Does Bacterial Vaginosis Cause Pelvic Inflammatory Disease?

Abstract

Pelvic inflammatory disease (PID), the infection and inflammation of the female genital tract, results in serious reproductive morbidity including infertility and ectopic pregnancy. Bacterial vaginosis (BV) is a complex alteration of the vaginal flora that has been implicated in PID. The role of BV in the etiology and pathogenesis of PID has not been studied extensively. Our objective was to extensively review data related to the relationship between BV and PID (n = 19 studies). Several studies found a link between BV and cervicitis, endometritis, and salpingitis. Furthermore, it seems that some BV-associated organisms are associated with PID, whereas others are not. However, studies demonstrating an independent association between BV-associated organisms and PID are sparse. In addition, a causal association between BV and PID has not been established. Prospective studies are needed to further delineate the role of BV in PID, with particular focus on individual BV-associated organisms.

In Brief

Studies have found a link between bacterial vaginosis and pelvic inflammatory disease. However, a causal association between bacterial vaginosis and pelvic inflammatory disease has not been established.

Pelvic inflammatory disease (PID) is the infection and inflammation of the female upper genital tract including the tubes and ovaries (salpingitis) and the uterine lining (endometritis),¹ which can cause serious reproductive sequelae including infertility, chronic pelvic pain, and ectopic pregnancy.² Unfortunately, the diagnosis of PID is often difficult because signs and symptoms vary, may be mild or absent, and are generally based on clinical findings.^1,3 Thus, many women may go undiagnosed.

Pelvic inflammatory disease generally occurs when microorganisms ascend from the lower genital tract to the upper genital tract. Various organisms have been implicated in the etiology of PID including Neisseria gonorrhoeae, Chlamydia trachomatis, Mycoplasma genitalium, and anaerobic and aerobic bacteria commonly associated with “bacterial vaginosis” (BV).^1,3 Although upper genital tract chlamydial and gonococcal infections have been studied extensively, the etiology and pathogenesis of nonchlamydial/nongonococcal PID have not been fully delineated.

The condition BV is characterized by an imbalance in vaginal microflora and occurs when hydrogen peroxide–producing lactobacilli decrease in concentration and are replaced by anaerobic and facultative aerobic bacteria including Gardnerella vaginalis and Mycoplasma organisms.⁴ Bacterial vaginosis is the most common genital infection among reproductive-aged women affecting up to a third of women in the United States.⁴ Prevalence varies widely, with reports ranging from 7% to 51% depending on the patient population.^4–6 Bacterial vaginosis and BV-associated organisms have been identified among women with cervicitis^7–9 and PID.^10–14 However, the role of BV as an etiologic cause of PID is not well understood. The primary objective of this study was to extensively review data that report relationships between BV, PID, and post-PID sequelae. We additionally describe methodological challenges in the study of BV and PID, including the diversity of microbial etiologies and diagnosis.

METHODS

Our primary literature search was conducted with Medline to identify original research articles, from 1950 through 2012, which examined the role of BV in PID. The term “bacterial vaginosis” was combined with “pelvic inflammatory disease.” This yielded a total of 210 articles. Additional searches were conducted using the terms “bacterial vaginosis” and “endometritis,” “upper genital tract infection,” or “salpingitis.” Studies were reviewed for relevance and further limited to human studies written in English. Additional articles were identified by cross-referencing bibliography lists. Last, studies among pregnant women, postabortion studies, case reports, studies among HIV-positive women, and review articles were excluded. This yielded 17 studies that examined the association between BV and PID.

BV IN THE LOWER GENITAL TRACT

The possibility that BV is sexually transmitted¹⁵ and has been associated with cervicitis^7–9 suggests a potential role for BV in PID because microbes infecting the cervix can ascend to the upper genital tract. Marrazzo et al.⁷ in a study of 424 women with BV found that 15% had cervicitis. In addition, cervicitis was associated with factors that were distinct from other sexually transmitted diseases (STDs) but consistent with risk factors of BV, including absence of H₂O₂ producing Lactobacillus (odds ratio [OR], 2.7; 95% confidence interval [CI], 3.3–5.9), female sex partner (OR, 6.2; 95% CI, 1.3–28.3), recent oral sex (OR, 2.3; 95% CI, 1.2–4.2), and increased age (P < 0.001, for trend). Similarly, a study of 100 women found that BV and cervicitis were both significantly associated with a decrease in Lactobacillus species.¹⁶ However, this study was not able to directly link BV to cervicitis. In a study of 297 women attending an STD clinic, cervicitis was associated with BV (P < 0.02).⁹ Schwebke and Weiss¹⁷ conducted a randomized trial of metronidazole gel for the treatment of BV and cervicitis among 75 women attending an STD clinic. Women who had resolved their BV after treatment were more likely to have cervicitis resolution compared with women with persistent BV (100% vs. 75% P = 0.04). However, in the multivariate analysis, this association did not reach significance after adjustment for number of partners in the past 30 days, history of gonorrhea, and douching.

