Introduction: Antenatal antibiotic exposure has been suggested as a risk factor for bronchopulmonary dysplasia (BPD). We aimed to summarize the evidence from randomized controlled trials (RCTs) and observational studies on this potential association. Methods: PubMed/Medline and Embase databases were searched. BPD was classified as BPD28 (supplemental oxygen during 28 days or at postnatal day 28), BPD36 (supplemental oxygen at 36 weeks postmenstrual age), BPD36 or death, and BPD-associated pulmonary hypertension (BPD-PH). Bayesian model-averaged (BMA) meta-analysis was used to calculate Bayes factors (BFs). The BF10 is the ratio of the probability of the data under the alternative hypothesis (H1) over the probability of the data under the null hypothesis (H0). Results: We included 6 RCTs and 27 observational studies (126,614 infants). Regarding BPD28, BMA showed that the evidence in favor of H0 (lack of association with antenatal antibiotics) was weak for the RCTS (BF10 = 0.506, 6 studies) and moderate for the observational studies (BF10 = 0.286, 10 studies). Regarding BPD36, the evidence in favor of H0 was moderate for the RCTs (BF10 = 0.127, 2 studies) and weak for the observational studies (BF10 = 0.895, 14 studies). Evidence in favor of H0 was also weak for the associations with BPD36 or death (BF10 = 0.429, 2 studies) and BPD-PH (BF10 = 0.384, 2 studies). None of the meta-analyses showed evidence in favor of H1. Conclusions: The currently available evidence suggests a lack of association between antenatal antibiotics and BPD. However, our results should not be interpreted as an argument for widespread use of antibiotics in the setting of preterm delivery.

Despite major developments, including the use of antenatal corticosteroids, surfactant, and new ventilation techniques, the incidence of bronchopulmonary dysplasia (BPD), the chronic lung disease of prematurity, has not decreased in the past decades. This is partly due to the increased survival of extremely preterm infants [1, 2].

BPD is increasingly recognized as the consequence of a pathological reparative response of the developing lung to both antenatal and postnatal injury [1, 2]. This results in abnormal lung growth with alveolar simplification and dysregulation of pulmonary vasculature [1, 2]. There is a growing body of evidence supporting the relevance of prenatal insults such as chorioamnionitis, intrauterine growth restriction, hypertensive disorders of pregnancy, or maternal smoking in developing BPD [1‒5]. Moreover, antenatal stress factors may alter susceptibility to postnatal lung injury, thereby impacting risk for BPD [1, 2].

An antenatal stressor that has recently been proposed to play a role in the pathogenesis of BPD is antibiotic exposure [6, 7]. As reviewed by Sola, the main indications for the use of antibiotics during pregnancy and around delivery are prevention of neonatal group B Streptococcus sepsis, preterm labor with intact or ruptured membranes, chorioamnionitis, endometritis, bacterial vaginosis, asymptomatic bacteriuria, urinary tract infections, and prevention of infection during cesarean section [8]. Consequently, it has been reported that about 40% of mothers receive antibiotics in the perinatal period, and this percentage is much higher in the case of preterm delivery [8]. In many of the aforementioned clinical situations, the use of antibiotics is fully justified, but antenatal antibiotics may also lead to short- and long-term adverse effects and increase the risk of antibiotic resistance in neonates [8‒11].

Recently, Willis et al. [6] reported that antenatal antibiotic exposure augmented lung injury in a hyperoxia-based experimental mouse model of BPD. In addition, several observational studies have shown that early and prolonged postnatal antibiotic exposure increased BPD risk in preterm infants [12, 13]. Antibiotic-induced microbiome dysbiosis has been suggested as a possible contributor to this increased risk [7, 14]. However, and to the best of our knowledge, the potential role of antenatal antibiotics on BPD has not yet been systematically evaluated. The aim of this systematic review and meta-analysis was to summarize the evidence from both randomized controlled trials (RCTs) and observational studies on the association between antenatal antibiotics and the risk of developing BPD in preterm infants.

Instead of the more commonly used frequentist statistics, we used a Bayesian approach to meta-analysis. A major advantage of Bayesian analyses over frequentist null hypothesis significance testing (NHST) is that the strength of evidence can be tested in terms of both the null hypothesis (H0) and the alternative hypothesis (H1) [15, 16]. In addition, Bayesian analysis can help distinguish between evidence of absence and absence of evidence [15, 16].

The methodology for this study was based on our recently published experience on performing meta-analyses to study the associations between antenatal and perinatal exposures and outcomes of prematurity [3, 4, 17]. The study was performed and reported according to PRISMA and MOOSE guidelines. Review protocol was registered in PROSPERO database (ID = CRD42021237460). The Population, Exposure, Comparison, and Outcome (PECO) question was: Do preterm infants (P) exposed to antenatal antibiotics (E) have a different risk of developing BPD (O) compared to preterm infants not exposed (C)? Therefore, H1 was formulated as “antenatal antibiotic exposure is associated with BPD risk,” and H0 was formulated as “antenatal antibiotic exposure is not associated with BPD risk.”

Data Sources and Search Strategies

A comprehensive literature search was conducted using PubMed/Medline, Embase, and Web of Science from their inception to December 1, 2022. The search strategy for PubMed used the following terms, including Mesh terms: (antibiotics OR anti-bacterial agents) AND (antenatal OR perinatal OR maternal) AND (preterm infant OR very low birth weight infant) AND (outcome OR bronchopulmonary dysplasia OR BPD OR chronic lung disease OR CLD). A similar strategy was used in the other databases. No language limit was applied. Translation was performed where necessary. RCTs and observational studies were included in the review. Narrative reviews, systematic reviews, case reports, letters, editorials, and commentaries were excluded but read to identify potential additional studies. Additional strategies to identify studies included use of “related articles” feature on PubMed and “cited by” tool in Web of Science and Google Scholar.

