Abstract
Background. We sought to assess the relationship between bacterial vaginosis (BV) assessed by Gram stain and incident trichomonal, gonococcal, and/or chlamydial genital infection.
Methods. This longitudinal study included 3620 nonpregnant women aged 15–44 years who presented for routine care at 12 clinics in Birmingham, Alabama. Participants were assessed quarterly for 1 year. Vaginal smears were categorized by the Nugent Gram stain score (0–3, normal; 4–6, intermediate state; 7–10, BV). Pooled logistic regression was used to estimate the hazard ratios for the comparison of trichomonal, gonococcal, and chlamydial infection incidence in participants by Nugent score at the prior visit. Participants were censored at their first visit with a positive test result for trichomonal, gonococcal, and/or chlamydial infection.
Results. Of the 10,606 eligible visits, 37.96% were classified by BV and 13.3% by positive detection of trichomonal, gonococcal, and/or chlamydial infection. An intermediate state or BV at the prior visit was associated with a 1.5–2-fold increased risk for incident trichomonal, gonococcal, and/or chlamydial infection (adjusted hazard ratio [AHR] for intermediate state, 1.41 [95% confidence interval {CI}, 1.12–1.76]; AHR for BV, 1.73 [95% CI, 1.42–2.11]; P=.058 for trend). Estimates were similar for trichomonal-only, gonococcal-only, and chlamydialonly infection outcomes.
Conclusion. BV microbiota as gauged by Gram stain is associated with a significantly elevated risk for acquisition of trichomonal, gonococcal, and/or chlamydial genital infection.
There is mounting evidence that vaginal bacterial communities play an important role in preventing colonization by pathogenic organisms, including those responsible for sexually transmitted infections (STIs), vulvovaginal candidiasis, and urinary tract infections [1–13]. Vaginal bacterial communities differ in species composition [14, 15], and therefore it is also likely that they differ in how they respond to pathogens.
Although the etiology of the clinical syndrome bacterial vaginosis (BV) remains unclear, it is traditionally described as a vaginal microbial community that lacks lactic acid-producing bacteria, mainly Lactobacillus species (spp.), with an overabundance of anaerobic bacteria (including Gardnerella vaginalis, Prevotella spp., Mobiluncus spp., Ureaplasma urealyticum, and Mycoplasma hominis) and elevated vaginal pH [16]. There appears to be a host-specific threshold at which the numbers of specific taxa increase and elicit a host response in the form of vaginal symptoms [17]. It is hypothesized that vaginal lactobacilli play a critical protective role in the vagina by producing bactericidal and virucidal agents, including lactic acid and bacteriocins, which prevent overgrowth of pathogens and other opportunistic organisms [18]. The prevalence rate of BV as determined by Gram stain analysis of vaginal fluid ranges from 29% in the United States [19] to >50% in Ugandan villages [20].
Several longitudinal observational studies have demonstrated that BV or absence of vaginal lactobacilli, as assessed by culture or Gram stain, is an independent risk factor for STIs such as human immunodeficiency virus (HIV), human papillomavirus (HPV), and herpes simplex virus (HSV) infections [6, 8, 11, 21, 22]. Most recently and notably, Schwebke et al [10] demonstrated in a small randomized trial that treatment and prophylaxis of asymptomatic women with atypical Gram stain smears resulted in a decreased risk for incident chlamydial genital infection compared with that of a group of control women who were under observation. Addition of this new clinical trial evidence to the existing observational studies strongly supports the hypothesis that BV microbiota is causally associated with the acquisition of STIs.
We are aware of only a few studies that have prospectively assessed the association of vaginal microbiota with the incidence of Neisseria gonorrhoeae, Chlamydia trachomatis, and Trichomonas vaginalis infection; however, these studies were conducted among relatively high-risk women, including sex workers, women attending STI clinics, or women at risk for unplanned pregnancies [10, 11, 22, 23]. Because prior studies were conducted among high-risk populations, we sought to evaluate the role of the vaginal bacterial environment as a biological risk factor for trichomonal, gonococcal, and chlamydial genital infection in a large, prospective study of 3620 reproductive-age women who were recruited during routine health clinic visits.
Methods
This is a secondary analysis from the Longitudinal Study of Vaginal Flora (LSVF), which has been described elsewhere [24]. Briefly, from 1999 through 2002, nonpregnant women aged 15-44 years (n=3620) who presented for routine care at 1 of 12 clinics in Birmingham, Alabama, were recruited to a longitudinal observational study. Participants were assessed quarterly for 1 year (5 study visits). Women were ineligible if they had significant medical or gynecological conditions, were planning to move from the area in the next 12 months, or had conditions precluding informed consent.
