The purpose of this study was to determine whether the administration of clindamycin to women with abnormal vaginal flora at <22 weeks of gestation reduces the risk of preterm birth and late miscarriage. We conducted a systematic review and metaanalysis of randomized controlled trials of the early administration of clindamycin to women with abnormal vaginal flora at <22 weeks of gestation. Five trials that comprised 2346 women were included. Clindamycin that was administered at <22 weeks of gestation was associated with a significantly reduced risk of preterm birth at <37 weeks of gestation and late miscarriage. There were no overall differences in the risk of preterm birth at <33 weeks of gestation, low birthweight, very low birthweight, admission to neonatal intensive care unit, stillbirth, peripartum infection, and adverse effects. Clindamycin in early pregnancy in women with abnormal vaginal flora reduces the risk of spontaneous preterm birth at <37 weeks of gestation and late miscarriage. There is evidence to justify further randomized controlled trials of clindamycin for the prevention of preterm birth. However, a deeper understanding of the vaginal microbiome, mucosal immunity, and the biology of BV will be needed to inform the design of such trials.
Spontaneous preterm labor and delivery is a syndrome caused by multiple pathologic processes that activate the common pathway of parturition. Mechanisms of disease involved in the “preterm labor syndrome” include infection/inflammation, vascular disease, uterine overdistension, abnormal allograft reaction (eg, rejection), an allergic-like phenomenon, a progesterone deficiency, and cervical disorders. The first mechanism of disease responsible for preterm labor and delivery for which a causal link was well-established is infection. Moreover, the mechanisms responsible for this process have been identified and involve pattern recognition receptors, chemokines, or inflammatory cytokines.
A conventional view is that most cases of intrauterine infection responsible for preterm labor and delivery result from ascending infection; therefore, multiple investigators have attempted to identify the patient at risk for preterm labor and delivery by assessing the microbiologic state of the lower genital tract. Several cohort studies have attempted to establish a relationship between a change in the lower genital tract flora and the risk of preterm delivery. Moreover, randomized clinical trials have been undertaken to test the effect of antibiotics in patients with bacterial vaginosis (BV), Trichomonas vaginalis , group B streptococci (GBS), Ureaplasma urealyticum , and Chlamydia trachomatis.
The initial expectation that assessing the lower genital tract for the presence of certain microorganisms or changes in the microbial flora (ie, BV) followed by treatment has not met initial hopes. We believe that the disappointing results are not due to any question about the importance of infection in the etiology of preterm labor and delivery (with intact or ruptured membranes), but rather the limitations of the experimental design of the randomized clinical trials implemented to test the effects of antimicrobial agents.
A fundamental principle of randomized clinical trials is that all patients included in a particular trial must potentially benefit from the intervention under study (in this case, antibiotics). Therefore, if the goal of antibiotic administration is to reduce preterm labor and delivery, this can only be accomplished by preventing infection-induced preterm labor and delivery. Consequently, randomized clinical trials designed to test hypotheses must identify patients at risk for infection, and such risks must be substantial in order for the trial to have a reasonable expectation of success.
Intraamniotic infection/inflammation can be readily identified by analysis of amniotic fluid. However, it would be convenient to identify these patients by examining changes in the microbial flora of the lower genital tract, because this would be relatively easy and non-invasive. BV is characterized by a change in the microbial ecosystem of the vagina. Its presence is associated with an increased risk for spontaneous preterm delivery and intraamniotic infection. Therefore, investigators reason that identification and treatment of BV would result in a decreased rate of preterm delivery. Many randomized clinical trials have been conducted to test this hypothesis, and the results have been largely negative; although, some trials have yielded positive results.
The contradictory results among trials of BV have been attributed to: (1) the definition of BV; (2) the gestational age at diagnosis and enrollment; (3) the choice of antimicrobial agent used, as well as the dose and route of administration; (4) whether antibiotic administration has been followed by a test of cure; (5) the primary outcome of the study (most have used delivery at <37 weeks rather than early preterm delivery, where most of the infection-related preterm births [PTBs] occur); (6) gene-environment interactions related to the inflammatory response; and (7) patient population, etc.
