Impact of maternal antibiotics on neonatal disease
Necrotizing enterocolitis (NEC) and late onset sepsis (TIS) cause significant morbidity and mortality among infants born prematurely in neonatal intensive care units (NICUs). Among children born at < 32 weeks gestational age, 7% develop NEC, in which case the mortality is ~ 30%.
SIT occurs in up to one-third of infants <32 weeks gestational age, with most preterm births experiencing the greatest risk. The high use of empirical post-natal antibiotics is an epidemiological correlate of both NEC and SIT.
Cotten et al. studied a national cohort of 5693 very low birth weight infants and found that infants who received prolonged initial antibiotic therapy were at increased risk of developing NEC or dying.
Subsequently, Kuppala et al. studied 365 preterm infants in Cincinnati, Ohio, and reported that initial prolonged use of empirical antibiotics was associated with an increased risk of NEC, SIT or death.
In a cohort of 328 preterm infants in Saudi Arabia, Abdel Ghany and Ali also reported that each day of treatment with empirical antibiotics was associated with an increased risk of death and a combined outcome of NEC or death.
Greenwood et al. examined the impact of high antibiotic use on the microbiome and found that children who received early antibiotics experienced an increase in Enterobacteriaceae as well as an increased risk of NEC, sepsis or death.
However, childhood exposure to antibiotics can begin prenatally as a result of maternal exposure to antibiotics, and this prenatal exposure could potentially influence neonatal disease. Bizzarro et al. noted an increase in ampicillin-resistant Escherichia coli infections in mothers exposed to ampicillin. Similarly, Didier et al. found that maternal exposure led to an increase in amoxicillin-resistant organisms.
However, the effects of maternal exposure to antibiotics before delivery on neonatal outcomes have not been well defined. With this concern in mind, the authors analyzed an existing cohort of preterm infants to test the hypothesis that infant exposure to prenatal antibiotics is associated with an increased incidence of NEC, SIT, or death.
A secondary analysis of an existing cohort of 580 infants < 32 weeks gestational age was conducted. All of the children in this study were part of a prospective study sponsored by the National Institutes of Health on preterm microbioma and risk of NEC, sepsis, and death. The infants in the study were enrolled in 2 Level III NICUs in Cincinnati, Ohio, and 1 Level III NICU in Birmingham, Alabama, from 2009 to 2012.
Enrollment criteria included delivery < 32 weeks gestation, absence of congenital anomalies, and NEC-free survival in the first week of life. One patient was excluded due to unknown maternal antibiotic status. Data collection followed the protocol of the Eunice Kennedy Shriver National Institute of Child Health and the Human Development Neonatal Research Network registry.
Neonates were enrolled immediately after delivery and were followed until discharge, referral, 120 days after delivery, or death. Maternal and infant data were summarized from medical records. The institutional review boards of the three participating hospitals approved the study.
Maternal exposure to antibiotics was defined as antibiotic treatment within 72 hours prior to delivery. Indications for maternal exposure to antibiotics included cesarean delivery, prophylaxis for Group B Streptoccoccus (GBS), premature rupture of membranes, chorioamnionitis and prolonged pregnancy (latency).
Maternal exposure to antibiotics did not include antibiotics given after the time of delivery or antibiotics initiated by surgeons during (but not before) the cesarean delivery.
Early empirical neonatal exposure to antibiotics was defined as antibiotic treatment initiated within the first postnatal day without culture-identified infection. The duration of early antibiotic therapy was defined as the total number of continuous days of antibiotic administration with sterile culture results.
The empirical childhood antibiotics used in this cohort were ampicillin and gentamicin, with empirical exposure to childhood antibiotics defined as low (< 5 days) or high (> 5 days).3,5,6,8 NEC was defined by Bell’s criteria as stage II or III. Spontaneous bowel perforation was excluded by including only cases of NEC that occurred after 7 days of age.
SIT was defined as a positive blood culture, cerebrospinal fluid culture, urine culture, or sterile site culture after the third day postnatal. Patients diagnosed with NEC or SIT or who died after day 7 of their NICU stay were considered positive for the combined result, NEC, sepsis or death (called MSD).