At each visit, women underwent a detailed interview in a private office with a female interviewer. The interview was focused on lower genital tract symptoms, personal hygiene, and sexual behavior; information on demographic factors, stress, and substance use was also obtained. Data on substance use were collected by self-report; participants were informed that we had obtained a Certificate of Confidentiality to assure that the data were not subject to subpoena. Additionally, stress was measured using the Cohen 10-item perceived stress scale [25, 26]. Participants underwent a standardized clinical assessment, a pelvic examination, and STI testing at each visit. Pelvic examinations were conducted by nurse practitioners who underwent specific training for this study. Participants self-reported racial classification using standard predefined categories. Racial information was captured because of known disparities in sexually transmitted disease rates between ethnic groups in the United States [27].
An endocervical swab was used to inoculate Thayer-Martin agar plates for culture of N. gonorrhoeae and C. trachomatis, and testing was performed by ligase chain reaction (Abbott Laboratories). The presence of T. vaginalis was determined by means of a positive finding after either culture (In-Pouch) or microscopic evaluation for trichomonads. Vaginal smears were assessed using the Gram staining criteria of Nugent et al [28]. The Nugent score reflects the diagnosis of BV [29] by evaluating the numbers of Lactobacillus spp. (large, gram-positive rods), Gardnerella spp. (gram-negative coccobacillary organisms), and Mobiluncus spp. (thin, curved gram-variable rods) morphotypes in each oil-immersion microscope field. By convention, a score of 0–3 is considered normal, a score of 4–6 indicates an intermediate state, and a score of 7–10 is regarded as indicating BV. Ten percent of samples were independently evaluated in a different laboratory; the k coefficient for the classification of BV was 0.81. The Nugent Gram stain test is a relatively objective research tool that is easily obtained in large cohort and field-based studies [19, 30]. Diagnosis of candidiasis was determined by potassium hydroxide smear and abnormal vaginal discharge and/or vulvovaginal erythema and/or edema. Treatment for candidiasis was determined by self-report of over-the-counter antimycotic treatments or prescribed medications at the clinical visit. The interviewers and laboratory personnel were blinded to results from other measures.
Pooled logistic regression was used to estimate the hazard ratio [31] for the comparison of trichomonal, gonococcal, and chlamydial infection incidence by vaginal microbiota state at the prior visit (Nugent scores of 7–10 and 4–6 vs 0–3). Pooled logistic regression is equivalent to a time-dependent Cox regression analysis [31]. Participants were considered censored at their first visit at which they tested positive for STI so as to evaluate the first observable infections in the study. Analyses evaluated outcome by combined trichomonal, gonococcal, and chlamydial infections and by a single STI (gonococcal only, trichomonal only, and chlamydial only) with respective censoring by each outcome infection status. A secondary analysis excluded intervals when participants reported no sex partners.
Factors collected from questionnaires that had been identified on the basis of previous findings, biological plausibility, and preliminary univariable analyses were evaluated as possible confounders. In addition, the regression analyses were performed with backward and forward selection procedures to verify the significant predictors of incident BV. Variables that were believed to be relevant, even if they did not achieve statistical significance, were retained in the final model. Data were analyzed using Stata/SE software for Windows (version 10.0; Stata).
The study was approved by the institutional review boards of the University of Alabama at Birmingham; the Jefferson County, Alabama, Department of Health; and the Eunice Kennedy Shriver National Institute of Child Health and Human Development. All participants provided written informed consent.
Results
Screening. The screened population (n=10,945) was 67% African American (7345 women), 26% white (2820 women), and 7% other (780 women), whereas the enrolled population was 82% African American (11,067 women), 17% white (2349 women), and 1% other (162 women). The majority of women screened were ineligible due to age (160 women [28%]) and pregnancy (161 women [28%]). However, we do not have additional information on screened women because of restrictions placed by the institutional review board.
Demographics. Participants were predominantly young (7841 [58%] were aged 15–25 years) and African American (11,067 [81.5%]); 5298 (42%) of participants had a monthly household income of $800–$3000. Table 1 displays the baseline demographics for women enrolled in the cohort by STI acquisition status. Women who did not acquire a trichomonal, gonococcal, or chlamydial infection during the study tended to be older, to report earning a high school diploma or General Educational Development (GED) credential, and to have higher monthly income and fewer numbers of sex partners.
