisms including species of peptostreptococcus, porphyomonas, prevotella, bacteroides and mobiluncus. 5 Recent additions to the multitudinous list of associated bacteria are Atopobium vaginae and previously unrecognized species identified by molecular analysis. 6 Theoreticians of the polymicrobial hypothesis acknowledge that G. vaginalis is essential to the occurrence of BV but hold that this generically unique bacterium is not intrinsically pathogenic. It is postulated that various host factors may promote alterations in the vaginal flora, with bacteria normally present in very low concentrations gaining dominance over lactobacilli. 5 According to this concept, the floral changes of BV are produced by a disruption of the vaginal microbial ecosystem. The ecological concept of pathogenesis though widely accepted, is recognized to be imperfect, as it does not adequately explain how the infection is acquired. 7
Unmasking the polymicrobial hypothesis
Two major events led to the hypothesis that BV has a polymicrobial aetiology, the first being demonstration of the role of anaerobic bacteria in the clinical manifestations of the disease. Careful microbiological evaluation of the vaginal flora in BV revealed increased concentrations of anaerobes. 8 Further work showed that the characteristic vaginal odour can be attributed to amine production as a byproduct of anaerobic metabolism. 9 It is believed that these amines raise the vaginal pH and are involved in creating the characteristic vaginal discharge.
It is clear that anaerobic activity is instrumental in producing the signs and symptoms of BV. It is less certain that anaerobes are capable of acting as primary pathogens in the intact, mature vagina. Cited as evidence that G. vaginalis requires the help of anaerobes to initiate development of BV are the results of vaginal inoculation experiments in monkeys. 10 However, the discharges produced in the monkeys by the combined inoculation of G. vaginalis and Mobiluncus spp. did not contain any clue cells. The presence of clue cells is distinctively indicative of the disease; 11 hence, the results of the monkey experiments cannot be regarded as valid evidence for polymicrobial causation of BV. More importantly, increased anaerobic concentrations are not specific to BV but are also associated with other vaginal conditions, particularly trichomoniasis. 12 It therefore is unjustified to assume a priori that the association of anaerobes with BV is one of cause and effect.
A recent study employing fluorescence in situ hybridization showed that a distinctive feature of BV is the presence of a bacterial biofilm adherent to the vaginal epithelium. 13 The discovery of this biofilm explains the appearance of clue cells in vaginal fluid: they are biofilm-coated epithelial cells desquamated from the epithelial surface. Although the biofilm was shown to contain high concentrations of a variety of bacterial groups, G. vaginalis was found to be the predominant constituent. This finding is in accordance with the results of a previous study of the bacteria associated with clue cells. 14 These studies confirm that G. vaginalis plays the central role in clue cell formation and raise the possibility that other bacteria in the amalgam are opportunistic secondary intruders.
The second event leading to the polymicrobial hypothesis was the finding that G. vaginalis could be cultured in women judged to lack signs of vaginitis. 15 The assumption was then made that G. vaginalis is not the specific causative agent of BV. The lack of a microbiological gold standard for diagnosis posed by that reasoning led to proposal of a syndromic definition of BV (Amsel criteria). 16 With the advent of large clinical trials relative to potential adverse sequelae, more broadly reproducible criteria were considered necessary, leading to adoption of the Nugent Gram stain scoring system 17 for research purposes. These methods, employed as diagnostic standards, have proved useful in the study of BV, particularly in identifying adverse sequelae and obtaining epidemiological data. Clearly, however, they are arbitrary schemes. They provide confirmation of the diagnosis but discount the likely existence of a spectrum of disease activity as in most genital tract infections, which often produce no signs or symptoms. Thus, the appropriateness of using an arbitrary clinical definition of BV to evaluate the specificity of G. vaginalis for the disease is open to question.
