Neonatal Infections: Insights from a Multicenter Longitudinal Research Collaborative

Abstract

For more than 30 years, the Neonatal Research Network (NRN) has conducted studies addressing the epidemiology of neonatal infections, including incidence, microbiology, maternal and neonatal risk factors, associated clinical findings, and outcomes. These studies have provided clinicians and policymakers critical data needed to inform national guidance for infection risk assessment and support daily practice. Further, NRN studies have prompted research into optimal approaches to infection diagnosis, treatment, and antimicrobial stewardship. In this article, we summarize the key findings of NRN infection-related studies, with an emphasis on those published in 2000 or later.

Introduction

Newborn infants are at high risk for infection, particularly when born preterm. Neonatal infections are associated with increased risks of mortality as well as short term medical morbidities and enduring neurodevelopmental impairment (NDI) among survivors. For more than 30 years, the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Neonatal Research Network (NRN) has conducted a broad range of studies addressing the incidence, microbiology, maternal and neonatal risk factors, associated clinical findings, and outcomes of early- and late-onset neonatal infections.¹ In contrast to data derived from laboratory-based surveillance or from administrative datasets, the multicenter NRN databases contain pre-specified maternal and neonatal information, linking microbial data to details of clinical management and to in-hospital and two-year post-discharge outcomes. Longitudinal data provides the opportunity to detect both secular and practice-driven changes in the incidence and microbiology of neonatal infection over time. Ultimately, NRN studies have informed national guidance on clinical practice, and have prompted research into optimal approaches to neonatal infection diagnosis, management, and treatment, as well as antimicrobial stewardship.

The NRN was founded in 1986 to support multicenter studies to advance neonatal clinical care and health outcomes.¹ Infection prevention, timely recognition, and optimal treatment are critical parts of neonatal care. As survival of the most vulnerable preterm infants improves, and newborn care includes infants previously considered pre-viable, the risks of infections and related morbidities continue to require close attention. In 2003, Stoll and Hansen published a review of NRN infection-related publications focused on very low birth weight (<1,500g; VLBW) infants.² Since that time, there have been significant contributions from the NRN in the arena of perinatal infectious disease research. In the current article, we summarize key findings of infection-related studies with an emphasis on those published in 2000 or later, highlighting contributions from the NRN to longitudinal epidemiology, laboratory-based infection research, infection prevention and treatment, and antimicrobial stewardship efforts. We will focus on research addressing bacterial and fungal infections of blood and cerebrospinal fluid (CSF), including early-onset sepsis (EOS, defined by isolation of pathogenic species in blood or CSF culture at ≤72 hours after birth) and late-onset sepsis (LOS, similarly defined but occurring >72 hours after birth and before hospital discharge).

Trends in Neonatal Infection among Very Preterm Infants

NRN studies have the advantage of consistent data collection over time that allows for rigorous detection of changes. Three database reports describe administered care and outcomes for more than 29,000 infants born at 22–28 weeks’ gestation from 2003–2018 at NRN centers.^3–5

Survival among these very preterm infants was 78.3% overall during 2013–2018; an increase from 76.0% in 2008–2012 (adjusted difference 2.0; 95% CI [1.0,2.9]) and from 71.6% during 2003–2007, with recent increases most notably among births at 22–24 weeks’ gestation. Perhaps reflecting the fundamental role that infection plays in promoting preterm birth, the incidence of EOS among these infants did not change over time (2.4%, 2003–2007; 2.1%, 2008–2012; 2.4% 2013–2018) despite statistically significant increases in administration of antenatal antibiotics to their mothers (67.4% during 2003–2007; 70.9% during 2008–2012 compared to 75% during 2013–2018 (adjusted difference 4.5, 95% CI [3.2,5.7]).^3–5 In contrast, LOS rates declined remarkably over this same time frame, highlighting the potential to prevent some types of late-onset bacterial infection. Among infants surviving >3 days after birth, 35.9% were diagnosed with LOS during 2003–2007; 24.4% during 2008–2012 compared to 19.9% from 2013–2018 (adjusted difference −4.9, 95% CI [−6.0, −3.7]).

Bacterial and fungal infections remain important causes of death among extremely preterm infants. An NRN report analyzing causes of death among 22,248 infants born 22–28 weeks’ gestation from 2000–2011 found that 6,075 infants died during this period (overall mortality rate 273 per 1,000 live births).⁶ The proportion of deaths caused by infection increased substantially from the first week after birth (3.5% of all deaths) to the second week after birth (15.3% of all deaths).⁶ Comparing 4 year epochs, the rate of death declined over the study period, largely due to decreases in death caused by acute or chronic respiratory disease. Although the rate of death due to infection (cause-specific mortality) was unchanged over this period, infection was the direct cause of death for approximately 22 infants per 1,000 live births in this gestational range while deaths complicated by infection declined over the time period.⁶

Early-Onset Bacterial Sepsis

Infection surveillance:

One of the most important contributions of the NRN to research focused on EOS has been longitudinal epidemiological surveillance. Key publications are listed in Table 1. Beginning in 1996, the Centers for Disease Control and Prevention (CDC) partnered with the American College of Obstetricians and Gynecologists and the American Academy of Pediatrics to publish consensus clinical care guidelines aimed at the prevention of perinatal Group B Streptococcal (GBS) disease. NRN investigators conducted a series of studies to measure the impact of the initial guidelines and revised guidelines published in 2002. Stoll et al published three epidemiological reports of EOS among VLBW infants born at NRN centers during distinct time periods: 1991–1993 (cohort 1)⁷, 1998–2000 (cohort 2)⁸, and 2002–2003 (cohort 3)⁹. These studies demonstrated that increasing use of intrapartum antibiotics did not significantly change the incidence of EOS among VLBW infants – in contrast to the remarkable decline in EOS among term-born infants that followed implementation of CDC guidance.¹⁰ Importantly, these studies also identified a shift in the microbiology of VLBW EOS from predominantly gram-positive to predominantly gram-negative bacteria between cohorts 1 and 2.^7,8 This finding persisted in cohort 3, where gram-negative pathogens remained the most common cause of EOS (52.9%) with Escherichia coli most frequently identified (41.2%).⁹ A concomitant decline in the rate of EOS caused by GBS was also noted between cohorts 1 and 2, which persisted for cohort 3.^7–9 These findings informed empiric antibiotic choice among VLBW infants at highest risk for EOS and highlight the persistent need for novel EOS preventative strategies among VLBW infants.

