Multiple studies have examined the frequency of preterm birth with the use of antibiotic therapy. These studies have varied in design and approach. Some studies have evaluated antibiotics targeted at specific pathogens,1–4 whereas others have investigated agents not specifically targeted at candidate pathogenic flora.5–8 The results of studies of antibiotic use to prevent preterm birth have been largely disappointing.3,6,9 The majority shows no benefit of therapy and some even suggest that antibiotic use is associated with an increased frequency of preterm birth.2,8,10–12 In fact, two separate randomized trials examining the use of metronidazole in pregnancy2,8 and one using it before conception5 showed an increase in preterm birth with the use of therapy.
If antibiotic use increases the frequency of preterm birth, the mechanism by which this occurs is unknown. One hypothesis holds that antibiotic therapy may alter vaginal flora and predispose to preterm birth. Carey and Klebanoff13 conducted a retrospective analysis of vaginal cultures collected as part of the Vaginal Infections and Prematurity (VIP) study. They found that alterations in vaginal flora increased the risk of preterm birth with an adjusted odds ratio (OR) of 2.99 (95% confidence interval [CI] 1.37–6.53. One mechanism by which this might occur is abolition of the most protective vaginal bacteria, hydrogen peroxide-producing lactobacilli. With loss of lactobacilli, there is an opportunity for growth and ascension of more pathogenic bacteria. Hauth et al,1 however, did not find significant disruption of flora in women treated with antibiotics. Bacterial vaginosis, diagnosed by Amsel's criteria, developed de novo in 5% of women who received antibiotics compared with 13% of those who received placebo.
The objective of this study was to evaluate whether systemic antibiotic exposure alters normal vaginal flora during pregnancy, as measured by the Nugent score, leading to increased risk of preterm birth compared with placebo among women with a positive fetal fibronectin test.
MATERIALS AND METHODS
We performed a secondary analysis of data collected during a randomized trial of metronidazole and erythromycin for the prevention of preterm birth in women with a positive cervicovaginal fetal fibronectin test.6 At each of 13 participating centers, asymptomatic women were enrolled between 21 0/7 and 25 6/7 weeks of gestation. All women enrolled in the study had a positive fetal fibronectin test (50 ng/mL or greater). At the screening visit, vaginal swabs were collected for Nugent Gram stain score classification of bacterial vaginosis and evaluated at a central laboratory.
Women were excluded from the parent trial if they had fetal death, major fetal anomalies, multiple gestation, allergy to study medication, antibiotic use at randomization, use of topical antifungals within 24 hours of screening, vaginal symptoms (nonelicited, spontaneously reported vaginal symptoms, including itching, burning, or malodor indicating need for specific treatment), major medical or obstetric complications, and anticipated prenatal care or delivery at a nonparticipating center.6 Women with prior preterm birth or other risk factors for preterm birth were not excluded.
The parent trial was approved by each of the Institutional Reviews Boards at the participating sites and all women provided written, informed consent. Women were randomized to either 250 mg metronidazole three times daily plus 250 mg erythromycin four times daily or to matching placebo. The parent trial used the simple urn method of randomization with stratification by clinical center. A follow-up visit was scheduled 14 days after randomization during which another swab was collected for Gram stain to assess Nugent score. For the purposes of this secondary analysis, women were included if they had available Nugent scores for both visits. The main focus of this analysis was those women who had normal vaginal flora and were antibiotic-exposed. The placebo group and the groups with bacterial vaginosis were included for comparison.
The primary outcome measure for this analysis was change in Nugent score among women with normal flora exposed to antibiotics. We also evaluated the relationship between this change and the subsequent preterm birth risk. The change in Nugent score was assessed for women with (score 7 or greater) or without (score less than 7) bacterial vaginosis in a stratified manner. Linear regression analysis was used to evaluate whether change in Nugent score was associated with preterm birth.
