Factors Associated with the Recurrence, Persistence and Clearance of Asymptomatic Bacterial Vaginosis among Young African American Women: A Repeated Measures Latent Class Analysis

Abstract

Background

While risk factors of recurrent and persistent bacterial vaginosis (BV) have been explored in the literature, the longitudinal incidence patterns of BV remain elusive.

Methods

We conducted a secondary analysis of longitudinal data from a randomized clinical trial of metronidazole treatment for asymptomatic BV. Repeated Measures Latent Class Analysis (RMLCA) was used to identify distinct longitudinal patterns of incident BV cases. Multinomial regression analysis was used to determine the predictors of class membership. The multivariable model included age, last BV treatment, douching frequency, birth control, sexual risk behavior and assignment to treatment arm.

Results

A total of 858 African American women, asymptomatic for BV were included in the analysis. Three emergent patterns of BV over 12-months were identified by RMLCA: persistent (55.9%), recurrent (30.5%) and clearance (13.5%). Participants who had douched at least once had significantly lower odds to be in the recurrent class versus the clearance class (adjusted Odds Ratio [adjOR]: 0.55; 95% Confidence Interval [CI]: 0.18–0.63). Women who had sex with women (WSW) had significantly lower odds of belonging to the persistent class versus the clearance class (adjOR: 0.38; 95% CI: 0.22–0.68) and the recurrent class (adjOR: 0.43; 95% CI: 0.23–0.81). Those who were assigned to the treatment arm had significantly increased odds of being in the recurrent class versus the clearance class (adjOR: 1.92; 95% CI: 1.22–3.03). Women >21 years were significantly more likely to be in the recurrent class (adjOR: 1.88; 95% CI: 1.17–3.00) than in the clearance class.

Conclusion

Assessment of BV cases revealed distinct patterns of recurrence and persistence of BV which were significantly associated with douching, being in the treatment arm, and being a woman who had sex with women.

Short Summary

Among young women with asymptomatic BV who were followed for one year, we identified factors associated with thse recurrence, persistence, or clearance of BV.

Introduction

Bacterial vaginosis (BV) is a polymicrobial condition with vaginal dysbiosis that was first documented by Gardner and Dukes in 1955.^1,2 BV is the most common cause of vaginal discharge,^3–5 and is associated with increased risk of preterm birth, pelvic inflammatory disease, and sexually transmitted diseases.⁴ The current recommended forms of treatment include metronidazole or clindamycin.^5–7 Though short term BV cure rates are high, approximately 80–90%, there are, however, high recurrence rates of BV post-treatment.⁵ There is no universally accepted definition of recurrent BV, and effective treatment for recurrent BV does not exist.⁷ Oral metronidazole is the recommended form of treatment for symptomatic BV; however, it may not be an effective treatment for women with recurrent BV who have failed metronidazole treatment in the past.^6,7 The effect of metronidazole on normalizing vaginal flora in women with asymptomatic BV is not known. It should be noted that approximately 25% of women with BV experience spontaneous resolution of BV.⁶ Furthermore, in some cases, BV continues to persist. Several longitudinal BV incidence patterns could be defined, and these patterns could be influenced by sexual behaviors or demographic characteristics.

The National Health and Nutrition Examination Survey (NHANES) data between 2001 and 2004 estimated the prevalence of BV in the United States (US) to be 29.2%.³ The prevalence of BV varies by race and ethnicity with Black and Hispanic women having a higher prevalence of BV compared to White and Asian women.³ Cultural backgrounds also influence the use of intravaginal practices such as vaginal douching which is associated with BV.⁸ Vaginal douching involves the use of cleansing solution in the vagina,⁹ and the effects of this practice have been disputed as some literature has found varying effects of douching.^10–12 Additionally, sexual activity increases the risk of incident and recurrent BV.^3,13 Women who have sex with women (WSW) and women who have sex with multiple male sex partners are at increased risk of BV.³ Furthermore, women who remain with their male sex partner after treatment are at increased risk of recurrent BV.³ Literature states that hormonal contraception and condom use are protective against BV.^1,3,13

Latent Class Analysis (LCA) has been used in research to assign participants to discrete, mutually exclusive groups known as latent classes through a probabilistic approach.^14,15 Latent classes are determined based on selected observed categorical variables.¹⁶ LCA has been increasingly popular in behavioral science research to investigate latent risk factors.¹⁷ Okeke and colleagues have implemented LCA techniques to investigate health care utilization behaviors and disengagement from HIV care.¹⁷ LCA has also been used in BV research by Jespers et al.¹⁸ They identified two types of “normal flora” and one “BV type flora” with LCA, which were significantly associated with various Lactobacillus bacteria.¹⁸ Though their research used LCA to examine BV and its relationship with vaginal flora, to the best of our knowledge, LCA has not been used to examine demographic and behavioral characteristics of BV and BV incidence patterns after treatment. Our study aimed to identify longitudinal patterns of incident BV patterns, and evaluate the demographic and behavioral risk factors associated with these patterns among women diagnosed with BV at baseline. This is of public health importance due to the high rates of recurrence of BV.

