Search and selection
Altogether 10,634 studies were identified by our search key (Fig. 1). A total of 9762 studies were checked by title and abstract, and 315 by full text. After excluding overlapping populations and conference abstracts, 13 studies were eligible for systematic review and meta-analysis10,12,19,20,27,28,29,30,31,32,33,34.
Baseline characteristics of included studies
Baseline characteristics of the enrolled studies are detailed in Table 1. Of the 13 included articles, nine were published from Europe, three from Asia, and one from North America. The screening methods used were different in the studies. The Nugent score evaluated the results in studies where Gram stain was used. In addition, the screening procedures’ timing and frequency among the studies were not uniform. In the intervention group, the preterm birth rate before 37 weeks ranged from 2.96 to 21.8%, while in the control group, between 5.1 and 22.3%. In most cases, oral or local Lactobacillus, Clindamycin, or Metronidazole was used to treat abnormal vaginal flora.
Two studies included multiple pregnancies19,20. However, in these studies, the author reported the same rate of multiple pregnancies in the intervention and control groups. Furthermore, we performed a leave-one-out (sensitivity) analysis to investigate whether these studies influenced our result in a significant way, and we found no differences (eFigs. 9–14). Further details are summarized in Tables S2 and S3.
Preterm birth before 37 weeks and birthweight under 2500 g
When pooling different types of screening, regular screening of abnormal vaginal flora significantly reduces the odds of preterm birth before 37 weeks compared to control (8.98% vs. 9.42%; OR 0.71, CI 0.57–0.87, Fig. 2). However, the analysis shows substantial heterogeneity (I2 = 97%, p < 0.001).
Forest plot representing the odds of preterm birth before 37 weeks.
Of different screening methods, Gram stain compared to control significantly reduces the odds of preterm birth before 37 weeks (10.68% vs. 17.38%; OR 0.61, CI 0.45–0.83; substantial heterogeneity). pH measurement based on two studies resulted in a similar effect (7.2% vs. 10.37%; OR 0.67, CI 0.52–0.87).
The subgroup analysis of RCTs, including all screening types, showed a tendency for lower preterm birth (OR 0.74, CI 0.53–1.03 eFig. 1), however, the difference was non-significant. The subgroup analysis of RCTs, including only Gram stain, showed similar results (OR 0.79, CI 0.54–1.16, eFig. 2).
On the other hand, screening of abnormal vaginal flora significantly reduces the odds of birthweight under 2500 g compared to control (6.53% vs. 7.24%; OR 0.64, CI 0.50–0.81, eFig. 11). However, the analysis shows substantial heterogeneity (I2 = 97%, p < 0.001).
Analyzing different types of screenings, Gram stain compared to control significantly reduces the odds of birthweight under 2500 g (7.94% vs. 14.79%; OR 0.55, CI 0.41–0.73; eFig. 11). Based on two studies, pH self-screening and based on one study, combined pH and Gram stain screening did not significantly differ from the control group.
The subgroup analysis of RCTs, including all screening types, showed statistically significant results for below 2500 g birthweight rate (OR 0.71, CI 0.54–0.93 eFig. 12). The subgroup analysis of RCTs, including only Gram stain, showed similar data (OR 0.71, CI 0.50–1.02 eFig. 13). However, the results were not significant.
Preterm birth before 34 weeks and birthweight under 2000 g
Regular screening of abnormal vaginal flora reduces the odds of preterm birth before 34 weeks (3.88% vs. 4.64%; OR 0.58, CI 0.31–1.08, eFig. 7) and birthweight under 2000 g (1.94% vs. 2.81%; OR 0.49, CI 0.31–0.75, eFig. 17) compared to control. However, both analyses show substantial heterogeneity.
Based on four studies analyzing the different types of screenings, Gram stain compared to no screening significantly reduces the odds of birthweight under 2000 g (2.1% vs. 7.7%; OR 0.33, CI 0.23–0.46; substantial heterogeneity).
Preterm birth before 32 weeks and birthweight under 1500 g
Screening of abnormal vaginal flora significantly reduces the odds of preterm birth before 32 weeks of gestation compared to control (1.35% vs. 2.03%; OR 0.51, CI 0.31–0.85, Fig. 3). Analyzing the different types of screenings, based on three studies, pH self-screening compared to control reduces the odds of preterm birth (0.93% vs. 1.02%; OR 0.77, CI 0.61–0.98; substantial heterogeneity). Based on one study, Gram stain and pH measurement showed a similar result.
Forest plot representing the odds of preterm birth before 32 weeks.
Similarly, regular screening of abnormal vaginal flora significantly reduces the odds of birthweight under 1500 g compared to control (1.17% vs. 1.78%; OR 0.43, CI 0.25–0.75 Fig. 4A). Analyzing the different types of screenings, Gram stain based on three studies reduces the odds of birthweight under 1500 g (1.91% vs. 7.58%; OR 0.23, CI 0.20–0.27; substantial heterogeneity). Based on two studies, pH self-screening compared to control reduces the odds of birthweight under 1500 g (0.83% vs. 1.00%; OR 0.81, CI 0.70–0.94; substantial heterogeneity).
Forest plots representing the odds of bithweight under 1500 g (A) and 1000 g (B).
Birthweight under 1000 g
Regular screening of abnormal vaginal flora significantly reduces the odds of birthweight under 1000 g compared to no screening (0.86% vs. 2.2%; OR 0.33, CI 0.19–0.57, Fig. 4B). However, the analysis shows substantial heterogeneity (I2 = 88%, p < 0.001).
Analyzing the different types of screenings, Gram stain based on three studies significantly reduces the odds of birthweight under 1000 g (1.21% vs. 5.27%; OR 0.22, CI 0.18–0.26; substantial heterogeneity). Based on one study, pH self-screening showed a similar effect.
Publication bias and heterogeneity
Outlier and influence analyses did not show a significant difference for any of the above mentioned results.
Egger’s test could only be performed in the case of preterm delivery before 37 weeks. For this, Eggers’ regression test shows no publication bias (t = − 0.45; df = 11; p = 0.661).
Risk of bias assessment and level of evidence
The risk of bias assessment results is summarized in Tables S4 and S5. The overall risk of bias shows some concerns. On the other hand, the quality of evidence was very-low for most results. A summary of the assessment is included in Table S6.