The univariable analyses were performed using chi square and Wilcoxon signed-rank test where appropriate. Logistic regression and calculation of ORs were conducted to assess the relationship of the risk of preterm birth and change of Nugent score. Multiple logistic regression was performed to control for potential confounders and did not change the findings (data not shown). Statistical significance was set at P<.05.
The parent trial was a randomized, double-masked, placebo-controlled trial and included 715 women in the intention-to-treat analysis. The flow of patients within the trial has been previously published.6 A total of 547 of 715 women had Nugent scores available from both visits and were included in this analysis. The age, body mass index, racial and ethnic, and demographic characteristics were similar between groups (Table 1). Like in previous studies, those with bacterial vaginosis were slightly more likely to report African American race. There were 388 women who did not and 159 women who did have bacterial vaginosis. Of the 388 women who did not have bacterial vaginosis, 200 were randomized to antibiotics. Of the 159 women with bacterial vaginosis, 69 were in the antibiotic arm. The median Nugent score for all women in the study declined from 4.0 to 2.0 after antibiotic therapy (P<.001). The primary outcome measure of the study was change in Nugent score among bacterial vaginosis-negative women exposed to antibiotics. The median score in this group was not worsened by antibiotic exposure, 2.0 to 1.5 (P=.11). The median Nugent score declined more dramatically within the bacterial vaginosis-positive group (from 8.0 to 3.0, P<.01) (Table 2).
To examine whether antibiotic exposure might influence one particular component of the Nugent score, we determined the change in score component in each of the subgroups. The components of the Nugent score that were affected by antibiotic exposure were similar among women with and without bacterial vaginosis and were primarily reductions in the scores for Gardnerella vaginalis and anaerobic Gram-negative rods and for Lactobacillus score. Of the different Gram stain components, Mobiluncus was the least affected by antibiotics. As anticipated, placebo had no apparent effect on the components of the score among women without bacterial vaginosis (Table 3).
Among women with normal flora, antibiotic exposure did not increase the risk of preterm birth (13% compared with 14%, P=.82) and change in Nugent score was not related to it with an OR of 1.07 (95% CI 0.9–1.2) for each one-point change in Nugent score. Logistic regression models were fit after stratification by bacterial vaginosis status at entry. Neither change in Nugent score nor antibiotic exposure was associated with preterm birth among women with or without bacterial vaginosis. As has been consistently demonstrated in other studies, prior spontaneous preterm birth and low prepregnancy body mass index were associated with preterm birth in this study (data not shown). The mean gestational age at delivery and frequency of preterm birth were not significantly different in any group (Table 4).
We did not find significant alteration of vaginal flora resulting from antibiotic exposure among women with normal flora. Although disruption of normal flora has been proposed as a possible mechanism for the recent studies demonstrating an increased risk of preterm birth among women exposed to certain antibiotics, our data do not support this theory. The theory would suggest that a lack of normal flora in the vagina would allow for ascent of pathogenic bacteria into the uterine cavity, thus leading to a cascade of inflammatory events culminating in preterm labor and delivery. We found a slight improvement in Nugent score among women with bacterial vaginosis who were exposed to antibiotics. This improvement was also demonstrated among women with normal flora, although less distinctly compared with those with bacterial vaginosis.
We did not find preterm birth to be more common among the women with normal flora exposed to antibiotics. In fact, the rate of preterm birth, ranging from 10% to 17% between the strata, was not as high as one might expect in a cohort of women with a positive fetal fibronectin test. The reasons for this are unclear and may be related to a “Hawthorne effect,” in which patients engaged in clinical trials have improved outcomes regardless of their study group. The mean gestational age at delivery was at term and was similar among the strata of bacterial vaginosis and therapy groups. We were unable to detect an association between change in Nugent score and the occurrence of preterm birth.