Materials and Methods

Study Design

We conducted a retrospective analysis of African American women of reproductive age who participated in a randomized clinical trial (RCT). The RCT aimed to determine whether regular screening and treatment for asymptomatic BV reduces the 1-year incidence of chlamydia and gonorrhea among women with asymptomatic BV. The RCT was conducted in six geographic locations (Baltimore, Maryland; Birmingham, Alabama; Durham and Raleigh, North Carolina; Pittsburgh, Pennsylvania; and San Francisco, California). The methodology used for the RCT has been previously described.¹⁹ Here we briefly summarize the key aspects of the study.

Women at increased risk for STDs who had asymptomatic BV were recruited from STD, family planning, obstetrics-gynecology and clinical research clinics.¹⁹ Participants were eligible for the study if they had vaginal pH >4.5 and >20% clue cells on vaginal wet preparation microscopy and did not complain of abnormal vaginal discharge or odor. Women who exhibited BV symptoms, such as increased vaginal discharge and vaginal odor, were not included in the study. Based on Nugent Score (NS) 87% of women had BV at baseline. Eligible participants were randomized to treatment and control groups and followed-up for a period of twelve months. Women randomized to the treatment group received 500mg of oral metronidazole for seven days upon enrollment. Participants received home testing kits every two months to collect vaginal samples for BV testing. Women in the treatment arm received 500mg of oral metronidazole for seven days after every positive BV result. Women in the control arm did not receive treatment for asymptomatic BV at any time during the study, in accordance with Centers for Disease Control and Prevention (CDC) guidelines.²⁰ All women, irrespective of the arm they were randomized to, could receive treatment if they developed symptoms during the follow-up period from their respective primary care physicians. Institutional Review Board (IRB) approvals were obtained by all sites approved to conduct the initial study. The current study was approved by the Florida International University IRB.

Measures

Participants obtained self-collected swabs. Swabs were rolled across a microscope slide and sent to a lab for evaluation. NS of vaginal smears were used to assess the presence of BV every two months. NS between 0–3 are considered healthy, 4–6 intermediate flora and 7–10 BV. For this study, a NS between 0–6 was considered as ‘no BV’ and 7–10 indicated ‘BV’.²¹ Baseline demographic, medical and sexual risk behavior characteristics were assessed using a self-administered questionnaire. Age, race, ethnicity and highest level of education were measured. Medical history assessed included prior antibiotic use (past 30 days), prior lifetime episodes of BV (never, once, 2–4 times, ≥ 5 times), prior lifetime treatment for BV (≤6 months, 7–12 months, ≥12 months ago), prior pregnancy (yes/no) and frequency of vaginal douching (ever/never). Types of birth control used in the past year were also assessed; birth control pills, birth control patch, Nuva-ring, condoms, spermicide cream, Depo-Provera shot, intra-uterine device (IUD) and other. Sexual risk behavior characteristics evaluated in the past year included; number of different sex partners, number of oral sex partners (receptive), number of unprotected anal sex partners, number of unprotected vaginal sex partners, women who have sex with women and new sex partners. Randomization arms were also assessed; treatment with metronidazole and no treatment.

Statistical Analysis

There were 1,365 women enrolled in the original RCT and 1,160 agreed to the future use of the data for research. For this study, we only included women who self-identified as African American (n=977). African American women were selected due to their increased risk of BV and unique cultural attributes. Participants with missing BV results for all follow-up timepoints were excluded from the analysis (119 women; 12.2%). Listwise deletion was selected since there was a large enough sample and BV results were missing completely at random. Consequently, 858 African American women were included in the analysis. There were no statistically significant differences with respect to age, education, ethnicity and race between women that were included compared to those who were excluded. Baseline participant characteristics were described using frequencies for categorical variables and means and standard deviations for continuous variables. Age was the only continuous variable and was categorized for easy comparison of groups. Chi-squared or Fisher exact tests for categorical variables were used as appropriate to assess bivariate differences in demographic characteristics or behaviors by latent class assignment.

