Screening for and Treatment of Bacterial Vaginosis Reduced Preterm Delivery in High-Risk Pregnant Women: A Systematic Review and Meta-Analysis

Bacterial vaginosis (BV) and its milder form, abnormal vaginal flora (AVF), are polymicrobial infections in which the normal lactobacilli-containing flora is replaced by Gardnerella vaginalis and anaerobic organisms [1]. It is the most common bacterial infection and affects approximately 30% of women worldwide [2], with the same prevalence among pregnant women [3]. Risk factors for the development of BV include smoking, sexual activity, vaginal douching, and ethnicity [4]. Symptoms, affecting 50% of women, include increased vaginal discharge and unpleasant smell [5]. While BV and AVF may be clinically suspected in physical examination, a definitive diagnosis requires laboratory evaluation.

BV during pregnancy was shown to be associated with obstetric complications, such as spontaneous abortions including late miscarriages, premature rupture of membranes, preterm labor and delivery, chorioamnionitis, and post-cesarean endometritis [6‒8]. Antibiotics are usually the treatment of choice. Although studies performed on pregnant women who were at low risk for preterm deliveries did not show any benefit in screening for and treating BV to prevent preterm delivery, studies performed on women who were at high risk for preterm birth, demonstrated conflicting results [3‒5, 9‒15]. While 4 [11‒14] studies demonstrated reduced risk for preterm delivery following BV treatment, 3 studies did not [3, 10, 16]. The systematic review and meta-analysis of randomized control trials presented below was conducted to determine whether screening for and treatment for BV effectively reduced the rate of preterm deliveries in high-risk populations.

Embase, PubMed, Ovid-Medline, and Web of Science were searched using the following keywords: “obstetric labor, premature,” “infant, premature,” “fetal membranes, premature rupture,” “pregnancy outcome,” “fetal death,” “stillbirth,” “preterm,” “premature,”“fetal loss,” “ abortion,” “prematurity,” “miscarriage,” “vaginosis, bacterial,” “reproductive tract infections,” “vaginitis,” “colpitis,” “vagina-flora,” “genital-tract-infection.” The search was restricted to English-language journals and full articles (no abstracts). All reference lists from the main reports and relevant reviews were hand-searched for additional eligible publications. In addition, when clarifications were needed, authors of the included studies were contacted. The search was until December 26, 2019.

Manuscripts were included if they described a randomized controlled trial, which compared antibiotic treatment versus no or placebo treatment for BV/AVF during pregnancy in women at high risk for preterm delivery. In studies that did not include only a high-risk population, we considered only the data of the high-risk patients. If a study included a subgroup of women with increased risk for preterm delivery, the subgroup was included as long as data regarding the rate for preterm delivery or late miscarriages was reported for this group in the manuscript. Only trials in which BV/AVF was diagnosed according to laboratory methods were included. The most common method for BV diagnosis is the Nugent criteria, a standardized method of Gram stain interpretation of the vaginal flora. Vaginal swab smears are graded on a 10-point scale based on the presence or absence of Lactobacillus morphotypes, gram-variable and gram-negative rods, and curved gram-negative rods. A score of 0–3 is representative of normal flora, a score of 4–6 is representative of AVF, and a score of 7–10 is consistent with BV flora. Additional methods were vaginal pH > 4.5, the presence of clue cells and presence of an amine odor (trimethylamine) when 10% potassium hydroxide was applied to vaginal secretions [5, 17, 18].

The credentials of the investigators are indicated in the authors’ list. Two independent reviewers (DM and AA) checked each full-text report for eligibility and extracted and tabulated all relevant data. There was no disagreement between the reviewers. All procedures conformed to the guidelines for systematic review and meta-analysis of randomized controlled trial in epidemiology – PRISMA checklist [19].

The primary outcome was the rate of preterm delivery before 37 weeks and/or late miscarriage (pregnancy loss ≥13 weeks). A sub-analyses was conducted to determine the rate of preterm delivery and/or miscarriage following BV treatment versus placebo according to the type of medication, multicenter versus single center and time of treatment initiation before versus after 20 gestational weeks. The 20 weeks was selected as the cutoff point, since this period represents the midpoint of the second trimester and serves as the delineation between miscarriage and delivery.

A pooled relative risk (RR) was calculated for the study outcomes. All reports were assigned a quality scores based on the CONSORT checklist [20]. The maximum score was 25. The quality of the body of evidence for the outcome (preterm delivery and/or miscarriage) was assessed according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system [21]. The body of evidence was assessed based on the following considerations: risk of bias, inconsistency, indirectness, imprecision, and publication bias. The overall rating reflects our certainty in the evidence.

