Study population
This study was performed at the Public health service (PHS) clinic in Amsterdam, the Netherlands. Ethical approval was obtained for this study from the Medical Ethics Council of the Amsterdam Medical Centre, reference number: METC 2014_413. We recruited women through word of mouth and through advertisements that were placed on the website of the PHS of Amsterdam and in the newsletters of two local universities. An incentive of 100 euros was rewarded at study completion. Women were eligible for the study if they: 1) were 18–36 years old; 2) had a regular menstrual cycle; and 3) tested negative for sexually transmitted infections (STI) (nucleic acid amplification tests for Chlamydia trachomatis, Neisseria gonorrhoeae, and Trichomonas vaginalis (APTIMA, Hologic, Marlborough, USA)), vulvovaginal candidiasis (based on microscopy), and BV by Amsel criteria at baseline. Women were excluded if they: 1) were pregnant, or planning to become pregnant within the next 3 months; 2) had an intra-uterine device (because some women with copper intra-uterine devices experience heavy menstrual bleeding and some women with Mirena intra-uterine devices experience amenorrhea and/or spotting; hormonal contraceptive pill and NuvaRing were allowed); 3) worked as a sex-worker; 4) had used antibiotics in the last 30 days; or 5) were allergic to any of the ingredients of the study product.
Intra-vaginal douching product
The ingredients of the Etos® over-the-counter intra-vaginal douche were listed on the package label in the following order: aqua, butylene glycol, lactic acid, caprylyl glycol, sodium pyroglutamic acid, Zea mays kernel extract, hydrolyzed milk protein, niacinamide, and adenosine triphosphate. Concentrations per ingredient were not listed. The manufacturer’s instructions for use were as follows: add Etos® liquid and lukewarm water to the Etos® flask in a 1:7 ratio, as indicated by markings on the flask, twist on the nozzle, shake, and squirt slowly into the vagina once; replace the flask after ten uses. We measured the lactic acid concentration (both isomers) of the product (both neat and 1:7 dilution) using a commercial enzyme-based assay for quantitation of lactic acid (Boehringer/Mannheim Cat. No. 11 112 821 035, R-Biopharm, Darmstadt, Germany), following the manufacturer’s instructions. We measured the pH using an electrode.
Clinical study design
The participants were followed over three menstrual cycles and were instructed to use the douching product three times weekly for the duration of cycle 2, starting on the first day of menses. We chose this frequency because the product insert advises a maximum of three times weekly douching. For each participant, the VM composition was assessed before initiation of product use (cycle 1), during product use (cycle 2), and after cessation of product use (cycle 3) (Additional file 1: Figure S1).
The participants visited the PHS clinic just before the start of each menstrual cycle where they were seen by a nurse who measured the vaginal pH, collected vaginal smears for Nugent scoring [49], and performed a ‘whiff’ test (fishy odor after addition of potassium hydroxide to vaginal smear). The participants were given instructions on how to take a vaginal swab and were instructed to swab every other day during the first and third cycle and daily during the second cycle. On days that the douche was used the participants were instructed to take a swab before douching and 1 h after douching. Self-collected vaginal swabs (Copan dry swabs (COPAN Diagnostics Inc., USA, Murrietta)) were stored in the participants’ home freezers (approximately − 18 °C) and these swabs were handed in during each clinic visit using a cooler box with ice-packs provided by the clinic. The participants were also instructed to take a pH measurement mid cycle using vaginal acidity test gloves (measures pH 4.0–7.0 at 0.3 increments; CarePlan VpH, UK, Bedford). Baseline characteristics were collected through self-administered questionnaires at the baseline visit. All participants kept a daily diary throughout the study period in which they recorded whether they douched, were menstruating, or were sexually active that day and if so, whether they used a condom and/or lubricants. Vaginal swabs were numbered sequentially and the corresponding number was recorded in the daily diary. At the last study visit, the participants were asked to complete an evaluation questionnaire on their perception and experience of using the intra-vaginal douche.
Vaginal microbiota (VM) composition analysis
Vaginal swabs were stored at − 20 °C at the Public Health Laboratory of the PHS of Amsterdam until further processing. The following selection of swabs were assessed for VM composition: every first and last swab per cycle and every second swab in between, and all swabs collected before and after douching during cycle 2. Additional swabs were included if events of interest were reported in the daily diary, such as vaginal symptoms. For DNA extraction, the swabs were thawed, eluted in 800 μL phosphate buffered saline, and shaken for 30 min at room temperature. Two-hundred microliters of the eluted sample was used for DNA extraction by isopropanol precipitation, and the pellet was dissolved in 50 μl of 10 mM Tris/HCL [50]. Negative extraction controls were included in each batch, which were also sequenced.
