Abstract
Gardnerella vaginalis is detected in women with and without bacterial vaginosis (BV). Identification of 4 G. vaginalis clades raised the possibility that pathogenic and commensal clades exist. We investigated the association of behavioral practices and Nugent Score with G. vaginalis clade distribution in women who have sex with women (WSW).
Longitudinal self-collected vaginal specimens were analyzed using established G. vaginalis species-specific and clade-typing polymerase chain reaction assays. Logistic regression assessed factors associated with detection of G. vaginalis clades, and multinomial regression assessed factors associated with number of clades.
Clades 1, 2, and 3 and multiclade communities (<2 clades) were associated with Nugent-BV. Clade 1 (odds ratio [OR], 3.36; 95% confidence interval [CI], 1.65–6.84) and multiclade communities (relative risk ratio [RRR], 9.51; 95% CI, 4.36–20.73) were also associated with Lactobacillus-deficient vaginal microbiota. Clade 4 was neither associated with Nugent-BV nor Lactobacillus-deficient microbiota (OR, 1.49; 95% CI, 0.67–3.33). Specific clades were associated with differing behavioral practices. Clade 1 was associated with increasing number of recent sexual partners and smoking, whereas clade 2 was associated with penile-vaginal sex and sharing of sex toys with female partners.
Our results suggest that G. vaginalis clades have varying levels of pathogenicity in WSW, with acquisition occurring through sexual activity. These findings suggest that partner treatment may be an appropriate strategy to improve BV cure.
Bacterial vaginosis (BV) is the most common vaginal condition of reproductive-aged women [1]. Current BV treatments are suboptimal with over 50% of women experiencing recurrence within 6–12 months after treatment [2, 3]. Improving BV treatment has been impeded by its complex aetiology and pathogenesis. Gardnerella vaginalis is thought to play a key role in BV pathogenesis, potentially as a founder organism [4–6]. Gardnerella vaginalis is almost always present in the vagina of women who have BV [7], and it possesses characteristics important for pathogenesis, including production of sialidase, an enzyme that degrades cervicovaginal mucus [8], and vaginolysin, a cytolysin that induces vaginal epithelial cell lysis [9]. Both factors may assist G. vaginalis adherence to host epithelial cells and form biofilms [10]. Gardnerella vaginalis is also detected in the vagina of women without BV, albeit at lower prevalence and abundance [11–15]. Substantial genetic diversity exists within G. vaginalis [16–18], and different genetic types/clades may have different virulence potential [18–22]. These findings, taken together, suggest that commensal and pathogenic G. vaginalis variants exist. Different methods have been used to define G. vaginalis clades including cpn60 gene analysis [17], whole-genome sequence analysis [16], detection of clade-specific genes [23], and ecotyping [18]. Comparison of methods has been discussed previously [18, 20].
Previous studies have explored the association of specific G. vaginalis clades and number of G. vaginalis clades with BV, and although results regarding specific clades have been mixed, detection of multiple G. vaginalis clades is consistently associated with BV [12, 21, 23–25]. Bacterial vaginosis prevalence is high amongst women who have sex with women (WSW), with estimates ranging from 25% to 52% [26–29]. Despite this, there are no published data about G. vaginalis clade distribution in WSW. In addition, there is limited information concerning associations between G. vaginalis clades and BV risk factors. Defining the role of G. vaginalis clades in BV pathogenesis and understanding how they are acquired may lead to the development of more effective treatment strategies. Using a multiplex real-time polymerase chain reaction (PCR) G. vaginalis clade-typing assay [23], we aimed to investigate the association of G. vaginalis clade distribution with sexual and behavioral practices and Nugent Score in a cohort of Australian WSW.
METHODS
Participants and Specimens
Participants and specimens were selected from the Women On Women’s (WOW) Health cohort, which examined factors associated with incident BV [27, 30]. In brief, nonpregnant premenopausal women reporting a female sexual partner (FSP) in the 18 months before enrollment and who had a Nugent Score (NS) <7 [31] on 3 consecutive vaginal smears were eligible for enrollment [27, 30]. Participants completed a questionnaire and self-collected a vaginal swab and smear at 3-monthly intervals until the study endpoint (incident Nugent-BV [NS = 7–10] or 24 months without Nugent-BV [NS = 0–6]). Women were asked to collect specimens avoiding the time during menses. Smears were scored using the NS method [31]: a score of 0–3 was defined as non-BV, a score of 4–6 was defined as intermediate-BV, and a score of 7–10 was defined as Nugent-BV [32].
Within WOW, we conducted a nested cohort study that included women who developed incident Nugent-BV and women who reached the study endpoint without Nugent-BV [30] (selected using a random sort command in Stata/IC [version 14.2; StataCorp LP, College Station, TX]). Because there were 7 women who developed incident Nugent-BV who coenrolled in WOW with their ongoing FSP [30], controls were frequency matched on coenrollment and age, to ensure a similar distribution of both variables. A baseline specimen and 1–4 longitudinal specimens (typically the last 3 specimens collected over the 24-month study period) were included from 101 women. As a result, 101 women contributed a total of 372 specimens to analyses, comprising 101 enrollment samples and 271 longitudinal samples, 48 of which were collected at incident Nugent-BV. Ethical approval was obtained from the Human Research Ethics Committees of Alfred Hospital, Melbourne (251/09) and the University of Melbourne (0932804). Participants provided informed written consent.
Laboratory Methods
Swabs were agitated in 1 mL RNAlater (Thermo Fisher Scientific, Waltham, MA) and stored at −80°C until analyzed. Deoxyribonucleic acid (DNA) was extracted as outlined in Fethers et al [13], using the MagNA Pure 96 instrument and the DNA and Viral NA small volume kit (Roche Diagnostics, Mannheim, Germany). Gardnerella vaginalis was detected using a species-specific real-time PCR assay previously described [13]. Presence of the 4 G. vaginalis clades (hereafter referred to as clades 1, 2, 3, and 4) was determined using a validated multiplex real-time PCR assay targeting sequences unique to each clade [23] as previously described [12]; the limit of detection for this assay is 10 copies per reaction [23]. Specificity and sensitivity of the primer set was assessed by BLAST [33] analysis; additional details are provided in Supplementary Methods.
Identification of G. vaginalis clades using clade-specific PCR provides biological information on top of what can be inferred from microbiota studies utilizing 16S ribosomal ribonucleic acid (rRNA) gene sequencing. Vaginal microbiota composition was previously determined for 360 specimens using 16S rRNA gene sequencing of the V3-V4 hypervariable regions [34]. Vaginal microbiota composition was categorized into 3 groups based on McKinnon et al [32]: optimal non-Lactobacillus iners microbiota (dominated by non-L. iners spp.), L. iners microbiota (L. iners was the most abundant taxon), and nonoptimal microbiota (deficient in Lactobacillus spp.); additional details are provided in Supplementary Methods.
Statistical Analysis
Univariable and multivariable logistic regression models fitted with generalized estimating equations were used to explore associations between covariates (including history of BV, self-reported symptoms [vaginal discharge and/or odor], behavioral and sexual practices, NS, and vaginal microbiota composition) and detection of G. vaginalis (present vs absent). Logistic regression models explored associations between covariates and detection of each G. vaginalis clade versus detection of any other clade/s. Univariable and multivariable multinomial logistic regression examined associations between covariates and number of G. vaginalis clades detected. This analysis determined the relative risk of having 1 clade or multiple clades in a specimen compared with having no G. vaginalis (ie, G. vaginalis was not detected), generating relative risk ratios (RRRs) and 95% confidence intervals (CIs). Sexual and behavioral practices deemed significant in univariable analyses (P < .05) were included in multivariable analyses. Self-reported symptoms and behavioral and sexual practices were recorded for the 3-month interval preceding each specimen collection. Self-reported symptoms, NS, and vaginal microbiota composition were not included in multivariable analyses because they are correlated with G. vaginalis presence. Regression models accounted for repeated measures from individuals. We assumed an exchangeable correlation structure and used a cluster-based variance estimate for standard error. Statistical analyses were performed using Stata/IC (version 14.2; StataCorp LP).