Of the 13,591 total number of visits prior to censoring, 5468 (40%) were classified by a Nugent score of 7–10 and 2454 (18%) were classified by a positive STI test result. After censoring at the first visit at which STI was diagnosed (10,659 visits remaining), 4026 visits (38%) were classified by BV as assessed by the Nugent score and 1410 visits (13%) were classified by positive test results for trichomonal, gonococcal, and/or chlamydial infection.
Loss to follow-up. Among the 3620 women who enrolled in the study, the median length of follow-up was 3 visits. Retention was 85% (3077 women), 75% (2710 women), 65% (2338 women), and 51% (1844 women) for study visits 2–5, respectively. There were missing data for the covariates—primarily number of sex partners (47 observations), vaginal douching (55 observations), and condom use (167 observations)—used in statistical modeling on 290 participants (328 observations). Because the outcome of interest was incident infection, participants were excluded from the analysis if they attended only a baseline visit (n=543), which resulted in a final sample size of 3077 participants. In addition, 679 women had missed scheduled visits in that they did not return for a follow-up visit within the scheduled time frame but did return for a visit during the following interval.
Incident STIs.Table 2 displays the number of incident STI cases and exposure intervals that were available for modeling after censoring at the first visit at which STI was diagnosed. Of the 2454 positive STI test results in the cohort prior to censoring, 729 were detected at baseline and excluded. The remaining cases that were excluded were among participants with another positive STI test result.
In an analysis that combined trichomonal, gonococcal, and chlamydial infection into 1 STI outcome, a Nugent score of 4– 6 and a score of 7–10 were both associated with incident STI diagnosis at the next study visit in univariable (Table 2, model 1) and multivariable models (Table 3, model 1). A Nugent score of 4–6 and a score of 7–10 were associated with a 1.4–1.7-fold increased risk for STI acquisition (adjusted hazard ratio [AHR] for Nugent score of 4–6, 1.41 [95% confidence interval {CI}, 1.12–1.76]; AHR for Nugent score of 7–10, 1.73 [95% CI, 1.42– 2.11]).
Trichomonal, gonococcal, and chlamydial infection were also considered as individual STI outcomes. Estimates were similar for all individual STI outcomes. A Nugent score of 7–10 and a score of 4–6 both conferred a 1–2-fold increased risk for incident trichomonal, gonococcal, and chlamydial infection (Tables 2 and 3). There was also a statistically significant doseresponse trend for increasing Nugent score in multivariable modeling for incidence of trichomonal-only infection (Table 3).
Age, African American ethnicity, and vaginal douching were significant covariates associated with STI outcomes and were included in multivariable modeling (Table 3). In addition, chlamydial and trichomonal infection were significant risk factors for gonococcal infection incidence, as was gonococcal infection for chlamydial infection incidence and both gonococcal and chlamydial infection for trichomonal infection incidence.
After intervals with no reported sexual activity were excluded, estimates remained similar (data not shown). In another secondary analysis in which pH at the prior visit was used as the exposure of interest (instead of Nugent score), a pH of >4.6 was found to be highly associated with incident trichomonal, gonococcal, and/or chlamydial infection in an adjusted analysis compared with a reference pH of <4.0 (AHR for pH of 4.1– 4.5, 1.22 [95% CI, 0.83–1.79]; AHR for pH of 4.6–5.0, 1.34 [95% CI, 0.97–1.85]; AHR for pH of 5.1–5.5, 1.96 [95% CI, 1.38–2.78]; AHR for pH of >5.6, 1.67 [95% CI, 1.20–2.33]). Similarly, BV diagnosed by the Amsel clinical criteria (also in place of Nugent score as the exposure of interest) at the prior visit was associated with a 25% increased risk for trichomonal, gonococcal, and/or chlamydial infection (AHR, 1.25 [95% CI, 1.05–1.49]).
Discussion
In this prospective, observational study of 3620 reproductiveage women, we found that BV diagnosed on the basis of Gram stain analysis of vaginal fluid was associated with a 1.4–2-fold increased risk for acquisition of gonococcal, chlamydial, or trichomonal genital infection. Our data are consistent with those from other studies of high-risk women.