Table 1:

Early-onset sepsis research

Reference	Study Population	Years	Key Points
Stoll, et al. J Pediatr 19967	7,861 VLBW infants	1991–1993 (cohort 1)	EOS rate: 1.9% of VLBW infants; rates varied by BW and GA GBS (31%) most common bacterial isolate E. coli (16%) and H. influenzae (12%) next most common 26% of VLBW infants with EOS died
Stoll, et al. N Engl J Med 20028	5,447 VLBW infants	1998–2000 (cohort 2)	Non-significant decline in EOS rate compared to 1991–1993 cohort (19.3 vs. 15.4 per 1000) E. coli (44%) and GBS (10%) most common bacterial isolates 85% of E. coli isolates resistant to ampicillin EOS associated with higher adjusted risk of death, RDS, severe IVH and BPD
Stoll, et al. Pediatr Inf Dis 20059	5,999 VLBW infants	2002–2003 (cohort 3)	Stable rate of EOS compared to 1998–2000 E. coli most common organism (41%) with stable rates of antibiotic resistance 35% of VLBW infants with EOS died
Stoll, et al. Pediatrics 201111	396,586 infants of all GA	2006–2009 (EOS I)	Most frequent pathogen was GBS (43%), followed by E. coli (29%) Incidence was 0.98 cases per 1000 live births and was highest among VLBWs
Weston, et al. Pediatr Infect Dis J 201112	CDC compared with NRN	2006–2008	Similar pathogen distribution, incidence trends, and case fatality rates comparing CDC and NRN cohorts For both cohorts, GBS isolated in most cases; E. coli isolated in most deaths
Wynn, et al. J Pediatric 201320	34,396 VLBW infants	1998–2009	Overall, risk of LOS or death was not different overall for infants with and without history of EOS Infants born <25 weeks’ GA with EOS had lower risk of LOS
Pappas, et al. JAMA Pediatr 201422	2,390 infants 22–26 weeks GA	2006–2008	Chorioamnionitis associated with lower GA at birth and higher rates of EOS and severe IVH Histologic plus clinical chorioamnionitis associated with increased risk of NDI as compared with no chorioamnionitis
Wortham, et al. Pediatrics 201616	396,586 infants of all GA	2006–2009	Of 229 infants with EOS exposed to chorioamnionitis, 13% had no symptoms by 6 hours after birth, and 9% had no symptoms by 72 hours Estimated 60 to 1400 newborns would receive empiric antibiotics for each infected asymptomatic newborn
Weissman, et al. J Pediatr Dis Soc 201614	28 infants with E. coli EOS	2008–2009	Among E. coli isolates: high rate of ampicillin resistance; low rate of fluoroquinolone resistance; no aminoglycoside resistance 1 isolate resistant to third-generation cephalosporins
Puopolo, et al. Pediatrics 201718	15,433 infants 22–28 weeks GA	2006–2014	37% met “low-risk” criteria based on delivery characteristics EOS incidence was 0.5% in low-risk group vs 2.5% in comparison group Prolonged antibiotics were administered to 34% of uninfected low-risk infants versus 47% of uninfected comparison infants
Stoll, et al. JAMA Pediatr 202013	217,480 infants of all GA	2015–2017 (EOS II)	Incidence was 1.08 per 1000 live births and varied significantly by GA Most frequent pathogen was E. coli (37%), followed by GBS (30%) Rate of E. coli among VLBW infants increased compared to EOS I cohort
Mukhopadhyay, et al. J Pediatr 202021	6,565 infants 22–26 weeks’ GA	2006–2014	EOS associated with increased risk of death/NDI Among matched infants without EOS, prolonged early antibiotic administration was not associated with death/NDI
Flannery, et al. Pediatr Infect Dis J 202215	217,480 infants of all GA	2015–2017 (secondary analysis)	8% of EOS isolates suboptimally treated with ampicillin and gentamicin Resistant EOS infections more common among preterm, VLBW infants

Footnotes: Studies listed in order of publication. BPD (bronchopulmonary dysplasia); BW (birthweight); CDC (Centers for Disease Control and Prevention); ELBW (extremely low birth weight); EOS (early onset sepsis); GA (gestational age); GBS (Group B Streptococcus); IVH (intraventricular hemorrhage); NDI (neurodevelopmental impairment); NEC (necrotizing enterocolitis); NRN (Neonatal Research Network); RDS (respiratory distress syndrome); VLBW (very low birth weight)