RESULTS
Specificity and Sensitivity of Clade-Specific Primers
No G. vaginalis isolate had more than 1 clade-specific amplicon detected, and detection of clade-specific amplicons corresponded with clade phylogeny (Supplementary Figure 1), demonstrating high specificity of the primers. Isolates JCP8481A, JCP8481B, and PSS_7772B did not have a clade-specific amplicon detected and appeared genetically unrelated to the 4 clades. These isolates may represent an additional clade, as previously suggested [35]. B482 and GED7275B clustered with clade 2 isolates, but the clade 2 amplicon was not detected in these isolates. Thus, the primers have reduced sensitivity for clade 2.
Participant Demographics
One hundred one women, contributing 372 specimens, were included in analyses. Participant median age was 28 (interquartile range, 23–36 years) and most women were Australian born (n = 87, 86%). At enrollment, 72 (71%) women had an FSP (median duration of relationship, 2 years; interquartile range, 0.3–4 years), 44 (44%) reported smoking, and 22 (22%) had a past history of BV.
Factors Associated With Gardnerella vaginalis Detection
Seventy-seven women (76%; 95% CI, 67%–84%) had at least 1 specimen in which G. vaginalis was detected. Gardnerella vaginalis was detected in 184 of 372 specimens (49%; 95% CI, 44%–55%); 130 of 306 specimens with NS = 0–3 (42%; 95% CI, 67%–48%), 15 of 18 specimens with NS = 4–6 (83%; 95% CI, 59%–96%), and 39 of 48 specimens with Nugent-BV (81%; 95% CI, 67%–91%). In univariable analyses, G. vaginalis detection was associated with self-reported symptoms; NS = 4–6 and Nugent-BV (Table 1). Gardnerella vaginalis was also associated with having a nonoptimal (Lactobacillus deficient) microbiota (odds ratio [OR], 4.35; 95% CI, 2.47–7.66; P < .001).
Factors Associated With Detection of Gardnerella vaginalis
| Risk Factor | G. vaginalis Absent n (%) (N = 188) | G. vaginalis Present n (%) (N = 184) | OR (95% CI) | P Valuea | aOR (95% CI) | P Valueb |
|---|---|---|---|---|---|---|
| Agec | ||||||
| <28 | 92 (49) | 84 (46) | 1 | |||
| ≥28 | 96 (51) | 100 (54) | 1.17 (.67–2.05) | .577 | ||
| Self-Reported History of BV | ||||||
| No | 154 (82) | 142 (77) | 1 | |||
| Yes | 34 (18) | 42 (23) | 1.43 (.68–2.99) | .346 | ||
| Baseline Sexual Practices | ||||||
| No. of lifetime FSPsc | ||||||
| <5 | 96 (51) | 70 (38) | 1 | |||
| ≥5 | 92 (49) | 114 (62) | 1.69 (.93–3.09) | .088 | ||
| No. of FSPs in Previous 12 Monthsc | ||||||
| ≤1 | 129 (69) | 101 (55) | 1 | |||
| >1 | 59 (31) | 83 (45) | 1.80 (.97–3.34) | .062 | ||
| Lifetime History of Penile-Vaginal Sex With a Man | ||||||
| No | 66 (35) | 36 (20) | 1 | 1 | ||
| Yes | 122 (65) | 148 (80) | 2.27 (1.14–4.54) | .020 | 2.03 (1.01–4.10) | .047 |
| Interval Characteristics in Prior 3 Monthsd | ||||||
| Any Smoking | ||||||
| No | 128 (68) | 94 (51) | 1 | 1 | ||
| Yes | 59 (32) | 90 (49) | 1.88 (1.15–3.09) | .012 | 1.76 (1.06–2.93) | .028 |
| Any Douching | ||||||
| No | 183 (98) | 179 (97) | 1 | |||
| Yes | 4 (2) | 5 (3) | 0.58 (.15–2.30) | .441 | ||
| Any Hormonal Contraceptive Use | ||||||
| No | 157 (84) | 166 (90) | 1 | |||
| Yes | 31 (16) | 18 (10) | 0.62 (.30–1.27) | .191 | ||
| Last Menstrual Period | ||||||
| ≤7 days | 31 (17) | 23 (13) | 1 | |||
| >7 days | 149 (83) | 155 (87) | 1.05 (.62–1.78) | .858 | ||
| Number of SP | ||||||
| 0 | 25 (13) | 18 (10) | 1 | |||
| 1 | 147 (79) | 129 (70) | 1.01 (.52–1.95) | .979 | ||
| ≥2 | 15 (8) | 37 (20) | 2.11 (.92–4.85) | .079 | ||
| Frequency of Sex | ||||||
| Several times/month or less | 143 (76) | 115 (63) | 1 | 1 | ||
| Several times/week | 45 (24) | 69 (37) | 1.70 (1.07–2.71) | .025 | 1.61 (.99–2.61) | .052 |
| Sex With New Partnere | ||||||
| No | 165 (88) | 147 (80) | 1 | |||
| Yes | 23 (12) | 37 (20) | 1.39 (.84–2.30) | .195 | ||
| Sexual Practices With FSP in Prior 3 Monthsf | ||||||
| Number of FSPsc | ||||||
| 0 | 37 (20) | 32 (17) | 1 | |||
| ≥1 | 151 (80) | 152 (83) | 1.16 (.68–1.99) | .581 | ||
| Receptive Oral Vaginal Sex | ||||||
| Nog | 77 (41) | 69 (38) | 1 | |||
| Yes | 110 (59) | 115 (63) | 1.03 (.68–1.56) | .874 | ||
| Sharing Sex Toys | ||||||
| No toys/washed/ condoms usedg | 142 (76) | 141 (77) | 1 | |||
| Unwashed | 45 (24) | 43 (23) | 0.87 (.53–1.42) | .575 | ||
| Current Partner With BV Symptoms | ||||||
| No/don’t knowg | 187 (100) | 177 (97) | ||||
| Yes | 0 | 7 (4) | … | |||
| Sexual Practices With an MSP in Prior 3 Monthsh | ||||||
| Number of MSPsc | ||||||
| 0 | 173 (92) | 156 (85) | 1 | |||
| ≥1 | 15 (8) | 28 (15) | 1.94 (.89–4.23) | .097 | ||
| Any Penile-Vaginal Sex | ||||||
| Noi | 173 (93) | 158 (86) | 1 | |||
| Yes | 14 (7) | 26 (14) | 1.86 (.81–4.24) | .143 | ||
| Self-Reported Symptoms And Microbiota Measures | ||||||
| Abnormal Vaginal Discharge and/or Odour | ||||||
| No | 175 (93) | 152 (83) | 1 | |||
| Yes | 13 (7) | 32 (17) | 2.09 (1.13–3.87) | .019 | ||
| Nugent Score | ||||||
| 0–3 | 176 (93) | 130 (71) | 1 | |||
| 4–6 | 3 (2) | 15 (8) | 3.96 (1.44–10.93) | .008 | ||
| 7–10 | 9 (5) | 39 (21) | 3.48 (1.95–6.22) | <.001 | ||
| Vaginal Microbiota Typej | ||||||
| Optimal non-Lactobacillus iners microbiota | 118 (66) | 79 (44) | 1 | |||
| L. iners microbiota | 49 (27) | 45 (25) | 1.10 (.69–1.75) | .692 | ||
| Nonoptimal microbiota | 13 (7) | 56 (31) | 4.35 (2.47–7.66) | <.001 |