Latent classes were modeled through Repeated Measures Latent Class Analysis (RMLCA) using PROC LCA in SAS.¹⁶ For RMLCA of BV patterns over a twelve-month period, we used the BV results (yes/no) at months two, four, six, eight, ten and twelve. Baseline BV results were excluded since all women had asymptomatic BV at baseline and inclusion of the baseline status may bias results. We assessed the model fit statistics of models which included two to five latent classes. The maximum number of iterations for each model was 9000 and the seed was 1000. Bayesian information criterion (BIC), adjusted Bayesian information criterion (aBIC), Akaike information criterion (AIC) and entropy were used to identify the best model.¹⁶ Participants were assigned to latent classes based on the highest posterior probability of membership.

We used multinomial logistic regression to examine the relationship between baseline demographic characteristics, medical history, sexual risk behavior in the past year and latent class membership. The clearance class was used as the reference group due to its clinical relevance as this is the most optimal outcome. The analysis considered demographic and behavioral characteristics that were assessed at baseline. Variables with Chi-square test p-values <0.2 were included in the multivariable multinomial logistic regression, as seen in literature.²² Age, frequency of douching, last BV treatment, prior pregnancy, birth control (Depo Provera shot), unprotected anal sex partners, sex with other women, new sex partners and assignment to the treatment arm were included in the final model. SAS version 9.4 (SAS Institute Inc, Cary, NC) was used to clean and analyze all data, and RMLCA was conducted using the SAS-based add on package PROC LCA, version 1.3.2 (University Park, PA).

Results

Participant Characteristics

The mean age of the sample was 21.25 (SD: ±2.12) years. All women were African American and 3.8% of women also self-identified as Hispanic/Latino. Most women were educated, with 48.4% completing more than a high school diploma or GED. Participant responses indicated that 94.8% of women did not use antibiotics in the past 30 days and that most women had never been diagnosed with BV (54.2%). Of the women who reported prior episodes of BV, 39.1% were treated ≤6 months ago. Additionally, 57.5% of women reported a prior pregnancy and 50.3% had douched at least once. Various forms of birth control were used by most women (93%) such as, birth control pills/patches (19.6%), Nuva Ring (4.2%), condoms (77.6%), Depo-Provera shot (14.7%), IUDs (9.8%) and spermicide (1.2%). All women engaged in sexual activity in the prior year. Most women had two or more different sex partners (76.6%) as well as two or more oral sex partners (52.4%) in the prior year. There were 150 (17.5%) women that engaged in unprotected anal sex, and 799 (93.1%) women that engaged in unprotected vaginal sex. Moreover, 97 (11.3%) women had sex with other women and 418 (48.7%) had new sex partners. Approximately half of the sample were assigned to the treatment arm (N=438). Table one details the participant characteristics of the study sample stratified by latent class. Classes were named based on item-response probabilities as follows: “persistent” (52.6%), “recurrent” (29.3%) and “clearance” (18.1%) (Table 2).

Table 2:

Crude and Adjusted Odds Ratios for Risk Factors Associated with Latent Class by Multinomial Logistic Regression Analysis

