Bacterial Vaginosis and Alcohol Consumption: A Cross-Sectional Retrospective Study in Baltimore, Maryland

Abstract

Background

Bacterial vaginosis (BV) is the most cited cause of vaginal complaints among women of reproductive age. Its etiology and associated risk factors are not entirely understood. Here we examined the association between BV and at-risk alcohol consumption in women attending two sexually transmitted infection (STI) clinics in Baltimore, Maryland.

Methods

This was a retrospective cross-sectional analysis utilizing data from first clinic visits from 2011–2016. At-risk alcohol use was defined as heavy episodic (“binge”) drinking within the last 30 days or self-report of having had vaginal or anal sex in the context of alcohol consumption. Pearson Chi-square test and Student’s t-test were used to assess baseline associations. Log binomial models were used to estimate prevalence ratios (PR) before and after adjustments for potential confounding factors.

Results

Of the 10,991 women included in the analysis, 2,173 (19.7%) met the clinical diagnostic criteria for BV. Having had vaginal or anal sex in the context of alcohol consumption was associated with an increased risk of BV (PR 1.25 (95% CI: 1.13–1.37)), as was binge drinking (PR 1.15 (95% CI: 1.04–1.27)) after adjustment for confounders.

Conclusions

In this population, at-risk alcohol consumption was associated with an increased risk of BV. The mechanisms remain uncertain. Future prospective studies are needed to verify and evaluate causality in these associations.

SUMMARY

A study of women attending sexually transmitted infection clinics in Baltimore, Maryland found a significant positive association between at-risk alcohol use and bacterial vaginosis.

INTRODUCTION

Bacterial vaginosis (BV) is a disorder characterized by a decrease in Lactobacillus spp. colonization of the vagina with a concomitant expansion of a variety of anaerobic bacteria (1, 2). In about 50% of affected women, BV is associated with symptoms such as vaginal discharge and odor (2, 3). The condition is common: prevalence amongst US women is approximately 30%, with African-American women (estimated prevalence 51.4%) disproportionately impacted (4). Symptomatic BV generally responds to antibiotic treatments, however recurrence is common. Approximately 40% of women may report symptomatic recurrence within 3 months of completing treatment and nearly 60% experience recurrence within 12 months (5). BV is associated with adverse outcomes such as recurrent urinary tract infections and an increased risk for acquiring sexually transmitted infections (STIs), including HIV (6–10). As such, BV continues to present research and treatment challenges in both epidemiologic studies and clinical settings. The causes of and risk factor profile for BV remain poorly understood, although associations with increased frequency of sexual activity, greater number of sexual partners, inconsistent condom use, hormonal contraception use and vaginal hygiene practices such as douching have been described (10–15).

Alcohol use is one factor which could plausibly contribute to an increased risk for BV, either through a biobehavioral or possibly a direct biological effect. At-risk alcohol use, (including frequent or heavy episodic ((i.e. “binge”) drinking and sexual activity in the context of alcohol use) has been linked to disinhibition and increased sexual risk behaviors such as condomless sex or sex with multiple partners (16–19), which in turn have been associated with increased risk of BV (13, 20, 21). In addition, evidence from animal studies suggests that very heavy alcohol consumption might contribute to BV susceptibility through direct alteration of the vaginal microbiome. In one study in female primates, after 3 months of daily binge level alcohol consumption there was a significant increase in vaginal colonization with BV-associated bacteria (22). In humans, relatively few studies have examined relationships between alcohol consumption and BV. Results from existing literature are conflicting, with some studies finding no association (23–29) and others finding increased BV risk (21, 30–35s). Only one study reported a decreased odds of BV in women reporting alcohol use (36s), though a second (37s) found that women who reported any alcohol use within the previous week showed a decreased risk of certain BV-associated bacteria, though not of BV diagnosis overall. Most studies did not specifically examine at-risk alcohol use, and these conflicting results may relate to heterogeneity in alcohol use definitions, populations studied, and study design (See Supplementary Table 1).

Given potential associations between at-risk alcohol use and BV and the limitations of the existing literature, we performed an analysis to assess the relationship between BV prevalence and at-risk alcohol use (defined as either 1. heavy episodic (“binge”) drinking, or 2. as having had vaginal or anal sex in the context of alcohol consumption within the last month), adjusting for relevant confounders, among women seen at two inner city STI clinics. We hypothesized that at-risk alcohol use would be associated with an increased risk of BV.