A limitation of the EOS reports from the VLBW registry was that more mature infants were not included, prompting a call for more comprehensive surveillance among infants of all birth weights (BW) and gestational ages (GA).⁹ Therefore, NRN investigators expanded EOS surveillance in the EOS I study which included 396,586 live births of all gestational ages, born at NRN centers from 2006–2009.¹¹ The EOS I study identified GBS as the predominant pathogen among all live births and among term infants born ≥37 weeks’ gestation, while E. coli was predominant among infants born <37 weeks’ gestation. The overall incidence of EOS was 0.98 cases per 1,000 live births and 15.7% of those infected died prior to hospital discharge - but both incidence and case fatality were higher among preterm infants. NRN EOS surveillance data from 2006–2008 were compared to CDC Active Bacterial Core EOS surveillance data from 2005–2008 in a 2011 report, yielding similar pathogen distributions, incidence trends, and case fatality rates.¹² Most recently, NRN investigators reported the results of the EOS II surveillance study, which included 217,480 newborns born 2015–2017 at NRN centers.¹³ This study found a similar overall EOS incidence of 1.08 per 1,000 live births, with marked variation in both incidence and case-mortality by gestational age. For the first time, a U.S. surveillance study found E. coli (36.6%) to be the dominant pathogen, followed by GBS (30.2%). Among all live births (0.40 vs. 0.27 per 1000), and among VLBW births (8.68 vs 5.07 per 1000), the absolute rate of E. coli infection increased significantly compared to EOS I surveillance.¹³ Among infants born ≥37 weeks, GBS remained the dominant pathogen, accounting for 51.9% of isolates. Remarkably, there were no deaths among term infants with EOS, yet mortality ranged from 14.0% among infants with EOS born at 29–36 weeks GA to 43.3% among those born at 22–28 weeks GA.

Antibiotic susceptibility:

EOS surveillance reports from the NRN have periodically addressed isolate susceptibility data to inform clinical decisions regarding empiric antibiotic therapy. NRN investigators collected E. coli isolates from 28 cases of EOS during 2008–2009 and utilized molecular techniques to determine clonal, virulence-associated, and antibiotic resistance-associated traits.¹⁴ Study isolates had high rates of ampicillin resistance but no aminoglycoside resistance; only a single isolate was resistant to third-generation cephalosporins. In the 2015–2017 EOS II study, investigators focused on susceptibility of GBS and E. coli isolates to ampicillin and gentamicin, as these are the recommended first-line antibiotics for empiric treatment of EOS.¹³ Only 25.9% of tested E. coli isolates were susceptible to ampicillin and 7.8% were resistant to both ampicillin and gentamicin.¹³ In a further analysis of EOS II investigators reviewed available susceptibility data for all pathogens and estimated that 7.9% of EOS cases were caused by isolates for which treatment with the empiric combination of ampicillin and gentamicin was likely suboptimal.¹⁵ Resistant cases were predominantly E. coli and most occurred in VLBW infants. Unfortunately, this study demonstrated that neither the substitution nor the addition of one single antimicrobial agent would provide adequate empiric coverage for all cases.¹⁵ The identification of changes in pathogen antibiotic susceptibility over time is a critically important clinical goal of ongoing NRN surveillance studies.

Risk assessment:

Determining which infants are at risk for EOS is challenging, leading to practice variation and high rates of empiric antibiotic administration, particularly among preterm infants. Beyond younger GA and lower BW, NRN studies have identified longer duration of rupture of membranes and spontaneous preterm rupture of membranes as risk factors for EOS. Concern for maternal intraamniotic infection – whether denoted by obstetric clinical concern for chorioamnionitis, occurrence of maternal intrapartum fever, or histologic evidence of chorioamnionitis identified on placental pathology – has been repeatedly associated with increased risk of EOS among both term and preterm infants. The sensitivity and specificity of these definitions varies by GA, and among term infants can result in high rates of empiric antibiotic administration to uninfected infants. To improve the utility of the obstetric diagnosis in neonatal risk assessment, NRN investigators reviewed records for 229 infants from the EOS I study who were exposed to chorioamnionitis to determine if clinical illness could be used in combination with chorioamnionitis to predict EOS. The study found that most infected infants had clinical illness, but 12.7% had no symptoms by 6 hours after birth, and 9.2% had no symptoms by 72 hours.¹⁶ The investigators estimated that 60 to 1,400 newborns would be administered empiric antibiotics to identify each infected, asymptomatic newborn, highlighting the clinical conundrum of EOS guidelines in place at the time of the study.¹⁶ Notably, most (79.3%) of the asymptomatic, chorioamnionitis-exposed infants were born at term. Subsequent to the publication of the study, significant progress has been made in EOS risk assessment among term and late preterm infants using other large datasets and national guidance has evolved.¹⁷ Risk assessment for EOS among preterm infants is extremely challenging, particularly for those of younger gestation among whom initial clinical instability is almost universally present regardless of infection status. NRN investigators made the first real progress in risk distinction among preterm infants by defining delivery characteristics associated with lowest risk of EOS.¹⁸ Among a dataset of 15,433 extremely preterm infants, 37.3% (5759/15,433) met hypothesized “low-risk” criteria, defined as cesarean delivery, with membrane rupture at delivery, and absence of clinical chorioamnionitis.¹⁸ These infants had significantly lower risk of EOS (0.5% vs 2.5%) – although the study found that one-third received prolonged empiric antibiotic courses. The study suggested that differential risk recognition based on delivery characteristics can guide decisions about empiric early antibiotic use among preterm infants, a concept later incorporated into revised American Academy of Pediatrics clinical guidance.^18,19

Clinical outcomes:

NRN studies have highlighted the impact of EOS on mortality among preterm infants, and more recently have addressed the role of EOS on outcomes among survivors. Both early infection and early antibiotic therapy could influence the development of immunity and constitution of the infant microbiome, motivating an NRN study that addressed whether EOS increases the subsequent risk of LOS. Using a dataset of 34,396 VLBW infants born 1998–2009, investigators did not observe an increased risk of LOS in adjusted analyses.²⁰ Among a cohort of 6,565 infants 22–26 weeks’ gestation born 2006–2014, NRN investigators found that those with EOS had higher risks of severe initial respiratory morbidity and severe brain injury.²¹ Using this same cohort, NRN investigators tested the hypothesis that EOS specifically increases the risk of death or NDI by 18–26 months corrected GA among extremely preterm infants. In addition, the authors hypothesized that among infants without EOS, prolonged early antibiotics alters the risk of death or NDI.²¹ This observational study utilized a propensity score for receiving prolonged antibiotics and matched nonbacteremic infants with and without prolonged antibiotics. Infants with EOS had higher adjusted risk of death or NDI compared to those who were uninfected with or without prolonged antibiotics. Among the nonbacteremic matched cohort, no significant difference in death or NDI with or without prolonged antibiotics was found.²¹ Related to EOS, exposure to chorioamnionitis was also found in a previous NRN study to be associated with death and NDI among extremely preterm infants.²² Together, these studies can inform counseling regarding EOS among preterm infants, and support the importance of research into novel preventative approaches for preterm EOS.