Eligibility Criteria and Study Selection

Studies were included if they examined infants with gestational age (GA) lower than 37 weeks and reported primary data that could be used to measure the association between exposure to antenatal antibiotics and the development of BPD. Both RCTs and observational studies were included. Outcomes considered in meta-analysis were (1) BPD28, defined as oxygen requirement during 28 days or at postnatal day 28; (2) BPD36, defined as oxygen requirement at the postmenstrual age (PMA) of 36 weeks; (3) BPD36 or death; (4) severe BPD, defined as need for ≥30% oxygen and/or positive pressure respiratory support at 36 weeks PMA; (5) BPD-associated pulmonary hypertension (BPD-PH), defined by any echocardiographic criteria as long as the evaluation was performed at a postnatal age >4 weeks. Using these definition criteria, BPD28 was considered to include all severities of BPD, whereas BPD36 was considered to include a combination of moderate and severe BPD [3, 4]. To identify relevant studies, two reviewers (K.V.M. and E.V.) independently screened the results of the searches and applied inclusion criteria using a structured form. Discrepancies were identified and resolved through discussion or in consultation with the other researchers.

Data Extraction and Assessment of Risk of Bias

Two investigators (K.V.M. and T.H.) extracted the data by using a data collection form designed for this review. The following information was collected: study type, number of patients, number and name of centers, study period, inclusion/exclusion criteria, patient characteristics, antenatal antibiotics, and incidence of BPD. Two other investigators (E.v.W.-K. and E.V.) independently validated the accuracy of the extracted data.

Two reviewers (E.V. and T.H.) independently assessed risk of bias in each study, using two predetermined tools. Risk of bias in RCTs was assessed by using the Cochrane “Risk of Bias Assessment Tool.” For each domain (random number generation, allocation concealment, blinding of intervention and outcome assessors, completeness of follow-up, selectivity of reporting, and other potential sources of bias), the risk was assessed as low, high, or unclear. Risk of bias in observational studies was assessed using the Newcastle-Ottawa scale for quality assessment of cohort and case-control studies. This scale uses a rating system (range: 0–9) that gives points for selection (0–4), comparability (0–2), and outcome/exposure (0–3). Discrepancies during the data extraction and assessment of risk of bias were resolved by discussion and consensus among all reviewers.

Statistical Analysis

The log odds ratio (logOR) and the corresponding standard error for each individual study were calculated using Comprehensive Meta-Analysis V4.0 software (Biostat Inc., Englewood, NJ, USA). The results were further pooled and analyzed by a Bayesian model-averaged (BMA) meta-analysis [15, 16]. We performed the BMA in JASP, which utilizes the metaBMA R package. BMA employs Bayes factors (BFs) and Bayesian model averaging to evaluate the likelihood of the data under the combination of models assuming the presence versus the absence of the meta-analytic effect and heterogeneity [15, 16]. As described by Gronau, Bartoš et al. [15, 16], BMA combines the results of four Bayesian meta-analysis models: (a) fixed-effects and H0, (b) fixed-effects and H1, (c) random-effects and H0, and (d) random-effects and H1. These models are combined according to their plausibilities given the observed data to address the two key questions, “Is the overall effect nonzero?” and “Is there between-study variability in effect size?” [15, 16].

The BF10 is the ratio of the probability of the data under H1 over the probability of the data under H0. The BF10 was interpreted using the evidence categories suggested by Lee and Wagenmakers [18]: <1/100 = extreme evidence for H0, from 1/100 to <1/30 = very strong evidence for H0, from 1/30 to <1/10 = strong evidence for H0, from 1/10 to <1/3 = moderate evidence for H0, from 1/3 to <1 weak/inconclusive evidence for H0, from 1 to 3 = weak/inconclusive evidence for H1, from >3 to 10 = moderate evidence for H1, from >10 to 30 = strong evidence for H1, from >30 to 100 = very strong evidence for H1, and >100 extreme evidence for H1. The BF01 (ratio of the probability of the data under H0 over the probability of the data under H0) is the inverse of BF10. The BFrf is the ratio of the probability of the data under the random-effects model over the probability of the data under the fixed-effects model. The BFrf was interpreted in the following way: <1/100 = extreme evidence for fixed-effects, from 1/100 to <1/30 = very strong evidence for fixed-effects, from 1/30 to <1/10 = strong evidence for fixed-effects, from 1/10 to <1/3 = moderate evidence for fixed-effects, from 1/3 to <1 weak/inconclusive evidence for fixed-effects, from 1 to 3 = weak/inconclusive evidence for random-effects, from >3 to 10 = moderate evidence for random-effects, from >10 to 30 = strong evidence for random-effects, from >30 to 100 = very strong evidence for random-effects, and >100 extreme evidence for random-effects. We used the empirical prior distributions based on the Cochrane Database of Systematics Reviews transformed to logOR: logOR ∼ Student t (µ = 0, σ = 0.78, ν = 5), tau = Inverse-Gamma (k = 1.71, θ = 0.73) [15, 16].

Description of Studies and Quality Assessment

The PRISMA flow diagram of the search process is shown in online supplementary Figure 1 (for all online suppl. material, see https://doi.org/10.1159/000536220). Of 1,615 potentially relevant studies, 33 (6 RCTs and 27 observational) were included. These studies included 126,614 infants. Characteristics of the RCTs [19‒24] are summarized in online supplementary Table 1 and data on risk of bias in online supplementary Table 2. Characteristics of the observational studies [25‒51] are summarized in online supplementary Table 3.