Characteristics	Persistent vs Clearance		Recurrent vs Clearance
	Crude OR (95% CI)	Adjusted ORa (95% CI)	Crude OR (95% CI)	Adjusted ORa (95% CI)
Demographic Variables
Age
>21 years	1.72 (1.15–2.59)	1.90 (1.23–2.95)	1.46 (0.94–2.26)	1.88 (1.17–3.00)
≤21 years	-	-	-	-
Education
HS	0.99 (0.55–1.80)	-	0.83 (0.43–1.58)	-
More than HS	0.95 (0.54–1.68)	-	1.09 (0.59–2.01)	-
No HS	-	-	-	-
Ethnicity
Hispanic/Latino	0.59 (0.22–1.56)	-	0.88 (0.32–2.4)	-
Not Hispanic/Latino	-	-	-	-
Medical History
Prior Antibiotic Useb
Yes	1.05 (0.42–2.61)	-	0.96 (0.36–2.58)	-
Noc	-	-	-	-
Prior Episodes of BV
Once	0.71 (0.42–1.19)	-	0.91 (0.52–1.60)	-
2–4 times	0.61 (0.36–1.03)	-	0.79 (0.45–1.38)	-
≥5 times	0.64 (0.28–1.49)	-	1.15 (0.48–2.73)	-
Neverd	-	-	-	-
Last BV treatment
≤6 months ago	0.77 (0.44–1.35)	0.74 (0.42–1.32)	0.97 (0.54–1.78)	0.98 (0.53–1.82)
7–12 months ago	0.37 (0.20–0.68)	0.34 (0.18–0.63)	0.70 (0.37–1.31)	0.68 (0.35–1.30)
>12 months ago	0.75 (0.43–1.31)	0.68 (0.38–1.21)	0.90 (0.49–1.64)	0.90 (0.48–1.68)
Nevere	-	-	-	-
Prior Pregnancy
Yes	1.18 (0.78–1.77)	1.08 (0.69–1.67)	0.81 (0.52–1.25)	0.70 (0.44–1.13)
No	-	-	-	-
Douching History
At least once	0.88 (0.59–1.33)	0.80 (0.53–1.22)	0.60 (0.38–0.93)	0.55 (0.35–0.87)
Never	-	-	-	-
Birth Control Methods
Pills/Patch
Yes	1.05 (0.63–1.77)	-	1.03 (0.59–1.80)	-
No	-	-	-	-
Nuva Ring
Yes	1.03 (0.34–3.12)	-	1.70 (0.55–5.24)	-
No	-	-	-	-
Condoms
Yes	1.22 (0.76–1.96)	-	1.12 (0.67–1.87)	-
No	-	-	-	-
Depo-Provera Shot
Yes	0.70 (0.40–1.23)	0.67 (0.37–1.20)	1.17 (0.66–2.10)	1.26 (0.69–2.30)
No	-	-	-	-
Intra-Uterine Device
Yes	1.81 (0.8–4.1)	-	1.79 (0.76–4.24)	-
No	-	-	-	-
None
Yes	0.86 (0.41–1.79)	-	0.60 (0.26–1.39)	-
No	-	-	-	-
Sexual Risk Behaviorf
Different Sex Partners
Two or More	0.83 (0.50–1.36)	-	0.84 (0.49–1.43)	-
One	-	-	-	-
Oral Sex Partnersg
One	1.01 (0.52–1.94)	-	0.88 (0.43–1.81)	-
Two or more	1.23 (0.65–2.33)	-	1.38 (0.69–2.76)	-
None	-	-	-	-
Unprotected Anal Sex Partners
One	2.23 (1.12–4.44)h	2.45 (1.21–4.96)h	1.68 (0.80–3.51)	1.80 (0.85–3.82)
Two or More	1.16 (0.33–4.16)	1.54 (0.41–5.69)	0.78 (0.18–3.31)	1.06 (0.24–4.71)
None	-	-	-	-
Unprotected Vaginal Sex Partners
One	1.31 (0.61–2.82)	-	1.32 (0.57–3.06)	-
Two or More	1.33 (0.61–2.89)	-	1.41 (0.60–3.29)	-
None	-	-	-	-
Sexual Activity with Other Women
Yes	0.45 (0.26–0.78)	0.38 (0.22–0.68)	0.45 (0.24–0.82)	0.43 (0.23–0.81)
No	-	-	-	-
New Sex Partners
Yes	0.69 (0.46–1.03)	0.71 (0.47–1.09)	0.76 (0.49–1.18)	0.76 (0.48–1.20)
No	-	-	-	-
Treatment Arm
Yes	0.81 (0.54–1.22)	0.79 (0.52–1.20)	1.91(1.23–2.98)	1.92 (1.22–3.02)
No	-	-	-	-

HS- High school

^aThe model was adjusted for age, last BV treatment, douching frequency, use of Depo-Provera shot, use of IUDs, women who have sex with women in the past year, unprotected anal sex in the past year, new sex partners in the past year and treatment.