Differences in clinical characteristics were assessed through Fisher’s exact test for categorical variables and ANOVA, t-tests or Kruskal-Wallis for continuous variables, as appropriate.
Maternal and neonatal baseline characteristics were compared in all groups based on maternal antibiotic exposure alone and infant antibiotic exposure alone, and in the 4 combined maternal and infant antibiotic exposure groups.
Individual outcomes of NEC, SIT, and death and the occurrence of any of these outcomes were also compared between the groups as noted above. Multivariate logistic regression models were used to assess independent associations between maternal and infant antibiotic exposure and these outcomes. A set of clinical predictors of NEC, SIT and death were defined a priori and examined in the models.
Gestational age, sex, race, mode of delivery, high exposure to human milk, maternal chorioamnionitis, cesarean delivery, preterm rupture of membranes, maternal pre-eclampsia, prenatal steroids, parity, multiple birth and hospital location were all included in the initial analysis.
High exposure to human milk was defined as feeding with human milk for more than 75% of the days from birth to 30 days of life or the occurrence of NEC or death, whichever occurred first.
Modeling was done by using reverse elimination, with factors systematically eliminated in order of highest (non-significant) P values. Factors not included as potential confounders were measures of postnatal infant health status that were considered to be in the same causal chain as (or were possible biomarkers of) the study outcomes (e.g., Apgar scores or respiratory status, which may be indicators of infant health).
The factors that were removed from the models did not significantly affect the association observed between maternal or infant use of antibiotics and study outcomes; the ORs for the impact of maternal antibiotics on study outcomes changed < 4% after the removal of possible covariates (maximum one digit change at the second decimal point).
The multiple logistic regression models reported here were standardized across the results by including a set of independent variables, including prenatal and postnatal antibiotic use and all factors with a P < 0.10 value in any of the final specific outcome models.
In this cohort, maternal administration of antibiotics was associated with a reduced likelihood of NEC and death
We compared the clinical and demographic characteristics of infants and their mothers in relation to the number of maternal antibiotics used.
Of the 580 children included in the study, 362 were from mothers who had received antibiotics in the 72 hours prior to delivery. The birth weights of the children were similar across all maternal antibiotic exposure groups, as were gestational age, infant sex and race.
The median gestational age at delivery in both maternal exposure groups (without antibiotics and ≥ 1 antibiotic) was 28 weeks. Pre-eclampsia was more common among women without antibiotic exposure (46.3% and 24.5% respectively, P < 0.001).
Three hundred and sixty-three children (62.5%) were born by caesarean section, which was more common among mothers who were not exposed to antibiotics (P < 0.001).
Prenatal corticosteroids were used more regularly in mothers who had exposure to antibiotics (P < 0.001). Clinical chorioamnionitis was higher among mothers who received antibiotics (P < 0.001). Human milk feeding did not differ because of maternal exposure to antibiotics. Overall, 135 (23.2%) children in the study received ≥ 5 days of postnatal empiric antibiotics. Infants born with prenatal exposure to antibiotics were more likely to experience high postnatal empirical exposure to antibiotics compared with infants without prenatal exposure to antibiotics (P = 0.002). Among the 362 children exposed to prenatal antibiotics, 100 (27.5%) had high postnatal exposure to antibiotics, while among the 218 children without prenatal antibiotic exposure, 35 (16.0%) had high postnatal exposure to antibiotics. Of the 580 children in the study, there were 106 twin births and 10 triplet births. There was no difference in prenatal antibiotic exposure between multiple and single births. However, the analysis of maternal antibiotic use was based on the 309 individual mothers who received prenatal antibiotics (rather than the 362 infants in the study who also received prenatal antibiotics through their mothers). In the 72 hours prior to delivery, a total of 534 antibiotic courses were given, which could be grouped into 55 different combinations of 14 different antibiotics. Sixty percent of mothers on antibiotics received > 1 type.
Of the 580 infants studied, 44 (7.5%) developed NEC, 65 (11.1%) developed SIT, 56 (9.6%) died before discharge from the NICU, and 124 (21%) experienced at least 1 of (combined result) NEC, SIT or death.
