Assessment of the vaginal microbiota before and after use of hyperosmolal lubricant during transvaginal ultrasound

Overall, there were no significant changes in vaginal microbiota composition immediately following lubricant exposure during TVUS. Significant decreases in vaginal microbiota stability were observed over the nine weeks following lubricant exposure compared to the week prior; however, that decreased stability did not reflect an overall increased odds in having a low- Lactobacillus (CST IV) vaginal microbiota after TVUS. In stratified analyses, peri/postmenopausal individuals and those with a previous history of bacterial vaginosis (BV) — both groups who may be more likely to experience destabilization in their vaginal microbiota — were more likely to develop and maintain a low- Lactobacillus vaginal microbiota post-TVUS. These findings highlight a possible increased vulnerability of certain subgroups to vaginal dysbiosis outcomes following lubricant exposure. Prior in vitro studies have found that some hyperosmolal lubricants suppress growth of Lactobacillus spp. in culture, 11 , 13 – 15 but there is a paucity of data on how lubricant exposure may affect the vaginal microbiota in vivo. One study reported little change in vaginal pH one day following lubricant use, 47 suggesting the abundance of Lactobacillus spp. may not be largely affected. Similarly, we observed that vaginal samples collected ~6–12 hours after TVUS were not compositionally different from those before TVUS. A limitation to our study is that 16S rRNA gene amplicon sequencing cannot distinguish between alive and dead organisms and cannot determine changes in bacterial activity or lactic acid production. 48 Over the entire ten weeks of follow-up, we identified a decrease in stability of the vaginal microbiota after exposure to hyperosmolal lubricant. The response of the microbiota to lubricant varied significantly by reproductive stage and prior history of BV. A recent study estimated that approximately 20% of microbiota were “multi-stable” and may fluctuate between composition types based on external factors like sexual and hygiene behaviors, while the majority appear “mono-stable” and exhibit minimal fluctuations. 49 There are few prior studies on the effects of behaviors on the vaginal microbiota in peri/postmenopausal women. Some studies have found no association between intravaginal practices 50 or sexual activity 25 with bacterial composition. Another study found no significant change in vaginal microbiota composition over 12 weeks among postmenopausal women with GSM using a hyperosmolal vaginal moisturizer, which is similar in composition to vaginal lubricants, 2 , 6 compared to controls using a placebo gel. 51 Our study found a sustained decrease in protective Lactobacillus spp. in peri and postmenopausal women after a single exposure. Peri/postmenopausal women without a Lactobacillus- dominated vaginal microbiota may experience more GSM symptoms or biomarkers of vaginal atrophy. 25 , 52 , 53 Among premenopausal participants, there was a decrease in the stability of the vaginal microbiota after TVUS but no changes in the odds of having CST IV microbiota. Instability, which can be descriptively visualized in the longitudinal relative abundance plots, may not always result in changes from Lactobacillus- dominated CSTs to a low- Lactobacillus CST (see Supplementary Figure 4 , IDs G227, G247, and G333) or vice versa (see Supplementary Figure 4 , IDs G310, and G262). Consistent with our finding of no overall change in the odds of having CST IV microbiota, Supplementary Figure 5 highlights that increases and decreases in the proportion of CST IV microbiota after TVUS are equally common in premenopausal participants. At study entry, there was no significant link between a participant’s self-reported lifetime history of BV and the presence of a low- Lactobacillus vaginal microbiota. However, only those with a history of BV experienced an increase in the odds of having CST IV microbiota after lubricant exposure, suggesting a single dose of lubricant was associated with a sustained loss of lactobacilli in women at higher risk for BV. Participants without a history of BV showed no changes in the stability of the vaginal microbiota after TVUS but did have significantly decreased odds of having CST IV microbiota in some follow-up windows. Supplementary Figure 5 highlights that while most of these participants had no change in the proportion of CST IV microbiota between study phases, a few experienced dramatic decreases in the proportion of samples harboring a CST IV microbiota after TVUS. Given that those without a history of recurrent BV may have a more stable vaginal microbiota, no sustained decrease in Lactobacillus spp. was expected. Premenopausal individuals without BV history may recover from minor destabilization; however, it is unknown if repeated exposures could lead to more persistent effects. While many individuals may not exhibit clinical consequences following a single exposure to vaginal lubricants, certain populations, including peri/postmenopausal and those with a history of bacterial vaginosis, may be at increased risk of microbiota disruptions, which may, in turn, increase risk for urogenital infections. Future research should explore how frequent use of hyperosmolal vaginal lubricants affect the entire vaginal microbiome and microenvironment, examining both bacterial activities and tissue responses. Clinical trials with iso-osmolar controls, with repeated exposures and extended follow-up periods, would provide valuable insights for both personal and clinical applications. The analyses minimized bias by comparing participants to themselves before and after lubricant exposure, controlling for individual-specific factors. Studying lubricant use in the context of TVUS ensured that all participants received the same hyperosmolal lubricant under similar conditions. A separate lubricant-unexposed control group could have provided support that the observed findings do not reflect the usual temporal variations in the vaginal microbiota. 54 We may have overlooked acute changes occurring between the evening of TVUS and the first follow-up visit. Furthermore, our sample included a limited number of peri/postmenopausal participants. While some participants were referred for TVUS due to heavy vaginal bleeding- a factor that could influence the microbiota composition 55 , 56 this was accounted for in our analyses. The GALE study aimed to analyze 117 participants, but recruitment was halted in March 2020, which may result in the CST IV analysis having less statistical power than the originally anticipated 80%. Lastly, these results may not be applicable to everyday personal use of lubricants.