BV AND PID

Bacterial vaginosis–associated bacteria G. vaginalis and Mycoplasma hominis have been shown to cause oviduct damage in animal models.^18,19 In human subjects, BV has been found to be associated with endometritis and salpingitis.^{10–14,20–25} Thus, there is a possible role for BV in PID. However, microbes such as N. gonorrhoeae and C. trachomatis frequently coinfect patients with BV,²⁶ and BV has been shown to increase the risk of incident gonococcal and chlamydial infection.²⁷ Lactobacilli generate products including lactic acid and bacteriocins, suggested to prevent bacterial overgrowth, and thus are important for host defense against sexually transmitted infections (STIs).^26,27 Because women with BV have decreased lactobacilli concentrations, they may be at increased risk for STI acquisition. Bacterial vaginosis–associated microbial products may also modify mucosal immunity, alter local cytokine responses, and increase vaginal PH resulting in increased susceptibility to STIs and possibly increased risk of bacterial ascension.^26,27 These studies suggest that BV may play a direct role in STI pathogenesis by altering mucosal immunity to pathogens. On the other hand, data from the Gynecologic Infection Follow-Through shows co-occurrence of BV and gonococcal/chlamydial infection (OR, 2.83; 95% CI, 1.81–4.42), but no association between baseline BV and incident gonococcal/chlamydial infection (relative risk [RR], 1.52; 95% CI, 0.74–3.13).²⁸ Therefore, it is unclear whether anaerobes and facultative bacteria can cause PID or whether they facilitate the ascension of N. gonorrhoeae or C. trachomatis infection or whether they ascend as a consequence of N. gonorrhoeae or C. trachomatis infection.^28–30 This adds to the complexity of determining a temporal association between BV and PID.

We found that studies examining the role of BV in PID were heterogeneous and used various definitions for PID and BV diagnosis. Table 1 shows a summary of the characteristics of these studies. Pelvic inflammatory disease research is often hampered by diagnostic methods, which are often based on clinical signs such as pelvic pain and are nonspecific. Visual inspection of the fallopian tubes by laparoscopy is considered the gold standard of PID diagnosis; however, this method is invasive, subjective, and not widely used in the United States.¹ Endometrial biopsy is an alternative to laparoscopy, with a sensitivity of 70% to 89% and a specificity of 67% to 92%.^22,31,32 Thus, although endometritis is a good marker of salpingitis, not all women with endometritis have tubal disease.

TABLE 1: Characteristics of Studies Examining BV in PID

Only 3 studies have examined BV in women with laparoscopically confirmed salpingitis (Table 2). Peipert et al.,¹² in a cross-sectional study of 116 women presenting with possible signs of upper genital tract infection, found that BV diagnosed by Amsel criteria was associated with histologically or laparoscopically diagnosed PID after adjusting for race and age (adjusted OR [OR_adj], 3.0; 95% CI, 1.2–7.6). In another study of 102 women presenting with clinically suspected PID, BV diagnosed by Gram stain was found to be a common concurrent condition in women with confirmed salpingitis, being diagnosed in 61.8% (34/55) of cases.¹³ However, BV or intermediate flora was identified in 100% of patients with normal tubes, suggesting that BV is common in women with clinically suspected PID, whether or not they have salpingitis. Last, a study of 45 women with clinical PID found that BV diagnosed by abnormal vaginal fluid succinate/lactate ratio (≥0.4) was associated with plasma cell endometritis and confirmed salpingitis or endometritis alone (P = 0.023).²² This study did not use a standard definition of BV. Furthermore, BV was not found to be associated with salpingitis without endometritis. However, this study was limited by power. None of these small cross-sectional studies were able to adjust for important risk factors for PID, including chlamydia or gonorrhea, which may have confounded the associations. Thus, they are unable to demonstrate if BV can independently ascend to the upper genital tract and cause tubal inflammation.