^bAntibiotic use in the past 30 days

^cNo prior and unsure of antibiotic use in the past 30 days

^dNever diagnosed and unsure if diagnosed

^eNever treated and unsure of treatment

^fIn the past year

^gReceptive oral sex

^hp value >0.05

Latent Class Description

BIC, aBIC, AIC and entropy scores were used to compare a five-class (AIC-99.79, BIC-261.45, aBIC-153.47, Entropy-0.63), four-class (AIC-97.05, BIC-225.43, aBIC- 139.68, Entropy-0.58), three-class (AIC-102.90, BIC-197.99, aBIC-134.47, Entropy-0.61), and two-class model (AIC-120.17, BIC-181.98, aBIC-140.69, Entropy-0.65). A three-class model was selected based on its optimal fit for the data. The persistent class was characterized by high probability (>77%) of being positive for BV (NS:7–10) at each time point (months two, four, six, eight, ten and twelve). This class had the highest probability of BV at every timepoint, with the highest probability (87%) at month ten. Women in the recurrent class alternated between decreasing (as low as 13%) and increasing (as high as 31%) probabilities of having BV at each timepoint. The probability of BV in this class decreased between months two and six and then steadily increased until month twelve. Finally, women in the clearance class had a consistently decreasing probability of BV from 72% to 8% from months two to ten and increased to 19% at month twelve. This class had the lowest probability (8%; month ten) of BV of all three classes.

Factors associated with Latent Class Membership

There were statistically significant differences in demographic and behavioral characteristics in women classed as recurrent vs those classed as clearance. Women >21 years old (vs. those ≤21 years) were significantly more likely to be in the persistent class (adjOR:1.90; 95% CI:1.23–2.95) compared to the clearance class and recurrent class (adjOR: 1.88; 95% CI: 1.17–3.00) compared to the clearance class. Participants who douched at least once (vs. those who never douched) had significantly lower odds of being in the recurrent class compared to clearance class (adjOR: 0.55; 95% CI: 0.35–0.87). Women who had experienced a prior episode of BV and received treatment seven to twelve months ago (vs. women who were not treated for prior episode of BV) had significantly lower odds of being in the persistent class compared to clearance class (adjOR: 0.34; 95% CI: 0.18–0.63). WSW (vs. women who did not have sex with women) had significantly lower odds of being in the recurrent class (adjOR: 0.43; 95% CI: 0.23–0.81) compared to clearance class and being in persistent class (adjOR:0.38; 95% CI:0.22–0.68) compared to the clearance class. Those who were randomized to the treatment arm (vs. control arm) were significantly more likely to be in the recurrent class versus the clearance class (adjOR: 1.92; 95% CI: 1.22–3.02).

Discussion

Our findings illustrate distinct patterns of multiple cases of BV over a twelve-month period. Through RMLCA, we were able to elucidate three distinct patterns; persistence, recurrence and clearance. Of these three groups, two groups (persistence and recurrence) were at significantly increased risk of multiple episodes of BV. Our evidence supports the hypothesis that older age and assignment to the treatment arm increase the risk of being classed as recurrent versus clearance while prior douching and WSW decreased the risk of being classed as recurrent versus clearance. We also determined that older age increases the risk of being classed as persistent while WSW and BV treatment 7–12 months ago decrease the risk of being classes as persistent versus clearance. Rates of BV recurrence have been evaluated in the literature; however, this is the first study to characterize BV patterns over a twelve-month period using RMLCA. The proportion of women assigned to the treatment arm classified as persistent was high and highlights the poor efficacy of oral metronidazole to reduce the long-term occurrence of multiple cases of BV.

Although women who were assigned to the treatment arm were less likely to be classed as persistent, this observation was not statistically significant (adjOR: 0.79; 95% CI: 0.52–1.20). Conversely, women who were assigned to the treatment arm were significantly more likely to be classed in the recurrent compared to clearance class (adjOR: 1.92; 95 % CI:1.22–3.02). Standard BV treatment includes a course of 500mg of oral metronidazole for seven days for symptomatic cases.⁷ Standard BV treatment was administered to all participants who tested positive for BV in the treatment arm; the follow-up questionnaire did not fully assess symptoms after baseline. It should be noted that clinical guidelines do not recommend the treatment of asymptomatic BV.²³ Furthermore, literature also does not support standard BV treatment for recurrent BV.⁷ If it is assumed that the mechanism of asymptomatic and symptomatic BV are the same, the increased likelihood of multiple episodes of BV may have been caused by ineffective treatment of those randomized to the treatment arm.⁷ Faught et al. present alternative treatment methods for recurrent BV such as metronidazole 750 mg/miconazole 200 mg for five nights each month for twelve months or metronidazole vaginal gel (0.75%) for three to six months.⁷ Furthermore, the mechanism for recurrent BV is unknown. It is hypothesized that recurrent BV may be caused by residual infection due to the persistence of Gardnerella vaginalis, which is resistant to metronidazole treatment.⁷ Women who received BV treatment 7–12 months ago were less likely to be classes as persistent (adjOR: 0.34; 95% CI: 0.18–0.63). This finding is unclear and may be due to the categories used to assess this variable or the distribution observed among the categories. This finding may be an artifact of the analysis.