The findings of the parent trial for this analysis showed that the rate of preterm birth among women receiving antibiotics was 14.4% compared with 12.4% among those who received placebo. Subsequent secondary analysis of the trial showed that among the subset of women in whom bacterial vaginosis actually resolved, the preterm birth rate was decreased.14 Nugent scores were examined as a dichotomous variable for the purposes of that analysis with bacterial vaginosis being defined as present or absent. Nugent et al15 commented in their original article that the score could be used as a continuum rather than the forced dichotomy of bacterial vaginosis positive or negative. Our analysis treated the score as a continuous variable with the intent of being able to detect more subtle changes among women with normal flora. Although we were able to detect more subtle differences in score than the typical dichotomous use of the Nugent score offers, our findings did not translate to clinically significant outcomes.
Strengths of our study include the high quality of data collection with regular edits and audits of data mandated by the study protocol used at all Maternal-Fetal Medicine Units Network centers. Gram stains and Nugent scores were read at a central laboratory with documented high reproducibility of results. A secondary analysis of the Network “bacterial vaginosis” study showed that 78% of women with bacterial vaginosis experience resolution with therapy and that there is also spontaneous regression of bacterial vaginosis in up to one-third of placebo-treated women.16 Our findings support those results. We observed spontaneous improvement of Nugent score in the placebo-treated arm, from a mean of 8.38 to 7.20. We did not, however, observe improvement in the Nugent score among women with normal flora in the placebo arm. The very slight improvement in Nugent score among those with normal flora exposed to antibiotics was significant, mean 2.37 to 2.05, and likely a reflection of a true improvement as a result of antibiotic exposure.
This study does not explain the mechanism of preterm birth associated with antibiotic exposure. A number of potential mechanisms are plausible. A commonly cited potential mechanism is the notion that bacteria, when lysed, actually cause an increase in the proinflammatory state, which might lead to uterine contractions. Another theory is that there could be disruption in normal flora that is not measurable by Nugent score. Our study is limited by its retrospective design and use of a surrogate marker for disruption of flora leading to preterm birth from exposure to antibiotics. More detailed vaginal flora information was not collected in the original study, so this was not possible given the nature of the secondary analysis. The analysis by Carey and Klebanoff found that in particular, acquisition of Escherichia coli or Klebsiella pneumoniae was an independent risk factor for preterm birth (adjusted OR 2.99; 95% CI 1.37–6.53).13 It has also been suggested that antibiotics may not be adequate therapy for upper tract disease. However, in a secondary analysis of a randomized trial of interconception antibiotics, neither endometrial microbial invasion nor plasma cell endometritis were risk factors for adverse pregnancy outcomes.12 Alternatively, preterm birth and infections may be markers of an underlying dysfunction in maternal immunity.13 Such disorders likely have a genetic or epigenetic predisposition. Macones et al17 found that there was an interaction among carriers of a tumor necrosis factor polymorphism exposed to bacterial vaginosis and the risk of preterm birth. Future study of the association between exposure to antibiotics, pathogens, and preterm birth outcomes is necessary to examine these relationships.
1. Hauth JC, Goldenberg RL, Andrews WW, DuBard MB, Copper RL. Reduced incidence of preterm delivery with metronidazole and erythromycin in women with bacterial vaginosis. N Engl J Med 1995;333:1732–6.
2. Klebanoff MA, Carey JC, Hauth JC, Hillier SL, Nugent RP, Thom EA, et al; National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units. Failure of metronidazole to prevent preterm delivery among pregnant women with asymptomatic Trichomonas vaginalis infection. N Engl J Med 2001;345:487–93.
3. Klebanoff MA, Regan JA, Rao AV, Nugent RP, Blackwelder WC, Eschenbach DA, et al. Outcome of the Vaginal Infections and Prematurity Study: results of a clinical trial of erythromycin among pregnant women colonized with group B streptococci. Am J Obstet Gynecol 1995;172:1540–5.
4. Thomsen AC, Morup L, Hansen KB. Antibiotic elimination of group-B streptococci in urine in prevention of preterm labour. Lancet 1987;1:591–3.