Late-Onset Bacterial Sepsis

Infection surveillance and risk profiles:

Similar to EOS surveillance reports, the NRN has periodically conducted prospective LOS surveillance, focusing on the preterm population. Key contemporary references are summarized in Table 2. NRN studies over time have revealed the magnitude of infection risk among preterm infants. In a report of 7,861 VLBW infants cared for at NRN centers from 1991 to 1993, one-quarter of those surviving >3 days suffered at least one episode of LOS.²³ Median age of LOS onset was 17 days, and infected infants were at higher risk for adverse outcomes including BPD, IVH, NEC, and mortality.²⁴ A subsequent study described LOS among a cohort of 6,956 VLBW infants born 1998–2000, with similar findings: 21.1% of those surviving beyond 3 days suffered LOS, with similar microbiology.²⁵ Infected infants were twice as likely to die compared with uninfected infants (18.4% vs 7.3%), especially those infected with gram-negative bacteria (36.2%) or fungi (31.8%). Importantly, this study demonstrated the potential role of neonatal intensive care practices in increasing the occurrence of LOS, demonstrating quantitatively that increased duration of both umbilical and percutaneous central catheters, longer duration of parenteral nutrition, later age at first enteral feeding, and longer duration of mechanical ventilation were associated with risk of LOS, even when controlling for GA at birth.²⁵ An NRN study focused on bacteremia associated with necrotizing enterocolitis (NEC) observed that surgical NEC with or without subsequent intestinal failure was associated with higher rates of LOS compared to medical NEC.²⁶

Table 2.

Late-onset sepsis research

Reference	Study Population	Years	Key Points
Stoll, et al. J Pediatr 199623	7,861 VLBW infants	1991–1993	One quarter of VLBW infants who survived >3 days developed LOS 73% caused by gram-positive bacteria; 55% were caused by CoNS Infants with LOS had higher rates of morbidities and mortality
Fanaroff, et al. Pediatr Infect Dis J 199824	2,416 VLBW infants	1988–1991 (secondary analysis)	Risk factors for LOS included lower BW, lower GA, male sex, respiratory distress, vaginal delivery, longer duration of umbilical catheters Median age of LOS onset was 17 days Infected infants at higher risk of BPD, IVH, NEC, and death
Stoll, et al. Pediatrics 200225	6,956 VLBW infants	1998–2000	Similar incidence and pathogen distribution compared to 1991–1993 cohort Infected infants twice as likely to die compared with uninfected infants Mortality highest with gram-negative bacterial or fungal infection
Stoll, et al. JAMA 200432	6,093 ELBW infants	1993–2001	Majority of ELBW survivors (65%) had at least 1 infection (EOS, LOS, NEC, and/or clinical infection) Infants with infection had increased risks of neurodevelopmental and growth impairment
Stoll, et al. J Pediatr 200431	10,660 VLBW infants	1998–2000	Of 31,777 blood and CSF cultures performed among 10,660 across the NRN, only one case of E. sakazakii was identified
Cole, et al. J Pediatr 201226	932 VLBW infants with NEC	2002–2005	Proportion of infants with bacteremia was higher among infants with surgical NEC and intestinal failure, as compared to those with medical NEC or surgical NEC without intestinal failure Recurrent bacteremia was common among infants with intestinal failure
Shane, et al. Pediatrics 201230	8,444 VLBW infants	2006–2008	MRSA infection occurred in 1% of VLBWs 99% of MRSA infections were LOS 28% of S. aureus cases were MRSA Mortality was similar for MRSA (26%) and MSSA (24%)
Greenberg, et al. Pediatr Infect Dis J 201727	10,131 ELBW infants	2000–2005 (Era 1) 2006–2011 (Era 2)	Incidence of LOS decreased significantly over time: 41% of infants in Era 1 and 34% of infants in Era 2 had LOS Fungal infections decreased in Era 2 9 centers had significantly higher odds of LOS compared with the center with the lowest incidence
Mukhopadhyay, et al. Arch Dis Child Fetal Neonatal Ed 202129	3,940 ELBW infants	2006–2014	Infants with culture-confirmed LOS had higher adjusted relative risks of death or NDI compared to uninfected infants and compared to those with culture-negative LOS Among survivors, risk for NDI did not differ significantly between the groups

Footnotes: Studies listed in order of publication. BPD (bronchopulmonary dysplasia); CoNS (coagulase negative staphylococci); CSF (cerebrospinal fluid); ELBW (extremely low birth weight); EOS (early onset sepsis); GA (gestational age); IVH (intraventricular hemorrhage); MSSA (methicillin-susceptible Staphylococcus aureus); MRSA (methicillin-resistant Staphylococcus aureus); NEC (necrotizing enterocolitis); LOS (late-onset sepsis); NRN (Neonatal Research Network); SA (Staphylococcus aureus); VLBW (very low birth weight)

After the early epidemiological reports, several interventions were implemented into the care of extremely preterm infants to mitigate the risk of LOS, including rigorous hand hygiene protocols, central line care bundles, antimicrobial stewardship, and antifungal prophylaxis. To determine the potential impact of infection prevention measures, NRN investigators assessed LOS incidence, center variation in incidence, and pathogen distribution among those infants at highest risk of LOS, including 10,131 extremely-low birth weight (ELBW; birth weight <1000 grams) infants cared for at NRN centers from 2000 to 2011.²⁷ This study assessed the data a priori comparing 2 eras: 2000–2005 (Era 1) and 2006–2011 (Era 2). Incidence of LOS decreased over time, occurring in 41.4% of infants in Era 1 and 33.9% of infants in Era 2 (p<.001). The overall pathogen distribution did not differ significantly between eras, except fungal infections decreased in Era 2. Marked center variation was noted, with 9 out of 12 centers having significantly higher odds of LOS compared with the lowest incidence center.²⁷ This study demonstrated the potential impact of infection prevention efforts in reducing LOS risk among the highest-risk infants.

