Smoking in women with chronic vaginal discomfort is not associated with decreased abundance of Lactobacillus spp. but promotes Mobiluncus and Gardnerella spp. overgrowth – secondary analysis of trial data including microbio­me analysis

Summary

Background: Smoking is considered a risk factor for bacterial vaginosis. It is currently unknown which parameters of the vaginal environment are affected and how smoking triggers the disease.

Aim of the study: The primary objective is to estimate the effect size of smoking on vaginal pH and the Nugent score in patients with chronic vulvovaginal discomfort prior to the development of episode of vaginosis. The secondary goal is to investigate the effect of smoking on individual microscopic parameters of the vaginal environment and on subjectively reported symptoms of vaginal discomfort.

Methods: Smoking reported by patients was tested as a predictor, using multivariate logistic and ordinal logistic regression analysis on a dataset from the first visit of a randomized trial NCT04171947, which enrolled patients with intermediate vaginal environment. We tested the primary hypothesis (odds ratio (OR) for vaginal pH > 4.5 and Nugent score > 3 in smokers) at the significance level á = 5%. For exploratory analyses of the effect of smoking on the parameters of the vaginal environment, á was corrected as per Bonferoni.

Results: In a cross-sectional sample of 250 women after adjusting for other risk factors, smoking had an impact on the Nugent score (OR = 3.3 (1.3–8.5), P = 0.011), while pH was not affected (OR = 1.2 (0.5–2.8), P = 0.698). Smoking was associated with the prevalence of clue cells (P < 0.000), Gardnerella spp. (P = 0.001) and Mobiluncus spp. (P = 0.001), while the prevalence of Lactobacillus remained unchanged (P = 0.049).

Conclusion: Contrarily to common assumptions, vaginal Lactobacillus is not directly affected by smoking, which rather promotes the growth of bacteria of Gardnerella and Mobiluncus spp. Given that other parameters remained unaffected, it appears that smoking leads to vaginal dysbio­sis by creating specific favourable conditions for these two opportunistic  pathogens.

Keywords:

Bacterial vaginosis – smoking – Nugent score – vaginal pH – vaginal flora – Lactobacillus – Gardnerella – Mobiluncus – human microbiome

Introduction

Vaginal microflora dominated by Lactobacillus species represent the main pillar of mucosal homeostasis [1–4]. Along with Lactobacillus, less frequent elements coexist on the vaginal mucosa including bacteria, yeast, protozoa and blood cells [2,5,6]. In healthy women, lactic bacteria actively maintain low pH by anaerobic glycolysis of mucosal glycogen deposits which, along with the production of hydrogen peroxide, restricts the spread of other strains [2,5]. Opportunistic pathogens are thus present in latent numbers unless the conditions favour their overgrowth [6]. The stability of such an ecosystem is inversely proportional to the community  diversity [3,5].

Contrarily to lay assumption, vaginal mucosa is a dynamic environment. Natu­ral fluctuations in the composition relate to pregnancy and menstruation. In the course of life, vaginal mucosa is also repeatedly exposed to secretions, hormonal changes, and external influences; some of which have immediate negative effect on the ecosystem stabi­lity [7]. In healthy women, however, the microenvironment presents a tendency to spontaneously normalize towards Lactobacillus-dominated microflora [5] and short episodes of bacterial infection resolve spontaneously [7].

The specific conditions allow other strains – typically Mobiluncus and Gardnerella species, to compete with symbio­tic Lactobacillus. Albeit often asymptomatic [8], roughly one third of patients with this intermediate flora progress to bacterial vaginosis [7,9] during which the bacteria exceed 100- to 1,000-fold the numbers seen in healthy women [1].

Globally, bacterial vaginosis is the most common vaginal infection among women of reproductive age [10] having a significant impact on their quality of life [11]. Bacterial vaginosis is also responsible for problematic conception [12,13] adverse pregnancy outcomes [6,8], and increased risk of sexually transmitted infections [14–16], which imposes a large economic burden on society [10].

The treatment strategies for bacterial vaginosis are few [17] and they fail in over one half of patients [17,18]. Moreover, the repeated use of antibio­tics represents a risk of the spread of resistant species. From the perspective of public health, the prevention of bacterial vaginosis is vital. Several risk factors have been identified and their avoidance is encouraged via health education [19]. In the common practice, smoking is considered one of them, although the epidemiological evi­dence is restricted to exploratory analyses burdened with multiple testing. Hellberg et al [20] showed roughly three-fold risk of bacterial vaginosis in smo­kers. The dia­gnosis was determined using Amsel criteria not allowing to uncover the impact of smoking on individual parameters. More detailed results were reported by Alnaif et al [21] describing the association between smoking and semiquantitative score developed by Nugent et al [22]. Again, smoking increased the risk about three-fold. Bradshaw et al [23] showed a dose-dependent relationship between the amount of cigarettes smoked and the risk of increased Nugent score in women and their female partners. Taken together, we know that smoking is associated with bacterial vagi­nosis in observational studies but we do not know what parameters of va­ginal environment are affected and how is vaginosis triggered. We do not know if smoking favours pathogenic bacteria, if it restricts the symbio­tic organisms and protective leukocytes, or if it acts directly on the mucosal pH.