Heterogeneity of the studies was tested using Cochrane’s Q test of heterogeneity (p < 0.1 was considered statistically significant). Inconsistency in study results was assessed by I². Random-effects model (DerSimonian and Laird) was chosen if Cochrane’s Q test p < 0.1 or I² ≥ 50%. Otherwise, the fixed-effects model (inverse variance methods) was chosen. The funnel plot and the Egger test were used to assess publication bias (p < 0.05 defined a statistically asymmetric funnel plot).

The statistical analysis and graphical presentation were performed using StatsDirect software, version 2.7.2 (StatsDirect Ltd.). Meta-analyses and review articles are exempt from IRB approval in our institutions.

The study selection process is shown in Figure 1. Of the 4,701 abstracts identified, 4,694 were excluded due to irrelevance. After a full review, 7 studies were deemed eligible according to the inclusion criteria [3, 10‒14, 16].

Study summaries are presented in Tables 1, 2. Among the participating women, 738 were at high risk for preterm delivery and included in the analysis. Overall, 397 women received antibiotic and 341 women were in control groups, which, in all cases, included placebo treatment. In five studies, metronidazole with or without erythromycin were prescribed to manage BV [3, 11‒13, 16], and in two studies, clindamycin was used to treat BV [10, 14]. Dosing was heterogonous (Table 1). No treatment was given to the partners. In six studies, the risk for preterm delivery was history of preterm delivery [3, 10‒14], and in one study, the risk factor was positive fibronectin test [16]. Quality score of the studies ranged from 14.5 to 20.5. Women with cervical cerclage were excluded in 5 studies [3, 12‒14, 16], and in 2 studies, cerclage was not specifically addressed, but no cases of cerclage were described in the background characteristics [10, 11]. Vaginal progesterone was assessed only in 2003 [22] and therefore was not relevant to the time frame of the analyzed studies.

Bias assessment plot (Fig. 2) and Egger test (p = 0.41) suggested a low probability for publication bias.

One study, in which clindamycin 2% vaginal cream was used (N = 16), was excluded from the analysis since no group had preterm deliveries [10]. The other studies used oral antibiotics (Table 1). BV treatment in high-risk women reduced the rate of preterm deliveries (pooled RR with 95% CI, 0.65 [0.44–0.98]; GRADE: moderate; Figure 3).

Sub-analyses for studies with previous preterm delivery as the risk factor, type of medication, multicenter versus single center, and time of treatment initiation are presented in Table 3. A statistically significant reduction in risk for preterm delivery following treatment for BV was demonstrated in studies with previous preterm delivery, in one study that used clindamycin, oral administration (6/7 studies), in single-center studies, and when treatment was started after 20 gestational weeks. In the other parameters, there was a positive trend toward a protective effect of treatment for BV against preterm delivery, which did not reach statistical significance (Table 3).

The present meta-analysis found that treatment for BV in pregnant women reduced the risk for preterm delivery in high-risk populations. This favorable effect was more apparent in studies that used clindamycin, in single-center studies, and when the treatment was initiated after 20 gestational weeks.

728 women were included in the analysis. Assuming that the rate of preterm delivery in the control group represented the baseline risk of this population, the sample size was sufficient to detect a RR of 0.67 between the groups (from 30% to 20%) with a power of 88%, 5% 2-sided alpha.

The role of vaginal microbiota is particularly important during pregnancy because vaginal dismicrobism is one of the most important mechanisms associated with preterm birth. The immunomodulation properties of vaginal microbiota may be based on the modification of bacterial colonies; in particular, the family of Lactobacilli seems to play a key role in this process [20]. BV is a known risk factor for preterm delivery [23]. It was suggested that BV is associated with an increased vaginal concentration of cytokines and that this inflammatory response may trigger early parturition [24].

Previous meta-analyses found that treatment of BV with clindamycin or metronidazole did not reduce the risk of preterm delivery [25, 26]. Yet, these meta-analyses were performed on unselected population and not specifically in women at high risk for preterm delivery. One meta-analysis [25] assessed 9 randomized controlled trials (N = 10,923). Clindamycin and metronidazole were administered in 7 and 2 studies, respectively. Seven studies were placebo controlled. No reduction in the incidence of preterm delivery was found for either clindamycin (OR, 1.01; 95% CI, 0.75–1.36) or metronidazole (OR, 0.94; 95% CI, 0.71–1.25). In the second meta-analysis [26], 13 randomized controlled trials were assessed. No significant association was observed between treatment and spontaneous delivery before 37 weeks (pooled absolute risk difference −1.44% [95% CI, −3.31%–0.43%]; 8 RCTs, N = 7,571) or any delivery before 37 weeks (pooled absolute risk difference, 0.20% [95% CI, −1.13%–1.53%]; 6 RCTs, N = 6,307).

