Douching cessation and molecular bacterial vaginosis: a re-analysis of archived specimens

Abstract

Objectives:

Observational studies demonstrate an association between vaginal douching and bacterial vaginosis characterized by Gram stain. We sought to describe the effect of a douching cessation intervention on the composition and structure of the vaginal microbiota and molecular-BV, a state defined by low levels of Lactobacillus spp. evaluated by molecular tools.

Methods:

33 women self-collected mid-vaginal swabs twice weekly (982 samples) during a douching observation phase (4 weeks) followed by a douching cessation phase (12 weeks) in a 2005 single crossover pilot study conducted in Baltimore, Maryland. Vaginal microbiota were characterized by 16S rRNA gene amplicon sequencing (V3-V4) and clustered into community state types (CSTs). Conditional logistic regression modeling allowed each participant to serve as their own control. Wilcoxon signed-rank tests were used to evaluate changes in microbiota between phases. Broad-range qPCR assays provided estimates of bacterial absolute abundance per swab in a sub-sample of seven participants before and after douching. A piecewise linear mixed effects model was used to assess rates of change in bacterial absolute abundance before and after douching.

Results:

There was no statistically significant change in the odds of molecular-BV versus Lactobacillus-dominated CSTs comparing the douching cessation interval to douching observation (aOR 1.77, 95% CI 0.89 to 3.55). Removal of L. iners-dominated CST III from the outcome did not affect the results. There were no significant changes in the relative abundance of four Lactobacillus spp. and no meaningful changes in other taxa investigated. There was no significant change in bacterial absolute abundance between a participant’s sample collected 3 days prior to and following douching (p=0.46).

Conclusions:

In this pilot study, douching cessation was not associated with major changes in vaginal microbiota. Douching cessation alone may not durably shift the vaginal microbiota and additional interventions may be needed to restore optimal vaginal microbiota among those who douche.

INTRODUCTION

Vaginal douching is one of many intravaginal practices (IVPs) commonly practiced worldwide. Data from the 2001 U.S. National Health and Nutrition Examination Survey indicated approximately 20% of women douche.¹ Douching prevalence was higher among non-Hispanic Black women (50%),¹ which may be attributed to targeted marketing and cultural norms.^{2 3} Commonly cited reasons for douching are to feel clean, remove menstrual blood, and reduce vaginal odor and discharge.^{4 5} Observational studies have suggested that douching is associated with increased risk for bacterial vaginosis (BV), HIV, pelvic inflammatory disease, and changes to the vaginal microbiota.^6–8 However, it is less clear whether douching cessation, particularly in the absence of other behavioral or biomedical interventions, will favor an optimal vaginal microbiota.

Prior interventions to reduce douching and IVPs have largely focused on the effect of cessation on BV diagnosed by Nugent’s Gram stain score (termed “Nugent-BV”⁹), with only some studies finding a significant reduction in Nugent-BV following cessation among certain groups,^10–12 such as those who douche to remove menstrual blood.¹³ Few studies have evaluated the effect of cessation on the vaginal microbiota using higher resolution molecular techniques, such as 16S rRNA gene amplicon sequencing.¹⁴ Nugent score uses microscopy to estimate the concentration of Lactobacillus, Gardnerella/Bacteroides, and Mobiluncus morphotypes, and is a useful and inexpensive tool for research studies; however, it is limited in its capacity to identify all bacterial species associated with BV.⁹ Molecular methods can provide higher resolution taxonomic assignment and assess BV outcomes defined by low-Lactobacillus states (termed “molecular-BV”).⁹ Molecular-BV broadly includes vaginal bacterial communities which are lacking Lactobacillus spp., and have higher abundance of facultative anaerobic bacteria such as Gardnerella vaginalis, Atopobium vaginae, and “Ca. Lachnocurva vaginae” (previously known as BVAB1), among others.