Accurate infection surveillance requires appropriate ascertainment. The optimal evaluation for suspected early and late-onset neonatal sepsis includes culture of both blood and CSF. Due to the challenges of performing lumbar puncture in very preterm and/or clinically unstable infants, this procedure is not always performed as part of the initial evaluation for infection. In a cohort of 9,641 VLBW infants from 1998–2001, NRN investigators evaluated the epidemiology of late-onset meningitis as well as the concordance of blood and CSF culture results.²⁸ Lower GA and prior episodes of sepsis increased likelihood of meningitis, and infants with meningitis had more complicated hospitalizations and increased rates of seizures and death. Most notably, one third of infants with meningitis had a sterile blood culture. The most common species isolated from CSF in the setting of sterile blood cultures included Enterococcus and Candida.²⁸ This study demonstrated the importance of obtaining CSF as part of LOS evaluation whenever feasible.

Microbiology:

Most cases of LOS (72.7%) in the NRN 1991–1993 study were caused by gram-positive bacteria, with coagulase-negative staphylococci (CoNS) most frequently (55%) isolated.²³ The predominance of gram-positive organisms was remarkably similar in the 1998–2000 (70.2%), 2000–2011 (73.3%) and 2006–2014 (70.9%) studies.^23,27,29 In the 1991–1993 surveillance, Enterobacter spp., E. coli and Klebsiella spp. were the most commonly-isolated gram-negative pathogens; in subsequent NRN reports, E. coli and Klebsiella spp. predominated, though Pseudomonas spp. and Enterobacter spp. were also frequently identified. Other gram-negative pathogens including Serratia spp. and Citrobacter spp. accounted for <5% of LOS cases in all reports. The specific impact of methicillin-sensitive and methicillin-resistant Staphylococcus aureus (MSSA and MRSA, respectively) strains was examined in an NRN study conducted among 8,444 VLBW infants born 2006–2008 and surviving >3 days after birth.³⁰ LOS due to MSSA or MRSA occurred in 3.7% of infants; 72% of these cases were due to MSSA. Mortality was high among infants with both MRSA (26.1%) and MSSA (24.1%) infections, and morbidities including mechanical ventilation and NEC did not differ between infants with MRSA and MSSA.³⁰ A 2004 report sought to determine if Enterobacter sakazakii (now reclassified as Cronobacter sakazakii), a gram-negative bacillus implicated in neonatal infection and NEC, had affected VLBW infants in the NRN.³¹ The organism has been linked with infection outbreaks in NICUs with powdered infant formula and additives as a potential source. Of 31,777 blood cultures and 5,433 CSF cultures performed among 10,660 across the NRN, only one case of E. sakazakii was identified, which was a case of bacteremia in a 28 weeks’ GA infant without meningitis or NEC. Other Enterobacter species, including E. cloacae and E. aerogenes, were more commonly identified.³¹

Long-term outcomes:

NRN surveillance demonstrates that LOS risk increases with earlier GA at birth and lower BW, but even in adjusted analyses LOS is associated with morbidities including respiratory distress syndrome, patent ductus arteriosus, bronchopulmonary dysplasia (BPD), and NEC.²³ In adjusted analyses, infected infants had longer periods of mechanical ventilation and longer length of hospitalization and higher rates of death (17.1% overall among those with LOS; 39.6% if infected with gram-negative organisms).²³ NRN investigators assessed growth and neurodevelopmental outcomes of 6,093 ELBW infants born 1993–2001 who survived and completed 18–22 month evaluations.³² Although infants with both EOS and LOS were included, 97.3% of those with culture-confirmed infection had LOS. Compared to uninfected children, those with infection were more likely to have poor linear and head growth, and more likely to have some form of NDI (48.4% vs. 29.1%).³² More recently NRN investigators determined the risk of death or NDI among 3,940 ELBW infants born 22–26 weeks’ GA from 2006–2014, comparing those with LOS, culture-negative, antibiotic-treated presumed LOS, and unaffected infants.²⁹ The adjusted risk of death or survival with NDI was higher among those with LOS compared to unaffected infants and compared to those with presumed, culture-negative infection. These and other studies emphasize the adverse consequences of late-onset infection, and justify continued efforts to reduce risk of infection among preterm infants.

Fungal Infection

Incidence and risk factors:

NRN LOS surveillance from 1991–1993 found that 9.3% of LOS cases were caused by fungal species, primarily Candida albicans and Candida parapsilosis.²³ In contrast, fungal species were isolated in 3.6% of LOS cases in 2006–2014 NRN data.²⁹ Key references specifically addressing fungal infection are provided in Table 3. NRN investigators focused on the epidemiology, risk factors, and outcomes of invasive fungal infection among two consecutive ELBW cohorts. One study included 4,579 ELBW infants born from 1998 to 2001; 7.0% of infants developed invasive fungal infection.³³ Delayed fungal clearance was common (10% had fungemia for 14 days or longer), and blood cultures were sterile in half of the infants with fungal meningitis. In adjusted analyses, risk factors for fungal infection included lower BW, third generation cephalosporin exposure, and lack of enteral feeding. Remarkably, three-quarters of infants with invasive fungal infection died or suffered NDI at 2-year evaluation; outcomes were even worse with delayed removal of central catheters.³³ A second study included 1,515 ELBW infants born from 2004 to 2007.³⁴ Nine percent of infants had invasive fungal infection, most with fungemia although fungi were also isolated from CSF, urine, or other sterile sites.³⁴ Incidence across centers ranged from 2% to 28%, and risk factors included antenatal antibiotics, postnatal broad-spectrum antibiotics (in particular third generation cephalosporin agents), intravenous lipids, and presence of central catheters and endotracheal tubes.³⁴

Table 3.