BMA Meta-Analysis

Table 1 summarizes the results of the BMA. To facilitate interpretation, the results in Table 1 are expressed as OR instead of logOR. In addition, to allow the reader to compare the frequentist and Bayesian approaches, we have included the p value of the corresponding random-effects frequentist meta-analysis in Table 1. Detailed data on heterogeneity are presented in the online supplementary Table 4. Six RCTs [19‒24] and 10 observational studies [30, 32, 36‒38, 43, 44, 47, 48, 50] reported on BPD28 (shown in Fig. 1). BMA showed that the evidence in favor of H0 (i.e., lack of association between antenatal antibiotics and BPD28) was weak for the RCTs (BF10 = 0.506, shown in Fig. 1a) and moderate for the observational studies (BF10 = 0.286, shown in Fig. 1b). When RCTs and observational studies were combined in one meta-analysis, the evidence in favor of H0 was moderate (BF10 = 0.246, Table 1). Two RCTs [21, 22] and 14 observational studies [25‒29, 31, 34, 39‒41, 47‒49, 51] reported on BPD36 (shown in Fig. 2). BMA showed that the evidence in favor of H0 (i.e., lack of association between antenatal antibiotics and BPD36) was moderate for the RCTs (BF10 = 0.172, shown in Fig. 2a) and weak for the observational studies (BF10 = 0.895, shown in Fig. 2b). When RCTs and observational studies were combined in one meta-analysis, the evidence in favor of H0 was weak (BF10 = 0.333, Table 1). In two RCTs [21, 22], pregnant women were randomized to amoxicillin-clavulanic acid, erythromycin, both, or a matching placebo regimen. When the three antibiotic regimens were analyzed separately, BMA showed that the evidence in favor of H0 was moderate for all three antibiotic regimens and BPD36 (amoxicillin-clavulanic acid BF10 = 0.256, erythromycin BF10 = 0.205, both antibiotics BF10 = 0.240), and for amoxicillin-clavulanic acid (BF10 = 0.162) or erythromycin (BF10 = 0.236) and BPD28 (online suppl. Table 5). The evidence for H0 was weak for the combination of both antibiotics and BPD28 (BF10 = 0.361) (online suppl. Table 5). Only one study [33] reported on severe BPD and therefore a meta-analysis of this outcome could not be performed. This study did not show an association between antenatal antibiotics and severe BPD (logOR −0.08, 95% CI −0.51 to 0.34, BF10 = 0.102). Two observational studies reported on BPD36 or death [35, 42] and BMA showed that the evidence in favor of H0 was weak (BF10 = 0.429, Table 1). Finally, two observational studies reported on BPD-PH [45, 46] and BMA showed that the evidence in favor of H0 was weak (BF10 = 0.384, Table 1).

Table 1.

BMA meta-analysis of the association between antenatal antibiotic exposure and BPD

GroupStudy typeKnAveraged effect (OR)Standard deviationCredible intervalBF10Evidence forp value frequentist analysisaBFrfEvidence for
lower limitupper limitH1H0random-effectsfixed-effects
BPD28 RCT 12,715 0.867 1.155 0.689 1.237 0.506  Weak 0.240 0.883  Weak 
OBS 10 107,576 1.170 1.239 0.815 1.839 0.286  Moderate 0.272 2.452 Weak  
All 16 120,291 0.986 1.164 0.791 1.404 0.246  Moderate 0.938 3.213 Moderate  
BPD36 RCT 11,050 0.988 1.219 0.673 1.516 0.172  Moderate 0.861 0.252  Moderate 
OBS 14 4,178 1.191 1.091 1.004 1.412 0.895  Weak 0.027 0.316  Moderate 
All 16 15,228 1.122 1.075 0.976 1.296 0.333  Weak 0.093 0.209  Moderate 
BPD36 or death OBS 1,534 0.879 1.499 0.370 1.891 0.429  Weak 0.269 4.417 Moderate  
BPD-PH OBS 298 0.877 1.354 0.486 1.602 0.384  Weak 0.556 0.441  Weak 
GroupStudy typeKnAveraged effect (OR)Standard deviationCredible intervalBF10Evidence forp value frequentist analysisaBFrfEvidence for
lower limitupper limitH1H0random-effectsfixed-effects
BPD28 RCT 12,715 0.867 1.155 0.689 1.237 0.506  Weak 0.240 0.883  Weak 
OBS 10 107,576 1.170 1.239 0.815 1.839 0.286  Moderate 0.272 2.452 Weak  
All 16 120,291 0.986 1.164 0.791 1.404 0.246  Moderate 0.938 3.213 Moderate  
BPD36 RCT 11,050 0.988 1.219 0.673 1.516 0.172  Moderate 0.861 0.252  Moderate 
OBS 14 4,178 1.191 1.091 1.004 1.412 0.895  Weak 0.027 0.316  Moderate 
All 16 15,228 1.122 1.075 0.976 1.296 0.333  Weak 0.093 0.209  Moderate 
BPD36 or death OBS 1,534 0.879 1.499 0.370 1.891 0.429  Weak 0.269 4.417 Moderate  
BPD-PH OBS 298 0.877 1.354 0.486 1.602 0.384  Weak 0.556 0.441  Weak 

BF, Bayes factor; BF10, ratio of the probability of the data under H1 over the probability of the data under H0; BFrf, ratio of the probability of the data under the random-effects model over the probability of the data under the fixed-effects model; BPD28, bronchopulmonary dysplasia defined as oxygen requirement during the first 28 days of life or at postnatal day 28; BPD36, bronchopulmonary dysplasia defined as oxygen or respiratory support requirement at the postmenstrual age of 36 weeks; BPD-PH, BPD-associated pulmonary hypertension; K, number of studies; OBS, observational; RCT, randomized controlled trial.

aRandom-effects frequentist meta-analysis.

Fig. 1.

BMA meta-analysis of the association between antenatal antibiotic exposure and BPD28 (bronchopulmonary defined as oxygen requirement during the first 28 days of life or at postnatal day 28). a RCTs. b Observational studies. The forest plot shows the observed effect size estimates from each study, the overall meta-analytic fixed-effects and random-effects estimates, and the corresponding model-averaged effect size estimate. LogOR>0 denotes increased risk of BPD in infants exposed to antenatal antibiotics.

Fig. 1.

BMA meta-analysis of the association between antenatal antibiotic exposure and BPD28 (bronchopulmonary defined as oxygen requirement during the first 28 days of life or at postnatal day 28). a RCTs. b Observational studies. The forest plot shows the observed effect size estimates from each study, the overall meta-analytic fixed-effects and random-effects estimates, and the corresponding model-averaged effect size estimate. LogOR>0 denotes increased risk of BPD in infants exposed to antenatal antibiotics.

Close modal
Fig. 2.

BMA meta-analysis of the association between antenatal antibiotic exposure and BPD36 (bronchopulmonary defined as oxygen or respiratory support requirement at the postmenstrual age of 36 weeks). a RCTs. b Observational studies. The forest plot shows the observed effect size estimates from each study, the overall meta-analytic fixed-effects and random-effects estimates, and the corresponding model-averaged effect size estimate. LogOR>0 denotes increased risk of BPD in infants exposed to antenatal antibiotics.

Fig. 2.

BMA meta-analysis of the association between antenatal antibiotic exposure and BPD36 (bronchopulmonary defined as oxygen or respiratory support requirement at the postmenstrual age of 36 weeks). a RCTs. b Observational studies. The forest plot shows the observed effect size estimates from each study, the overall meta-analytic fixed-effects and random-effects estimates, and the corresponding model-averaged effect size estimate. LogOR>0 denotes increased risk of BPD in infants exposed to antenatal antibiotics.