Death was attributable to NEC in 19 cases (34%) and SIT in 12 cases (21%). The bacterial organisms causing SIT were coagulase negative Staphylococcus (48%), Klebsiella (10%), E. coli (8%), Enterococcus (6%), Group B Streptococcus (5%), methicillin-resistant Staphylococcus aureus (5%) and S. aureus (4%), Streptococcus viridans (2%) and 1% each for Pseudomonas, Enterobacter and Bacillus species.
Fungal species accounted for 1% of SIT cases. Neither probiotic therapy nor fluconazole prophylaxis was routinely used in the study NICUs when this cohort was recruited.
To adjust for possible confounding factors, disease rates were corrected for gestational age using infants who did not receive empirical postnatal or prenatal antibiotics as the comparison group. NEC was less frequently diagnosed among infants with prenatal antibiotic exposure (3.2%) compared to infants without prenatal antibiotic exposure (10.5%, P < 0.001). Mortality rates followed a similar pattern to that of NEC (P = 0.002). SIT was not significantly different in infants with prenatal exposure (8.2%) compared to infants without prenatal exposure to antibiotics (6.8%, P = 0.159). On the other hand, the combined outcome of NEC, SIT or death was more common in infants without prenatal antibiotic exposure compared to those with prenatal antibiotic exposure (P = 0.012). Because maternal indication for antibiotic use may be a critical factor, we assessed neonatal outcomes in relation to maternal circumstances of use. Among the 580 infants in the study, 33 (5.6%) were born to mothers who received antibiotics for cesarean section prophylaxis only; 83 (14.3%) were born to mothers who were given antibiotics for GBS prophylaxis only; 107 (18.4%) were born to mothers who received antibiotics for premature rupture of membranes only; and the remaining 139 (23.9%) were born to mothers who received antibiotics for a combination of indications. Comparing results based on these maternal indication groups, NEC was more common in infants without exposure to maternal antibiotics (11.4%) and less common in infants whose mothers received antibiotics for GBS prophylaxis only (3.6%), but NEC rates were low in all infants exposed to prenatal antibiotics regardless of maternal indication. Similar to NEC, death occurred more frequently in infants without maternal exposure to antibiotics (11.9%) and was lower among all infants exposed prenatally regardless of maternal indication. In contrast, SIT was less frequent in children without maternal exposure to antibiotics (7.3%), with higher SIT rates regardless of maternal indication for antibiotics. A multiple logistic regression model was performed to adjust for possible confounding factors. Maternal use of antibiotics (yes/no) and high empirical use of infant antibiotics (yes/no) were included in all models, regardless of their significance. Repeated data analysis using only single birth patients was compared with analysis of the whole study population, and results were not affected by inclusion of the whole cohort. Overall, the authors found that maternal antibiotics were associated with better outcomes. In these multiple logistic regression models, high empirical postnatal exposure of infants to antibiotics was not significantly associated with risk of NEC or SIT. However, according to previous studies,5,6 high receipt of empirical infant antibiotics was associated with a significantly increased risk of death (OR 3.18, P = 0.002). The combined outcome of NEC, SIT or death was not significantly different in infants with high empirical antibiotic exposure (OR 1.44, P = 0.17). The combination of maternal antibiotic exposure and high infant empirical antibiotic exposure was then examined. In infants with low empirical antibiotic exposure, maternal antibiotic exposure remained protective for NEC (OR = 0.24, P = 0.001). Maternal antibiotics in combination with high infant exposure to antibiotics were not significantly associated with a reduction in NEC (OR = 0.36, P = 0.09). Maternal antibiotic exposure was not a risk factor for the development of SIT in infant groups with high or low antibiotic exposure (OR 1.06, P = 0.9 and OR 2.13, P = 0.07, respectively). Similar to NEC, maternal receipt of antibiotics was protective for death in the group of children with high exposure (OR 0.33, P = 0.048) and for death and MSE in the group with low exposure (OR 0.19, P = 0.003 and OR 0.53, P = 0.04) to postnatal antibiotics. Maternal antibiotics in combination with high postnatal antibiotic exposure were not statistically different for MSM (OR 0.48, P = 0.12). Discussion Contrary to the original hypothesis, the authors found that prenatal exposure to antibiotics within 72 hours before delivery was significantly associated with triple protection against NEC and similar protection against death, based on multiple logistic regression models that controlled for multiple confounders. In contrast, prenatal exposure to antibiotics was associated with a 1.6-fold increased risk of SIT in the same infants. Although the association with SIT was not statistically significant, the potentially disparate effect of maternal antibiotics on these outcomes is intriguing. The use of prenatal antibiotics remained significantly associated with protection against combined NEC, SIT, or death metrics. Prepartum antibiotic prophylaxis has become commonplace as therapy for a number of potential illnesses. Some estimates report that > 40% of all pregnant women receive antibiotics before delivery for GBS prophylaxis or by cesarean section. When other indications for antibiotics are considered, such as prevention of preterm birth and chorioamnionitis, the percentage of mothers receiving antibiotics is even higher.