| Risk Factor | G. vaginalis Absent n (%) (N = 188) | G. vaginalis Present n (%) (N = 184) | OR (95% CI) | P Valuea | aOR (95% CI) | P Valueb |
|---|---|---|---|---|---|---|
| Agec | ||||||
| <28 | 92 (49) | 84 (46) | 1 | |||
| ≥28 | 96 (51) | 100 (54) | 1.17 (.67–2.05) | .577 | ||
| Self-Reported History of BV | ||||||
| No | 154 (82) | 142 (77) | 1 | |||
| Yes | 34 (18) | 42 (23) | 1.43 (.68–2.99) | .346 | ||
| Baseline Sexual Practices | ||||||
| No. of lifetime FSPsc | ||||||
| <5 | 96 (51) | 70 (38) | 1 | |||
| ≥5 | 92 (49) | 114 (62) | 1.69 (.93–3.09) | .088 | ||
| No. of FSPs in Previous 12 Monthsc | ||||||
| ≤1 | 129 (69) | 101 (55) | 1 | |||
| >1 | 59 (31) | 83 (45) | 1.80 (.97–3.34) | .062 | ||
| Lifetime History of Penile-Vaginal Sex With a Man | ||||||
| No | 66 (35) | 36 (20) | 1 | 1 | ||
| Yes | 122 (65) | 148 (80) | 2.27 (1.14–4.54) | .020 | 2.03 (1.01–4.10) | .047 |
| Interval Characteristics in Prior 3 Monthsd | ||||||
| Any Smoking | ||||||
| No | 128 (68) | 94 (51) | 1 | 1 | ||
| Yes | 59 (32) | 90 (49) | 1.88 (1.15–3.09) | .012 | 1.76 (1.06–2.93) | .028 |
| Any Douching | ||||||
| No | 183 (98) | 179 (97) | 1 | |||
| Yes | 4 (2) | 5 (3) | 0.58 (.15–2.30) | .441 | ||
| Any Hormonal Contraceptive Use | ||||||
| No | 157 (84) | 166 (90) | 1 | |||
| Yes | 31 (16) | 18 (10) | 0.62 (.30–1.27) | .191 | ||
| Last Menstrual Period | ||||||
| ≤7 days | 31 (17) | 23 (13) | 1 | |||
| >7 days | 149 (83) | 155 (87) | 1.05 (.62–1.78) | .858 | ||
| Number of SP | ||||||
| 0 | 25 (13) | 18 (10) | 1 | |||
| 1 | 147 (79) | 129 (70) | 1.01 (.52–1.95) | .979 | ||
| ≥2 | 15 (8) | 37 (20) | 2.11 (.92–4.85) | .079 | ||
| Frequency of Sex | ||||||
| Several times/month or less | 143 (76) | 115 (63) | 1 | 1 | ||
| Several times/week | 45 (24) | 69 (37) | 1.70 (1.07–2.71) | .025 | 1.61 (.99–2.61) | .052 |
| Sex With New Partnere | ||||||
| No | 165 (88) | 147 (80) | 1 | |||
| Yes | 23 (12) | 37 (20) | 1.39 (.84–2.30) | .195 | ||
| Sexual Practices With FSP in Prior 3 Monthsf | ||||||
| Number of FSPsc | ||||||
| 0 | 37 (20) | 32 (17) | 1 | |||
| ≥1 | 151 (80) | 152 (83) | 1.16 (.68–1.99) | .581 | ||
| Receptive Oral Vaginal Sex | ||||||
| Nog | 77 (41) | 69 (38) | 1 | |||
| Yes | 110 (59) | 115 (63) | 1.03 (.68–1.56) | .874 | ||
| Sharing Sex Toys | ||||||
| No toys/washed/ condoms usedg | 142 (76) | 141 (77) | 1 | |||
| Unwashed | 45 (24) | 43 (23) | 0.87 (.53–1.42) | .575 | ||
| Current Partner With BV Symptoms | ||||||
| No/don’t knowg | 187 (100) | 177 (97) | ||||
| Yes | 0 | 7 (4) | … | |||
| Sexual Practices With an MSP in Prior 3 Monthsh | ||||||
| Number of MSPsc | ||||||
| 0 | 173 (92) | 156 (85) | 1 | |||
| ≥1 | 15 (8) | 28 (15) | 1.94 (.89–4.23) | .097 | ||
| Any Penile-Vaginal Sex | ||||||
| Noi | 173 (93) | 158 (86) | 1 | |||
| Yes | 14 (7) | 26 (14) | 1.86 (.81–4.24) | .143 | ||
| Self-Reported Symptoms And Microbiota Measures | ||||||
| Abnormal Vaginal Discharge and/or Odour | ||||||
| No | 175 (93) | 152 (83) | 1 | |||
| Yes | 13 (7) | 32 (17) | 2.09 (1.13–3.87) | .019 | ||
| Nugent Score | ||||||
| 0–3 | 176 (93) | 130 (71) | 1 | |||
| 4–6 | 3 (2) | 15 (8) | 3.96 (1.44–10.93) | .008 | ||
| 7–10 | 9 (5) | 39 (21) | 3.48 (1.95–6.22) | <.001 | ||
| Vaginal Microbiota Typej | ||||||
| Optimal non-Lactobacillus iners microbiota | 118 (66) | 79 (44) | 1 | |||
| L. iners microbiota | 49 (27) | 45 (25) | 1.10 (.69–1.75) | .692 | ||
| Nonoptimal microbiota | 13 (7) | 56 (31) | 4.35 (2.47–7.66) | <.001 |
Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; BV, bacterial vaginosis; FSP, female sexual partner; OR, odds ratio; MSP, male sexual partner; SP, sexual partner (refers to total number of sexual partners in a study interval, female and male).
NOTE: Includes 372 specimens from 101 participants.
aUnivariate logistic regression fitted with generalized estimating equations (GEE) clustered for multiple specimens from each participant.
bMultivariable logistic regression fitted with GEE, clustered for multiple specimens from each participant.
cVariables were dichotomized at median value.
dInterval characteristics were measured as any exposure over the prior follow-up interval (~90 days).
eSex with a new partner with whom first sexual contact was within 90 days. May represent a new FSP or new MSP.
fThe following characteristics/sexual practices with an FSP were left out of the table for simplicity: digital vaginal sex, receptive oral anal sex, digital anal sex. No significant associations between G. vaginalis and these sexual practices were identified.
gOr did not have an FSP.
hThe following sexual practices with an MSP were left out of the table for simplicity: receptive oral sex, digital vaginal sex, and penile-anal sex. No significant associations between G. vaginalis and these sexual practices were identified.
iOr did not have an FSP.
jVaginal microbiota type available for 360 specimens from 100 women. Optimal non-L. iners microbiota includes specimens predominately consisting of non-L. iners spp., L. iners microbiota includes specimens predominately consisting of L. iners, and nonoptimal microbiota includes specimens predominately consisting of non-Lactobacillus spp.
P-values < .05 are bolded to indicate statistically significant associations.