TABLE 2: Articles Examining the Role of BV in Women With Salpingitis

Several studies have investigated the role of BV in endometritis or clinically suspected PID (Tables 3 and 4).^{10,11,20,21,23–25,29,30,33–35} Only one prospective study has been conducted to date. Ness et al.²⁹ reported, among 1179 women from the Gynecologic Infection Follow-Through study, no increase in the risk of developing incident PID or histologic endometritis over a 3-year period among women with BV, diagnosed by Nugent criteria, after adjustment for chlamydia and gonorrhea (adjusted hazard ratio, 0.89; 95% CI, 0.55–1.45). However, acute carriage of pigmented, anaerobic gram-negative rods was associated with PID. Because bacterial colonies vary among women with BV and it is likely that not all BV-associated bacteria are pathogenic, this may explain the null findings. In fact, further analysis in this cohort found that women with the highest growth of a cluster of BV-associated microorganisms (G. vaginalis, M. hominis, pigmented and nonpigmented anaerobic gram-negative rods, and Ureaplasma urealyticum) were significantly more likely to develop PID (adjusted RR_adj, 2.03; 95% CI, 1.16–3.53).³⁰ Some of these BV-associated organisms would not be recognized by Gram stain. Hillier et al.,³³ in a study of 117 women with clinically suspected PID, found that BV was not independently associated with histologic endometritis. However, anaerobic gram-negative rods (M. hominis, Peptostreptococcus, G. vaginalis, and Mobiluncus) were significantly associated with endometritis independent of chlamydia and gonorrhea (OR_adj, 2.6; 95% CI, 1.1–5.7). Haggerty et al.,¹⁰ in a study of 278 women with complete endometrial histology from the PID Evaluation and Clinical Health study, found an association between BV, diagnosed by Nugent criteria, and acute endometritis independent of chlamydia and gonorrhea (OR_adj, 2.4; 95% CI, 1.3–4.3). Similar to the studies of Ness et al. and Hillier et al., significant associations between endometrial diphtheroids, black-pigmented gram-negative rods, anaerobic gram-positive cocci, and acute endometritis were observed. All 3 studies were conducted in similar populations using the same criteria to diagnose BV and endometritis. Wiesenfeld et al.,¹⁴ in a cross-sectional study of 556 women who were at risk for lower genital infections but did not have symptoms of clinically suspected PID, found that BV diagnosed by Amsel criteria was associated with histologic endometritis (OR_adj, 2.7; 95% CI, 1.02–7.2) after adjustment for proliferative phase of the menstrual cycle, gonorrhea, chlamydia, Trichomonas vaginalis, previous pregnancy, and black race. In a subsequent study from this group that examined 736 women from the same population, BV was significantly more frequent among women with histologic endometritis (P = 0.03).³⁶ After controlling for race, BV was not retained in the multinomial logistic model. Collectively, these cross-sectional studies consistently show an association between BV and endometritis.

TABLE 3: Articles Examining the Role of BV in Histologically Confirmed Endometritis

TABLE 4: Studies Examining BV in Clinical PID

Because BV is a polymicrobial condition and not all microbes may be pathogenic, it may be optimal to examine the relationships between individual BV-associated bacteria and PID. In a pilot study of 50 women from the PID Evaluation and Clinical Health study, Haggerty et al.³⁷ found that fastidious bacteria Leptotrichiasanguinegens/amnionii (62%), Atopobium vaginae (54%), Ureaplasma urealyticum biovar 2 (32%), Ureaplasmaparvum (32%), and BV-associated bacteria 1 (BVAB1) (28%) were common among women with nongonococcal, nonchlamydial endometritis. This small pilot study did not have a control group to examine the association between these fastidious bacteria and PID. Hebb et al.³⁸ examined bacterial phylotypes among 45 women with salpingitis and 44 controls who were seeking tubal ligation. Bacteria were not detected by polymerase chain reaction (PCR) in any of the control patients (0/44) without salpingitis but were identified in 24% (11/45) of patients with salpingitis.

Overall most studies using different patient populations and different diagnostic criteria show that BV is frequently identified among women with PID. However, cross-sectional studies cannot determine if abnormal vaginal flora preceded PID or if BV mediates the ascension of other microbes to the upper genital tract. Thus, there is no strong evidence that BV can independently cause PID. A handful of cross-sectional studies controlling for important confounders have found significant associations between BV determined by Nugent score and acute endometritis. However, a large prospective study was unable to confirm these findings. Additional prospective studies are needed to delineate the relationship between BV and PID. Studies must take into account important confounders such as C. trachomatis, N. gonorrhoeae, and race. Furthermore, because the vaginal microflora may change over time, sampling should be conducted several times throughout the study period. Future studies should focus on BV-associated organisms rather than the more broad definition of BV because it seems that some but not all BV organisms are pathogenic.