Fungal infection research

Reference	Study Population	Years	Key Points
Fungal infection
Benjamin, et al. Pediatrics 200633	4,579 ELBW infants	1998–2001	7% of ELBW infants developed invasive fungal infection Risk factors included BW, cephalosporin exposure, lack of enteral feeding 73% of infants with fungal infection suffered death/NDI; increased risk with delayed removal of infected central catheters
Benjamin, et al. Pediatrics 201034	1,515 ELBW infants	2004–2007	9% of ELBW infants developed invasive candidiasis (center range 2–28%) Risk factors included antenatal and postnatal antibiotics, intravenous lipids, central catheters, and endotracheal tubes
Greenberg, et al. J Pediatr 201239	136 ELBW infants	2004–2007	Incidence of death/NDI was lower in infants who received empiric antifungal therapy versus those who did not No significant difference between groups for time to clearance of infection, retinopathy, BPD, and NDI
Wynn, et al. Clin Infect Dis 201236	1,515 ELBW infants	2004–2007 (secondary analysis)	2.2% of infants had Candiduria only Most urine isolates were C. albicans (62%) or C. parapsilosis (29%) No difference in death and composite of death or NDI between infants with candiduria versus candidemia.
Bliss, et al. J Pediatr 201237	1,515 ELBW infants	2004–2007 (secondary analysis)	Enhanced virulence was detected in 61% of invasive isolates of C. albicans and 42% of invasive isolates of C. parapsilosis Infants with more virulent isolates were younger at the time of infection and had higher serum creatinine levels
Adams-Chapman, et al. J Pediatr 201335	1,515 ELBW infants	2004–2007 (secondary analysis)	31% with Candida and 31% with late-onset non-Candida sepsis had NDI Compared to infants with non-Candida sepsis or no sepsis, infants with Candida sepsis were at the highest risk of death or NDI
Autmizguine, et al. Pediatr Infect Dis J 201838	1,515 ELBW infants	2004–2007 (secondary analysis)	Antifungal agent resistance was rare among infecting Candida isolates (1% resistant to fluconazole, none were resistant to amphotericin B or micafungin) MIC level was not associated with increased risk of death or NDI

Footnotes: Studies listed in order of publication. BPD (bronchopulmonary dysplasia); BW (birth weight); ELBW (extremely low birth weight); MIC (minimum inhibitory concentration); NDI (neurodevelopmental impairment)

Outcomes and factors affecting outcomes:

A number of other NRN studies utilized the 2004–2007 cohort of 1,515 infants to address outcomes among infants with invasive fungal infection. NDI or death was compared among infants with Candida sepsis, infants with non-Candida sepsis and those who were never evaluated for sepsis; infants with Candida sepsis were at the highest risk of death or NDI.³⁵ In a study focused on candiduria, rate of death or NDI was higher among infants with candiduria than those with suspected but not proven infection, but was not different between infants with candiduria compared to candidemia.³⁶ These findings might be explained by failure to detect fungemia among those infants in whom only the urine culture yielded Candida or might suggest that urinary tract infection alone could result in an inflammatory cascade involving the central nervous system. Another study addressed variation in virulence properties, including phenotypic switching, adhesion, and cytotoxicity, among infecting and colonizing Candida species.³⁷ Strains with higher virulence were more commonly isolated when infants suffered both candidemia and candiduria compared to candidemia alone, and were associated with younger age at infection and higher serum creatinine levels. Minimal Candida resistance to fluconazole (only 1% of isolates), and no resistance to amphotericin B or micafungin was observed in a study focused on antifungal susceptibility.³⁸ No difference in death or NDI was found between groups with low and high minimum inhibitory concentrations.³⁸

Given the significant burden of invasive fungal infection among ELBW infants, empiric antifungal therapy emerged as a successful preventative measure in small studies. NRN investigators reported on 136 ELBW infants with invasive candidiasis from the 2004 to 2007 cohort³⁴, and found that the incidence of death or NDI was lower in infants who received empiric antifungal therapy versus those who did not (50.0% vs 64.0%; adjusted odds ratio 0.27; 95% CI [0.08–0.86]).³⁹ No significant difference was found between the groups for other outcomes, including time to clearance of infection, retinopathy, BPD, and NDI. Based on their findings, the investigators advocated for a prospective randomized trial of empiric antifungal therapy among at-risk infants.³⁹

Prevention of Neonatal Infection

Although the NRN conducts both epidemiologic studies and clinical intervention trials, few trials have focused directly on infection prevention. Key references are shown in Table 4. Two randomized trials addressing infection were previously described in the 2003 summary of NRN infection research by Stoll and Hansen.² A randomized, placebo-controlled trial of intravenous immunoglobulin to prevent LOS was conducted among 2,416 ELBW infants born 1988–1991, but found no difference in infection outcomes.⁴⁰ Another trial conducted among 1,433 ELBW infants in 1999–2001 randomized infants to receive parenteral glutamine supplementation with parenteral nutrition or parenteral nutrition alone, but found no difference in the primary outcome of LOS or death.⁴¹ Other clinical trials not focused on infection as a primary outcome have also provided meaningful data regarding infection. For example, a multicenter trial conducted by NRN to determine the effect of dexamethasone therapy among VLBW infants at risk of BPD found an increased risk of infection in the treatment group.⁴² A secondary study analyzed infections among the trial cohort in detail, and confirmed this finding.⁴³ Moreover, H2-receptor blocker therapy was also associated with increased risk of infection. Another NRN study found that H2-receptor blocker therapy was associated with increased risk of NEC (P<0.0001).⁴⁴

Table 4.