MATERIALS AND METHODS

Study design and setting:

A retrospective cross-sectional analysis was performed using data from a clinical database associated with two STI clinics located in Baltimore, Maryland. This study was approved by the Johns Hopkins School of Medicine Institutional Review Board under Protocol NA_00002614.

Data collection, timeframe and participants:

As part of routine care, clinically relevant sexual history and demographic data were collected by clinicians on each patient (See Table 1). Clinicians entered these data into a standardized electronic medical record form at the time of visit. Women attending the clinics were asked about STI symptoms (including BV symptoms), STI history, reason for visit, and medical and behavioral factors such as hormonal contraception use, substance use, condom use (without differentiating site or mechanism of exposure), number of sexual partners in the last 6 months and alcohol use. We restricted our analysis to the first clinic visit for each cis-gender woman older than age 15. Analyses were further restricted to data from February 2011 to December 2016, as alcohol use variables were most consistently collected during this period. During this timeframe, all patients were asked 1. whether they had “vaginal or anal sex with alcohol use” within the past 30 days and 2. whether they had any episodes of binge drinking (defined as consuming >4 alcoholic drinks in any one drinking occasion from 2011–2013, and >=4 drinks in any one drinking occasion from late 2013 onwards) within the past 30 days. Women were then evaluated with physical exams and laboratory testing as necessary. On exam, a clinical diagnosis of BV was made if at least three of the four following (Amsel’s) criteria were met: (1) a thin, homogenous vaginal discharge, (2) a vaginal pH of >4.5, (3) the presence of clue cells on saline microscopy, and (4) a positive whiff test, where the addition of 10% KOH to vaginal discharge produces a fishy odor (3). Starting in 2013, asymptomatic women could be eligible for ‘express testing’ without physical exam, utilizing microscopy for trichomonas, and nucleic acid amplification testing for gonorrhea and chlamydia.

Table 1.

Prevalence of bacterial vaginosis by demographic and reproductive and sexual risk factors among 10,991 Baltimore women, 2011–2016

Characteristic	BV-Negative N=8818 (%)	BV-Positive N=2173 (%)	P value
Mean Age (SD)	28.3 (11.21)	28.1 (9.78)	0.30
Race			<0.0001
White	913 (10.35)	144 (6.63)
Black/African American	7201 (81.66)	1924 (88.54)
Other	704 (7.98)	105 (4.83)
Patient reported vaginal or anal sex in the context of alcohol consumption in last 30 days			<0.0001
Yes	1087 (12.33)	367 (16.89)
No	7731 (87.67)	1806 (83.11)
Patient reported heavy episodic (“binge”) drinking in last 30 days ‡			<0.01
Yes	1179 (13.37)	349 (16.06)
No	7639 (86.63)	1824 (83.94)
Contraception Type			<0.0001
Condom	1607 (18.22)	481 (22.14)
Diaphragm	4 (0.05)	0 (0.00)
Spermicidal foam	1 (0.01)	3 (0.14)
IUD	389 (4.41)	111 (5.11)
Pill	607 (6.88)	93 (4.28)
Ring	69 (0.78)	11 (0.51)
Tubal ligation	415 (4.71)	152 (6.99)
DMPA	424 (4.81)	81 (3.73)
Emergency contraceptive pill	7 (0.08)	1 (0.05)
Patch	37 (0.42)	18 (0.83)
Other	438 (4.97)	104 (4.79)
None	3817 (43.29))	1028 (47.31)
Unknown	1003 (11.37)	90 (4.14)
Number of male sexual partners in last six months			<0.0001
0	696 (7.89)	88 (4.05)
1	4689 (53.18)	1131 (52.05)
≥2	3433 (38.93)	954 (43.90)
Patient reported average condom use			<0.0001
Never	2910 (33.00)	766 (35.25)
Sometimes	2718 (30.82)	689 (31.71)
Often	1630 (18.48)	449 (20.66)
Always	1098 (12.45)	232 (10.68)
Not reported	462 (5.24)	37 (1.70)

^‡Defined as consuming >4 alcoholic drinks in any one drinking occasion from 2011–2013, and >=4 drinks in any one drinking occasion from late 2013 onwards.