Since its introduction, the polymicrobial hypothesis has governed most reported studies on BV and its public health impact. As yet, however, no host factors have been identified that initiate BV. In recent years, a number of studies have been conducted on vaginal douching as a possible predisposing factor. However, the data on the relationship between douching and BV are limited to cross-sectional studies and a causal effect has not been demonstrated. 18 Although BV is characterized in part by a reduction of H2O2-producing lactobacilli, the loss of lactobacilli may be a consequence of the floral changes rather than a cause, as the anaerobic vaginal environment of BV is not conducive to lactobacillary dominance. 1
Epidemiology of BV
A major conceptual problem with the polymicrobial hypothesis is its inconsistency with epidemiological data. Studies of risk factors show that the risk profile of BV mirrors that of a sexually transmitted disease (STD). Although hormonal associations and other possible factors have been explored, the dominant theme in studies of risk factors is sexual activity. In a study comparing behavioural factors associated with BV with those associated with vaginal candidiasis, women with BV reported sexual behaviours that characterize women with STDs whereas women with candidiasis did not. 19 Behavioural risk factors associated with BV include a high number of recent male sex partners19–23 and a high number of lifetime partners.19,21,24,25 Studies of whether condom use is protective against BV have shown inconsistent results, perhaps explained by recall bias in self-reported condom use. 19 Strongly suggesting that BV is sexually transmitted, exposure to a new partner is a major risk factor for acquiring a new infection 26 and continued exposure to the current partner is a major risk factor for recurrent infection. 27 In keeping with the idea that BV is an STD, co-infection with STDs such as trichomoniasis, gonorrhoea and chlamydial infection is quite common.2,28 Further, recently reported data from the National Health and Nutrition Examination Survey, in addition to verifying the relation to sexual activity, confirmed the previously observed association of BV with non-white race and low socioeconomic status, 29 sociodemographic variables that are characteristic of STDs. 20
Other studies support sexual transmission of BV. In particular, studies of lesbian couples show high concordance rates of BV between partners.30,31 Also, a significant relationship between non-gonococcal urethritis in males and BV in their sex partners has been reported. 32
Frequently cited as evidence against BV being a true STD, are studies that failed to show a decrease in recurrence rates from treatment of male partners. However, in view of the high recurrence rates following antibiotic therapy and the lack of a generally accepted aetiological agent it is difficult to interpret such findings. 19 The other main barrier to acceptance of BV as an STD is its reported occurrence in ostensibly sexually inexperienced women.29,33,34 However, those studies suffer from lack of confirmation of a virginal state in the subjects. Contravening those findings, complete absence of BV in virgins has been reported. 16
Analogies Between Bvand Trichomoniasis
BV and trichomoniasis exhibit a number of features in common. The remarkable similarities between the two diseases argue against the ecological view of the pathogenesis of BV. Epidemiologically, the risk profile of BV is that of an STD, as discussed in the preceding section. Clinically, both infections are characterized by a malodorous, though visibly dissimilar, vaginal discharge. The vaginal pH typically is elevated in trichomoniasis 35 as it is in BV. Reduction of lactobacilli is also a feature of both infections. As previously noted, both infections are associated with increased vaginal concentrations of anaerobes, 12 leading to high concentrations of amines. 36 The ‘whiff’ test for amine production, a component of the Amsel criteria for BV diagnosis, may also be positive in women with trichomoniasis. 35
The most striking potential clinical manifestation shared by the two infections is the rare occurrence of vaginitis emphysematosa in immunocompromised women. 37 This condition is characterized by gas-filled blebs in the vaginal mucosa. Impairment of immunity may allow superficial penetration of the mucosa by anaerobes in some women, giving rise to the mucosal gas pockets.
Finally, both infections are associated with certain serious complications. Like BV, trichomoniasis has been implicated in preterm birth and acquisition of HIV. 35 An association of both diseases with postoperative infections has also been observed. 38 Lower genital tract infections including BV and trichomoniasis are associated with elevated vaginal levels of cytokines, which have been linked to chorioamnionitis and premature birth. 35
Conclusion
A critical assessment of the polymicrobial hypothesis of BV causation reveals it to be founded on premises of questionable validity. In view of the compelling evidence that BV is sexually transmitted, the associated anaerobic bacteria, rather than being causative agents, may simply be opportunistic pathogens activated by the presence of a specific aetiological agent as in trichomoniasis.