Factors Associated With Detection of Gardnerella vaginalis
| Risk Factor | G. vaginalis Absent n (%) (N = 188) | G. vaginalis Present n (%) (N = 184) | OR (95% CI) | P Valuea | aOR (95% CI) | P Valueb |
|---|---|---|---|---|---|---|
| Agec | ||||||
| <28 | 92 (49) | 84 (46) | 1 | |||
| ≥28 | 96 (51) | 100 (54) | 1.17 (.67–2.05) | .577 | ||
| Self-Reported History of BV | ||||||
| No | 154 (82) | 142 (77) | 1 | |||
| Yes | 34 (18) | 42 (23) | 1.43 (.68–2.99) | .346 | ||
| Baseline Sexual Practices | ||||||
| No. of lifetime FSPsc | ||||||
| <5 | 96 (51) | 70 (38) | 1 | |||
| ≥5 | 92 (49) | 114 (62) | 1.69 (.93–3.09) | .088 | ||
| No. of FSPs in Previous 12 Monthsc | ||||||
| ≤1 | 129 (69) | 101 (55) | 1 | |||
| >1 | 59 (31) | 83 (45) | 1.80 (.97–3.34) | .062 | ||
| Lifetime History of Penile-Vaginal Sex With a Man | ||||||
| No | 66 (35) | 36 (20) | 1 | 1 | ||
| Yes | 122 (65) | 148 (80) | 2.27 (1.14–4.54) | .020 | 2.03 (1.01–4.10) | .047 |
| Interval Characteristics in Prior 3 Monthsd | ||||||
| Any Smoking | ||||||
| No | 128 (68) | 94 (51) | 1 | 1 | ||
| Yes | 59 (32) | 90 (49) | 1.88 (1.15–3.09) | .012 | 1.76 (1.06–2.93) | .028 |
| Any Douching | ||||||
| No | 183 (98) | 179 (97) | 1 | |||
| Yes | 4 (2) | 5 (3) | 0.58 (.15–2.30) | .441 | ||
| Any Hormonal Contraceptive Use | ||||||
| No | 157 (84) | 166 (90) | 1 | |||
| Yes | 31 (16) | 18 (10) | 0.62 (.30–1.27) | .191 | ||
| Last Menstrual Period | ||||||
| ≤7 days | 31 (17) | 23 (13) | 1 | |||
| >7 days | 149 (83) | 155 (87) | 1.05 (.62–1.78) | .858 | ||
| Number of SP | ||||||
| 0 | 25 (13) | 18 (10) | 1 | |||
| 1 | 147 (79) | 129 (70) | 1.01 (.52–1.95) | .979 | ||
| ≥2 | 15 (8) | 37 (20) | 2.11 (.92–4.85) | .079 | ||
| Frequency of Sex | ||||||
| Several times/month or less | 143 (76) | 115 (63) | 1 | 1 | ||
| Several times/week | 45 (24) | 69 (37) | 1.70 (1.07–2.71) | .025 | 1.61 (.99–2.61) | .052 |
| Sex With New Partnere | ||||||
| No | 165 (88) | 147 (80) | 1 | |||
| Yes | 23 (12) | 37 (20) | 1.39 (.84–2.30) | .195 | ||
| Sexual Practices With FSP in Prior 3 Monthsf | ||||||
| Number of FSPsc | ||||||
| 0 | 37 (20) | 32 (17) | 1 | |||
| ≥1 | 151 (80) | 152 (83) | 1.16 (.68–1.99) | .581 | ||
| Receptive Oral Vaginal Sex | ||||||
| Nog | 77 (41) | 69 (38) | 1 | |||
| Yes | 110 (59) | 115 (63) | 1.03 (.68–1.56) | .874 | ||
| Sharing Sex Toys | ||||||
| No toys/washed/ condoms usedg | 142 (76) | 141 (77) | 1 | |||
| Unwashed | 45 (24) | 43 (23) | 0.87 (.53–1.42) | .575 | ||
| Current Partner With BV Symptoms | ||||||
| No/don’t knowg | 187 (100) | 177 (97) | ||||
| Yes | 0 | 7 (4) | … | |||
| Sexual Practices With an MSP in Prior 3 Monthsh | ||||||
| Number of MSPsc | ||||||
| 0 | 173 (92) | 156 (85) | 1 | |||
| ≥1 | 15 (8) | 28 (15) | 1.94 (.89–4.23) | .097 | ||
| Any Penile-Vaginal Sex | ||||||
| Noi | 173 (93) | 158 (86) | 1 | |||
| Yes | 14 (7) | 26 (14) | 1.86 (.81–4.24) | .143 | ||
| Self-Reported Symptoms And Microbiota Measures | ||||||
| Abnormal Vaginal Discharge and/or Odour | ||||||
| No | 175 (93) | 152 (83) | 1 | |||
| Yes | 13 (7) | 32 (17) | 2.09 (1.13–3.87) | .019 | ||
| Nugent Score | ||||||
| 0–3 | 176 (93) | 130 (71) | 1 | |||
| 4–6 | 3 (2) | 15 (8) | 3.96 (1.44–10.93) | .008 | ||
| 7–10 | 9 (5) | 39 (21) | 3.48 (1.95–6.22) | <.001 | ||
| Vaginal Microbiota Typej | ||||||
| Optimal non-Lactobacillus iners microbiota | 118 (66) | 79 (44) | 1 | |||
| L. iners microbiota | 49 (27) | 45 (25) | 1.10 (.69–1.75) | .692 | ||
| Nonoptimal microbiota | 13 (7) | 56 (31) | 4.35 (2.47–7.66) | <.001 |
| Risk Factor | G. vaginalis Absent n (%) (N = 188) | G. vaginalis Present n (%) (N = 184) | OR (95% CI) | P Valuea | aOR (95% CI) | P Valueb |
|---|---|---|---|---|---|---|
| Agec | ||||||
| <28 | 92 (49) | 84 (46) | 1 | |||
| ≥28 | 96 (51) | 100 (54) | 1.17 (.67–2.05) | .577 | ||
| Self-Reported History of BV | ||||||
| No | 154 (82) | 142 (77) | 1 | |||
| Yes | 34 (18) | 42 (23) | 1.43 (.68–2.99) | .346 | ||
| Baseline Sexual Practices | ||||||
| No. of lifetime FSPsc | ||||||
| <5 | 96 (51) | 70 (38) | 1 | |||
| ≥5 | 92 (49) | 114 (62) | 1.69 (.93–3.09) | .088 | ||
| No. of FSPs in Previous 12 Monthsc | ||||||
| ≤1 | 129 (69) | 101 (55) | 1 | |||
| >1 | 59 (31) | 83 (45) | 1.80 (.97–3.34) | .062 | ||
| Lifetime History of Penile-Vaginal Sex With a Man | ||||||
| No | 66 (35) | 36 (20) | 1 | 1 | ||
| Yes | 122 (65) | 148 (80) | 2.27 (1.14–4.54) | .020 | 2.03 (1.01–4.10) | .047 |
| Interval Characteristics in Prior 3 Monthsd | ||||||
| Any Smoking | ||||||
| No | 128 (68) | 94 (51) | 1 | 1 | ||
| Yes | 59 (32) | 90 (49) | 1.88 (1.15–3.09) | .012 | 1.76 (1.06–2.93) | .028 |
| Any Douching | ||||||
| No | 183 (98) | 179 (97) | 1 | |||
| Yes | 4 (2) | 5 (3) | 0.58 (.15–2.30) | .441 | ||
| Any Hormonal Contraceptive Use | ||||||
| No | 157 (84) | 166 (90) | 1 | |||
| Yes | 31 (16) | 18 (10) | 0.62 (.30–1.27) | .191 | ||
| Last Menstrual Period | ||||||
| ≤7 days | 31 (17) | 23 (13) | 1 | |||
| >7 days | 149 (83) | 155 (87) | 1.05 (.62–1.78) | .858 | ||
| Number of SP | ||||||
| 0 | 25 (13) | 18 (10) | 1 | |||
| 1 | 147 (79) | 129 (70) | 1.01 (.52–1.95) | .979 | ||
| ≥2 | 15 (8) | 37 (20) | 2.11 (.92–4.85) | .079 | ||
| Frequency of Sex | ||||||
| Several times/month or less | 143 (76) | 115 (63) | 1 | 1 | ||
| Several times/week | 45 (24) | 69 (37) | 1.70 (1.07–2.71) | .025 | 1.61 (.99–2.61) | .052 |
| Sex With New Partnere | ||||||
| No | 165 (88) | 147 (80) | 1 | |||
| Yes | 23 (12) | 37 (20) | 1.39 (.84–2.30) | .195 | ||
| Sexual Practices With FSP in Prior 3 Monthsf | ||||||
| Number of FSPsc | ||||||
| 0 | 37 (20) | 32 (17) | 1 | |||
| ≥1 | 151 (80) | 152 (83) | 1.16 (.68–1.99) | .581 | ||
| Receptive Oral Vaginal Sex | ||||||
| Nog | 77 (41) | 69 (38) | 1 | |||
| Yes | 110 (59) | 115 (63) | 1.03 (.68–1.56) | .874 | ||
| Sharing Sex Toys | ||||||
| No toys/washed/ condoms usedg | 142 (76) | 141 (77) | 1 | |||
| Unwashed | 45 (24) | 43 (23) | 0.87 (.53–1.42) | .575 | ||
| Current Partner With BV Symptoms | ||||||
| No/don’t knowg | 187 (100) | 177 (97) | ||||
| Yes | 0 | 7 (4) | … | |||
| Sexual Practices With an MSP in Prior 3 Monthsh | ||||||
| Number of MSPsc | ||||||
| 0 | 173 (92) | 156 (85) | 1 | |||
| ≥1 | 15 (8) | 28 (15) | 1.94 (.89–4.23) | .097 | ||
| Any Penile-Vaginal Sex | ||||||
| Noi | 173 (93) | 158 (86) | 1 | |||
| Yes | 14 (7) | 26 (14) | 1.86 (.81–4.24) | .143 | ||
| Self-Reported Symptoms And Microbiota Measures | ||||||
| Abnormal Vaginal Discharge and/or Odour | ||||||
| No | 175 (93) | 152 (83) | 1 | |||
| Yes | 13 (7) | 32 (17) | 2.09 (1.13–3.87) | .019 | ||
| Nugent Score | ||||||
| 0–3 | 176 (93) | 130 (71) | 1 | |||
| 4–6 | 3 (2) | 15 (8) | 3.96 (1.44–10.93) | .008 | ||
| 7–10 | 9 (5) | 39 (21) | 3.48 (1.95–6.22) | <.001 | ||
| Vaginal Microbiota Typej | ||||||
| Optimal non-Lactobacillus iners microbiota | 118 (66) | 79 (44) | 1 | |||
| L. iners microbiota | 49 (27) | 45 (25) | 1.10 (.69–1.75) | .692 | ||
| Nonoptimal microbiota | 13 (7) | 56 (31) | 4.35 (2.47–7.66) | <.001 |
Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; BV, bacterial vaginosis; FSP, female sexual partner; OR, odds ratio; MSP, male sexual partner; SP, sexual partner (refers to total number of sexual partners in a study interval, female and male).