The Gynecology and Lubricant Effects (GALE) study evaluated the association between hyperosmolal lubricant use (GLIDE, formerly known as E-Z lubricating jelly, Athena Medical Products: ~2,400 mOsm/kg, pH 5.5) and vaginal health over 10 weeks in non-pregnant women referred for TVUS at the University of Maryland Medical Center between 2017–2020. 34 , 35 Participants attended three clinical visits (approximately one week before TVUS [“baseline” visit], 2–5 days after TVUS [“post-TVUS follow-up” visit], 10 weeks after baseline [“final” visit] where a clinician evaluated them for Amsel-BV 36 and collected a mid-vaginal ESwab (Copan Diagnostics). Participants were instructed to self-collect mid-vaginal samples using Eswabs daily between baseline and TVUS, immediately prior to TVUS (“pre-TVUS” sample), approximately 6–12 hours after TVUS (“post-TVUS” sample), and then twice-weekly for eight weeks ( Supplementary Figure 1 ). Self-collected vaginal swabs have been validated against clinician-collected vaginal swabs for microbiota analysis. 37 , 38 Participants stored samples in their home freezer (−20°C), and frozen samples were transported by participants weekly to the University of Maryland Baltimore (UMB) and stored at −80°C. The study was approved by the UMB Institutional Review Board. DNA extraction, library construction and sequencing for these samples have been previously described. 35 , 39 Genomic DNA was extracted from vaginal ESwabs with the MagAttract Microbiome DNA/RNA Kit (Qiagen, Germantown, MD) using a custom automated protocol on the Hamilton Microlab STAR. The V3-V4 regions of the 16S rRNA gene were amplified and sequenced on an Ilumina HiSeq 2500 using Rapid Run chemistry or Ilumina MiSeq (Illumina, San Diego, CA). 40 Sequencing data were processed as described in Holm et al . 40 Amplicon sequence variants (ASVs) of major vaginal taxa were assigned species-level annotations using speciateIT . 41 Taxa present at less than 10 −5 across all samples were removed and samples with fewer than 1000 reads were removed from analysis. Community state types (CSTs) were assigned using the nearest centroid-based tool VALENCIA, which assigns CSTs based on their similarity to reference CSTs using the Yue-Clayton theta index. 42 Five primary CSTs were identified in this study ( Supplementary Figure 2 ); four dominated by Lactobacillus species ( L. iners, L. crispatus, L. gasseri, L. jensenii ), and one characterized by low levels of Lactobacillus spp. and higher proportions of strict and facultative anaerobic bacteria (CST IV, including CST IV-A, IV-B, and IV-C). We previously reported complete within-subject agreement in vaginal CST assignment comparing samples sequenced on HiSeq or MiSeq Ilumina instruments. 40 We evaluated changes in similarity of vaginal microbiota composition immediately following TVUS using Yue-Clayton theta indices ( θ YC ) calculated in the vegan package. 43 This index considers the relative abundance of shared and non-shared taxa between samples (0: completely dissimilar, 1: completely similar). For samples taken before TVUS, we calculated the average similarity of each sample with the sample taken one day prior for each participant. Then, we calculated the similarity of the post-TVUS sample taken 6–12 hours after TVUS with the sample taken the day before TVUS. The Wilcoxon signed-rank test and matched-pairs rank-biserial correlation were used to determine within participants if the similarity of daily successive samples was different immediately after TVUS versus the average daily similarity in the week before ( Supplementary Figure 3 ). Statistical analyses were conducted using the ggstatsplot package. Analysis of Compositions of Microbiomes II (ANCOM II) 44 , 45 was used to determine differences in the relative abundance of all individual taxa comparing the post-TVUS and pre-TVUS sample, accounting for repeated samples by participant. We evaluated longer-term changes in stability and vaginal microbiota composition approximately nine weeks after TVUS versus one week before. The average stability of the vaginal microbiota in three-week intervals after TVUS was compared to the average stability of the microbiota before TVUS within each participant using the Durbin-Conover test, and p-values were adjusted for multiple pairwise comparisons (Holm-Bonferroni method). Stability within each interval was measured by comparing the similarity of all samples within an interval to each other using θ YC . Three-week intervals were chosen because they allowed for intervals with an equal number of weeks after TVUS, and a similar number of samples in all intervals (median 6–7 in each). Only participants contributing at least two samples in all intervals were included. Statistical analyses were conducted using the ggstatsplot package. In an analysis of all 10 weeks of follow-up, CSTs were dichotomized as CST IV (low Lactobacillus spp. relative abundance) versus Lactobacillus- dominated (CSTs I/II/III/V) and generalized linear mixed effects models with a random intercept were used to model the subject-specific odds of having CST IV sampling days after versus before TVUS. Time-varying factors self-reported on daily diaries, including sexual behaviors, use of vaginal hygiene products and menses, were evaluated as potential