Close modal

This is the first systematic review and meta-analysis investigating the association between antenatal antibiotic exposure and risk of developing BPD. The Bayesian approach allowed us to assess the strength of the evidence in favor of H1 and H0. We found that none of the meta-analyses showed evidence in favor of an association between antenatal antibiotic exposure and BPD risk (H1). Furthermore, some of the meta-analyses showed moderate evidence in favor of an absence of association between antenatal antibiotics and BPD (H0).

Frequentist meta-analysis with p values and confidence intervals remains the most widely used analytical approach. However, overconfidence in what significant p values can reveal and negative attitudes toward p values that fail to pass the significance threshold have contributed to the so-called reproducibility crisis in scientific research [52]. An important limitation of frequentist NHST is that failing to reject H0 (absence of effect) when the p value is below a predetermined threshold (usually 0.05) does not mean evidence for H0. Conversely, the rejection of H0 when the p value is above the threshold does not necessarily mean that we have found evidence in support of H1 (presence of effect) [15, 52]. As an alternative to this and other limitations of NHST, Bayesian statistics is increasingly accepted and used in biomedicine. The BF is the way to quantify the relative degree of support for a hypothesis in a data set and is the primary tool used in Bayesian inference for hypothesis testing [15, 16]. For example, a BF10 of 5 allows the researcher to make statements of the form “compared with the effect-absent hypothesis, the data have made the effect-present hypothesis 5 times more likely than it was before” [15, 16]. In addition, in BMA meta-analysis, multiple models are considered simultaneously, and inference is proportioned to the support that each model receives from the data. This eliminates the need for stage-wise, multistep inference procedures that first identify a single preferred model (i.e., a fixed-effects model or a random-effects model) and then interpret the model parameters without acknowledging the uncertainty inherent in the model selection stage [16]. The multi-model approach also decreases the potential impact of model misspecification [16]. In the following paragraphs, we will discuss the advantages of the Bayesian approach over the classical frequentist approach using several illustrative examples from our meta-analysis.

RCTs and meta-analyses based on RCTs are cornerstones of evidence-based medicine [53]. Several RCTs have compared systemic antibiotics versus placebo in situations considered to be associated with increased risk of intrauterine infection, such as preterm premature rupture of the membranes (PPROM) and preterm labor with intact membranes [19‒24]. When these RCTs were pooled, meta-analyses showed that antibiotics for women with PPROM was associated with prolongation of pregnancy and improvements in a number of short-term neonatal morbidities, such as early-onset sepsis, respiratory distress syndrome, or cranial ultrasound abnormalities [54‒57]. Therefore, antibiotic administration has become the standard of care for patients with PPROM. In contrast, the evidence supports not giving antibiotics routinely to women in preterm labor with intact membranes in the absence of overt signs of infection [54]. Unfortunately, very few RCTs considered BPD as one of the outcomes to be investigated. The present meta-analysis included six RCTS reporting on BPD28 and two RCTs reporting on moderate to severe BPD (BPD36). Frequentist analysis yielded a p value of 0.240 for BPD28 and 0.861 for BPD36 (see Table 1). These p values would lead us to the conclusion that we cannot reject H0. However, as mentioned above, the fact that H0 cannot be rejected does not necessarily mean that it has to be accepted.

The problem of interpreting non-significant frequentist results is compounded by the fact that most meta-analyses are based on small numbers of trials, which limits the statistical power of the analyses [16, 53]. As mentioned above, Bayesian hypothesis testing aims to quantify the relative plausibility of H1 and H0 [58, 59]. In our analysis of RCTs on BPD36 (p = 0.861), the Bayesian approach allows us to estimate that the observed data are 5.8 times (BF10 = 0.172; BF01 = 5.814) more likely under no association (H0) with antenatal antibiotics than under the presence of association (H1). We can therefore conclude that there is moderate evidence for absence of effect of antenatal antibiotics on the development of moderate-to-severe BPD. In contrast, in the case of RCTs reporting on BPD28 (p = 0.240), the evidence in favor of H0 is less conclusive because it is only 2 times greater than the evidence in favor of H1 (BF10 = 0.506; BF01 = 1.976).

The ORACLE I [21] and II [22] trials stand out among the RCTs included in this meta-analysis. These RCTs include the majority of the population in the various meta-analyses of the effect of maternal antibiotics on PPROM (ORACLE I) and preterm labor with intact membranes (ORACLE II) [54‒57]. In the ORACLE trials, women were randomized to erythromycin, amoxicillin-clavulanic acid, both, or a matching placebo regimen. This allowed us to conduct separate meta-analyses for the different antibiotic regimens. As shown in online supplementary Table 5, the evidence in favor of H0 was weak to moderate for the three antibiotic regimes, suggesting that the lack of association with BPD is not dependent on the anti-microbial agent. However, it should be noted that a limitation of the ORACLE trials, as well as other RCTs on maternal antibiotics, is that the population of newborns with a GA of less than 32 weeks as very small (∼10%). As BPD only affects infants with the lowest GA, this small percentage of neonates at risk of developing BPD may have influenced the results of our meta-analysis.

Due to the low number of RCTs, we have also included observational studies in the meta-analysis. Observational evidence must be interpreted cautiously due to the risk of bias by indication [60]. Thus, the infants at greatest risk for BPD may be also the infants who receive the most antenatal antibiotics, which may lead to a false association between this exposure and the adverse outcome. Intrauterine infection is very frequently the trigger for preterm birth, particularly at early GAs where neonatal morbidity and mortality is at its greatest [3, 61‒64]. Moreover, prenatal infection may have an effect on the developmental trajectory of the fetal lung, thereby altering the susceptibility to develop BPD [3, 62‒64]. Finally, if prenatal infection is suspected as the cause of preterm birth, the likelihood of continuing antibiotic treatment after birth increases [65, 66]. Therefore, it is not possible to distinguish between the effects of antenatal and postnatal antibiotics in observational studies. Of note, early and prolonged exposure to antibiotics in preterm infants without confirmed infection has been associated with increased mortality as well as increased risk of complications, including BPD [12, 13, 65, 66]. Although all of these factors may contribute to an association between antenatal antibiotics and BPD, at least in observational studies, the BMA data do not provide support for the hypothesis of an association.