The first randomized controlled trial that assessed women at high risk for preterm delivery was conducted by Morales et al. [12]. Eighty women with a history of spontaneous preterm delivery received metronidazole or placebo to treat BV. Compared with the placebo group, women in the metronidazole group had significantly fewer preterm births (18% versus 39%), hospital admissions for preterm labor (27% versus 78%), births of infants weighing less than 2,500 grams (14% versus 33%), and premature rupture of membranes (5% versus 33%). Later on, other studies addressed women at high risk for preterm delivery either as the whole study population [10, 11, 16] or as a subgroup of the entire cohort [3, 13, 14]. Based on the previous meta-analyses [25, 26] and the current one, treatment for BV may be ineffective in pregnant women at low risk for preterm delivery; however, in a high-risk population, treatment for BV reduces the risk for preterm delivery. Clindamycin to treat BV demonstrated a lower risk for preterm delivery compared to placebo. We suggest to give oral clindamycin 300 mg twice daily for 5–7 days based on the American Centers for Disease Control and Prevention (CDC) recommendations and the study of Ugwumadu et al. [14] that has shown a significant effect [14, 27]. Studies that used metronidazole also demonstrated a trend toward reduced rates of preterm delivery, which did not reach statistical significance, perhaps due to a smaller sample size. Alternatively, the high treatment failure rate reported for metronidazole (45%) [28] may explain its lower impact on preterm delivery rates. As both clindamycin and metronidazole are considered to be effective treatments for BV, more studies are required to examine whether clindamycin is superior to metronidazole in reducing the risk for preterm delivery.

The risk for preterm delivery was lower in BV patients treated in single-center studies but not among those treated in multicenter studies. These observations may be the result of the smaller sample size, and differences between medical centers in the study population, laboratory evaluation of BV, patient compliance, and management of preterm labor.

The effect of BV treatment in preventing preterm delivery was greater when treatment was initiated after 20 gestational weeks. Considering the fact that BV has a high recurrence rate [29], this finding may have been a function of the shorter period during which BV recurrence could have occurred. Rescreening and treatment for BV was shown to be beneficial for BV eradication [30]. It should be noted, however, that the information regarding the proportion of women recruited before 20 gestational weeks is not available, and additional studies are required in order to evaluate the preferable time for BV screening. Since BV may lead to late miscarriages and can recur following eradication, it is reasonable to screen for and treat BV in the early second trimester and for positive women to repeat screening after the 20th gestational weeks. Yet, further studies are needed to determine the most appropriate schedule for BV screening and treatment in women at risk for preterm delivery.

A possible alternative or supplement for antibiotic treatment against BV is a probiotic supplementation as shown in several reports. It can be related to a global anti-inflammatory effect on vaginal immunity, with potential implications in preventing preterm birth [31].

The strengths of this meta-analysis lay in its incorporation of several high-quality, randomized, controlled trials, and sufficient data for performing multiple sub-analyses in order to evaluate reasons for heterogeneity. Its limitations included the fact that most studies were not specifically designed to assess high-risk patients for preterm delivery, the incorporation of small-sample-size cohorts, and the inclusion of subgroups of women from the cited studies. In addition, the studies are highly heterogeneous with respect to study protocol, type of treatment, and time of recruitment. Data regarding ascending infection as an etiology for preterm delivery were lacking. Finally, in most studies, the risk for preterm delivery was a previous preterm delivery; therefore, no conclusions could be made with respect to the effectiveness of BV treatment in women with other risk factors for preterm delivery, such as short cervix or multiple gestations.

Antibiotic treatment for BV reduced the risk for preterm delivery in high-risk populations, particularly women with history of preterm delivery. Additional studies are required to confirm the results of this study, determine the optimal screening schedule, the type of treatment and to assess the impact of BV treatment on additional high-risk populations.

We would like to thank Mrs. Tal Kaminski-Rosenberg and Mrs. Leora Mauda, the Alfred Goldschmidt Medical Sciences Library, Rappaport Faculty of Medicine, Technion – Israel Institute of Technology, for their assistance in the literature search.

Statement of Ethics is not applicable because this study is based exclusively on published literature. Meta-analyses and review articles are exempt from IRB approval in our institution. Informed consent was not required.

No funding was provided to any research relevant to the study, and no funding source was involved in the study design, execution, analysis, manuscript conception, planning, writing, and decision to publish.

The authors have no conflicts of interest to declare.

E.Y., D.M., A.A., M.M., and Z.N. made substantial contributions to the concept and design of the study. E.Y. and Z.N. analyzed the data and drafted the article, and D.M. and A.A. provided critical revisions of the article for intellectual content. All authors approved the final manuscript version.

Trial registration: this study was registered on PROSPERO (CRD42020162621)

All relevant data generated or analyzed during this study are included in this published article. Further inquiries can be directed to the corresponding author.