No studies of douching cessation and BV have been conducted in U.S. populations using high-throughput molecular techniques. The use of specific intravaginal products, frequency, and motivations for douching vary between countries,^{4 5} and a recent observational study by Sabo et al found differences in the effect of vaginal washing on the concentration of specific vaginal bacteria across populations.¹⁵ Among 85 Kenyan participants, any vaginal washing was associated with increased detection of several bacterial taxa associated with BV including “Ca. Lachnocurva vaginae”, BVAB2, G. vaginalis, and A. vaginae, while no changes were observed for 26 American participants.¹⁵ This highlights the need to evaluate douching cessation interventions in multiple populations. We leveraged data and samples from a 2005 pilot crossover study^{13 16} and sought to estimate and describe the effect of 12 weeks of douching cessation on the vaginal microbiota.

MATERIALS AND METHODS

Sample collection

Thirty-three women with any reported vaginal douching in the prior two months were recruited at the Johns Hopkins Bayview Medical Center and completed the study between December 2005 and March 2007. The study began on day one of the menstrual cycle with a four-week douching observation phase (“D phase”), followed by a 12-week douching cessation phase (“DC phase”). Participants were instructed to continue with typical douching behaviors in the D phase, and to discontinue the use of all feminine hygiene products in the DC phase. Samples following a final four-week interval, in which participants decided whether or not to resume douching, were not available for analysis. Study procedures, including eligibility criteria and baseline screening (pregnancy, Chlamydia trachomatis, and Neisseria gonorrhoeae), trial enrollment, and loss to follow-up, have been previously detailed in Brotman et al.¹³ The protocol was approved by the Institutional Review Board of the Johns Hopkins University School of Medicine. All participants provided written informed consent.

At baseline, participants completed demographic and behavioral questionnaires. Throughout the study, participants self-collected vaginal swabs twice weekly, created vaginal smears for Nugent scoring (0–3: Normal, 4–6: Intermediate, 7–10: BV), and completed daily health and sexual behavior diaries. Slides and diaries were sent by postal mail to the study site weekly. Vaginal swabs were stored in the participant’s home freezer and transported to the clinic at the end of each phase. Swabs were archived at the study site at −80° C.

Sample processing

Whole genomic DNA was extracted from vaginal swabs as described in Gajer and Brotman et al,¹⁶ and DNA extracts were stored at −80°C until 2018. A two-step PCR reaction was used to amplify the V3-V4 regions of the 16S rRNA gene and amplicon pooling, sequencing on the Illumina HiSeq 2500, and sequence data processing were carried out using methods described in Holm et al.¹⁷ Amplicon sequence variants (ASVs) generated by DADA2 were classified using the RDP Naïve Bayesian Classifier trained with the SILVA v128 16S rRNA gene database as implemented in the DADA2 R package. ASVs of major vaginal taxa were assigned species-level annotations using speciateIT (http://ravel-lab.org/speciateit/). Bacterial taxa present at less than 10⁻⁵ across all samples were removed and samples with fewer than 1,000 reads were removed from analysis. Community state types (CSTs) were assigned using VALENCIA,¹⁸ an algorithm which assigns samples to a CST based on their similarity to the centroid of each reference CST as determined from a set of over 13,000 vaginal microbiota profiles.

Five primary CSTs were identified in this study; four dominated by Lactobacillus species (L. iners, L. crispatus, L. gasseri, L. jensenii), and one lacking significant numbers of lactobacilli and characterized by higher proportions of strict and facultative anaerobic bacteria (termed CST IV, including IV-A, IV-B, and IV-C). Total bacterial absolute abundance per swab was determined using broad-range qPCR for a subset of samples from seven participants who reported douching once.