Material and methods

Our primary hypothesis is that, among patients visiting outpatient gynaecology clinics for vulvovaginal discomfort, smo­kers have higher odds of Nugent score of > 3 and vaginal pH > 4.5. We further test the relationship between self-reported current smoking, individual subjective symptoms and individual microscopic parameters of the vaginal microenvironment.

This is a secondary analysis of the data from the first visit of the clinical trial MAT072017 [24] (ClinicalTrials.gov Identifier: NCT04171947) which took place in 14 outpatient gynaecology clinics in the Czech Republic in 2018. We included premenopausal patients aged 18 to 55 attending outpatient gynaecology for vulvovaginal discomfort. The subjects had negative pregnancy test, were not breastfeeding, with no antibio­tics during the previous month, no bleeding of unknown aetiology, no severe infection requiring antibio­tic treatment, and without diabetes mellitus or any oncological condition. For more details, refer to the Clinical Trials registry.

At the first visit, the patients signed informed consent and consent to the processing of personal information. The patients then filled in questionnaires, rated their symptoms on the scale 0 to 4, and underwent initial general examination, their medical history was assessed, pH measurements were taken, and wet mounts were prepared. In all patients, pH was determined using uniform gynaecology test strips (MColorpHast, Merck, Germany) according to a predefined procedure. Wet mounts were prepared from the posterior vaginal wall using a cotton swab. Fixed and blinded slides (4-digit codes) from all study centers were transported to a central accredited laboratory (private accredi­ted laboratory AeskuLab k.s., Prague). One slide was stained by Giemsa-Romanowsky and the other by Gram stain. The blinded slides were assessed by a trained microbio­logist using an oil immersion lens, providing total magnification of 1,000×. The following elements were quantified by a microbio­logist: squamous epithelial cells; clue cells; mixed flora; yeast as pseudomycelia; Gram neg. diplococci; fibrous Lactobacillus; Gardnerella; spirochetes; parabasal epithelial cells; leukocytes; yeast as blastospores; Gram pos. cocci in chains; Lactobacillus; Mobiluncus; Leptotrichia and Trichomonas vaginalis. The Nugent score was determined, according to Nugent et al [22] (Fig. 1). Pointwise semiquantitative rating was performed according to the laboratory standardized procedure. Briefly, the number of elements per field were categorized into 5 points. For Mobiluncus and Gardnerella, 0 = none; 1~tens; 2~hundreds, 3~thousands; and 4 means abnormal infection. For Lactobacillus, 0 = none; 1~10 to 15; 2~15 to 50, 3~50 to 100; and 4 means over 100 bacteria per field. All ratings were done by the same microscopists. Unclear findings were resolved by another independent expert.

Current smoking was reported with the patient by the gynaecologist via an electronic case-report form based on the interview with the patient. The smoking status recorded in the electronic case report forms was verified during monitoring visits against the patient’s medical record; any inconsistency was reconciled with the gynaecologist.

The hypothesis was tested in the cross-sectional sample using multivariate logistic regression (binary outcomes) and ordinal logistic regression (pointwise outcomes) at á level 0.05. The same was used to test the relationship between smoking and individual symptoms. The association between smoking and individual microscopic parameters was tested at á level 0.005 (corrected as per Bonferoni). The estimates were adjusted to the prevalence of patient-reported psychological distress, BMI, age, recent menstruation, recent coitus, multiple male sexual partners, soap with normal pH, regular usage of tampons and presence of a female sexual partner, using multivariate regression analysis. These factors were considered potential confounders based on a) the baseline differences between smokers and non-smokers and b) risk factors previously published in the literature. The data was complete thanks to thorough monitoring of the trial MAT072017. No imputation was done.

We report in line with the STROBE statement [25].

Results

Of 274 patients enrolled in 14 outpatient gynaecology clinics, the data of 24 were excluded from the analysis due to repeated non-compliance of one centre. Of 250 patients attending out-patient gynaecology for vulvo­­- va­ginal discomfort fulfilling the abovementioned criteria, 136 had vaginal pH > 4.5 (107/201 non-smokers vs. 29/49 smokers). Seventy-two subjects had Nugent score > 3 (51/201 non-smokers vs. 21/49 smokers) (Fig. 2). Smokers reported psychological distress (depression, anxiety, bereavement, divorce, job difficulties) roughly twice more often. The remaining characteristics did not differ significantly between smokers and non-smokers (Tab. 1).