In a 2015 study from Canada, the authors noted that 39% of 449 full-term mothers had received intrapartum antibiotics, indirectly exposing their infants to antibiotics. In the present study, maternal indications for antibiotic therapy included but were not limited to prophylaxis of GBS, cesarean section, and premature rupture of membranes.
In this study, 62.3% of mothers received antibiotics before delivery, most commonly ampicillin (37.6%) and azithromycin (26.4%). The prevalence of maternal antibiotic use (62.3%) found in this study may be higher than previously reported (40%) due to the different study populations (preterm vs. full term).
The authors’ finding regarding prenatal antibiotic exposure and protection against NEC is in stark contrast to the findings reported by Weintraub et al. In that retrospective case-control study, it was observed that in 97 crossover pairs of preterm infants, there was more than twice the likelihood of NEC among infants with prenatal ampicillin exposure.
A Cochrane review of prenatal antibiotic administration for preterm rupture of membranes noted that among the various antibiotics given, amoxicillin-clavulanic was the only antibiotic that increased the neonatal risk for NEC.
The findings of the present study are also contrary to expectations because of an increased risk of NEC reported in previous studies associated with empirical postnatal antibiotic use.5-7,18,19 Speculation that prenatal antibiotics would have a similar association with neonatal disease as postnatal antibiotics appears to be incorrect.
The trend the authors observed toward an increased risk of SIT associated with prenatal exposure to antibiotics was not statistically significant. However, they considered the findings to be of concern, given previous reports.
Lin et al. noted that in one hospital, after the initiation of a standardized treatment protocol of intrapartum antibiotic prophylaxis for GBS, maternal antibiotic use increased from 40% to 90% over 4 years in mothers with GBS-positive, during which time early-onset sepsis associated with GBS in neonates decreased (45% to 20%) and early-onset sepsis associated with E. Coli increased proportionately (41%-70%).
Similarly, in a 2002 study of > 5400 very low birth weight infants by the Eunice Kennedy Shriver National Institute of Child Health and Human Development, researchers reported a change in the pathogens causing early-onset sepsis from predominantly gram-positive to predominantly gram-negative organisms, which correlated with increased maternal antibiotic prophylaxis during labor and delivery.
A retrospective review of early-onset sepsis, SIT, and ampicillin resistance showed that increased use of intrapartum antibiotics was an independent risk factor for early-onset sepsis associated with ampicillin-resistant E. coli and for increased SIT associated with E. coli.9 Similarly, Didier et al. found that maternal exposure to antibiotics was significantly associated with the risk of amoxicillin-resistant E. coli infections.
Coagulase negative Staphylococcus was the causative organism in almost half of the SIT cases. Although this organism is generally considered to have low pathogenic potential, this is not always the case in children born extremely prematurely. In addition, in this study 2 deaths attributable to SIT were caused by coagulase negative Staphylococcus.
The use of antibiotics would be expected to alter the maternal microbiome, thus changing the infant’s initial exposure to the microbial environment, which could lead to an abnormal sequence of microbial colonization in the child.
Alterations in early microbial transfer and succession could have significant effects, as the maternal microbiome has been shown to promote the development of neonatal immunity.
For example, Deshmukh et al. showed that neonatal mice from females exposed to antibiotics had reduced numbers of intestinal microbes, altered colonization and structure of the microbiota, decreased neutrophils and granulocyte/macrophage progenitor cells in the bone marrow, and decreased interleukin-producing cells in the intestine, resulting in the production of granulocyte colony-stimulating factor.