Statistical analysis

Conditional logistic regression with cluster-robust standard errors was used to model the within-woman odds of having molecular-BV (CST IV-A, IV-B, or IV-C) versus a Lactobacillus-dominated microbiota (CST I, II, III, or V) in DC phase compared to D phase. Models were adjusted for any same-day sexual activity and menses within 3 days of sampling reported on daily diaries, and participants with missing data on these covariates were excluded from adjusted models. The conditional model allows each participant to serve as their own control, eliminating confounding due to time-invariant covariates.

We conducted a number of secondary analyses, including repeating some models described in the original Nugent score analysis.¹³ First, we restricted the analysis to participants who reported that their primary reason for douching was to remove menstrual blood, as douching and menstrual bleeding could present interacting effects on the vaginal microbiota. Next, because L. iners may represent a Lactobacillus species which may not fully acidify the vagina or produce high D-lactic acid concentrations as other vaginal Lactobacillus species, and is associated with BV,¹⁹ all samples classified as L. iners-dominated (CST III) were removed from our Lactobacillus-dominated outcome and the analysis was repeated. To evaluate washout periods and assess whether changes to the microbiota occurred early or late in the intervention period, and to compare equal lengths of observation in DC and D phases, the model was repeated using nine four-week sliding periods in the DC phase, with each period starting one week after the start of the last, and extending one week further. Lastly, the first sample was used to define baseline microbiota composition, and the analysis was repeated to evaluate whether the association between douching cessation and molecular-BV differed between participants starting the study with molecular-BV versus a Lactobacillus-dominated microbiota.

Wilcoxon signed-rank tests were used to evaluate paired changes in the median relative abundance of individual key taxa between phases. For seven participants, a piecewise linear mixed effects model was used to assess differences in rates of change in bacterial absolute abundance (log-transformed) in samples taken three days before and after douching. Data were analyzed using STATA/SE 15.0 (Stata Corp, College Station, TX), SAS University Edition (SAS Institute, Inc., Cary, NC), and R version 4.0.5.

RESULTS

As previously reported,¹³ a majority of participants in the study were over 30 years old (81%), had a history of tubal ligation (47%), and reported Black (48%) or White (42%) race (Table 1). While there were many reasons participants douched, common motivations included removing menstrual blood, and to generally feel clean or fresh. Patterns in douching during the D phase were similar to douching frequencies reported at baseline – most participants reported douching one to two times during the four weeks of douching observation (70%), suggesting study participation did not meaningfully affect D phase douching behaviors.

Table 1:

Demographic Characteristics and Vaginal Douching Behaviors Among Women Enrolled in a Pilot Douching Cessation Study, 2005–2007, N=33

	N (%)
Age (Years)
18–29	6 (18)
30–39	14 (42)
40+	13 (39)
Race
White	14 (42)
Black	16 (48)
Hispanic	1 (3)
Other	2 (6)
Contraceptive Methods Used
Oral contraceptive pill	7 (22)
Condoms	10 (31)
Tubal ligation	15 (47)
Intrauterine device (IUD)	2 (6)
Other	2 (6)
Primary Reason for Douching
To remove menstrual blood	12 (36)
To feel good and fresh	9 (27)
To feel clean	7 (21)
Other	5 (15)
Douching Frequency at Baseline
Less than once per month	5 (15)
Once or twice per month	23 (70)
Three or four times per month	4 (12)
More than four times a month	1 (3)
Douching Products Used
Store-bought solution	29 (88)
Homemade vinegar and water	1 (3)
Baking soda	4 (12)
Betadine	1 (3)
Water only	3 (9)
Number of Douches Used in Douching Phase
1–2	23 (70)
3–4	5 (15)
5 or more	5 (15)
Community Class in Douching Phase
L. crispatus/L. gasseri-dominated	8 (24)
L. iners-dominated	13 (39)
Low-Lactobacillus	12 (36)

In total, 33 participants provided 982 vaginal samples. The distribution of CSTs and Nugent scores are shown in Figure 1. A majority of samples were dominated by Lactobacillus spp. (59%) and had a normal Nugent score (0–3, 72%), while 41% of samples were categorized as molecular-BV (CST IV-A, IV-B, or IV-C). There was near perfect concordance between findings of molecular-BV and Nugent-BV among samples assigned to CSTs I (100%), II (98%), III (98%), and V (100%), while 70% (N=14), 41% (N=121), and 4% (N=3) of samples assigned CST IV-A, IV-B, and IV-C, respectively, had high Nugent scores (7–10, BV).