Interestingly, the meta-analysis of observational studies on BPD36 provides a good example of the differences between the frequentist and Bayesian approaches. The frequentist analysis shows a “statistically significant” result (p = 0.027) that would allow rejecting H0 and accepting H1 by default. However, the frequentist NHST is not directly testing H1. The BMA analysis showed that the BF10 is close to 1 (0.89), indicating that the probabilities of the data under H1 or under H0 are almost the same. Therefore, the frequentist conclusion could be “meta-analysis of observational studies showed a significant association between antenatal exposure to antibiotics and risk of developing BPD36,” whereas the Bayesian conclusion is that there is no evidence from observational studies to support or refute the association between antenatal antibiotics and BPD risk.

Since BPD is a condition with a multifactorial pathogenesis, both hypotheses of increased and decreased risk of BPD in association with perinatal antibiotic exposure are biologically plausible. Antenatal antibiotics have the potential to alter the microbial colonization process in infants [11]. Disruption of normal microbiome can lead to abnormal inflammatory responses, which may play a role in the pathogenesis of BPD [10, 11, 14]. On the other hand, antenatal antibiotics reduce fetal inflammation, delay preterm delivery, and decrease the incidence of early-onset sepsis [54‒57]. All these actions may lead to a decreased the risk of BPD. However, it should be noted that the main objective of antenatal antibiotics is the prevention of early-onset sepsis by group B Streptococcus or gram-negative bacteria such as E. coli [8]. The antibiotics commonly used for this purpose are not effective against Ureaplasma. Ureaplasma is recognized as a major pathogen in prenatal infection triggering preterm birth and is frequently claimed to be involved in the pathogenesis of BPD [64]. Nevertheless, there is no current evidence on a positive effect of treating maternal Ureaplasma infection/colonization on the development of BPD [64].

A relevant limitation of our meta-analysis is that a significant number of studies were conducted more than 25 years ago and used definitions of BPD based on the 28-day time point (i.e., BPD28). These definitions predict pulmonary outcome worse than those based on oxygen requirements at 36-week PMA [67]. In addition, there was confusion because some studies used oxygen need at day 28 of life as the defining criterion, while others used 28 consecutive days of supplemental oxygen. Most infants with the so-called “new BPD” may have intervals during the first weeks after birth when they do not need supplemental oxygen [67]. Therefore, the use of different criteria (for 28 days vs. at 28 days) has resulted in marked variability in the incidence of BPD between centers when applied to the same population of preterm infants [67]. Nevertheless, definitions based on the 36-week PMA time point have been also widely criticized for failing to capture the complexity of the various BPD phenotypes [1].

In conclusion, the currently available evidence suggests a lack of association between antenatal antibiotics and risk of developing BPD. However, our results should not be interpreted as an argument for widespread use of antibiotics in the setting of preterm delivery. Reducing the use of perinatal antibiotics in very preterm infants is challenged by the relatively high incidence of both the risk factors for early-onset sepsis as well as sepsis itself, the difficulty in differentiating between clinical instability due to prematurity and that due to sepsis, and the potentially disastrous consequences of delayed initiation of antibiotic treatment [65]. Despite these difficulties, efforts for rational antibiotic use are increasing, and a growing number of NICUs have developed antibiotic stewardship programs [65]. These programs are essential to prevent antibiotic resistance and the short- and long-term adverse health effects of unnecessary antibiotics [8‒11].

We would like to thank Dr. Andrews for providing us with additional information from his study.

An ethics statement is not applicable because this study is based exclusively on published literature.

The authors have no conflicts of interest to declare.

This research did not receive any specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

E.V. conceived and designed the study, with input from the other authors. K.V.M. and T.M.H. executed the literature search, screened and reviewed the search results, abstracted the data, and assessed the quality of the included studies. E.V. and E.v.W.-K. checked data extraction for accuracy and completeness. F.B. designed and conducted the Bayesian analysis with input from E.V. All authors contributed to the interpretation of analyses, reviewed the manuscript, provided important intellectual content, and have read and agreed to the published version of the manuscript. K.V.M. and E.V. made the figures and tables, drafted the manuscript with input from the other authors, and took responsibility for the article as a whole.

Data and materials can be found in the supplementary materials. Additional information can be requested from the corresponding author.