Though BV is not considered an STI, as there is no known pathogen, it is associated with sexual activity.^3,7 Specifically, irregular condom use, multiple sex partners and having a consistent sex partner are associated with recurrent BV.^3,7 We determined that the only significant associations between sexual activity and class membership occurred among WSW. WSW were less likely to be classed as persistent (adjOR: 0.38; 95% CI: 0.22–0.68) or recurrent class (adjOR: 0.43; 95% CI: 0.23–0.81) than to clearance class. Our findings are similar to those of Muzny et al. who determined that WSW were less likely to be diagnosed with BV (adjOR: 0.08; 95% CI: 0.01–0.74).²⁴ Contrary to these findings, Falconi-McCahill et al. present data which concludes that there is an increased risk of BV among WSW.³ The discrepancies observed may be due to varying practices among WSW such as vaginal lubricant use, oral-anal sex and improper cleaning of vaginal sex toys.²⁴ Women engaging in unprotected anal sex with one partner were more likely to be classed as persistent (adjOR: 2.45; 95% CI: 1.21–4.98) versus clearance class compared to women with no anal sex partners. This is consistent with literature which concludes that unprotected anal sex increases the risk of BV.²⁵ Though the CI did not contain one and thus would indicate significance, the associated p-value corresponding to women engaging in anal sex with one partner and persistent class membership was not significant (p=0.15). Though women who had one unprotected anal sex partner were more than twice as likely to be in the persistent class compared to the clearance class, women with two or more partners did not have a significant association with class membership. Of note, the association between the anal sex variable and latent class membership did not yield a significant type III p-value (p=0.12). This may suggest that the overall effect of anal sex partners may not be significant, however, this does not negate the observed association between women who had one anal sex partner and class membership.

Intravaginal practices such as douching are also associated with the risk of BV.^8,26 However, the effects of douching vary.^8,10–12,27 Our results show that women who douched at least once were less likely to be classed as recurrent versus clearance (adjOR: 0.55; 95% CI: 0.35–0.87) as compared to women who had never douched. This association may be the result of the postulated beneficial effects of vaginal douching such as removal of semen after sex, and alleviation of vaginal irritation.^8,10 The removal of semen decreases the load of sexually transmitted pathogens.¹⁰ Furthermore, the temporal association between BV and vaginal douching is also disputed.¹¹ It is possible that women douche to alleviate the symptoms of BV as well as that vaginal douching causes BV.¹¹ Hutchinson et al. determined through prospective longitudinal analyses that douching was not associated with BV,¹¹ however, cross-sectional studies have determined that douching was associated with BV.^11,13 Our analysis did not assess the bidirectional nature of the association between BV and vaginal douching since the exposures were only assessed at baseline. There were no statistical differences in terms of douching between the persistent and clearance class. This may suggest that douching may only be beneficial for women with recurrent BV. There may be inherent differences in the vaginal microbiota of women classed as recurrent versus persistent which could account for this observation. Additionally, behavioral practices among women classed as persistent versus those classed as clearance may differ. Older women (>21 years) were more likely to be members of the persistent and recurrent classes compared to the clearance class, indicating that older women were more likely to experience multiple episodes of BV over a twelve-month period. BV is the most common vaginal dysbiosis among women of reproductive age.¹³ Our findings are consistent with those of Ranjit et al. who determined that BV was least prevalent among women <21 years.¹³

