Introduction

Necrotizing enterocolitis (NEC) is a gastrointestinal emergency which mainly affects preterm infants. The incidence of NEC varies highly, not only from country to country, but also among different hospitals in the same country [3, 4, 11]. Despite the progress in overall neonatal care, NEC is still a leading cause of mortality and morbidity in preterm infants, especially those with very low birth weight (< 1500 g, VLBW) [14]. NEC is associated with serious complications, including short- and long-term gastrointestinal (GI) morbidities and neurodevelopmental problems [11, 19].

The pathogenesis of NEC is multifactorial, and it has frequently been associated with intestinal microbiota, prematurity and enteral feeding [20]. In clinical neonatology, antibiotic regimens (including indications for treatment, duration of the antibiotic therapy, type of antibiotics for late onset sepsis) are highly variable and are usually initiated empirically in infants where there is a clinical suspicion of sepsis or in the presence of risk factors [17]. The widespread use of broad-spectrum antimicrobial agents in the neonatal intensive care unit (NICU) has many obvious drawbacks. The initiation of antibiotic therapy leads to a perturbation of the neonatal gut microbiome, which may then predispose a preterm infant to NEC [10, 15].

Although numerous studies have been conducted to explore the pathogenesis, prophylaxis and treatment of NEC, only minor reductions in the incidence of NEC have been achieved [2, 9, 13, 18]. The aim of our study was to investigate whether the widespread use of broad-spectrum antibiotics in VLBW infants is an associated risk factor for the development of NEC.

Material and methods

We designed a single-centre, retrospective, 1:2 case-control study of VLBW infants admitted to the NICU of the Coombe Women and Infants University Hospital (Dublin, Ireland) between January 1st, 2012 and December 31st, 2014. The cases were VLBW infants with a confirmed NEC stage ≥ IIA according to modified Bell’s criteria [5]. We intended to match each case for gender, gestational age (± 1 week), birth weight (± 150 g), small for gestational age (SGA) status, mode of delivery and maternal chorioamnionitis to two controls. Infants with clinical or radiological evidence of major congenital anomalies and congenital heart diseases apart from patent ductus arteriosus (PDA) were excluded from the study. For the control group, we excluded infants who died before the postnatal age of onset of NEC in the corresponding case. Data were collected for each case and the controls from birth up to the day before documented onset of NEC in the case and corresponding controls. For example, if the case developed NEC on day of life 10, data for the case and its assigned controls were collected from birth up to day of life 9.

We defined early-onset sepsis (EOS) and late-onset sepsis (LOS) according to National Institute for Health and Care Excellence (NICE) guidelines [6]. Confirmed EOS was defined as a blood culture-positive bacterial infection within the first 3 days of life. LOS was defined as sepsis presenting after 72 h of life. Cumulative days of antibiotic exposure were defined as the total number of days an infant was on any antibiotic prior to diagnosis of NEC.

SGA was defined as weight less than the 10th percentile for a baby’s gestational age, and the growth charts that were used in the study have been adopted as policy by the Department of Health and have been adapted for Ireland from materials originally developed by the WHO and Royal College of Paediatrics and Child Health in the UK. [21]

Mechanical ventilation days were defined as cumulative number of days an infant was intubated and needed mechanical ventilation prior to diagnosis of NEC. Non-invasive oxygen treatment days were defined as the total number of days an infant was on oxygen therapy without using an invasive artificial airway (endotracheal tube).

PDA was diagnosed on the basis of echocardiographic findings and was considered hemodynamically significant in our study if the measured diameter was greater than 2 mm and diagnosed prior to the onset of NEC.

Full feeds were defined as a successful enteral intake of at least 130 mls/kg/day.

The primary outcome of the study was the difference between cases and controls in relation to antibiotic usage prior to NEC diagnosis in the study group (data were collected for each case and the controls from the birth up to the day before documented onset of NEC in each case and corresponding controls). Our secondary aim was to explore possible association between potential risk factors and development of NEC (incidence of proven EOS/LOS, duration of mechanical ventilation, surfactant therapy, oxygen treatment/exposure, significant PDA, use of umbilical catheters, requirement of red cell transfusion and feeding history).