Numerous longitudinal studies have found that BV or absence of vaginal Lactobacillus spp. are associated with incident STI regardless of whether vaginal microbiota were measured by Gram stain or culture [6, 11, 22]. In a longitudinal study of 657 female sex workers in Kenya, Martin et al [11] reported that intermediate and BV states as assessed by Gram stain were associated with incidence of trichomoniasis (AHR, 1.8 [95% CI, 1.3–2.4]), and absence of vaginal lactobacilli on culture was associated with the acquisition of gonococcal infection (AHR, 1.7 [95% CI, 1.3–2.4]) and HIV infection (AHR, 2.0 [95% CI, 1.2–3.5]). BV-associated microorganisms have also been found to be associated with a 2-fold increased risk for both incident HSV-2 infection (hazard ratio, 2.1 [95% CI, 1.0–4.5]) [6] and development of pelvic inflammatory disease (rate ratio, 2.03 [95% CI, 1.16–3.53]) [32]. Schwebke et al [10] demonstrated in a randomized trial of 107 women that treatment and prophylaxis of asymptomatic women who had atypical Gram stain smears resulted in a significantly decreased incidence of chlamydial infection compared with that among a group of control women who were under observation. Wiesenfeld et al [5] also found in a cross-sectional study of women who reported recent exposure to a male partner with urethritis that BV was a strong predictor of gonococcal infection (odds ratio, 4.1 [95% CI, 1.7–9.7]) and chlamydial infections (odds ratio, 3.4 [95% CI, 1.5–7.8]).
In contrast to these studies, a large prospective study, the GYN Infections Follow-Through (GIFT) Study, found the association between BV and incident gonococcal and chlamydial infection was not statistically significant (adjusted relative risk, 1.52 [95% CI, 0.74–3.13]) [23]. However, the point estimate (1.52) is similar to those noted in the present study for gonococcal and chlamydial infection. There may have been increased power to detect significant associations with sexually transmitted disease outcomes in the LSVF, as the GIFT study observed 46 cases of incident gonococcal and chlamydial infection and the LSVF modeled the analysis on 176 cases of gonococcal infection and 678 cases of chlamydial infection.
The hazard ratios for incident trichomonal infection also demonstrated a dose-response trend for increasing risk with higher Nugent scores. A study of 7918 pregnant women reported by Hillier et al [33] found trichomonal infection was more strongly associated with intermediate Gram stain smears in a cross-sectional analysis. The data suggest the epidemiology of intermediate status may be unique. We hypothesize that T. vaginalis may influence its environment. In vitro data suggest that T. vaginalis grows better in an elevated pH [34], and BV provides this high-pH environment. Both the intermediate state and the BV state are associated with decreased abundance of protective lactobacilli [15]. Recent data from Ravel et al [15] also suggest that intermediate Nugent scores may represent a bacterial community of unique taxa as compared to higher Nugent scores.
Molecular studies of the vagina based on the analysis of 16S ribosomal RNA gene sequences have shown that in both women presenting with BV and those without BV, the diversity, composition, and relative abundance of microbial species in the vagina varies dramatically between women [14, 35] and that a significant proportion of women lack appreciable numbers of Lactobacillus spp. [36]. Lactobacilli have historically been considered keystone species of vaginal communities in reproductive- age women. Lactobacilli benefit the vagina by producing lactic acid as a fermentation product, and it accumulates in the vagina and is associated with a reduced pH [18]. Low pH and high lactic acid concentration are hypothesized to inhibit BVassociated bacteria and STIs [37]. Lactobacilli also produce bacteriocins— low-molecular-weight proteins that can inhibit the growth of a variety of bacteria, thus reinforcing the protective role of lactobacilli [38–40]. Vaginal communities that lack Lactobacillus spp. may have high numbers of other lactic acid- producing bacteria such as Atopobium spp., Megasphaera spp., or Leptotrichia spp. [36]. Despite the ability of these bacteria to produce lactic acid, the community resilience and infection susceptibility profiles may not be equivalent to those of lactobacilli, which may account for the different microbial community states that are observed between women [15]. Adding complexity to the problem is the fact that the functional differences between the many different species (or strains) of lactobacilli, the most dominant of which include Lactobacillus crispatus, Lactobacillus iners, Lactobacillus gasseri, and Lactobacillus jensenii, are not known and our study based on Gram stain was not able to decipher them. As part of the Human Microbiome Project (http://nihroadmap.nih.gov/hmp/) [41], genomic studies of the vagina are expected to describe the structure of the complex microbial communities and how they contribute to disease susceptibility. Finally, BV may predispose to acquisition of STIs upon exposure because the local cytokine production associated with BV may facilitate the acquisition of STIs [42].