IUD=intrauterine device, DMPA=depot medroxyprogesterone acetate

Statistical analysis and variables:

Conventional descriptive statistics such as the Pearson Chi-square test for ordinal or categorical variables and Student’s t-test for continuous variables were used to assess baseline variables of interest. Race/ethnicity was collected by self-report. For the univariable and multivariable analyses, race/ethnicity variables were dichotomized to Black/African American or White/Other due to a small number of patients in remaining categories. Hormonal contraception use was defined to include contraceptive pill, patch, ring, injections and oral emergency contraception. BV was defined based on the diagnosis variable recorded by clinicians at the end of each visit.

Factors which were considered clinically relevant, known to be associated with BV, or had a P value <0.2 in the univariable analysis were included in the multivariable analysis. Due to the high prevalence of BV (19.8%) in the study population at the time of clinic visit, we used a log binomial model with robust variance to estimate prevalence ratios. This regression model was used to estimate prevalence ratios and 95% confidence intervals (CI) before and after adjustments for potential confounding factors. We explored potential interactions between alcohol use by race by including interaction terms in the multivariable log binomial model.

We carried out sensitivity analyses to determine the effect of symptoms such as discharge, dysuria, genital odor, genital itching, rashes, lesions, and/or abdominal pain on BV diagnosis. Log binomial models limited to those who reported STI related symptoms (either all symptoms, or only those symptoms most characteristic of BV: i.e. genital discharge and odor) and controlling for relevant confounders were fitted to assess associations with BV diagnosis. Additionally, log binomial models using all subjects and including a symptoms variable in addition to relevant confounders (again, either all symptoms or only genital discharge and odor) were fitted to assess associations with BV diagnosis. In another sensitivity analysis, we included women with IUDs in the hormonal contraception category in multivariable models as there was no distinction between copper or hormonal IUDs on clinic questionnaires. We determined the effect of missing contraception use data, first by classifying women for whom hormonal contraception (HC) data were missing as using HC and secondly by classifying them as not using HC in multivariable models. Finally, microscopy results were less reliably recorded than BV diagnosis by clinicians. We conducted a sensitivity analysis excluding patients who did not have data recorded on whiff test, clue cells, trichomonas or yeast on microscopy of vaginal smears (similar to previous definitions (38s)) to assess whether this could have biased or impacted our results. All analyses were conducted using STATA v.15 (College Park, TX).

We summarized studies reporting on alcohol consumption and bacterial vaginosis including data on study design, location, population, BV measurement, alcohol intake measurement and study results in Suppl. Table 1.

RESULTS

From February 2011–2016, there were 11,803 first visits from women over age 15 years. We excluded N=812 (6.9%) of visits due to missing information on either age, race, or number of sexual partners, leaving visits from N=10,991 available for analysis. Of these women, N=2,173 (19.7%) were diagnosed with BV. Overall, women with BV were more likely to self-identify as Black or African American, were less likely to report using hormonal contraception, had more sexual partners and were more likely to report never using condoms than women without BV (See Table 1). N=501 (23.1%) of those with BV and N=1531 (17.4%) of those without BV were diagnosed with urogenital chlamydia, gonorrhea, trichomonas or with herpes simplex virus at the visit (p<0.01). N=27 (1.2%) of those with BV and N=190 (2.2%) of those without BV (p=0.01) were positive for HIV (either established or based on new result from current visit). Overall, N=1528 (13.9%) of women reported binge drinking in the last 30 days and N=1454 (13.2%) reported vaginal or anal sex in the context of alcohol use. Of women reporting vaginal or anal sex in the context of alcohol use, 99.5% reported vaginal sex and 13.7% reported anal sex.

In multivariable modelling, there was a statistically significant association between self-reported vaginal or anal sex in the context of alcohol use in the last 30 days and increased risk for BV with an observed prevalence ratio (PR) of 1.25 (95% CI: 1.13–1.37, p=<0.01) (See Table 2). In this model, Black/African American race, and increased number of sexual partners in the last 6 months were also associated with increased risk for BV, while hormonal contraception use and “always” using condoms were associated with decreased risk. In multivariable modelling there was also a statistically significant association between binge drinking in the last 30 days and increased risk of BV (PR 1.15 (95% CI: 1.04–1.27, p=0.01), with similar results for race, increased number of sexual partners, hormonal contraception and condom use (See Table 3).