The software used for analysis was SPSS version 22.0 and StatsDirect v.3.1.12 (StatsDirect Ltd., Cheshire, UK). Descriptive statistics were used for all demographic variables of interest, and comparisons between case group and control group were made using frequency distribution and percentage for categorical variables. Mean and standard deviation (SD) were used for parametric numeric data, while non-parametric numeric data were summarised using median and interquartile range (IQR). Univariate analysis was conducted using Pearson’s chi-square test or Fisher exact test for categorical data. Parametric numeric data were evaluated using independent sample t test. When the numeric data were non-parametric, the Mann–Whitney test was used for two-sample comparisons. Our results were summarised using p value and 95% confidence intervals. P values < 0.05 were considered statistically significant.

The research ethics committee of the Coombe Women and Infants University Hospital approved the study (Study No. 2-2015).

Results

A total of 371 VLBW infants were admitted to the NICU of the Coombe Women and Infants University Hospital during the study period. Of these, 33 infants (8%) were diagnosed with NEC, and 22 infants (6%) met the criteria for NEC stage ≥ IIA according to the modified Bell’s criteria. We identified 32 controls who met the criteria for matching. We sought to match each case with two controls according to several variables, including the gender, gestational age, sex, maternal chorioamnionitis, growth parameters and mode of delivery. However, because of strict guidelines which were used in choosing the two controls for each case, a number of cases either could not be matched to any control or could be matched only to one control. Four cases could not be matched to any control and another four cases were matched to only one control. The mean birth weight (±SD) of the NEC cohort was 728 g (± 162) vs. 778 g (± 163) for the control group and mean gestational age at birth (±SD) was 25.3 weeks (± 1.9) of postmenstrual age vs. 25.9 weeks (± 1.7) respectively. Demographic characteristics of both groups are presented in Table 1.

Table 1 Demographic characteristic of the study cohort
Full size table

Of the 22 cases, 14 received the standard feeding regimen (starting at 20–30 mls/kg/day and increased after 24 h by 20–30 mls/kg/day) and four received trophic feeds (up to 20 mls/kg/day started on day 1 and continued at 20 ml/kg/day for 3 days, indicated in the presence of antenatal risk factors). In two infants, enteral feeding was advanced quicker than per the hospital protocol and in two infants, enteral feeding was not started and full TPN was provided. Full enteral feeding was achieved at a median age of 9 days [IQR 7–13]. NEC was diagnosed at a median postnatal age of 16 days [IQR 9–26] (median postconceptional age of 28.2 weeks [IQR 25.3–30.7]). Although the median age at the diagnosis of NEC was lower in the medical cases (13 days) compared to the surgical cases (22 days), there was no statistically significant association between the age of NEC diagnosis and management of NEC (p = 0.6).

The total number of days an infant was on any antibiotic (prior to diagnosis of NEC) was higher in the NEC group with a median of 5 days [IQR 3–10] in comparison to controls which were 3 days [IQR 2–6] (Table 2). Prolonged duration of initial empirical antibiotic exposure was significantly associated with an increased risk of NEC (cases 5 [IQR 3–6] days vs. controls 3 [IQR 0–5] days, p = 0.05) (Table 2). The infants who developed NEC were exposed to a more frequent number of antibiotic courses than those who did not develop NEC (cases 1 [IQR 1–2] course of ATB vs. controls 1 [IQR 0.3–1] course of ATB, p = 0.03) (Table 2).

Table 2 Comparison between cases and controls according to antibiotic exposure (primary outcomes)
Full size table

There were significant differences between cases and controls with respect to the duration of exposure to gentamicin (cases 6 [IQR 0–6] days vs. controls 3 [IQR 3–8] days, p = 0.03) and meropenem (cases 0 [IQR 0–4] days vs. controls 0 [IQR 0–0] days, p = 0.03). However, no significant differences were noticed between the two groups regarding the duration of exposure to other antibiotics.