Relative granulocytopenia contributed to an increased susceptibility of neonatal mice exposed to antibiotics to sepsis from E. coli K1 and Klebsiella pneumoniae.
The transplacental passage of maternal antibiotics may also alter the neonatal microbiome, reducing microbial diversity and creating a preponderance of pathogenic bacteria. It is well established that maternal antibiotics cross the placenta to reach the fetus.
Mazzola et al. showed that infants whose mothers received intrapartum antibiotic prophylaxis against GBS were abundant in Enterobacteriaceae compared to control infants. As previous studies have shown, neonatal microbial dysbiosis is increasingly associated with disease development. The authors previously reported that early dysbiosis strongly predicts the presence of NEC.
Fecal samples from NEC patients were less bacterially diverse, lacked Propionibacterium, and were dominated by Firmicutes or Proteobacteria, compared to controls. Based on a meta-analysis of cohort studies from various sites, Pammi et al. reported a higher relative abundance of Proteobacteria and decreased Firmicutes and Bacteroidetes in NEC patients.
However, NEC and SIT are multifactorial diseases, and risk factors should not be considered in a vacuum. Different microbes or forms of dysbiosis can be distinguished in the etiologies of NEC11,27 and SIT28 that may explain the opposite effects of prenatal antibiotics in NEC and SIT seen in this study.
The strengths of this study were the inclusion of a relatively large number of participants and that the cohort was representative of the high-risk neonatal reference center population in two large U.S. areas of influence.
Because the authors previously reported temporal, regional, and inter-institutional differences in neonatal microbioma in this cohort,29 the findings may not be completely generalizable to other centers for the management of high-risk children. The results of this study are compelling enough to warrant further research.
The main limitation of this study was that no relevant maternal samples were collected; therefore, the simultaneous study of the maternal and infant microbiome was not possible. As noted above, these results may not be generalizable due to geographic and population differences in the microbial environment.
In addition, there were too many combinations of maternal exposure to antibiotics to separate results by the antibiotic administered, so the antibiotic indication was used as a surrogate. Due to limitations of data availability and accessibility, mothers in the exposure group were not excluded if their only exposure to antibiotics was at the time of cesarean section.
Although this was a potential limitation, we completed an analysis of the data as if all mothers who had a caesarean section were exposed to antibiotics and noted that this did not significantly change the findings.
In addition, it was felt that if antibiotics were given only at the time of the cesarean section, they would not have enough time to significantly affect the baby.
Another limitation is that as a retrospective review, the study is at risk of reporting bias. In addition, maternal antibiotics could suppress bacterial growth in cultures causing false negative results and limiting the identification of SIT cases.
Understanding the interactions between risk factors for NEC and SIT may help physicians provide better patient care in the future. For example, if patterns of resistance in the neonatal microbiome are identified based on maternal exposure to antibiotics, knowledge of maternal exposure to antibiotics could help clinicians choose the most appropriate antibiotics as empirical therapy.
Maternal exposure to antibiotics before delivery occurs in a large proportion of mothers; however, there is a paucity of information on the possible effects this exposure might have on the health of the baby. The present study indicates that in this cohort the administration of maternal antibiotics was associated with protective effects, reducing the possibility of NEC, death and the combined outcome of NEC, SIT and/or death.
In contrast, maternal use of antibiotics was found to be associated with a potentially harmful effect where the rate of SIT was higher among exposed children. Research incorporating the concomitant study of maternal and infant microbiota will be needed to understand these disparate results.
Necrotizing enterocolitis and late-onset sepsis cause significant morbidity and mortality among preterm infants, and are subject to high use of empirical post-natal antibiotics. However, infant exposure to antibiotics may begin prenatally as a result of maternal exposure to them, and this prenatal exposure could potentially influence neonatal disease.
The present study indicated that in this cohort maternal administration of antibiotics was associated with a reduced likelihood of NEC and death, but also with a potentially harmful effect with a higher rate of SIT among exposed children.
Studying the interactions of risk factors for NEC and SIT and the role of maternal and infant microbioma in antibiotic resistance may help to better understand the course of preterm infants with these clinical pictures and to optimize maternal use of antibiotics prior to delivery.