Infection prevention and management research

Reference	Study Population	Years	Key Points
Fanaroff, et al. N Engl J Med 199440	2,416 VLBW infants	1988–1991	17% of IVIG group and 19% of the control group developed LOS IVIG had no effect on RDS, BPD, IVH, duration of hospitalization, or death
Stoll, et al. Pediatrics 199943	371 VLBW infants	1992–1995 (secondary analysis)	Infants randomized to dexamethasone were significantly more likely than controls to have LOS
Poindexter, et al. Pediatrics 200441	1,433 ELBW infants	1999–2001	51% of infants receiving glutamine supplementation died or had LOS compared to 48% of controls (not significant difference) Glutamine had no effect on feeding tolerance, NEC, or growth.
Stoll, et al. Pediatrics 200428	9,641 VLBW infants	1998–2001	One third of infants with meningitis had a sterile blood culture Infants with meningitis had longer times on mechanical ventilation, longer hospitalization, and were more likely to have seizures and to die
Guillet, et al. Pediatrics 200644	11,072 VLBW infants	1998–2001	Antecedent H2-blocker use was associated with an increased incidence of NEC

Footnotes: Studies listed in order of publication. BPD (bronchopulmonary dysplasia); ELBW (extremely low birth weight); IVH (intraventricular hemorrhage); IVIG (intravenous immune globulin); LOS (late-onset sepsis); NEC (necrotizing enterocolitis); RDS (respiratory distress syndrome); VLBW (very low birth weight)

Laboratory-based Research

NRN infection-related studies have also addressed infection diagnosis, inflammatory mediators, and genetic susceptibility. Key references are provided in Table 5. A study of 1,700 blood cultures from infants cared for at three NRN centers measured the potential for a rapid chemiluminescent DNA probe assay to detect bacteremia before detection by continuous-monitoring blood culture systems.⁴⁵ Of 130 cultures that were positive by continuous-monitoring culture systems, 29% of CoNS and 78% of other organisms were detected by the assay, which the authors concluded was not sufficiently sensitive to be used in clinical practice. However, 98% of all positive cultures were detected by the conventional system within 48 hours, demonstrating that even conventional systems are sufficiently sensitive to aid in clinical decision-making regarding antibiotic duration.⁴⁵

Table 5.

Laboratory infection research

Reference	Study Population	Years	Key Points
Hertz, et al. J Perinatol 199945	1700 blood cultures	Not provided	DNA probe not sufficiently sensitive to be clinically useful Automated blood culture systems are sufficiently sensitive to aid in clinical decision-making regarding discontinuation of antibiotics after 48 hours
Carlo, et al. J Pediatr 201146	1,067 ELBW infants	1999–2002	Infants with cerebral palsy had altered cytokine profiles early in life, suggesting the condition may have late perinatal and/or early neonatal inflammatory origin
Schelonka, et al. Cytokine 201147	996 ELBW infants	2000–2001 (secondary analysis)	Infants with bacteremia had lower levels of IL-17, and higher levels of IL-6 and IL-10 Ratio of immune regulatory cytokines to inflammatory cytokines was associated with development of bacteremia
Sood, et al. Pediatr Res 201148	1,066 ELBW infants	1999–2002 (secondary analysis)	Significant differences between infants with fungal sepsis, bacterial sepsis, or no sepsis for interferon-γ, IL-10 and IL-18, transforming growth factor-β, and tumor necrosis factor-α
Morrow, et al. J Pediatr 201149	410 infants ≤32wks GA	2000–2004	15% of 135 infants with low secretor phenotype died, compared with 2% of 248 infants with high secretor phenotype Low secretor phenotype was associated with NEC, and non-secretor genotype was associated with gram negative sepsis
Srinvasan, et al. Arch Dis Child Fetal Neonatal Ed 201750	757 ELBW infants	Not provided	No SNPs reached genome-wide significance levels Areas of potential association and pathways meriting further study were identified

Footnotes: Studies listed in order of publication. DNA (deoxyribonucleic acid); ELBW (extremely low birth weight); GA (gestational age); IL (interleukin); NEC (necrotizing enterocolitis); SNP (single-nucleotide polymorphism)

NRN investigators have addressed the role of the host immune response to infection among preterm newborns. A study of 1,067 ELBW infants sought to determine if pro-inflammatory and anti-inflammatory cytokines were related to neurodevelopmental outcomes in ELBW infants.⁴⁶ Infants who were ultimately diagnosed with cerebral palsy had altered early cytokine profiles, suggesting that the condition may have a late perinatal and/or early neonatal inflammatory origin.⁴⁶ In another study, NRN investigators measured levels of multiple immune regulatory interleukins (ILs) and inflammatory cytokines among a cohort of ELBW infants in the three weeks after birth.⁴⁷ Infants who developed infection had lower levels of the inflammatory cytokine IL-17 and higher levels of the regulatory cytokines IL-6 and IL-10, and the ratio of regulatory to inflammatory cytokines was associated with development of infection.⁴⁷ In a hypothesis-generating exploratory study, significant differences in cytokine profiles were observed between infants with fungal sepsis, bacterial sepsis, and no sepsis.⁴⁸

Two NRN studies utilized genetic testing to identify risk for infection. In a study conducted at a single NRN site, the salivary secretor gene fucosyl transferase 2, secretor genotype and phenotype were associated with adverse outcomes, including NEC, sepsis and death, in a cohort of premature infants enrolled in the NRN’s very low birthweight registry.⁴⁹ Using data from the NRN’s anonymized genomic database biorepository, NRN investigators identified genetic variants associated with EOS and LOS using a genome-wide association analysis in a cohort of extremely premature infants.⁵⁰ Comparing 351 infants with sepsis and 406 infants without sepsis, more than 1 million single-nucleotide polymorphisms were identified, but none reached genome-wide significance. The presence of multiple polymorphisms, however, did meet target significance levels for fungal and gram-negative sepsis, including pathways involving gap junctions, fibroblast growth factor receptors, regulators of cell division and interleukin-1-associated receptor kinase 2.⁵⁰