For antibiotics to be effective in reducing the rate of preterm delivery, several criteria must be met: (1) antimicrobials must be effective against the target organism or the clinical condition under study (eg, BV); (2) antimicrobials should be used only in women who can benefit because they are at substantial risk for infection or an infection-related condition; and (3) antimicrobials must be used early enough so that eradication of the microorganisms would be followed by resolution of any inflammatory response and its unintended consequences (eg, damage of the chorioamniotic membranes, microbial invasion of the amniotic cavity, fetal microbial invasion, and fetal inflammation).
The importance of the timing of antibiotic administration has recently become more apparent because there is evidence that exposure to either bacterial products (eg, endotoxin or lipopolysaccharide) or bacteria itself, which is not enough to signal the onset of preterm labor, may predispose to a subsequent viral infection, and this, in turn, leads to both preterm labor and fetal damage.
Several systematic reviews and metaanalyses have been conducted to determine the effect of antimicrobial agents for the prevention of PTB in women with BV. Such efforts need to be revisited to focus both on clindamycin and early treatment.
The purpose of this study was to conduct a systematic review and metaanalysis to determine whether treatment of patients with BV with clindamycin before 22 weeks of gestation can reduce the rate of spontaneous PTB. The justification for the study is that several trials have suggested a beneficial effect of this antibiotic in reducing the rate of PTB in patients at risk when administered early during pregnancy. Most studies aimed at reducing the rate of PTB have used metronidazole, and such studies have yielded negative results.
Materials and methods
The systematic review was conducted using a prospectively prepared protocol and reported with the use of the Preferred Reporting Items for Systematic Reviews and Meta-analyses.
Literature search
A computerized search was performed with PubMed, Embase, Cinahl, and Lilacs (all from inception of database to July 31, 2011), ISI Web of science ( http://www.isiknowledge.com.easyaccess1.lib.cuhk.edu.hk ; 1960 to July 31, 2011), the Cochrane Central Register of Controlled Trials ( http://www.mrw.interscience.wiley.com.easyaccess1.lib.cuhk.edu.hk/cochrane/cochrane_clcentral_articles_fs.html ; 1960 to July 31, 2011), and Research Registries of ongoing trials ( www.clinicaltrials.gov , www.controlled-trials.com , www.centerwatch.com , www.actr.org.au , www.nrr.nhs.uk , and www.umin.ac.jp/ctr ) with a combination of key and text words that are related to antibiotics, BV, prevention, or PTB. Proceedings of the Society for Maternal-Fetal Medicine and international meetings on PTB, reference lists of identified studies, textbooks, previously published systematic reviews, and review articles were also explored. No language restrictions were applied. All searches were carried out independently by 2 authors (R.F.L. and C.N.), and the results were merged. For studies that were reported in multiple publications, the data from the publication with the report of the primary outcomes and the largest sample size were used.
Study selection
We included randomized controlled trials that compared early treatment of women at <22 completed weeks of gestation with abnormal vaginal flora and asymptomatic BV that was diagnosed using objective criteria, with either oral clindamycin or clindamycin vaginal cream (CVC) vs placebo or no intervention. Only studies whose primary aim was to prevent PTB (delivery at <37 weeks of gestation) were included in the analysis. Quasirandomized studies were excluded. Other inclusion criteria consisted of pregnant women with a gestational age of <22 weeks of gestation at screening and commencement of treatment who were not in labor, had no vaginal bleeding, and did not report symptoms of lower genital tract infection. The diagnosis of BV was based on first trimester or early second trimester screening programs through the identification of abnormal lower genital tract flora and BV with the use of screening Gram stains scored by the Spiegel et al, Nugent et al, Hay et al, or Ison and Hay criteria. Studies with rescreening and re-treatment protocols were included in the final analysis. All studies that were deemed appropriate were retrieved and reviewed independently by the 2 screening authors to determine inclusion. Disagreements were resolved through consensus discussions.
Outcome measures
The primary outcomes of interest were spontaneous PTB at <37 completed weeks of gestation and late miscarriage (LM; birth between 16 and 23 completed weeks of gestation). Spontaneous PTB <37 completed weeks of gestation was chosen because this primary outcome was used in most metaanalyses that evaluated preventative strategies for PTB. Secondary outcomes included birthweight, low birthweight, very low birthweight, gestational age at delivery, early PTB (<30 completed weeks of gestation), admission to the neonatal intensive care unit (NICU), success of treatment according to degree of abnormal flora, persistent or recurrent BV, peripartum infections, and long-term follow-up evaluation.