The VM compositions were analyzed by targeted sequencing of the 16 s rRNA gene (V3-V4 region), as described previously [8, 51]. In short, pooled and normalized amplified DNA (using universal primers 319F and 806R) was sequenced on the illumina MiSeq platform (illumina, USA, San Diego) using the V3 reagent kit. High quality sequences (> 99% base call accuracy; Trimmomatic [52]) were retained and aligned using PandaSeq [53]. Operational taxonomic units (OTUs) were picked using the Usearch tool in QIIME (version 1.9) [54] and aligned to a vaginal reference package developed by Srinivisan et al. [55] using PPLACER [56]. Samples with less than 100 reads were excluded from further analyses. The Shannon diversity index was calculated per sample using the alpha diversity tool in QIIME [54]. Based on the dissimilarity matrix of relative abundances per OTU per sample (ggplot2 package [57] in R (version 3.2.1) together with microbiological characteristics, clinical meaningful VM groups were formed. We subsequently inspected their ordination using non-metric multidimensional scaling (nMDS) based on the Bray-Curtis dissimilarity metric that considers both taxa presence/absence and relative abundance of species (i.e. diversity) using the vegan and plotly packages in R (version 3.2.1).
Candida albicans SYBR green PCR
An in-house validated SYBR green PCR based on the assay originally described by Zhang et al. [58] was used for the molecular detection of Candida albicans. In short, DNA was amplified using the 5.8S-1F and 28S-1R primers (RotoGene, Qiagen, Hilden, Germany), and PCR amplicons that melted in the range of 82.6 °C - 83.6 °C were considered positive for C. albicans. Samples from the following four study time points were tested for C. albicans: start of cycle 1, start of cycle 2, end of cycle 2 and end of cycle 3.
Statistical analyses
Univariable logistic regression using generalized estimating equations (GEE) was used to determine the association between douching and vaginal pH (dichotomized to pH < 4.5 and pH > 4.5). Univariable multinomial logistic regression models were constructed to assess the association between douching and VM composition. In these models, the VM composition during cycle 1 (before douching) acted as reference to which the VM compositions during cycle 2 (douching) and cycle 3 (wash-out period) were compared. The ‘healthiest’ VM composition, L. crispatus-dominated VM, was set as the reference category to which the three other VM composition categories (L. iners dominated, diverse lactobacilli, and diverse anaerobes) were compared. The models accounted for multiple measurements within the same individual by using GEE. We assessed the association between VM composition and the variables: study cycle, combined oral contraceptive use at baseline, vaginal intercourse on the day of sampling, having menses on the day of sampling, and, to capture the effect of douching while menstruating, an interaction term for menses and douching on the day of sampling. Variables significantly associated with VM composition (p < 0.05) and that did not display collinearity were included in the multivariable model. All logistic regression analyses were performed using SPSS Statistics software version 21 (IBM, New York, USA).
To assess whether any short-term changes in VM composition occurred after douching, a linear discriminant effect size analysis (LefSe) [59] was performed on samples taken before and 1 h after douching. This analysis separated independent features (i.e. bacterial abundances) that were significantly discriminative between classes (i.e. before or after douching), by subclass (i.e. subject id). The threshold for the logarithmic effect size score for discriminative features was set to 2.0, with alpha set to 0.05.
A linear mixed effects model (LMER) in R using lme4 [60] was constructed to study whether within-sample bacterial species diversity, i.e. alpha diversity, as defined by the Shannon diversity index, is affected by douching. Douching, menses, and sex on the day of sampling, and combined oral contraceptive use at baseline, were modeled as fixed effects and subject id as a random effect. The model also included by-subject random slopes for each of these variables. Visualization of residual plots showed no obvious deviations from normality. P-values were obtained by likelihood ratio tests of the full model with douching against a null-model without douching. A similar model was built to assess the effect of menses on VM species diversity. Median Shannon’s diversity index scores per participant per cycle were visualized using box plots.
The relative abundances of single bacterial species was visualized over time per participant with polynomial curves using locally weighted smoothed regression (loess) with 95% confidence intervals using the ggplot2 package in R [57]. Plots were made for the following species: Atopobium vaginae, BVAB 1, BVAB 2, Gardnerella vaginalis, L. crispatus, L. gasseri, L. iners, L. jensennii, Leptotrichia amnionii, Megaspheara, Prevotella bivia, Prevotella genogroup 1, and Prevotella genogroup 2.