5. Andrews WW, Goldenberg RL, Hauth JC, Cliver SP, Copper R, Conner M. Interconceptional antibiotics to prevent spontaneous preterm birth: a randomized clinical trial. Am J Obstet Gynecol 2006;194:617–23.
6. Andrews WW, Sibai BM, Thom EA, Dudley D, Ernest JM, McNellis D, et al; National Institute of Child Health & Human Development Maternal-Fetal Medicine Units Network. Randomized clinical trial of metronidazole plus erythromycin to prevent spontaneous preterm delivery in fetal fibronectin-positive women. Obstet Gynecol 2003;101:847–55.
7. Goldenberg RL, Mwatha A, Read JS, Adeniyi-Jones S, Sinkala M, Msmanga G; Hptn024 Team. The HPTN 024 Study: the efficacy of antibiotics to prevent chorioamnionitis and preterm birth. Am J Obstet Gynecol 2006;194:650–61.
8. Shennan A, Crawshaw S, Briley A, Hawken J, Seed P, Jones G, et al. A randomised controlled trial of metronidazole for the prevention of preterm birth in women positive for cervicovaginal fetal fibronectin: the PREMET Study. BJOG 2006;113:65–74.
9. Thinkhamrop J, Hofmeyr GJ, Adetoro O, Lumbiganon P. Prophylactic antibiotic administration in pregnancy to prevent infectious morbidity and mortality. The Cochrane Database of Systematic Reviews 2002, Issue 4. Art. No.: CD002250. DOI: 10.1002/14651858.CD002250.
10. Anderson BL, Simhan HN, Simons K, Wiesenfeld HC. Additional antibiotic use and preterm birth among bacteriuric and nonbacteriuric pregnant women. Int J Gynaecol Obstet 2008;102:141–5.
11. Kigozi GG, Brahmbhatt H, Wabwire-Mangen F, Wawer MJ, Serwadda D, Sewankambo N, et al. Treatment of Trichomonas in pregnancy and adverse outcomes of pregnancy: a subanalysis of a randomized trial in Rakai, Uganda. Am J Obstet Gynecol 2003;189:1398–400.
12. Tita AT, Cliver SP, Goepfert AR, Conner M, Goldenberg RL, Hauth JC, et al. Clinical trial of interconceptional antibiotics to prevent preterm birth: subgroup analyses and possible adverse antibiotic-microbial interaction. Am J Obstet Gynecol 2007;197:367.e1–6.
13. Carey JC, Klebanoff MA. Is a change in the vaginal flora associated with an increased risk of preterm birth? Am J Obstet Gynecol 2005;192:1341–6.
14. Hendler I, Andrews WW, Carey CJ, Klebanoff MA, Noble WD, Sibai BM, et al; National Institute of Child Health and Human Development, Maternal-Fetal Medicine Units Network. The relationship between resolution of asymptomatic bacterial vaginosis and spontaneous preterm birth in fetal fibronectin-positive women. Am J Obstet Gynecol 2007;197:488.e1–5.
15. Nugent RP, Krohn MA, Hillier SL. Reliability of diagnosing bacterial vaginosis is improved by a standardized method of gram stain interpretation. J Clin Microbiol 1991;29:297–301.
16. Klebanoff MA, Hauth JC, MacPherson CA, Carey JC, Heine RP, Wapner RJ, et al; National Institute for Child Health and Development Maternal Fetal Medicine Units Network. Time course of the regression of asymptomatic bacterial vaginosis in pregnancy with and without treatment. Am J Obstet Gynecol 2004;190:363–70.
17. Macones GA, Parry S, Elkousy M, Clothier B, Ural SH, Strauss JF. A polymorphism in the promoter region of TNF and bacterial vaginosis: preliminary evidence of gene-environment interaction in the etiology of spontaneous preterm birth. Am J Obstet Gynecol 2004;190:1504–8.