Table 2.

Sex in the context of alcohol use and other select sociodemographic and sexual risk factors associated with BV among 10,991 Baltimore women, 2011–2016

	Univariable Model		Multivariable Model
	PR	95% CI	PR*	95% CI*
Patient reported vaginal or anal sex in the context of alcohol use in last 30 days
No	1.00	Ref	1.00	Ref
Yes	1.33	1.21–1.47	1.25	1.13–1.37
Black/African American §	1.60	1.42–1.81	1.59	1.41–1.80
Hormonal contraception use †
No	1.00	Ref	1.00	Ref
Yes	0.69	0.60–0.79	0.72	0.63–0.82
Unknown	0.37	0.31–0.46	0.42	0.34–0.52
Number of sexual partners in last six months
0	1.00	Ref	1.00	Ref
1	1.73	1.41–2.12	1.37	1.11–1.68
≥2	1.94	1.58–2.38	1.51	1.22–1.85
Patient reported average condom use
Never	1.00	Ref	1.00	Ref
Sometimes	0.97	0.89–1.06	0.94	0.86–1.03
Often	1.04	0.93–1.15	1.00	0.90–1.11
Always	0.84	0.73–0.96	0.83	0.73–0.95
Not reported	0.36	0.26–0.49	0.48	0.35–0.66

^*Adjusted for age (Age non significant, p=0.08).

^§For this analysis, race was dichotomized. Black/African American women were compared to a single reference category including all other races. This was due to small sample sizes in some race categories.

^†Pill, patch, ring, depot medroxyprogesterone acetate, and emergency contraceptive pill were included as hormonal contraception.

PR= prevalence ratio, CI= confidence interval, BV= bacterial vaginosis

Table 3.

Heavy episodic drinking and other select sociodemographic and sexual risk factors associated with BV among 10, 991 Baltimore women, 2011–2016

	Univariable Model		Multivariable Model
	PR	95% CI	PR*	95% CI*
Patient reported heavy episodic (“binge”) drinking in last 30 days ‡
No	1.00	Ref	1.00	Ref
Yes	1.18	1.07–1.31	1.15	1.04–1.27
Black/African American §	1.60	1.42–1.81	1.60	1.41–1.81
Hormonal contraception use †
No	1.00	Ref	1.00	Ref
Yes	0.69	0.60–0.79	0.72	0.63–0.82
Unknown	0.37	0.31–0.46	0.42	0.34–0.51
Number of sexual partners in last six months
0	1.00	Ref	1.00	Ref
1	1.73	1.41–2.12	1.38	1.13–1.70
≥2	1.94	1.58–2.38	1.54	1.25–1.90
Patient reported average condom use
Never	1.00	Ref	1.00	Ref
Sometimes	0.97	0.89–1.06	0.94	0.86–1.03
Often	1.04	0.93–1.15	0.99	0.89–1.11
Always	0.84	0.73–0.96	0.83	0.73–0.95
Not reported	0.36	0.26–0.49	0.47	0.34–0.65

^*Adjusted for age, (Age non significant, p=0.07).

^‡Defined as consuming >4 alcoholic drinks in any one drinking occasion from 2011–2013, and >=4 drinks in any one drinking occasion from late 2013 onwards.

^§For this analysis, race was dichotomized. Black/African American women were compared to a single reference category including all other races. This was due to small sample size in some race categories.

^†Pill, patch, ring, depot medroxyprogesterone acetate and emergency contraceptive pill were included as hormonal contraception.

PR= prevalence ratio, CI=confidence interval, BV= bacterial vaginosis

Interactions explored between race and binge drinking, and race and sex with alcohol to assess modification of alcohol intake by race were non-significant. Interactions between binge drinking and sex with alcohol were also non-significant. Sensitivity analyses to assess missing contraception data, hormonal contraception categorization, and the impact of symptomatology on overall BV diagnosis did not meaningfully change results. Finally, excluding patients without data recorded on whiff test, clue cells, trichomonas or yeast on microscopy of vaginal smears left N=7824 patients available for analysis, 27.7% of whom were diagnosed with BV. Multivariate analyses restricted to these patients yielded consistent results for associations between reported vaginal or anal sex in the context of alcohol use (PR: 1.14, 95% CI 1.05–1.26, p=0.01) and BV as well as binge drinking (PR: 1.11, 95% CI 1.01–1.23, p=0.03) and BV.