confounders. We conducted several sensitivity analyses, including censoring participants after antibiotic or antifungal use, and dropping samples with CST assignment scores <5% of VALENCIA assignment scores for each CST as these samples may not bear sufficient similarity to any of the reference CSTs. To determine when lubricant effects may occur or how long they might persist, the models were repeated using six four-week sliding intervals after TVUS with each interval starting one week after the previous and extending one week further. Participants were censored at any additional vaginal lubricant use reported on daily diaries. We tested for effect modification by race, reproductive stage, and self-reported history of BV. Reproductive stage was defined using an adapted Stages of Reproductive Aging Workshop system 46 and included self-reported menstrual cycle regularity and hormonal contraceptive use. 35 Data were analyzed using SAS Studio version 3.81 (SAS Institute, Inc., Cary, NC), and RStudio version 2022.7.2.576.

Most participants were premenopausal (82%), self-reported Black race (62%), and were referred for TVUS to evaluate pelvic pain (42%), pelvic mass (32%), or abnormal uterine bleeding (24%) ( Table 1 ). Many participants had no findings from TVUS (35%), but fibroids (32%) and cysts (12%) were commonly identified. Over half of participants reported previously being diagnosed with BV; however, there was no association between history of BV and having a baseline CST-IV vaginal microbiota (p=0.42). Sexual activity was frequently reported throughout the study. The 104 participants provided a total of 2,582 mid-vaginal samples. After excluding samples in which additional lubricant use was reported (N=140/N=13 participants) as well as those with <1,000 reads (N=58 samples/N=32 participants), 2,384 samples remained. Longitudinal profiles of vaginal CSTs and phylotype relative abundances among participants in the GALE study are displayed in Figure 1 and Supplementary Figure 4 , respectively. The median of participants’ average consecutive daily similarity ( θ YC ) before TVUS, and the median similarity ( θ YC ) of participants’ post-TVUS samples compared with the sample taken one day prior were both 0.92 ( Figure 2 ). There was moderate evidence that the distribution of daily θ YC scores was different immediately after TVUS (p=0.06) with more participants experiencing decreases in daily similarity following TVUS rather than increases r ^ b i s e r i a l r a n k > 0 , though θ YC scores remained high. Findings were similar in analyses stratified by reproductive stage and BV history. Using ANCOM II, there were no differentially abundant taxa comparing the pre-TVUS samples and the samples taken 6–12 hours following TVUS. In the period before TVUS, the stability of the vaginal microbiota within individuals was not different comparing premenopausal and peri/postmenopausal participants ( θ YC =0.87 vs θ YC =0.93, p=0.35), or participants with and without prior BV ( θ YC =0.90 vs θ YC =0.86, p=0.62). The average stability of the vaginal microbiota was decreased within participants in all three-week intervals after TVUS compared to the before TVUS ( Figure 3 ). In stratified analysis, significant decreases in average stability after TVUS were seen for premenopausal participants and those with prior BV. There were similar reductions in average stability after TVUS for peri/postmenopausal participants, though these were not statistically significant. Supplementary Figure 5 shows the proportion of CST IV samples in each phase of the study for each participant, and the within-participant changes in the proportion of CST IV samples. There were significant interactions between lubricant use with reproductive stage (p=0.02) and prior history of BV (p=0.03), but not with self-reported Black race (p=0.84). The odds of a CST IV microbiota after exposure to lubricant during TVUS among all participants and stratified by reproductive stage and prior history of BV are presented in Table 2 . Overall, participants had a 28% non-significant increase in the odds of having CST IV microbiota after TVUS (95% CI:0.87–1.87). There were no changes in the odds of having CST IV versus Lactobacillus- dominated microbiota after TVUS among premenopausal participants while peri/postmenopausal participants had 3.2-fold higher odds of CST IV microbiota after TVUS (95% CI:1.16–8.98). Among participants with no history of BV, there was some evidence suggesting less CST IV after TVUS (aOR:0.54, 95% CI:0.25–1.19) while participants with a history of BV had a 1.7-fold higher odds of CST IV samples after TVUS (95% CI:1.10–2.72). Effect estimates were slightly increased when censoring participants after antibiotic or antifungal use and slightly decreased when samples that did not bear sufficient similarity to reference CSTs were not included, though overall conclusions were generally unchanged. The association between lubricant exposure and CST IV did not appear to consistently increase or decrease throughout the follow-up period for most participants but was often greatest in Weeks 4–7 ( Table 3 ). Models were adjusted for menses (excepting the peri/postmenopausal model), sexual activity, and feminine hygiene product use in the 24 hours preceding sampling.