REPRODUCTIVE MORBIDITY AFTER BV

Although chlamydia and gonorrhea have been studied extensively in PID, little is known about the role of BV in the development of reproductive sequelae after PID. No studies have directly examined infertility or ectopic pregnancy after BV-associated PID. However, PID causes damage to the fallopian tubes resulting in tubal factor infertility, and several studies have found BV to be associated with tubal infertility.^39–43 Because most studies have been conducted among women undergoing in vitro fertilization, they may not be generalizable. Still, there seems to be associations between BV and tubal factor infertility, although temporality cannot be determined. Prospective studies among women with BV-associated PID are needed to delineate this relationship.

MICROBIOLOGICAL ETIOLOGY AND DIAGNOSIS OF BV

Variation in or absence of symptoms,⁴⁴ together with the diversity of the vaginal microbiota, among women with BV creates difficulty in BV diagnosis. In clinical practice, Amsel criteria are the gold standard for the diagnosis of BV. Positive diagnosis of BV by Amsel criteria requires 3 of the following 4 criterion: (1) homogenous, thin, white discharge; (2) presence of clue cells on microscopic examination (>20% of epithelial cells with adherent bacteria); (3) pH level higher than 4.5; and (4) upon slide preparation, a positive “whiff test” (production of a fishy odor when 10% potassium hydroxide is added).⁴ An alternative commonly used in research studies is Gram stain interpreted using the Nugent score to determine the relative concentration of Lactobacilli, gram-negative and gram-variable rods and cocci (G. vaginalis, Prevotella, Porphyromonas, and peptostreptococci), and curved gram-negative rods (Mobiluncus).⁴⁵ Gram stain is highly reliable and reproducible and has good correlation with clinical signs. However, it requires highly experienced personnel, and although it is the gold standard for research studies, it is not commonly used in clinical practice. Furthermore, this method cannot detect organisms, which cannot be gram stained such as ureaplasmas and mycoplasmas.³⁷

The limitations of clinically diagnosed BV have led to searches for other diagnostic methods. Molecular methods provide a new opportunity for BV diagnosis. Cultivation-independent studies using PCR, amplified from vaginal swabs, have identified a range of novel bacteria associated with BV including A. vaginae, Leptotrichia spp., and 3 new species called BVABs 1, 2, and 3.^37,46,47 Fredricks et al.⁴⁷ found that PCR detection of a combination of BVAB2 or Megasphaera phylotype 1 had a sensitivity of 99% and a specificity of 89% compared with Amsel criteria and a sensitivity of 96% and specificity of 94% compared with Nugent score. In this study, G. vaginalis was a poor predictor of BV because it was also commonly found in women without BV. This suggests that PCR detection of 1 or more of these fastidious bacteria may be better predictors of BV compared with Amsel criteria and Nugent score.⁴⁷ However, these methods remain costly and have not been widely validated for the diagnosis of BV in large populations.¹⁵ Still, molecular techniques could aid in the identification and better understanding of vaginal communities associated with BV. Because inflammatory responses differ depending on the microbe or microbes present, these methods may identify subgroups of women at the highest risk for upper genital tract inflammation.

CONCLUSIONS

Bacterial vaginosis is a prevalent and complex condition that remains poorly understood. Studies suggest that BV may play a role in serious reproductive complications. Bacterial vaginosis has been found to be associated with cervicitis, endometritis, salpingitis, and tubal factor infertility. However, several factors have limited current research examining the role of BV in PID and reproductive tract sequelae and should be considered in the design of future studies. First, studies have been largely cross sectional in nature, and few have controlled for other STIs, making it difficult to determine if BV can independently cause PID or PID-associated sequelae. Second, because BV is not a condition of a single pathogen but a complex alteration of the vaginal ecosystem, studies using molecular techniques for the detection of individual BV-associated bacteria are needed to further delineate the relationship between BV and long-term sequelae. Last, because signs and symptoms of both PID and BV vary widely and are often asymptomatic, better markers of both are needed. Clinically diagnosed BV may miss some potentially pathogenic microbes, and molecular methods are costly and not yet used in clinical settings.

Bacterial vaginosis is frequent among women with PID. However, only one prospective study exists, and it did not confirm this relationship. Additional prospective studies are needed to determine if BV-associated organisms can independently cause PID. Although BV seems to be associated with infertility, temporality has never been determined. Furthermore, infertility has never been examined after BV-associated PID.