Comparing the differences in outcomes between smokers and non-smokers, we identified higher mean Nugent score and more severe discharge in smokers. The vaginal microenvironment in smokers was characterized by higher abundance of clue cells, Gardnerella species and Mobiluncus species (Tab. 2). The multivariate adjustment showed that smoking was associated with a 3.3-fold (1.3–8.5) higher odds of abnormal Nugent score (adjusted to psychological distress, BMI, age, recent menstruation, recent coitus, multiple male sexual partners, soap with normal pH, usage of tampons and presence of a female sexual partner) (Tab. 3). The estimate remains significant even after further adjustment to previous pregnancy, vaginal douching, allergy to pessary or condoms, use of hormonal contraception, antibio­tics, antimycotics or corticosteroids (P = 0.04).

Increased vaginal pH was not associated with subjectively reported smoking (P = 0.577). None of the subjective symptoms are associated with smoking. Among 15 microscopic parameters, only the presence of clue cells, Gardnerella spp. and Mobilluncus spp. are associated with smoking. The reduction in Lactobacillus content is not associated with smoking.

Discussion

Smoking is known to largely affect human microbio­me, notably with respect to oral, airway and gut health [26]. We show that, in addition to bacterial vaginosis [20], smoking is also associated with abnormal microflora during chronic recurrent vulvovaginal discomfort.

To date, a specific role of smoking in the aetiology of bacterial vaginosis has not been described. A pilot project using 16S rRNA gene sequencing in vaginal swabs of 20 smokers [27] aimed to identify the species that are most influenced by smoking. The study suggests that smoking correlates with decreased mucosal colonization by symbio­tic Lactobacillus spp. and smoking cessation may help restore normal microflora in women already suffering from bacterial vaginosis. Contrarily to this finding, smoking was not related with vaginal pH, nor with the decrease in Lactobacillus content in our cohort of women suffering from recurrent vaginal discomfort. Our results indicate that smoking generates favourable conditions for pathogenic bacteria rather than unfavourable conditions for symbio­tic Lactobacillus. Adjusted regression analysis based on the semi-quantitative microscopy shows that Gardnerella and Mobiluncus content increases in smokers along with the frequency of clue cells. Our result complements previous findings and shows that smoking promotes Gardnerella and Mobiluncus growth prior to the development of the clinical episode of bacterial vaginosis.

Our cohort was selected with the aim to represent a typical patient consulting the outpatient gynaecology clinic. Although roughly 70% women in our cohort had suffered with previous episodes of bacterial vaginosis, the swabs were taken prior to the development of the clinical episode or between the episodes. Thus, the conclusions drawn should not be extrapolated to bacterial vaginosis patients. Another limitation is that our sample consisted exclusively of Caucasian women which may somehow limit generalisation in other ethnic groups in which the microbio­me dynamic follows different rules [4,28]. Third, inherent to observational research, our conclusion relies on the assumption of no unmeasured confounding which cannot be proved. A randomized study of smoking, however, is and will not be achievable with respect to the ethical considerations.

Conclusion

To conclude, we hope to have presented a piece of evidence on the role of smoking in the development of vagi­nal dysbio­sis. A recent metabolomic study [29] shows increased concentration of nicotine metabolites and other amine compounds on the vaginal mucosa of smokers. Biogenic amine compounds have the ability to increase the virulence of latent opportunistic strains that are present even in healthy women. Further research should focus on the molecular mechanism and triggers responsible for Gardnerella and Mobiluncus overgrowth, notably with respect to smoking-cessation strategies such as nicotine-replacement therapy.

Declaration of interests

J.T. and B.F. are employees of Value Outcomes, a consultancy and research organization that designed, monitored and conducted the primary randomized trial on behalf of the sponsor Nature Laboratories Ltd., T.D. is the director of Value Outcomes. J.M. is the director of the la­boratory that performed blinded analysis of the vaginal swabs, J.K. is a co-owner of the Nature Laboratories Ltd. that provided data from the randomized trial MAT072017. There are no competing interests relevant to the analysis, conclusion or interpretation of the data.

Submitted/Doručeno: 27. 10. 2020

Accepted/Přijato: 8. 1. 2021

Jan Tužil

Value Outcomes Ltd.

Václavská 12

120 00 Prague

Czech Republic

jan.tuzil@valueoutcomes.cz