Vaginal lubricant use is widely reported by individuals of all ages. 1 They are used as a first-line non-hormonal treatment for the genitourinary syndrome of menopause (GSM) due to decreased estrogen resulting in reduced lubrication, 2 , 3 enhance satisfaction during sexual activity, 4 and are commonly utilized during pelvic exams, labor, and transvaginal ultrasound (TVUS). However, mounting evidence suggests lubricants may be toxic to cervicovaginal and rectal tissues and their associated microbiota. 5 Lubricant cytotoxicity is thought to result from factors including hyperosmolality, high pH and the use of preservatives, microbicides, and antimicrobial parabens. 6 Water-based lubricants are hyperosmolal due to glycols acting as humectants. Higher osmolality lubricants induced more severe irritation in slug mucosal irritation assays. 7 , 8 Studies using three-dimensional vaginal epithelial models and ex vivo tissue samples found that hyperosmolal, but not iso-osmolal, lubricants result in increased cellular stress, 9 reduced epithelial barrier integrity, 10 and sloughing of ectocervical epithelia. 11 Lastly, hyperosmolal lubricant excipients, including glycerin and polyethylene glycol, increased susceptibility to HSV-2 in mice. 12 There is little data on the effects of hyperosmolal lubricants on vaginal microbiota composition, though in vitro studies suggest Lactobacillus spp. may be affected. 11 , 13 – 15 One study comparing five hyperosmolal lubricants reported no changes in vaginal bacterial relative abundance over four weeks; however, this study did not present within-participant changes in vaginal microbiota following lubricant exposure. 16 Epidemiologic studies 17 – 19 have found lubricant use is associated with an increased risk for bacterial vaginosis (BV), a clinical syndrome characterized by a vaginal microbiota with low levels of Lactobacillus spp. and an increased abundance of strict and facultative anaerobic bacteria. Interactions between lubricant and the vaginal microenvironment warrant attention given the vital role of the microbiota in reproductive tract health. Lactobacillus spp. provide protection against urogenital pathogens in reproductive-age women by producing lactic acid and maintaining low pH, along with other mechanisms. 20 Conversely BV, which is twice as common in Black versus white women, 21 is associated with bothersome symptoms and increased risk for STI acquisition and transmission. 22 – 24 The relative abundance of lactobacilli also often declines during menopause; a recent study reported 36% of postmenopausal women had a Lactobacillus- dominated vaginal microbiota. 25 Emerging data indicate protection by Lactobacillus spp. against urinary tract infections in menopause. 26 – 28 There is a need to better understand how hyperosmolal lubricants may affect the vaginal microbiota in vivo . Observational studies frequently generalize lubricant use without considering type, frequency, and timing relative to outcome measures. 17 – 19 , 29 – 32 Personal lubricants vary in composition and osmolality, 2 and results may be affected by heterogeneous exposure definitions. 33 Using a single-crossover study design, we examined shifts in the vaginal microbiota over 10 weeks in participants exposed to a hyperosmolal lubricant once during TVUS.

While a single vaginal exposure to lubricant typically poses no clinical consequences, it may negatively affect the vaginal microbiota of peri/postmenopausal individuals and those with a history of BV. Further research may indicate that reformulating lubricants is necessary to mitigate adverse effects on the vaginal microenvironment.

Supplementary Figure 1: Study weeks and sampling scheme for the Gynecology and Lubricant Effects Study Supplementary Figure 2: Heatmap of vaginal Community State Types (CSTs) in samples taken before and after TVUS Supplementary Figure 3: Example schematic for comparing similarity of daily successive samples before and after TVUS. Legend: Within-participant change in similarity of daily successive samples after TVUS: θ YC 6 versus average(θ YC 1 − θ YC 5) Supplementary Figure 4: Individual-level CST and phylotype relative abundance plots. Legend: Participants collected samples daily before TVUS, immediately before TVUS, the evening after TVUS, and twice weekly following the post-TVUS follow-up visit. Samples taken after TVUS are not spaced by calendar time. Supplementary Figure 5: Proportion of CST-IV samples in each study phase and change in proportion of CST-IV samples after TVUS within participants.