NOTE: Includes 372 specimens from 101 participants.
aUnivariate logistic regression fitted with generalized estimating equations (GEE) clustered for multiple specimens from each participant.
bMultivariable logistic regression fitted with GEE, clustered for multiple specimens from each participant.
cVariables were dichotomized at median value.
dInterval characteristics were measured as any exposure over the prior follow-up interval (~90 days).
eSex with a new partner with whom first sexual contact was within 90 days. May represent a new FSP or new MSP.
fThe following characteristics/sexual practices with an FSP were left out of the table for simplicity: digital vaginal sex, receptive oral anal sex, digital anal sex. No significant associations between G. vaginalis and these sexual practices were identified.
gOr did not have an FSP.
hThe following sexual practices with an MSP were left out of the table for simplicity: receptive oral sex, digital vaginal sex, and penile-anal sex. No significant associations between G. vaginalis and these sexual practices were identified.
iOr did not have an FSP.
jVaginal microbiota type available for 360 specimens from 100 women. Optimal non-L. iners microbiota includes specimens predominately consisting of non-L. iners spp., L. iners microbiota includes specimens predominately consisting of L. iners, and nonoptimal microbiota includes specimens predominately consisting of non-Lactobacillus spp.
P-values < .05 are bolded to indicate statistically significant associations.
Gardnerella vaginalis detection was associated with smoking, history of penile-vaginal sex, and frequent sex (several times per week) in univariable analyses (Table 1). Multivariable analysis of behavioral practices found that G. vaginalis detection was associated with smoking in the previous 3 months (adjusted OR [aOR], 1.76; 95% CI, 1.06–2.93; P = .028) (Table 1) and lifetime history of penile-vaginal sex (aOR, 2.03; 95% CI, 1.01–4.10; P = .047). There was a borderline association between G. vaginalis and frequent sex (aOR, 1.61; 95% CI, 0.99–2.61; P = .052).
Factors Associated With Number of Gardnerell vaginalis Clades
Three G. vaginalis-positive specimens did not belong to any clade detectable by clade-specific PCR. These specimens were excluded and, consequently, 369 specimens from 101 women contributed to the following analyses.
Clade 4 was the most prevalent clade (n = 136 of 369; 37%; 95% CI, 32%–42%), followed by clade 1 (n = 116 of 369; 31%; 95% CI, 27%–36%), clade 2 (n = 76 of 369; 21%; 95% CI, 17%–25%), and clade 3 (n = 17 of 369; 5%; 95% CI, 3%–7%). Clade 4 was detected in 62 women (n = 62 of 101; 61%), clade 1 was detected in 55 women (n = 55 of 101; 54%), clade 2 was detected in 44 women (n = 44 of 101; 44%), and clade 3 was detected in 14 women (n = 14 of 101; 14%). Of the 181 specimens positive for a G. vaginalis clade, 63 had 1 clade detected (35%; 95% CI, 28%–42%) and 118 had multiple (ie, 2 or more) clades (65%; 95% CI, 58%–72%). Five specimens contained all 4 clades. Clade 3 was only detected in multiclade communities (Figure 1).
Distribution of Gardnerella vaginalis clades. Venn diagram showing the distribution and co-occurrence of G. vaginalis clades in vaginal specimens. Total number of specimens: 181 from 77 women.
In univariable analyses, women reporting symptoms were more likely than asymptomatic women to have multiple G. vaginalis clades relative to no G. vaginalis (RRR, 4.19; 95% CI, 1.85–9.49; P = .001) (Supplementary Table 1). Women with intermediate-BV (RRR, 8.72; 95% CI, 2.32–32.76; P = .001) or Nugent-BV (RRR, 8.72; 95% CI, 4.05–18.78; P < .001) were more likely than women with NS = 0–3 to have multiple clades relative to no G. vaginalis. Having a single clade was not associated with NS. Women with a nonoptimal microbiota were more likely than women with an optimal microbiota to have a single clade (RRR, 3.03; 95% CI, 1.20–7.61; P = .019) or multiple clades (RRR, 9.51; 95% CI, 4.36–20.73; P < .001) relative to no G. vaginalis (Supplementary Table 1).
In univariable analyses, having multiple G. vaginalis clades was associated with smoking in the previous 3 months and sexual practices including increased number of lifetime FSPs, history of penile-vaginal sex, increased frequency of sex, current sexual practices with a male, and sex with a new partner (predominantly representing new FSPs) (Supplementary Table 1). No significant associations were observed between practices and single clade communities. Smoking, lifetime number of FSPs, frequency of sex, and sex with a new partner were included in multivariable analyses (Table 2); sexual practices with a male partner were rare and were omitted to prevent overfitting the model due to their correlation with sex with a new partner. Women reporting smoking in the previous 3 months were more likely than nonsmokers to have multiple G. vaginalis clades relative to no G. vaginalis (adjusted RRR, 2.38; 95% CI, 1.19–4.74; P = .014). No other variable was significant in adjusted analyses.
Multinomial Adjusted Logistic Regression Investigating Behavioral Practices Associated With Number of Gardnerella Vaginalis Clades Detected
| Risk Factor | Single Clade (n = 63) vs G. vaginalis Not Detected | Multiple Clades (n = 118) vs G. vaginalis Not Detected | ||
|---|---|---|---|---|
| aRRR (95% CI) | P Valuea | aRRR (95% CI) | P Valuea | |
| Baseline Characteristics | ||||
| No. of Lifetime FSPsb | ||||
| <5 | 1 | 1 | ||
| ≥5 | 0.94 (.46–1.91) | .859 | 2.04 (.94–4.43) | .072 |
| Interval Practices in Prior 3 Monthsc | ||||
| Smoking | ||||
| No | 1 | 1 | ||
| Yes | 1.15 (.59–2.25) | .674 | 2.38 (1.19–4.74) | .014 |
| Sex With a New Partnerd | ||||
| No | 1 | 1 | ||
| Yes | 0.86 (.31–2.40) | .768 | 1.79 (.91–3.52) | .091 |
| Frequency of Sex | ||||
| Several times/month or less | 1 | 1 | ||
| Sex several times per week | 1.46 (.68–3.14) | .331 | 1.77 (.93–3.37) | .082 |
| Risk Factor | Single Clade (n = 63) vs G. vaginalis Not Detected | Multiple Clades (n = 118) vs G. vaginalis Not Detected | ||
|---|---|---|---|---|
| aRRR (95% CI) | P Valuea | aRRR (95% CI) | P Valuea | |
| Baseline Characteristics | ||||
| No. of Lifetime FSPsb | ||||
| <5 | 1 | 1 | ||
| ≥5 | 0.94 (.46–1.91) | .859 | 2.04 (.94–4.43) | .072 |
| Interval Practices in Prior 3 Monthsc | ||||
| Smoking | ||||
| No | 1 | 1 | ||
| Yes | 1.15 (.59–2.25) | .674 | 2.38 (1.19–4.74) | .014 |
| Sex With a New Partnerd | ||||
| No | 1 | 1 | ||
| Yes | 0.86 (.31–2.40) | .768 | 1.79 (.91–3.52) | .091 |
| Frequency of Sex | ||||
| Several times/month or less | 1 | 1 | ||
| Sex several times per week | 1.46 (.68–3.14) | .331 | 1.77 (.93–3.37) | .082 |
Abbreviations: aRRR, adjusted relative risk ratio; CI, confidence interval; FSP, female sexual partner; MSP, male sexual partner.