Figure 1:

Distribution of community state types (CSTs) and Nugent Scores, N=33 participants, 982 samples

Figure 2 shows the total proportion of samples classified as molecular-BV in the D phase and DC phase for each participant. Following douching cessation, 36% (12/33) of participants experienced an increase in molecular-BV, 27% (9/33) experienced a decrease in molecular-BV, and 36% (12/33) experienced little to no change (+/− 1%) in the proportion of samples classified as molecular-BV. Individual and aggregate CST trajectories during D and DC phases, and stacked phylotype relative abundance charts are shown in Supplemental Figure 1 and Supplemental Figure 2, respectively. When evaluating changes between Lactobacillus-dominated and molecular-BV CSTs, CST I (L. crispatus-dominated) appeared to most often shift to CST IV-C (a low-Lactobacillus state not dominated by either G. vaginalis, A. vaginae, or “Ca. Lachnocurva vaginae”), while CST III (L. iners-dominated) more often shifted to CST IV-B (dominated by G. vaginalis and A. vaginae).

Figure 2: Molecular-BV in douching phase and douching cessation phase.

Proportion of samples assigned to molecular-BV (mBV) in douching observation phase (D phase, 4 weeks) and douching cessation phase (DC phase, 12 weeks) among participants with no change, an increase, and a decrease in molecular-BV following douching cessation.

Modeling revealed no statistically significant change in the odds of having molecular-BV versus a Lactobacillus-dominated CST in a participant’s douching cessation phase compared to their douching observation phase (aOR 1.77, 95% CI 0.89 to 3.55) (Table 2). Conclusions were unchanged when L. iners-dominated CST III samples were removed from the model (aOR 1.21, 95% CI 0.61 to 2.42) and when considering participants who primarily douche to remove menstrual blood (aOR 1.72, 95% CI 0.87 to 3.39, interaction p-value = 0.40).

Table 2:

Conditional Logistic Regression: Odds of molecular-BV compared to a Lactobacillus-dominated community state type (CST) in douching cessation phase versus douching phase

	N	OR	95% CI	p-value	N	aOR§	95% CI	p-value
All samples	681	1.51	0.81, 2.86	0.20	676	1.77	0.89, 3.55	0.10
L. iners-dominated CST III removed	286	1.06	0.56, 2.03	0.85	283	1.21	0.61, 2.42	0.58
Stratified by primary motivation for douching
	N	OR	95% CI	p-value	N	aOR§	95% CI	p-value
Remove menstrual blood	293	1.41	0.71, 2.77	0.32	289	1.72	0.87, 3.39	0.12
Other	388	1.64	0.54, 4.93	0.38	387	1.93	0.54, 6.92	0.31

^§Conditional model adjusted for any sexual activity in the 24 hours prior to sampling and menses within 3 days of sampling. Participants who do not experience any changes in the outcome (molecular-BV vs Lactobacillus-dominated) are uninformative and do not contribute to the model.

We stratified by baseline microbiota composition to evaluate the effect of douching cessation separately for participants who were in a molecular-BV state versus those who were dominated by Lactobacillus during the start of douching observation (Supplemental Table 1). Among participants who started the study with molecular-BV, douching cessation was not associated with a change in the odds of molecular-BV (aOR 1.07, 95% CI 0.58 to 1.97). Similarly, there was not a statistically significant increase in molecular-BV among participants starting with a Lactobacillus-dominated microbiota (aOR 3.58, 95% CI 0.89 to 14.34).