Antimicrobial Stewardship Considerations

NRN research among preterm infants has highlighted the relatively high risk of infection and the significant consequences of infection in this population that is, at baseline, already at high risk for both morbidity and mortality. Traditionally, the risk/benefit balance for empiric antibiotic treatment is thought to fall on the side of benefit. Multiple NRN studies have assessed empiric antibiotic prescribing practices and related outcomes at both the infant- and center-level, with key references shown in Table 6. In a cohort of 3,702 ELBW infants born 1998–2001, NRN investigators found that early empiric third-generation cephalosporin exposure was associated with subsequent fungal infection among ELBW infants.⁵¹ Further, center variation in the incidence of fungal infection (ranging from 2.4% to 20.4%) correlated with third-generation cephalosporin use.⁵¹ Both multicenter and single center studies of infants cared for at NRN centers subsequently demonstrated that prolonged early antibiotic exposure, defined as 5 or more day of antibiotics, increased risk of LOS as well as the combined outcome of LOS, NEC, and/or death. A multicenter study of 5,693 ELBW found that 53.2% (center range, 27.3–85.0%) received prolonged early antibiotic therapy, which was associated with increased risk of NEC or death.⁵² Kuppala et al studied 365 VLBW infants from a single NRN center who survived 7 days free of sepsis or NEC, and found that 36% received prolonged initial empirical antibiotics, which was independently associated with LOS (OR, 2.45 [95% CI, 1.28–4.67]) and the combination of LOS, NEC, or death (OR, 2.66 [95% CI, 1.12–6.3]).⁵³

Table 6.

Antimicrobial stewardship research

Reference	Study Population	Years	Key Points
Cotten, et al. Pediatrics 200651	3702 ELBW infants	1998–2001	Third-generation cephalosporin exposure associated with fungal infection for individual infants Center incidence of fungal infection correlated with third-generation cephalosporin use per infant
Cotten, et al. Pediatrics 200952	5693 ELBW infants	1998–2001	Adjusted for GA, prolonged early antibiotic therapy was associated with increased odds of NEC/death and death
Kuppala, et al. J Pediatr 201153	365 VLBW infants	2000–2004	Prolonged initial empirical antibiotics associated with LOS and the combination of LOS, NEC, or death
Greenberg, et al. Pediatr Res 201954	5,730 ELBW infants	2008–2014	Proportion of infants receiving prolonged early antibiotics varied from 30–69% among centers; declined overall from 49% in 2008 to 35% in 2014 Prolonged early antibiotics associated with death but did not reach statistical significance (aOR 1.17 [95% CI: 0.99–1.40], p = 0.07) Prolonged early antibiotics also not associated with LOS or NEC

Footnotes: Studies listed in order of publication. ELBW (extremely low birth weight); LOS (late-onset sepsis); NEC (necrotizing enterocolitis); VLBW (very low birth weight)

With wide attention given to the potential risk of early, prolonged empiric antibiotic administration to ELBW infants, NRN investigators evaluated a cohort of 5,730 ELBW infants born 2008–2014 to determine how practice may be changing. Proportions of infants receiving prolonged early antibiotics varied from 29.6–68.9% among centers in this study, but overall decreased by almost a quarter from 48.9% in 2008 to 35.1% in 2014.⁵⁴ Notably, with lower exposure rates, prolonged early antibiotic administration was not significantly associated with death, LOS, or NEC, although an association was noted with death that did not quite reach statistical significance (OR 1.17; 95% CI [0.99,1.40]; P=0.07).⁵⁴ The authors speculated that multiple practice changes not assessed in the study may have contributed to the findings, and that more targeted antibiotic utilization may have contributed to decreased risk of harm.

Conclusions and Future Directions

In studies conducted among infants with neonatal invasive infections born in and/or cared for at NRN centers from 1993–2017, investigators have provided neonatal clinicians critically important information to inform decision-making, to guide clinical care of both term and preterm infants, to motivate care improvement, to educate clinical learners, to generate hypotheses for further research, and ultimately to counsel families. Incidence data combined with risk profiling informs risk assessment and targeting of empiric antibiotic therapies, while microbiologic data aids in determining antibiotic choice. The very large NRN datasets provide the opportunity to derive meaningful associations between specific care practices and infection-related outcomes of the most vulnerable extremely low-birth weight, very low gestation infants. The identification of factors such as the impact of prolonged, broad-spectrum empiric antibiotic therapy on the occurrence of fungal infection, prolonged antibiotic therapy on risk of LOS, NEC and death, the description of preterm infants at very low risk of EOS, and the longitudinal assessment of pathogen antibiotic susceptibility patterns, to cite a few examples, has provided neonatal clinicians with information that has been applied within a quality-improvement framework to prospectively evaluate clinical impact. Perhaps most important, NRN infection-related studies have highlighted the profoundly negative impact of neonatal infection on the most preterm infants, providing motivation for care improvement efforts as well as the moral imperative to conduct novel research aimed at infection prevention.

Abbreviations

BPD

Bronchopulmonary dysplasia

BW

Birth weight

CDC

Centers for Disease Control and Prevention

CoNS

Coagulase-negative staphylococci

CSF

Cerebrospinal fluid

ELBW

Extremely low birth weight

EOS

Early-onset sepsis

GA

Gestational age

GBS

Group B Streptococcus

IL

Interleukin

LOS

Late-onset sepsis

MRSA

Methicillin-resistant Staphylococcus aureus

MSSA

Methicillin-susceptible Staphylococcus aureus

NDI

Neurodevelopmental impairment

NEC

Necrotizing enterocolitis

NICHD

Eunice Kennedy Shriver National Institute of Child Health and Human Development

NRN

Neonatal Research Network

VLBW

Very low birth weight