NOTE: Variables included in the adjusted analysis were lifetime number of FSPs, smoking in prior 3 months, sex with a new partner in prior 3 months, and frequency of sex.
aMultinomial logistic regression with no G. vaginalis (ie, G. vaginalis not detected) as the referent group. Analysis clustered for multiple specimens from participants (101 clusters).
bVariable was dichotomized at median value.
cInterval practices were measured as any exposure over the prior study interval (~90 days).
dSex with a new partner with whom first sexual contact was within 90 days. May represent a new FSP or new MSP.
P-values < .05 are bolded to indicate statistically significant associations.
Multinomial Adjusted Logistic Regression Investigating Behavioral Practices Associated With Number of Gardnerella Vaginalis Clades Detected
| Risk Factor | Single Clade (n = 63) vs G. vaginalis Not Detected | Multiple Clades (n = 118) vs G. vaginalis Not Detected | ||
|---|---|---|---|---|
| aRRR (95% CI) | P Valuea | aRRR (95% CI) | P Valuea | |
| Baseline Characteristics | ||||
| No. of Lifetime FSPsb | ||||
| <5 | 1 | 1 | ||
| ≥5 | 0.94 (.46–1.91) | .859 | 2.04 (.94–4.43) | .072 |
| Interval Practices in Prior 3 Monthsc | ||||
| Smoking | ||||
| No | 1 | 1 | ||
| Yes | 1.15 (.59–2.25) | .674 | 2.38 (1.19–4.74) | .014 |
| Sex With a New Partnerd | ||||
| No | 1 | 1 | ||
| Yes | 0.86 (.31–2.40) | .768 | 1.79 (.91–3.52) | .091 |
| Frequency of Sex | ||||
| Several times/month or less | 1 | 1 | ||
| Sex several times per week | 1.46 (.68–3.14) | .331 | 1.77 (.93–3.37) | .082 |
| Risk Factor | Single Clade (n = 63) vs G. vaginalis Not Detected | Multiple Clades (n = 118) vs G. vaginalis Not Detected | ||
|---|---|---|---|---|
| aRRR (95% CI) | P Valuea | aRRR (95% CI) | P Valuea | |
| Baseline Characteristics | ||||
| No. of Lifetime FSPsb | ||||
| <5 | 1 | 1 | ||
| ≥5 | 0.94 (.46–1.91) | .859 | 2.04 (.94–4.43) | .072 |
| Interval Practices in Prior 3 Monthsc | ||||
| Smoking | ||||
| No | 1 | 1 | ||
| Yes | 1.15 (.59–2.25) | .674 | 2.38 (1.19–4.74) | .014 |
| Sex With a New Partnerd | ||||
| No | 1 | 1 | ||
| Yes | 0.86 (.31–2.40) | .768 | 1.79 (.91–3.52) | .091 |
| Frequency of Sex | ||||
| Several times/month or less | 1 | 1 | ||
| Sex several times per week | 1.46 (.68–3.14) | .331 | 1.77 (.93–3.37) | .082 |
Abbreviations: aRRR, adjusted relative risk ratio; CI, confidence interval; FSP, female sexual partner; MSP, male sexual partner.
NOTE: Variables included in the adjusted analysis were lifetime number of FSPs, smoking in prior 3 months, sex with a new partner in prior 3 months, and frequency of sex.
aMultinomial logistic regression with no G. vaginalis (ie, G. vaginalis not detected) as the referent group. Analysis clustered for multiple specimens from participants (101 clusters).
bVariable was dichotomized at median value.
cInterval practices were measured as any exposure over the prior study interval (~90 days).
dSex with a new partner with whom first sexual contact was within 90 days. May represent a new FSP or new MSP.
P-values < .05 are bolded to indicate statistically significant associations.
Factors Associated With Detection of Each Gardnerell vaginalis Clade
Seventy-two women changed clade at least once over the study period, accounting for 120 instances of change. Acquisition of new clade/s was the most frequent change observed (n = 71 of 120, 59%), followed by loss of clade/s (n = 38 of 120, 32%), and a combination of loss and acquisition of clade/s (n = 11 of 120, 9%). Five women had stable clade distribution over time.
Four univariable analyses were conducted to assess factors associated with detection of each specific clade versus detection of any other clade/s (Supplementary Table 2). Because multiclade specimens were common and we wanted to examine factors associated with detection of each individual clade rather than detection of G. vaginalis, we excluded specimens in which no G. vaginalis clade was detected. Seventy-seven women contributed 181 specimens to each of the 4 analyses.
Clade 1 detection was associated with Nugent-BV (OR, 3.55; 95% CI, 1.76–7.18; P < .001) and nonoptimal vaginal microbiota (OR, 3.36; 95% CI, 1.65–6.84; P = .001). Clades 2 and 3 were both associated with intermediate-BV (clade 2 OR = 3.49, 95% CI = 1.17–10.36, P = .025; and clade 3 OR = 4.90, 95% CI = 1.11–21.57, P = .035) and Nugent-BV (clade 2 OR = 1.90, 95% CI = 1.02–3.55, P = .043; and clade 3 OR = 3.67, 95% CI = 1.22–11.04, P = .020), but they were not associated with vaginal microbiota composition. Clade 4 was not associated with NS or vaginal microbiota composition (Supplementary Table 2).
For each clade, behavioral practices significantly associated by univariable analysis were included in a clade-specific multivariable analysis. Detection of clades 1, 2, and 3 were associated with interval practices (ie, those performed in the 3 months before specimen collection). Having clade 1 versus any other clade/s was associated with smoking (aOR, 2.42; 95% CI, 1.12–5.25; P = .025) (Table 3) and ≥1 sexual partner of any gender (aOR, 4.03; 95% CI, 1.16–14.01; P = .028), after adjusting for sex frequency. Having clade 2 versus any other clade/s was associated with sharing of unwashed sex toys with an FSP (aOR, 2.59; 95% CI, 1.22–5.51; P = .013) and recent penile-vaginal sex (aOR, 5.67; 95% CI, 1.74–18.51; P = .004). Having clade 3 versus any other clade/s was associated with sex with a new partner (aOR, 5.02; 95% CI, 1.25–20.18; P = .023) and having a current FSP with BV symptoms (aOR, 24.73; 95% CI, 3.37–181.27; P = .002). Having clade 4 versus any other clade/s was associated with having ≥5 lifetime FSPs (aOR, 3.35; 95% CI, 1.45–7.75; P = .005).