Though our study was novel in its approach, it was also subject to some limitations. There was no test of cure for BV done by the RCT from which data for this study was acquired. Therefore, it is possible that participants did not take the medication as prescribed. Furthermore, the treatment administered has been shown in literature to be ineffective. Additionally, all responses received on the baseline questionnaire were self-reported and subject to recall and social desirability bias. Consequently, sexual risk behavior may have been underreported. The analysis was further limited by the lack of inclusion of time varying predictors though the outcome is based on a twelve-month assessment. Our study only included women who were asymptomatic for BV at baseline. The presence or absence of BV symptoms was not assessed after baseline and therefore could not be accounted for in the analysis. As a result, the duration of asymptomatic BV was unknown. Similarly, douching practices were not assessed after baseline and it was assumed that behaviors assessed at baseline may persist during the follow up period. The frequency of douching prior to enrollment only was assessed. Additionally, the reasons for douching among the young women in this sample were not assessed, and despite the longitudinal design, temporality could be an issue. There is also limited generalizability of results since only African American women were included in the analysis. Our analysis also failed to compare persistent and recurrent classes. This is a potential area of exploration for future research. Additionally, Latent Transition Analysis could be explored to assess the risk factors associated with transitions between BV states at each time point and include time varying predictors. BV was assessed by gram stain only, which provided a consistency across all participants, but could not identify whether specific bacteria were involved in differentiating recurrent vs. cleared BV. Future studies might use 16S sequencing to better understand the longitudinal changes in specific vaginal microbiota that influence BV recurrence and response to metronidazole treatment. Furthermore, LCA can be used to create classes which combine epidemiologic and microbiome data to comprehensively assess the multiple predictors of repeated cases of BV. Considering the lack of knowledge of the mechanisms of recurrent and persistent BV, novel analytical approaches are needed to examine these phenomena. Examination of epidemiologic and microbiome data in silo is not sufficient.

Here we present novel findings that classify women based on BV persistence, recurrence and clearance over a twelve-month period. Our findings illustrate high incident cases of BV. Women aged 21 years and over, and women with one anal sex partner were significantly more likely to be in the persistent class compared to the clearance class. Women who were treated for BV 7–12 months ago and women who engaged in sexual activity with other women were less likely to be classed as persistent versus clearance. Women aged 21 years and over and women who received metronidazole treatment were significantly more likely to be in the recurrent class compared to the clearance class. Women who douched at least once and women who engaged in sexual activity with other women in the past year were less likely to be in the recurrent versus clearance class. Despite adhering to the recommended BV treatment guidelines, there was a high proportion of persistent and recurrent cases of BV according to RMLCA. Efforts should be made to evaluate alternative treatment for recurrent and persistent cases of asymptomatic BV.

Table 1:

Participant Characteristics stratified by Latent Class

Characteristics	Categories		Latent Classes (N=858)			p valuea
	Levels	Overall	Persistent (N=480; 55.9%)	Recurrent (N=262, 30.5%)	Clearance (N=116; 13.5%)
Demographic Variables
Age, n (%)						0.03
	≤21 years	355 (41.4%)	184 (38.3%)	111 (42.4%)	60 (51.7%)
	>21 years	503 (58.6%)	296 (61.7%)	151 (57.6%)	56 (48.3%)
Education, n (%)						0.48
	No HS	151 (17.6%)	85 (17.7%)	46 (17.6%)	20 (17.2%)
	HS	292 (34.0%)	173 (36.0%)	78 (29.8%)	41 (35.3%)
	More than HS	415 (48.4%)	222 (46.3%)	138 (52.7%)	55 (47.4%)
Ethnicity, n (%)						0.45
	Hispanic/Latino	33 (3.8%)	15 (3.1%)	12 (4.6%)	6 (5.2%)
	Not Hispanic/Latino	825 (96.2%)	465 (96.9%)	250 (95.4%)	110 (94.8%)
Medical History
Prior Antibiotic Useb, n (%)						0.96
	Yes	45 (5.2%)	26 (5.4%)	13 (5.0%)	6 (5.2%)
	Noc	813 (94.8%)	454 (94.6%)	249 (95.0%)	110 (94.8%)
Prior Episodes of BV, n (%)						0.25
	Once	176 (20.5%)	93 (19.4%)	57 (21.8%)	26 (22.4%)
	2–4 times	161 (18.8%)	83 (17.3%)	51 (19.5%)	27 (23.3%)
	5 or more times	56 (6.5%)	26 (5.4%)	22 (8.4%)	8 (6.9%)
	Neverd	465 (54.2%)	278 (57.9%)	132 (50.4%)	55 (47.4%)
Last BV treatment, n (%)						0.04
	≤6 months ago	154 (18.0%)	83 (17.3%)	50 (19.1%)	21 (18.1%)
	7–12 months ago	92 (10.7%)	38 (7.9%)	34 (13.0%)	20 (17.2%)
	≥12 months ago	148 (17.3%)	81 (16.9%)	46 (17.6%)	21 (18.1%)
	Nevere	464 (54.1%)	278 (57.9%)	132 (50.4%)	54 (46.6%)
Prior Pregnancy, n (%)						0.05
	Yes	493 (57.5%)	292 (60.8%)	135 (51.5%)	66 (56.9%)
	No	365 (42.5%)	188 (39.2%)	127 (48.4%)	50 (43.1%)
Douching History, n (%)						0.02
	At least once	432 (50.3%)	254 (52.9%)	113 (43.1%)	65 (56.0%)
	Never	426 (49.7%)	226 (47.1%)	149 (56.9%)	51 (44.0%)
Birth Control Methods
Pills/Patch, n (%)						0.98
	Yes	168 (19.6%)	95 (19.8%)	51 (19.5%)	22 (19.0%)
	No	690 (80.4%)	385 (80.2%)	211 (80.5%)	94 (81.0%)
Nuva Ring, n (%)						0.33
	Yes	36 (4.2%)	17 (3.5%)	15 (5.7%)	4 (3.5%)
	No	822 (95.8%)	463 (96.5%)	247 (94.3%)	112 (96.5%)
Condoms, n (%)						0.69
	Yes	666 (77.6%)	377 (78.5%)	202 (77.1%)	87 (75.0%)
	No	192 (22.4%)	103 (21.5%)	60 (22.9%)	29 (25.0%)
Depo-Provera Shot, n (%)						0.04
	Yes	126 (14.7%)	58 (12.1%)	49 (18.7%)	19 (16.4%)
	No	732 (85.3%)	422 (87.9%)	213 (81.3%)	97 (83.6%)
Intra-Uterine Device, n (%)						0.34
	Yes	84 (9.8%)	50 (10.4%)	27 (10.3%)	7 (6.0%)
	No	774 (90.2%)	430 (89.6%)	235 (89.7%)	109 (94.0%)
Spermicide, n (%)						0.42
	Yes	10 (1.2%)	7 (1.5%)	3 (1.2%)	0 (0.0%)
	No	848 (98.8%)	473 (98.5%)	259 (98.8%)	116 (100.0%)
No Birth Control, n (%)						0.42
	Yes	60 (7.0%)	36 (7.5%)	14 (5.3%)	10 (8.6%)
	No	798 (93.0%)	444 (92.5%)	248 (94.7%)	106 (91.4%)
Sexual Risk Behaviorf
Different Sex Partners, n (%)						0.75
	One	201 (23.4%)	115 (24.0%)	62 (23.7%)	24 (20.7%)
	Two or more	657 (76.6%)	365 (76.0%)	200 (76.3%)	92 (79.3%)
Oral Sex Partnersg, n (%)						0.39
	None	101 (11.8%)	56 (11.7%)	30 (11.5%)	15 (12.9%)
	One	307 (35.8%)	177 (36.9%)	83 (31.7%)	47 (40.5%)
	Two or more	450 (52.4%)	247 (51.5%)	149 (56.9%)	54 (46.6%)
Unprotected Anal Sex Partners, n (%)						0.16
	None	708 (82.5%)	384 (80.0%)	221 (84.4%)	103 (88.8%)
	One	129 (15.0%)	83 (17.3%)	36 (13.7%)	10 (8.6%)
	Two or More	21 (2.4%)	13 (2.7%)	5 (1.9%)	3 (2.6%)
Unprotected Vaginal Sex Partners, n (%)						0.95
	None	59 (6.9%)	32 (6.7%)	17 (6.5%)	10 (8.6%)
	One	424 (49.4%)	239 (49.8%)	128 (48.9%)	57 (49.1%)
	Two or more	375 (43.7%)	209 (43.5%)	117 (44.7%)	49 (42.2%)
Sexual Activity with Other Women, n (%)						0.01
	Yes	97 (11.3%)	48 (10.0%)	26 (9.9%)	23 (19.8%)
	No	761 (88.7%)	432 (90.0%)	236 (90.1%)	93 (80.2%)
New Sex Partners, n (%)						0.19
	Yes	418 (48.7%)	224 (46.7%)	129 (49.2%)	65 (56.0%)
	No	440 (51.3%)	256 (53.3%)	133 (50.8%)	51 (44.0%)
Treatment Arm, n (%)						<0.01
	Yes	438 (51.0%)	211 (44.0%)	170 (64.9%)	57 (49.1%)
	No	420 (49.0%)	269 (56.0%)	92 (35.1%)	59 (50.9%)

HS- High school

BV- Bacterial vaginosis

^ap values compare latent classes

^bAntibiotic use in the past 30 days

^cNo prior and unsure of antibiotic use in the past 30 days

^dNever diagnosed and unsure if diagnosed

^eNever treated and unsure of treatment

^fIn the past year

^gReceptive oral sex