Although there was no significant association between sepsis and NEC, the infants who had proven early-onset sepsis with a positive culture were removed from the analysis to eliminate the confounding potential effects. Antibiotic exposure was still significantly associated with NEC after removal of these cases (Table 3).

Table 3 Association between antibiotic exposure and necrotizing enterocolitis after removing cases with proven early onset sepsis
Full size table

There were no statistical differences in our secondary outcomes except for red cell transfusion and mortality (Table 4).

Table 4 Secondary outcomes of the study
Full size table

Discussion

The main findings of our study support the hypothesis that antibiotic exposure is a risk factor for the development of NEC in VLBW infants. Various studies have investigated the relationship between antibiotic exposure and NEC. Cotten et al. (2009) examined initial empirical antibiotic practices for 4039 extremely-low-birth-weight (< 1000 g, ELBW) infants between 1998 and 2001. They reported that exposure to prolonged initial empirical antibiotic for ≥ 5 days, despite sterile blood cultures, was associated with an increased risk of developing NEC (OR 1.34, 95% CI 1.04–1.73) [7]. Our results are in agreement with this finding, as in our study, prolonged exposure to initial antibiotics for more than 5 days was associated with a 3.6-fold increase in the risk of a baby developing NEC (OR 3.6; 95%CI 1.13–11.47). Unlike the previous study, we have included in our case-control study the cumulative total number of days of antibiotic exposure prior to the diagnosis of NEC, and this was also significantly associated with an increased probability of developing NEC. The association between the prolonged initial empirical antibiotic treatment (≥ 5 days) and composite outcome of death, LOS, and NEC after day 7 of life was observed by Kuppala et al. [16]. The association between the prolonged initial empirical antibiotic treatment and NEC alone was not significant in their study. Alexander et al. (2011) performed a 1:2 case-control study to examine whether the antibiotic use is a risk factor for NEC [1]. Unlike our result, they reported that the cumulative days on antibiotics was a non-significant risk factor for NEC (OR 1.02; 95% CI 0.98–1.07).

Another notable finding in our study was that gentamicin and meropenem had a significant association with the incidence of NEC. Infants exposed to gentamicin or meropenem for a prolonged duration of time were at higher risk of having NEC compared to other antibiotics. We speculate that the broad spectrum antibacterial effect of gentamicin and meropenem on the intestinal flora leads to a reduction in bacteria from beneficial genera such as Bifidobacterium and the Lactobacilli which may have a key role in prevention of NEC development [8, 12].

There are several limitations in this study that should be taken into consideration when seeking to generalise the results. Firstly, our study is a retrospective analysis of many clinical associations, which can be considered to be limited by incomplete data in some charts, in addition to being influenced by confounding related variables. In addition, we sought to match each case with two controls according to several variables, including the gender, gestational age, maternal chorioamnionitis, growth parameters and mode of delivery. However, because of strict guidelines which were used in choosing the two controls for each case, a number of cases either could not be matched to any control or could only be matched to one control. Lastly, the sample size of our study was relatively small and this might lead to unidentified important differences in the measures between the two groups.

The strength of our study relates to the fact that it utilised a recent dataset obtained from a neonatal intensive care unit with strict guidelines for the management of VLBW infants, in particular for the administration of antibiotics and nutritional management.

The data from our study demonstrates that prolonged exposure to antibiotic therapy is associated with an increased risk of NEC among VLBW infants. Therefore, to prevent the potential serious side effects of antibiotics, such as NEC, clinicians should prescribe antibiotic agents that have optimal effects, and for the shortest possible duration of time in infants with suspected or proven sepsis. We believe that further prospective trials are required in this area to examine whether judicious limitation of empirical antibiotics might decrease the risk of NEC in VLBW infants.