Practices Associated With Detection of Specific Gardnerella vaginalis Clades by Logistic Regressiona
| G. vaginalis Clade 1b | OR (95% CI) | P Value | aOR (95% CI) | P Value |
|---|---|---|---|---|
| Any Smoking in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 2.29 (1.40–6.10) | .004 | 2.42 (1.12–5.25) | .025 |
| Number of SP in Prior 3 Months | ||||
| 0 | 1 | 1 | ||
| 1 | 5.09 (1.55–16.66) | .007 | 4.03 (1.16–14.01) | .028 |
| ≥2 | 8.19 (2.05–32.75) | .003 | 5.25 (1.22–22.50) | .026 |
| Frequency of Sex With Any SP | ||||
| Several times/month or less | 1 | 1 | ||
| Several times/week | 2.03 (1.07–3.88) | .031 | 1.60 (.79–3.25) | .190 |
| G. vaginalis Clade 2c | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Any Sharing of Sex Toys in Prior 3 Months | ||||
| No toys/washed/condoms used | 1 | 1 | ||
| Unwashed | 2.21 (1.08–4.52) | .030 | 2.59 (1.22–5.51) | .013 |
| Any Penile-Vaginal Sex in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 4.80 (1.52–15.17) | .007 | 5.67 (1.74–18.51) | .004 |
| G. vaginalis Clade 3d | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Sex With a New Partner in Prior 3 Monthse | ||||
| No | 1 | 1 | ||
| Yes | 3.09 (1.10–8.69) | .032 | 2.90 (.95–8.92) | .063 |
| Current Partner With BV Symptoms | ||||
| No/don’t know | 1 | 1 | ||
| Yes | 17.40 (3.36–90.11) | .001 | 16.95 (3.11–92.23) | .001 |
| G. vaginalis Clade 4f | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| No. of Lifetime FSPs | ||||
| <5 | 1 | 1 | ||
| ≥5 | 3.93 (1.74–8.87) | .001 | 3.35 (1.45–7.75) | .005 |
| Lifetime History of Vaginal Sex With a Mang | ||||
| No | 1 | 1 | ||
| Yes | 2.88 (1.13–7.32) | .026 | 2.09 (.82–5.33) | .125 |
| G. vaginalis Clade 1b | OR (95% CI) | P Value | aOR (95% CI) | P Value |
|---|---|---|---|---|
| Any Smoking in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 2.29 (1.40–6.10) | .004 | 2.42 (1.12–5.25) | .025 |
| Number of SP in Prior 3 Months | ||||
| 0 | 1 | 1 | ||
| 1 | 5.09 (1.55–16.66) | .007 | 4.03 (1.16–14.01) | .028 |
| ≥2 | 8.19 (2.05–32.75) | .003 | 5.25 (1.22–22.50) | .026 |
| Frequency of Sex With Any SP | ||||
| Several times/month or less | 1 | 1 | ||
| Several times/week | 2.03 (1.07–3.88) | .031 | 1.60 (.79–3.25) | .190 |
| G. vaginalis Clade 2c | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Any Sharing of Sex Toys in Prior 3 Months | ||||
| No toys/washed/condoms used | 1 | 1 | ||
| Unwashed | 2.21 (1.08–4.52) | .030 | 2.59 (1.22–5.51) | .013 |
| Any Penile-Vaginal Sex in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 4.80 (1.52–15.17) | .007 | 5.67 (1.74–18.51) | .004 |
| G. vaginalis Clade 3d | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Sex With a New Partner in Prior 3 Monthse | ||||
| No | 1 | 1 | ||
| Yes | 3.09 (1.10–8.69) | .032 | 2.90 (.95–8.92) | .063 |
| Current Partner With BV Symptoms | ||||
| No/don’t know | 1 | 1 | ||
| Yes | 17.40 (3.36–90.11) | .001 | 16.95 (3.11–92.23) | .001 |
| G. vaginalis Clade 4f | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| No. of Lifetime FSPs | ||||
| <5 | 1 | 1 | ||
| ≥5 | 3.93 (1.74–8.87) | .001 | 3.35 (1.45–7.75) | .005 |
| Lifetime History of Vaginal Sex With a Mang | ||||
| No | 1 | 1 | ||
| Yes | 2.88 (1.13–7.32) | .026 | 2.09 (.82–5.33) | .125 |
Abbreviations: aOR, adjusted odds ratio; BV, bacterial vaginosis; CI, confidence interval; FSP, female sexual partner; MSP, male sexual partner; OR, odds ratio; SP, sexual partner (may refer to female or male partner).
aIncludes specimens positive for 1 or more G. vaginalis clade (181 specimens from 77 women).
bLogistic regression fitted with generalized estimating equations (GEE) with absence of clade 1 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
cLogistic regression fitted with GEE with absence of clade 2 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
dLogistic regression fitted with GEE with absence of clade 3 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
eSex with a new partner with whom first sexual contact was within 90 days. May represent a new FSP or new MSP.
fLogistic regression fitted with GEE with absence of clade 4 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
gVariable was dichotomized at median value.
P-values < .05 are bolded to indicate statistically significant associations.
Practices Associated With Detection of Specific Gardnerella vaginalis Clades by Logistic Regressiona
| G. vaginalis Clade 1b | OR (95% CI) | P Value | aOR (95% CI) | P Value |
|---|---|---|---|---|
| Any Smoking in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 2.29 (1.40–6.10) | .004 | 2.42 (1.12–5.25) | .025 |
| Number of SP in Prior 3 Months | ||||
| 0 | 1 | 1 | ||
| 1 | 5.09 (1.55–16.66) | .007 | 4.03 (1.16–14.01) | .028 |
| ≥2 | 8.19 (2.05–32.75) | .003 | 5.25 (1.22–22.50) | .026 |
| Frequency of Sex With Any SP | ||||
| Several times/month or less | 1 | 1 | ||
| Several times/week | 2.03 (1.07–3.88) | .031 | 1.60 (.79–3.25) | .190 |
| G. vaginalis Clade 2c | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Any Sharing of Sex Toys in Prior 3 Months | ||||
| No toys/washed/condoms used | 1 | 1 | ||
| Unwashed | 2.21 (1.08–4.52) | .030 | 2.59 (1.22–5.51) | .013 |
| Any Penile-Vaginal Sex in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 4.80 (1.52–15.17) | .007 | 5.67 (1.74–18.51) | .004 |
| G. vaginalis Clade 3d | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Sex With a New Partner in Prior 3 Monthse | ||||
| No | 1 | 1 | ||
| Yes | 3.09 (1.10–8.69) | .032 | 2.90 (.95–8.92) | .063 |
| Current Partner With BV Symptoms | ||||
| No/don’t know | 1 | 1 | ||
| Yes | 17.40 (3.36–90.11) | .001 | 16.95 (3.11–92.23) | .001 |
| G. vaginalis Clade 4f | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| No. of Lifetime FSPs | ||||
| <5 | 1 | 1 | ||
| ≥5 | 3.93 (1.74–8.87) | .001 | 3.35 (1.45–7.75) | .005 |
| Lifetime History of Vaginal Sex With a Mang | ||||
| No | 1 | 1 | ||
| Yes | 2.88 (1.13–7.32) | .026 | 2.09 (.82–5.33) | .125 |
| G. vaginalis Clade 1b | OR (95% CI) | P Value | aOR (95% CI) | P Value |
|---|---|---|---|---|
| Any Smoking in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 2.29 (1.40–6.10) | .004 | 2.42 (1.12–5.25) | .025 |
| Number of SP in Prior 3 Months | ||||
| 0 | 1 | 1 | ||
| 1 | 5.09 (1.55–16.66) | .007 | 4.03 (1.16–14.01) | .028 |
| ≥2 | 8.19 (2.05–32.75) | .003 | 5.25 (1.22–22.50) | .026 |
| Frequency of Sex With Any SP | ||||
| Several times/month or less | 1 | 1 | ||
| Several times/week | 2.03 (1.07–3.88) | .031 | 1.60 (.79–3.25) | .190 |
| G. vaginalis Clade 2c | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Any Sharing of Sex Toys in Prior 3 Months | ||||
| No toys/washed/condoms used | 1 | 1 | ||
| Unwashed | 2.21 (1.08–4.52) | .030 | 2.59 (1.22–5.51) | .013 |
| Any Penile-Vaginal Sex in Prior 3 Months | ||||
| No | 1 | 1 | ||
| Yes | 4.80 (1.52–15.17) | .007 | 5.67 (1.74–18.51) | .004 |
| G. vaginalis Clade 3d | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| Sex With a New Partner in Prior 3 Monthse | ||||
| No | 1 | 1 | ||
| Yes | 3.09 (1.10–8.69) | .032 | 2.90 (.95–8.92) | .063 |
| Current Partner With BV Symptoms | ||||
| No/don’t know | 1 | 1 | ||
| Yes | 17.40 (3.36–90.11) | .001 | 16.95 (3.11–92.23) | .001 |
| G. vaginalis Clade 4f | OR (95% CI) | P Value | aOR (95% CI) | P Value |
| No. of Lifetime FSPs | ||||
| <5 | 1 | 1 | ||
| ≥5 | 3.93 (1.74–8.87) | .001 | 3.35 (1.45–7.75) | .005 |
| Lifetime History of Vaginal Sex With a Mang | ||||
| No | 1 | 1 | ||
| Yes | 2.88 (1.13–7.32) | .026 | 2.09 (.82–5.33) | .125 |
Abbreviations: aOR, adjusted odds ratio; BV, bacterial vaginosis; CI, confidence interval; FSP, female sexual partner; MSP, male sexual partner; OR, odds ratio; SP, sexual partner (may refer to female or male partner).
aIncludes specimens positive for 1 or more G. vaginalis clade (181 specimens from 77 women).
bLogistic regression fitted with generalized estimating equations (GEE) with absence of clade 1 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
cLogistic regression fitted with GEE with absence of clade 2 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
dLogistic regression fitted with GEE with absence of clade 3 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
eSex with a new partner with whom first sexual contact was within 90 days. May represent a new FSP or new MSP.
fLogistic regression fitted with GEE with absence of clade 4 as the referent category. Analysis clustered for multiple specimens from participants (77 clusters).
gVariable was dichotomized at median value.