In the sliding windows analyses, assessing four-week intervals of DC phase compared to the four-week D phase, douching cessation was only found to be associated with a significantly increased odds of molecular-BV in the final five weeks of DC phase (Supplemental Table 2).

With regard to analyses of individual taxa, there was no significance difference in the median relative abundance of four species of Lactobacillus spp. comparing DC phase to D phase, and, while statistically significant, the median relative abundance of other taxa investigated differed by less than one percent (Supplemental Table 3). Further, there was no difference in the rate of change in bacterial absolute abundance comparing samples collected less than or more than three days after douching among seven participants (Supplemental Figure 3).

DISCUSSION

In a 2008 analysis of this cohort¹³, we reported that douching cessation was associated with a significant reduction in Nugent-BV among participants who reported primarily douching following menses (aOR 0.23, 95% CI 0.12 to 0.44). This result was not confirmed in the current analysis using the outcome of molecular-BV. Discrepancies between BV defined via Nugent score and molecular methods may offer an explanation for these disparate findings as it is well-documented that molecular-BV often encompasses Nugent-intermediate samples.¹⁸ In our study, although it was rare for Lactobacillus-dominated CSTs to be scored as Nugent-BV, 65% of the molecular-BV CSTs were not classified as Nugent-BV (Supplemental Table 4), including 30% of CST IV-A (dominated by “Ca. Lachnocurva vaginae”), 59% of CST IV-B (dominated by G. vaginalis and A. vaginae), and 96% of CST IV-C (dominated by other anaerobes including Streptococcus). Instead, these molecular-BV samples were classified as normal (31%), Nugent-intermediate (23%), or were missing Nugent score (11%). Further, 92% of Nugent-intermediate samples were classified as molecular-BV, and the previous analysis did not include Nugent-intermediate samples in its BV outcome. Additionally, the presence of some L. crispatus, and the low relative abundance of Gardnerella spp., (Figure 1) may have contributed to low Nugent scores in CST IV-C samples.

Stacked phylotype relative abundance charts (Supplemental Figure 2), though descriptive, reveal how CST assignment may change in response to small compositional shifts, or may not change despite large compositional shifts. For example, there appears to be variability in the relative abundance of G. vaginalis and A. vaginae among CST III samples for participants 15 and 19, and the relative abundance of L. iners among CST IV-B samples for participants 2 and 9. Still, CSTs remain a useful data reduction tool for summarizing the composition of the vaginal microbiota, particularly when repeated samples are available to define temporal profiles, and facilitate statistical modeling in studies assessing reproductive tract health. Future personal hygiene intervention studies could collect clinical data such as Amsel’s clinical criteria for BV, pelvic inflammatory disease, and additional microenvironmental measurements such as lactic acid concentrations or host immune response,²⁰ which would provide information on meaningful changes in vaginal health occurring outside of CST transitions. This data serves to highlight the need to use multiple approaches to measure the vaginal microenvironment.

The study design was not randomized, and there are limitations to interpreting single cross-over studies. The cross-over could lend evidence if it demonstrated an increase in Lactobacillus-dominated microbiota among participants with molecular-BV at baseline, and conversely, no increase in molecular-BV among participants in an already optimal, Lactobacillus-dominated state. While results from the stratified model suggest a non-significant greater odds of molecular-BV in the DC phase among participants starting with a Lactobacillus-dominated microbiota, we know from other studies that L. iners is frequently found in the vaginal microbiota of women who have, or go on to develop, Nugent-BV.^{21 22} Considering many participants in this study with a baseline microbiota dominated by Lactobacillus were dominated by L. iners, this finding may reflect physiological temporal variation in the microbiota. Additionally, the wide confidence intervals indicate a high degree of variability in the association between douching cessation and molecular-BV between participants, even among participants with similar baseline microbiota compositions.