1.
Thébaud
B
,
Goss
KN
,
Laughon
M
,
Whitsett
JA
,
Abman
SH
,
Steinhorn
RH
, et al
.
Bronchopulmonary dysplasia
.
Nat Rev Dis Primers
.
2019
;
5
(
1
):
78
. .
2.
Abman
SH
,
Bancalari
E
,
Jobe
A
.
The evolution of bronchopulmonary dysplasia after 50 years
.
Am J Respir Crit Care Med
.
2017
;
195
(
4
):
421
4
. .
3.
Villamor-Martinez
E
,
Álvarez-Fuente
M
,
Ghazi
AMT
,
Degraeuwe
P
,
Zimmermann
LJI
,
Kramer
BW
, et al
.
Association of chorioamnionitis with bronchopulmonary dysplasia among preterm infants: a systematic review, meta-analysis, and metaregression
.
JAMA Netw Open
.
2019
;
2
(
11
):
e1914611
. .
4.
Pierro
M
,
Villamor-Martinez
E
,
van Westering-Kroon
E
,
Alvarez-Fuente
M
,
Abman
SH
,
Villamor
E
.
Association of the dysfunctional placentation endotype of prematurity with bronchopulmonary dysplasia: a systematic review, meta-analysis and meta-regression
.
Thorax
.
2022
;
77
(
3
):
268
75
. .
5.
González-Luis
GE
,
van Westering-Kroon
E
,
Villamor-Martinez
E
,
Huizing
MJ
,
Kilani
MA
,
Kramer
BW
, et al
.
Tobacco smoking during pregnancy is associated with increased risk of moderate/severe bronchopulmonary dysplasia: a systematic review and meta-analysis
.
Front Pediatr
.
2020
;
8
:
160
. .
6.
Willis
KA
,
Siefker
DT
,
Aziz
MM
,
White
CT
,
Mussarat
N
,
Gomes
CK
, et al
.
Perinatal maternal antibiotic exposure augments lung injury in offspring in experimental bronchopulmonary dysplasia
.
Am J Physiol Lung Cell Mol Physiol
.
2020
;
318
(
2
):
L407
18
. .
7.
Casado
F
,
Morty
RE
.
The emergence of preclinical studies on the role of the microbiome in lung development and experimental animal models of bronchopulmonary dysplasia
.
Am J Physiol Lung Cell Mol Physiol
.
2020
;
318
(
2
):
L402
4
. .
8.
Sola
A
.
Abuse of antibiotics in perinatology: negative impact for health and the economy
.
Neoreviews
.
2020
;
21
(
8
):
e559
70
. .
9.
Pronovost
GN
,
Hsiao
EY
.
Perinatal interactions between the microbiome, immunity, and neurodevelopment
.
Immunity
.
2019
;
50
(
1
):
18
36
. .
10.
Dang
AT
,
Marsland
BJ
.
Microbes, metabolites, and the gut-lung axis
.
Mucosal Immunol
.
2019
;
12
(
4
):
843
50
. .
11.
Tirone
C
,
Pezza
L
,
Paladini
A
,
Tana
M
,
Aurilia
C
,
Lio
A
, et al
.
Gut and lung microbiota in preterm infants: immunological modulation and implication in neonatal outcomes
.
Front Immunol
.
2019
;
10
:
2910
. .
12.
Cantey
JB
,
Huffman
LW
,
Subramanian
A
,
Marshall
AS
,
Ballard
AR
,
Lefevre
C
, et al
.
Antibiotic exposure and risk for death or bronchopulmonary dysplasia in very low birth weight infants
.
J Pediatr
.
2017
;
181
:
289
93.e1
. .
13.
Ting
JY
,
Roberts
A
,
Sherlock
R
,
Ojah
C
,
Cieslak
Z
,
Dunn
M
, et al
.
Duration of initial empirical antibiotic therapy and outcomes in very low birth weight infants
.
Pediatrics
.
2019
;
143
(
3
):
e20182286
. .
14.
Pammi
M
,
Lal
CV
,
Wagner
BD
,
Mourani
PM
,
Lohmann
P
,
Luna
RA
, et al
.
Airway microbiome and development of bronchopulmonary dysplasia in preterm infants: a systematic review
.
J Pediatr
.
2019
;
204
:
126
33.e2
. .
15.
Gronau
QF
,
Heck
DW
,
Berkhout
SW
,
Haaf
JM
,
Wagenmakers
E-J
.
A primer on bayesian model-averaged meta-analysis
.
Adv Methods Pract Psychol Sci
.
2021
;
4
(
3
):
251524592110312
. .
16.
Bartoš
F
,
Gronau
QF
,
Timmers
B
,
Otte
WM
,
Ly
A
,
Wagenmakers
EJ
.
Bayesian model-averaged meta-analysis in medicine
.
Stat Med
.
2021
;
40
(
30
):
6743
61
. .
17.
Villamor
E
,
van Westering-Kroon
E
,
Gonzalez-Luis
GE
,
Bartoš
F
,
Abman
SH
,
Huizing
MJ
.
Patent ductus arteriosus and bronchopulmonary dysplasia-associated pulmonary hypertension: a Bayesian meta-analysis
.
JAMA Netw Open
.
2023
;
6
(
11
):
e2345299
. .
18.
Lee
M
,
Wagenmakers
E-J
.
Bayesian data analysis for cognitive science: a practical course
.
New York
:
Cambridge University Press
;
2013
. Available from: https://bayesmodels.com/.
19.
Andrews
WW
,
Sibai
BM
,
Thom
EA
,
Dudley
D
,
Ernest
JM
,
McNellis
D
, et al
.
Randomized clinical trial of metronidazole plus erythromycin to prevent spontaneous preterm delivery in fetal fibronectin-positive women
.
Obstet Gynecol
.
2003
;
101
(
5 Pt 1
):
847
55
. .
20.
Cox
SM
,
Leveno
KJ
,
Sherman
ML
,
Travis
L
,
De Plama
R
.
Ruptured membranes at 24 to 29 weeks: a randomized double blind trial of antimicrobials versus placebo
.
Am J Obstet Gynecol
.
1995
;
172
:
412
.
21.
Kenyon
SL
,
Taylor
DJ
,
Tarnow-Mordi
W
;
ORACLE Collaborative Group
.
Broad-spectrum antibiotics for preterm, prelabour rupture of fetal membranes: the ORACLE I randomised trial. ORACLE Collaborative Group
.
Lancet
.
2001
;
357
(
9261
):
979
88
. .
22.
Kenyon
SL
,
Taylor
DJ
,
Tarnow-Mordi
W
,
ORACLE Collaborative Group
.
Broad-spectrum antibiotics for spontaneous preterm labour: the ORACLE II randomised trial. ORACLE Collaborative Group
.
Lancet
.
2001
;
357
(
9261
):
989
94
. .
23.
Mercer
BM
,
Miodovnik
M
,
Thurnau
GR
,
Goldenberg
RL
,
Das
AF
,
Ramsey
RD
.