P-values < .05 are bolded to indicate statistically significant associations.
Discussion
We explored the distribution of G. vaginalis clades in WSW and found that clade 1 was the only clade associated with both Nugent-BV and a nonoptimal (Lactobacillus deficient) vaginal microbiota. Also of note, clade 4 was the most prevalent clade but was not associated with these 2 vaginal states. Factors associated with detection of G. vaginalis included smoking, frequent sex, and a history of penile-vaginal sex, and clades were associated with a range of differing sexual practices and behaviors in adjusted analysis. Clades 1, 2, and 3 were associated with recent behavioral and sexual practices, and clade 4 was associated with sexual practices before enrollment. These findings support the sexual exchange of G. vaginalis, and they suggest that different G. vaginalis clades may have varying levels of pathogenicity, differ in mode of acquisition and duration of infection, and may circulate in different populations or sexual networks.
Previous studies investigating G. vaginalis clade distribution that use the same clade typing applied in our study have inconsistently associated individual clades with BV [12, 21, 23–25]. However, most studies agree that clade 1 is associated with BV [21, 23, 25]. Our finding that clade 1 was associated with both Nugent-BV and nonoptimal vaginal microbiota suggests that it may have increased pathogenicity compared with other clades. A recent comparative genomic and ecotyping analysis of 35 G. vaginalis strains highlighted key differences between ecotype 1 (corresponds to clade 1) and other G. vaginalis ecotypes that may contribute to its pathogenicity [18]. Ecotype 1 uniquely encodes glycosidases that may aid cervicovaginal mucus degradation, and it has enriched galactose and pentose sugar metabolism pathways, which may provide ecotype 1 with an advantage when cocolonizing the vagina with lactic acid-producing bacteria [18].
Our finding that clade 4 was not associated with Nugent-BV or nonoptimal vaginal microbiota is supported by 2 studies reporting no association of clade 4 with BV [23, 25]. Clade 4 strains lack sialidase activity [19, 20], which is associated with mucin degradation in BV, further supporting reduced pathogenicity of clade 4. However, Vodstrcil et al [12] found that clade 4 was associated with Nugent-BV in a cohort of 52 young, sexually inexperienced women, and 2 additional studies noted an association between Nugent-BV and increasing prevalence and/or load of clade 4 [21, 24]. Inconsistencies between studies may be due to population differences including sexual practices and networks, behavioral practices such as smoking, past history of BV, and ethnicity. Previous associations of clade 4 and BV may also be a result of high load infections [21, 24]. Another consideration is that differences between studies are a result of spurious findings due to unmeasured confounding. Gardnerella vaginalis clades do not often occur in isolation, and associations between specific clades and BV may be due to high abundance and/or presence of other BV-associated bacteria or due to depleted levels of optimal Lactobacillus spp. in the vagina. Given the polymicrobial nature of BV, future studies should consider interactions between G. vaginalis clades and other inhabitants of the vaginal microbiome that may play an integral role in BV pathogenesis.
Multiclade G. vaginalis communities are common and are associated with BV [12, 21, 23, 25]. In our study, detection of multiple clades was associated with intermediate-BV, Nugent-BV, and nonoptimal vaginal microbiota. Multiple clades may act synergistically to supress Lactobacillus spp. or form biofilms. Alternatively, there may be one G. vaginalis clade driving disease as well as passenger clades; in specimens in which clades co-occur, it is difficult to identify which specific clade/s is driving disease because other clades may rapidly cohabit after an initiation event. To investigate whether there were independent associations between individual G. vaginalis clades and BV, we conducted an additional analysis that included all 4 clades with Nugent-BV as the outcome (Supplementary Table 3). The results of the adjusted analysis were consistent with our univariable findings, with clade 1 having the strongest association with Nugent-BV (Supplementary Table 2). However, this analysis was limited by the correlation between clades so was not investigated further. The high prevalence of multiclade communities may explain contradictory associations of specific clades with BV in previous studies.
Clades 1, 2, and 3 were significantly associated with current sexual practices, and we observed a nonsignificant trend between detection of multiple G. vaginalis clades and sexual practices including increased sex frequency and sex with a new partner. Several studies document exchange of G. vaginalis between sexual partners. Heterosexual couples share identical G. vaginalis strains [36] and demonstrate high concordance for G. vaginalis biofilm [37]. A study using culture methods reported increased G. vaginalis colonization with increasing frequency of digital-vaginal sex and sex-toy use in WSW [38]. Using PCR, G. vaginalis prevalence has been shown to increase after sexual debut and with increasing numbers of sexual partners [13]. Differing associations seen with individual clades does not necessarily suggest that specific clades are only transmitted by specific sexual practices, but rather provides support for the sexual exchange of G. vaginalis clades across a range of different practices.
Smoking was associated with clade 1 detection, and women who reported smoking were more likely than nonsmokers to have multiple G. vaginalis clades detected. This is consistent with a previous report of increased G. vaginalis detection by culture amongst smokers compared to nonsmokers [39], and it is supported by the frequent association of smoking with BV [26, 27, 40]. The mechanism by which smoking may increase G. vaginalis in the vagina is unknown; however, smokers have an altered vaginal metabolome (including high concentrations of nicotine and its derivatives) [41] and reduced oestradiol levels [42], which may impact vaginal microbiota composition. In addition, smoking negatively impacts immune function [43], which may result in reduced clearance of G. vaginalis, similar to what has been observed for human papillomavirus [44]. It is also possible that our observed association between smoking and both clade 1 and multiclade communities is a result of unmeasured confounding.
Three specimens that tested positive for G. vaginalis by species-specific PCR did not have any clade detected by clade-specific PCR. This raises the possibility that more than 4 G. vaginalis clades exist or that genetic subgroups may exist within the 4 clades [18, 20]. In early 2019, Vaneechoutte et al [45] proposed the existence of 13 different species within the Gardnerella genus, and additional clades/subgroups are likely to be revealed as we analyze larger numbers of G. vaginalis isolates.
Our study has limitations. Nugent Score was used to facilitate self-collection of vaginal specimens at home. Participants were not examined by a clinician, which limited our ability to assess G. vaginalis clade distribution in Amsel-BV. Longitudinal specimens included in the analysis typically represented the last 3 specimens collected from a participant and were not selected randomly or from specified study time points, which may have biased results. Specimens were collected at 3-month intervals so the immediate impact of practices on G. vaginalis clade distribution is unclear. Frequent sampling would address this and clarify whether G. vaginalis clades persist or are transient. Because we did not use the PCR assay to quantitate clades, clade associations are based on detection rather than quantity, and we were unable to assess associations between covariates and quantity of each clade. Finally, the study population comprised highly educated women who were predominately Australian born, and our findings may not be generalizable to other populations including sexually inexperienced women or women who exclusively have sex with men.
Conclusions
We report that multiclade G. vaginalis communities and clades 1, 2, and 3 were associated with Nugent-BV in Australian WSW. That multiclade communities and clade 1 were both also associated with nonoptimal (Lactobacillus deficient) vaginal microbiota indicates increased pathogenicity. The finding that clade 4 was not associated with either BV or nonoptimal vaginal microbiota strongly suggests it may be a commensal clade. Detailed comparative analyses of commensal and pathogenic G. vaginalis genetic types may help to identify potential mechanisms of G. vaginalis pathogenesis. Individual G. vaginalis clades were associated with differing sexual practices, adding to the growing evidence supporting the sexual exchange of G. vaginalis. These data have implications for BV treatment and suggest that partner treatment may be an effective strategy to improve BV cure.
Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Notes
Acknowledgments. We thank Glenda Fehler, Susan Peterson, Matthew Law, Dr. Marcus Chen, Dr. Sandra Walker, Dr. Jade Bilardi, and Clare Bellhouse for their contributions to the original Women On Women’s (WOW) Cohort study, from which this study arose. We also thank Prof Sepehr Tabrizi and Dr. Jimmy Twin for contributions to the laboratory work.
Financial support. This work funded by The Australian National Health and Medical Research Council Project Grant (1020457; to C. S. B.) and the Australian National Health and Medical Research Council Program Grant (1071269; to S. M. G. and C. K. F.). E. L. P. was supported by an Australian Government Research Training Program Scholarship.
Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.