Overall, this study confirms prior reports. One 2006 pilot study did not find a difference in Nugent scores over eight weeks between 48 participants randomized to continue or cease douching; however, this study was not powered to detect any effects of douching cessation and was a study to establish that women would enroll in the intervention.¹² Sivapalasingam et al found, in an intervention of IVP cessation in Kenyan participants, that IVPs were associated with a L. iners-dominated microbiota at baseline, but there were no significant changes in the microbiota 3 months after cessation.¹⁴ A prior study detailed that douching increased the acquisition of BV among participants with intermediate, but not low, Nugent scores, suggesting douching may increase incidence of BV among those with an already less optimal vaginal microbiota.²³

Further studies should evaluate the effect of douching cessation on the vaginal microbiota using frequent longitudinal sampling and should take into consideration the sample size needed to stratify by baseline composition of the vaginal microbiota. It is well-documented that not all microbiota are equally optimal nor resilient,²⁴ and it is important to understand which microbiota may be more affected by douching or douching cessation. Quantitative bacteriology studies suggests there is a large wash out of bacteria immediately following douching,²⁵ and so it would also be of interest to determine the immediate, short, and long-term effects of douching cessation on the absolute abundance of specific taxa, particularly taxa characteristic of molecular-BV.¹⁵

The study was a pilot and the sample size was limited at 33. Participants tended to douche infrequently in the douching phase (median:1, range:1–15), and results may not be generalizable to populations who douche more often, or who do not use commercial products to douche. Prior observational studies suggest a dose-response relationship between frequency of douching and increased likelihood of Nugent-BV.^{26 27} In addition, although this study was the first study to evaluate the effect of cessation on molecular-BV using multiple samples taken during douching and non-douching periods, twice-weekly sampling may have been inadequate to capture acute and transient effects to the vaginal microbiota following douching or douching cessation, as indicated by Onderdonk et al which documented more immediate effects by quantitative culture in less than 24 hours.²⁵ Lastly, because many participants douched in response to menstrual bleeding, it is challenging to separate effects due to douching cessation from menstrual cycle effects, or other unmeasured confounders.

These pilot data are too preliminary to inform clinical practice, and results should not be interpreted as evidence against douching cessation. This study only evaluated one component of lower reproductive tract health, the composition of the microbiota, and did not evaluate changes in other outcomes such as absolute abundance of specific bacteria, pH, inflammation, Amsel criteria, or evaluate environmental exposures such as endocrine-disrupting phthalates which have been associated with BV and are found in commercial douche products.²⁸ Healthcare providers should continue to discuss vaginal douching with patients as there is evidence that people may alter douching behaviors in response to guidance from healthcare providers,² and because people who use vaginal douches and other over-the-counter treatments may be doing so to alleviate or mask symptoms from treatable infections such as BV, vulvovaginal candidiasis, or trichomonas.²⁹ Recalcitrant BV, particularly among those who douche, may require multipronged approaches that include douching cessation, but with the addition of other interventional measures such as antibiotics, and other behavior modifications thought to affect BV, such as lubricant use,³⁰ and smoking, diet, and inconsistent condom use.²⁴

KEY MESSAGES.

What is already known on this topic:

This vaginal douching cessation cross-over study previously reported that douching cessation was associated with reduced Nugent-BV among participants who douched during menses. We conducted a re-analysis of archived specimens to evaluate the effect of a douching cessation on the composition of the vaginal microbiota and molecular-BV outcomes.

What this study adds:

Douching cessation was not associated with reduced molecular-BV or changes in absolute bacterial abundance.

How this study might affect research, practice, or policy:

This pilot study of 33 women provides important preliminary data and a framework for studying douching cessation. Additional interventions may be required to restore optimal vaginal microbiota, particularly in those with recalcitrant BV.

Data availability statement

The datasets are available at the National Center for Biotechnology Information (NCBI) Database of Genotypes and Phenotypes (dbGaP) accession numbers phs000261.v2.p1.