DISCUSSION

Our study suggests that there is a significant positive association between at-risk alcohol consumption (here defined as binge drinking and sex in the context of alcohol use) and BV, when controlling for race, hormonal contraception use, number of sexual partners, and condom use. There is a relative paucity of studies that assess the relationship between alcohol consumption and BV (see Suppl. Table 1), and fewer still that specifically assess the relationship between at-risk alcohol consumption and BV. Our results are in line with four other studies which reported associations between heavy/ frequent alcohol use and BV (21, 30, 32s, 33s). Only one other study (Jain et al.) has assessed associations between BV and binge alcohol use in women. While this did not find a significant association, it was conducted in a predominantly Latina population of female sex workers who inject drugs, very different from our study population (in whom <1.5% reported exchanging sex for money or drugs and <1% reported IV drug use in the last year)(25).

Two previous studies did not find an association between sex in the context of alcohol use and BV. However, the first, (Jain et al. (25)) involved a very different study population as outlined above. The other study (Rugpao et al. (28)) examined a variable that combined sex in the context of drug with alcohol use, and also involved a very different study population-i.e. young Thai women seeking family planning services. Our study is significantly larger (N=10,991) than either of these two studies (N=1522 (Rugpao) and N=584 (Jain)).

Mechanisms by which binge or heavy alcohol use or sex with alcohol use may lead to enhanced risk for BV are understudied. It is possible that the effect is mediated through sexual risk behaviors. Specifically, at risk alcohol consumption is known to be associated with sexual risk behavior such as forgoing condom use and increased sexual partners, both associated with an increased risk for BV (13, 20, 21). Non-human primate models suggest the possibility for a direct biologic mechanism governing the relationship between heavy alcohol consumption and BV as well. In a study of rhesus macaques, females were treated daily for three months to a blood alcohol concentration between 50 and 60 mM, consistent with intoxication concentrations in humans, showed decreased Lactobacillus colonization of the vagina and an increase in gram-positive cocci compared to controls treated with isocaloric sucrose (22). In humans, alcohol consumption has been previously associated with a decrease in oral lactobacilli (39s), while animal studies have shown that significant alcohol consumption can lead to an increase in gut lactobacilli (40s). It is possible that alcohol-induced changes in the gut microbiome could influence the composition of the rectal microbiome. Some studies have cited possible concordance between the gut or rectal microbiome and the microbiome of the vagina (41s–43s), indicating that women with increased rectal carriage of Lactobacilli have a decreased risk for BV (41s). Alcohol consumption has also been shown to negatively impact the immune system and gut barrier function (44s–46s). Gut-rectal-vaginal concordance could contribute to the explanation of these study findings, but possible pathogenic mechanisms need further exploration.

Strengths of our study include its large population size, as well as the ability to control for multiple potential confounders. However, the study has several limitations. Due to its cross-sectional nature, causal associations cannot be assessed, nor reverse causality excluded. Clinic questionnaire responses are subject to recall and social desirability bias as well as selection bias such as misclassification and underreporting. Alcohol use was ascertained by a clinician based on self-report which could have led to underreporting; no biological markers of alcohol use were available. Furthermore, neither clinician alcohol use query further quantified the amount of alcohol use so it was not possible to assess associations between heavier versus lighter alcohol use associated with sex or with binge drinking and BV. The majority of the study population comprised of African American women living in the inner city, and results may not be generalizable to other settings. Clinical (Amsel’s) criteria were used to diagnose BV, and it is possible that using other methods e.g. Nugent scoring or molecular methods would diagnose more asymptomatic BV yielding additional insights. Associations with same sex activity were not explored and impacts of progesterone only versus combined estrogen and progesterone contraceptives were not analyzed. Finally, no data was available on smoking or vaginal hygiene practices such as douching, which could confound or modify, respectively, the relationship between at risk alcohol consumption and BV. Prior literature has found positive correlations between both smoking and douching and an increased risk for BV(15, 23).

Despite these limitations, our study contributes to an existing body of literature that suggests that heavy/frequent or binge alcohol use may contribute to an increased risk for BV, and suggests that sex in the context of alcohol use may also contribute to increased BV risk. Future larger and prospective studies are needed to assess potential dose-response relationships, underlying biological mechanisms, and interpersonal factors.