Strengths of our study include the prospective design, which enabled us to evaluate temporal ordering of Gram stained vaginal smears and subsequent incidence of STIs, collection of detailed data on sexual behaviors, and demonstration of a doseresponse relationship between Nugent score and STI outcome. Unlike the high-risk populations in other studies, this study population was drawn from women who presented for routine health care, which may allow the findings to be more generalizable than those of previous studies. In addition, the prevalence of BV among African American and white women in this study was similar to rates reported from nationally representative surveys in the United States [19].
A limitation regarding the geneA limitation regarding the generalizability of the study population, however, is the fact that it was composed of women who predominantly were young, were African American, had lower incomes, and were not high school graduates. A critical limitation of this study was the wide sampling interval of 3 months between study visits. Several longitudinal studies have shown that vaginal microbiota fluctuate rapidly and over the time course of the menstrual cycle [43–46]. More frequent sampling is necessary to address this time-sensitive sampling issue; however, the interval-censored sampling presented in this study is suggestive, because women with BV episodes often have recurrent episodes [47]. In addition, although the characterization of the vaginal microbiota was not immediately proximal to the STI acquisition, a woman who lacks vaginal Lactobacillus spp. is likely to persist or enter that community state again over the time course of the menstrual cycle [44]. The hazard ratios obtained in this epidemiological study are indeed only estimates, but they are highly indicative of a strong association between BV microbiota and subsequent STI acquisition. It would also be very challenging to achieve the necessary sample size to observe STI outcomes in a study of daily vaginal microbiota sampling, and this interval-censored cohort provides a reasonable estimate of the risk. The results of this analysis were based on Gram stain and would be appreciably bolstered by additional molecular studies, such as information of the bacteria, fungi, and viruses that compose the vaginal microbiome.
Another limitation of this study is the fact that one of the strongest epidemiological risk factors for BV is sexual activity [48]; therefore, BV may be a marker for having sex partners who are also harboring STIs. Unmeasured confounding may account for the association between BV and STIs. We adjusted for number of sex partners and use of condoms, but high-risk sexual activity or membership in a core group of STI may account for our findings.Women who reported consistent condom use were also observed to have increased risk for incident STIs, confirming that condom use may be subject to self-reporting bias [49], error in use [50], or misuse in high-risk situations or that the condom or its components have an effect on the vaginal microbiome.
In summary, vaginal microbiota consistent with BV by Gram stain analysis is significantly associated with increased risk for STI acquisition. Given the strength of evidence in this large, prospective study combined with the findings from a randomized trial that treatment and prophylaxis of asymptomatic women with atypical Gram stain smears was associated with decreased rates of incident chlamydial genital infection [10], consideration and future research should focus on whether intervention should be recommended to women who are at high risk for STIs and who present with BV or BV-associated microbiota. There also remains an urgent need to developmore effective interventions for BV because the recurrence following current treatments is disappointingly high [47]. Future research that utilizes molecular phylogenetic technologies will facilitate our understanding of microbial communities and species that are more protective of the vagina and confer protection from STIs. We envision that better knowledge of vaginal microbial communities will lead to the development of interventions, such as antibiotic regimens and probiotic or prebiotic treatments, to drive the vaginal microbiota toward healthier states that are more resilient to pathogens.
References
Figures and Tables
Participant Characteristics at Study Entry by Sexually Transmitted Infection (STI) Acquisition during Observation in the Longitudinal Study of Vaginal Flora in Birmingham, Alabama, 1999–2003 (n=3077)
Number of Incident Sexually Transmitted Infection (STI) Cases and Unadjusted Hazard Ratios (HRs) for Risk of Incident STI by Nugent Score at Prior Visit
Adjusted Hazard Ratios (AHRs) between Bacterial Vaginosis at Prior Visit and Incident Sexually Transmitted Infection (STI)
Potential conflicts of interest: none reported.
Presented in part: 18th Biennial Congress of the International Society for Sexually Transmitted Disease Research, London, UK, 28 June to 1 July 2009 (poster P3.43).
Financial support: National Institute of Allergy and Infectious Diseases (grant K01-AI080974); Eunice Kennedy Shriver National Institute of Child Health and Human Development (contract NO-1-HD-8-3293 and intramural fund Z01 HD002535-10).