Antibiotic therapy for reduction of infant morbidity after preterm premature rupture of the membranes. A randomized controlled trial. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network
.
JAMA
.
1997
;
278
(
12
):
989
95
. .
24.
Romero
R
,
Sibai
B
,
Caritis
S
,
Paul
R
,
Depp
R
,
Rosen
M
, et al
.
Antibiotic treatment of preterm labor with intact membranes: a multicenter, randomized, double-blinded, placebo-controlled trial
.
Am J Obstet Gynecol
.
1993
;
169
(
4
):
764
74
. .
25.
Baker
CD
,
Balasubramaniam
V
,
Mourani
PM
,
Sontag
MK
,
Black
CP
,
Ryan
SL
, et al
.
Cord blood angiogenic progenitor cells are decreased in bronchopulmonary dysplasia
.
Eur Respir J
.
2012
;
40
(
6
):
1516
22
. .
26.
Baud
O
,
Zupan
V
,
Lacaze-Masmonteil
T
,
Audibert
F
,
Shojaei
T
,
Thebaud
B
, et al
.
The relationships between antenatal management, the cause of delivery and neonatal outcome in a large cohort of very preterm singleton infants
.
BJOG
.
2000
;
107
(
7
):
877
84
. .
27.
Botet
F
,
Figueras-Aloy
J
,
Miracle-Echegoyen
X
,
Rodríguez-Miguélez
JM
,
Salvia-Roiges
MD
,
Carbonell-Estrany
X
.
Trends in survival among extremely-low-birth-weight infants (less than 1,000 g) without significant bronchopulmonary dysplasia
.
BMC Pediatr
.
2012
;
12
:
63
. .
28.
Cullen
A
,
Van Marter
LJ
,
Allred
EN
,
Moore
M
,
Parad
RB
,
Sunday
ME
.
Urine bombesin-like peptide elevation precedes clinical evidence of bronchopulmonary dysplasia
.
Am J Respir Crit Care Med
.
2002
;
165
(
8
):
1093
7
. .
29.
Dessardo
NS
,
Dessardo
S
,
Mustać
E
,
Banac
S
,
Petrović
O
,
Peter
B
.
Chronic lung disease of prematurity and early childhood wheezing: is foetal inflammatory response syndrome to blame
.
Early Hum Dev
.
2014
;
90
(
9
):
493
9
. .
30.
Duan
J
,
Kong
X
,
Li
Q
,
Hua
S
,
Zhang
S
,
Zhang
X
, et al
.
Association between anemia and bronchopulmonary dysplasia in preterm infants
.
Sci Rep
.
2016
;
6
:
22717
. .
31.
Erdemir
G
,
Kultursay
N
,
Calkavur
S
,
Zekioğlu
O
,
Koroglu
OA
,
Cakmak
B
, et al
.
Histological chorioamnionitis: effects on premature delivery and neonatal prognosis
.
Pediatr Neonatol
.
2013
;
54
(
4
):
267
74
. .
32.
Eriksson
L
,
Haglund
B
,
Odlind
V
,
Altman
M
,
Kieler
H
.
Prenatal inflammatory risk factors for development of bronchopulmonary dysplasia
.
Pediatr Pulmonol
.
2014
;
49
(
7
):
665
72
. .
33.
EXPRESS Group
.
Incidence of and risk factors for neonatal morbidity after active perinatal care: extremely preterm infants study in Sweden (EXPRESS)
.
Acta Paediatr
.
2010
;
99
(
7
):
978
92
. .
34.
Hansen
AR
,
Barnés
CM
,
Folkman
J
,
McElrath
TF
.
Maternal preeclampsia predicts the development of bronchopulmonary dysplasia
.
J Pediatr
.
2010
;
156
(
4
):
532
6
. .
35.
Kandasamy
J
,
Roane
C
,
Szalai
A
,
Ambalavanan
N
.
Serum eotaxin-1 is increased in extremely-low-birth-weight infants with bronchopulmonary dysplasia or death
.
Pediatr Res
.
2015
;
78
(
5
):
498
504
. .
36.
Koroglu
OA
,
Yalaz
M
,
Levent
E
,
Akisu
M
,
Kültürsay
N
.
Cardiovascular consequences of bronchopulmonary dysplasia in prematurely born preschool children
.
Neonatology
.
2013
;
104
(
4
):
283
9
. .
37.
Mao
X
,
Qiu
J
,
Zhao
L
,
Xu
J
,
Yin
J
,
Yang
Y
, et al
.
Vitamin D and IL-10 deficiency in preterm neonates with bronchopulmonary dysplasia
.
Front Pediatr
.
2018
;
6
:
246
. .
38.
Marom
R
,
Mimouni
FB
,
Lubetzky
R
,
Deutsch
V
,
Mandel
D
.
Absolute nucleated red blood cells counts do not predict the development of bronchopulmonary dysplasia
.
J Matern Fetal Neonatal Med
.
2016
;
29
(
10
):
1603
6
. .
39.
McGowan
EC
,
Kostadinov
S
,
McLean
K
,
Gotsch
F
,
Venturini
D
,
Romero
R
, et al
.
Placental IL-10 dysregulation and association with bronchopulmonary dysplasia risk
.
Pediatr Res
.
2009
;
66
(
4
):
455
60
. .
40.
Lodha
A
,
Sauvé
R
,
Bhandari
V
,
Tang
S
,
Christianson
H
,
Bhandari
A
, et al
.
Need for supplemental oxygen at discharge in infants with bronchopulmonary dysplasia is not associated with worse neurodevelopmental outcomes at 3 years corrected age
.
PLoS One
.
2014
;
9
(
3
):
e90843
. .
41.
Novitsky
A
,
Tuttle
D
,
Locke
RG
,
Saiman
L
,
Mackley
A
,
Paul
DA
.
Prolonged early antibiotic use and bronchopulmonary dysplasia in very low birth weight infants
.
Am J Perinatol
.
2015
;
32
(
1
):
43
8
. .
42.
Oh
W
,
Poindexter
BB
,
Perritt
R
,
Lemons
JA
,
Bauer
CR
,
Ehrenkranz
RA
, et al
.
Association between fluid intake and weight loss during the first ten days of life and risk of bronchopulmonary dysplasia in extremely low birth weight infants
.
J Pediatr
.
2005
;
147
(
6
):
786
90
. .
43.
Ohyama
M
,
Itani
Y
,
Yamanaka
M
,
Goto
A
,
Kato
K
,
Ijiri
R
, et al
.
Re-evaluation of chorioamnionitis and funisitis with a special reference to subacute chorioamnionitis
.
Hum Pathol
.
2002
;
33
(
2
):
183
90
. .
44.
Prendergast
M
,
May
C
,
Broughton
S
,
Pollina
E
,
Milner
AD
,
Rafferty
GF
, et al
.
Chorioamnionitis, lung function and bronchopulmonary dysplasia in prematurely born infants
.
Arch Dis Child Fetal Neonatal Ed
.
2011
;
96
(
4
):
F270
4
. .