Study quality assessment
We developed a quality assessment tool for studies of clindamycin effectiveness using the recommendations of Health Technology Assessment for the identification of systematic reviews. The format for the quality assessment tool previously has been used in a systematic review of the efficacy of nifedipine as a tocolytic. The quality of study methods was assessed with a tailored quality checklist that structured items in 2 broad categories: topic-specific or method-specific items. Furthermore, these items were divided into 3 general subcategories: selection bias, performance bias, and measurement bias. All quality determinations were scored as (1) adequate, (2) inadequate, or (3) not stated in the article. Consensus among the authors was reached through discussion and reevaluation.
Data abstraction
Two reviewers (R.F.L. and C.N.) scanned the abstracts and titles. Hard copies for all potentially relevant articles were acquired and evaluated independently by the 2 reviewers; authorship of the articles was not blinded. All outcome data were extracted independently, in duplicate form, by 2 reviewers. Information was extracted on study methods, study group demographics, intervention details (dose, type, placebo, no treatment), the use of rescreening and re-treatment in the studies, and the outcomes (number of outcome events). Consensus among the authors was reached through discussion and reevaluation.
Statistical analysis
Data from each of the studies considered in this study were organized in 2 × 2 contingency tables that contained the number of patients who were classified according to pregnancy outcome and the clindamycin treatment (treated, not treated). We calculated the summary relative risk (RR) for dichotomous data and weighted mean difference for continuous data with associated 95% CIs. Two prespecified subgroup analyses were performed to compare clindamycin with placebo/no treatment according to the route of administration (vaginal vs oral) and re-treatment (yes vs no). In addition, we conducted sensitivity analyses to explore the robustness of the findings for the primary outcome according to a statistical model (fixed-effects vs random-effects), use of placebo (yes vs no), and study quality (high vs not high). However, the sensitivity analysis according to study quality was not performed because all trials that were included were considered as high quality.
Heterogeneity of the results among studies was tested with the quantity I 2 , which describes the percentage of total variation across studies that is due to heterogeneity rather than chance. A value of 0% indicates no observed heterogeneity, whereas I 2 values of ≥50% indicate a substantial level of heterogeneity. We planned to pool data across studies using the fixed-effects models if substantial statistical heterogeneity was not present. We used random-effects models to pool data across studies if I 2 values were ≥50%.
We assessed publication and related biases visually by examining the symmetry of funnel plots and statistically by using the test of Egger et al. The larger the deviation of the intercept of the regression line from zero, the greater the asymmetry and the more likely it was that the metaanalysis would yield biased estimates of effect. As suggested by Egger et al, we considered a probability value of < .1 to indicate significant asymmetry.
We also calculated the number that was needed to treat (NNT) for an additional beneficial outcome with its 95% confidence intervals (CIs) for outcomes in which the treatment effect was significant at the 5% level (the 95% CI for the absolute risk difference did not include zero). NNT was computed from the results of metaanalysis of RRs: NNT = 1/control group event rate × (1-RR).
In this review, NNT for an additional beneficial outcome is the number of women with abnormal vaginal flora who need to be treated with clindamycin at <22 weeks of gestation rather than with placebo or no treatment to prevent 1 case of spontaneous PTB at <37 weeks of gestation or 1 LM between 16 and 23 completed weeks of gestation. Analyses were performed with the Review Manager (RevMan; software version 5.0.23; The Nordic Cochrane Centre, Copenhagen, Denmark).
Results
Literature identification
The flow of the electronic literature search is shown in Figure 1 . Of the 428 potentially relevant citations that were identified, 414 were excluded based on the title or on review of the abstract. Based on abstract review, hard copies of 14 articles were obtained. After a detailed review, 5 studies fulfilled the inclusion criteria and are included in the analysis. Of the 9 studies that were excluded, 2 were subgroup analyses of larger studies; 1 article was not a randomized controlled trial; 4 trials included patients who were screened and treated at >22 weeks of gestation; 1 trial was a conference abstract of 1 of the included studies, and 1 article was an observational study. We contacted the first authors of 3 trials to request further information. None of these authors were able to provide data for the subgroup of women who were treated at <22 weeks of gestation, so these studies were excluded from the primary analysis. The 5 trials comprised a total of 2346 women.