45.
Sheth
S
,
Goto
L
,
Bhandari
V
,
Abraham
B
,
Mowes
A
.
Factors associated with development of early and late pulmonary hypertension in preterm infants with bronchopulmonary dysplasia
.
J Perinatol
.
2020
;
40
(
1
):
138
48
. .
46.
Slaughter
JL
,
Pakrashi
T
,
Jones
DE
,
South
AP
,
Shah
TA
.
Echocardiographic detection of pulmonary hypertension in extremely low birth weight infants with bronchopulmonary dysplasia requiring prolonged positive pressure ventilation
.
J Perinatol
.
2011
;
31
(
10
):
635
40
. .
47.
Stichel
H
,
Bäckström
E
,
Hafström
O
,
Nilsson
S
,
Lappalainen
U
,
Bry
K
.
Inflammatory cytokines in gastric fluid at birth and the development of bronchopulmonary dysplasia
.
Acta Paediatr
.
2011
;
100
(
9
):
1206
12
. .
48.
Streubel
AH
,
Donohue
PK
,
Aucott
SW
.
The epidemiology of atypical chronic lung disease in extremely low birth weight infants
.
J Perinatol
.
2008
;
28
(
2
):
141
8
. .
49.
Todd
DA
,
Jana
A
,
John
E
.
Chronic oxygen dependency in infants born at 24-32 weeks’ gestation: the role of antenatal and neonatal factors
.
J Paediatr Child Health
.
1997
;
33
(
5
):
402
7
. .
50.
Uberos
J
,
Jimenez-Montilla
S
,
Molina-Oya
M
,
García-Serrano
JL
.
Early energy restriction in premature infants and bronchopulmonary dysplasia: a cohort study
.
Br J Nutr
.
2020
;
123
(
9
):
1024
31
. .
51.
Van Marter
LJ
,
Dammann
O
,
Allred
EN
,
Leviton
A
,
Pagano
M
,
Moore
M
, et al
.
Chorioamnionitis, mechanical ventilation, and postnatal sepsis as modulators of chronic lung disease in preterm infants
.
J Pediatr
.
2002
;
140
(
2
):
171
6
. .
52.
Cohen
HW
.
P values: use and misuse in medical literature
.
Am J Hypertens
.
2011
;
24
(
1
):
18
23
. .
53.
Shrier
I
,
Boivin
JF
,
Steele
RJ
,
Platt
RW
,
Furlan
A
,
Kakuma
R
, et al
.
Should meta-analyses of interventions include observational studies in addition to randomized controlled trials? A critical examination of underlying principles
.
Am J Epidemiol
.
2007
;
166
(
10
):
1203
9
. .
54.
Hutzal
CE
,
Boyle
EM
,
Kenyon
SL
,
Nash
JV
,
Winsor
S
,
Taylor
DJ
, et al
.
Use of antibiotics for the treatment of preterm parturition and prevention of neonatal morbidity: a metaanalysis
.
Am J Obstet Gynecol
.
2008
;
199
(
6
):
620.e1
6208
. .
55.
Mercer
B
.
Antibiotics in the management of PROM and preterm labor
.
Obstet Gynecol Clin North Am
.
2012
;
39
(
1
):
65
76
. .
56.
Chatzakis
C
,
Papatheodorou
S
,
Sarafidis
K
,
Dinas
K
,
Makrydimas
G
,
Sotiriadis
A
.
Effect on perinatal outcome of prophylactic antibiotics in preterm prelabor rupture of membranes: network meta-analysis of randomized controlled trials
.
Ultrasound Obstet Gynecol
.
2020
;
55
(
1
):
20
31
. .
57.
Lin
LL
,
Hung
JN
,
Shiu
SI
,
Su
YH
,
Chen
WC
,
Tseng
JJ
.
Efficacy of prophylactic antibiotics for preterm premature rupture of membranes: a systematic review and network meta-analysis
.
Am J Obstet Gynecol MFM
.
2023
;
5
(
7
):
100978
. .
58.
Keysers
C
,
Gazzola
V
,
Wagenmakers
EJ
.
Using Bayes factor hypothesis testing in neuroscience to establish evidence of absence
.
Nat Neurosci
.
2020
;
23
(
7
):
788
99
. .
59.
Brydges
CR
,
Bielak
AAM
.
A Bayesian analysis of evidence in support of the null hypothesis in gerontological psychology (or lack thereof)
.
J Gerontol B Psychol Sci Soc Sci
.
2020
;
75
(
1
):
58
66
. .
60.
Sjoding
MW
,
Luo
K
,
Miller
MA
,
Iwashyna
TJ
.
When do confounding by indication and inadequate risk adjustment bias critical care studies? A simulation study
.
Crit Care
.
2015
;
19
(
1
):
195
. .
61.
Hundscheid
TM
,
Villamor-Martinez
E
,
Villamor
E
.
Association between endotype of prematurity and mortality: a systematic review, meta-analysis, and meta-regression
.
Neonatology
.
2023
;
120
(
4
):
407
16
. .
62.
Pierro
M
,
Van Mechelen
K
,
van Westering-Kroon
E
,
Villamor-Martínez
E
,
Villamor
E
.
Endotypes of prematurity and phenotypes of bronchopulmonary dysplasia: toward personalized neonatology
.
J Pers Med
.
2022
;
12
(
5
):
687
. .
63.
Parsons
A
,
Netsanet
A
,
Seedorf
G
,
Abman
SH
,
Taglauer
ES
.
Understanding the role of placental pathophysiology in the development of bronchopulmonary dysplasia
.
Am J Physiol Lung Cell Mol Physiol
.
2022
;
323
(
6
):
L651
58
. .
64.
Van Mechelen
K
,
van Westering-Kroon
E
,
Hütten
M
,
Mahieu
L
,
Villamor
E
.
Placing Ureaplasma within the context of bronchopulmonary dysplasia endotypes and phenotypes
.
Children
.
2023
;
10
(
2
):
256
. .
65.
Rajar
P
,
Saugstad
OD
,
Berild
D
,
Dutta
A
,
Greisen
G
,
Lausten-Thomsen
U
, et al
.
Antibiotic stewardship in premature infants: a systematic review
.
Neonatology
.
2020
;
117
(
6
):
673
86
. .
66.
Flannery
DD
,
Ross
RK
,
Mukhopadhyay
S
,
Tribble
AC
,
Puopolo
KM
,
Gerber
JS
.
Temporal trends and center variation in early antibiotic use among premature infants
.
JAMA Netw Open
.
2018
;
1
(
1
):
e180164
. .
67.
Ibrahim
J
,
Bhandari
V
.
The definition of bronchopulmonary dysplasia: an evolving dilemma
.
Pediatr Res
.
2018
;
84
(
5
):
586
8
. .