Multiple perinatal investigators have identified phenotypes of preterm birth based on reason of delivery.1,2,3,4,5,6,7,8,9 Conceptually such a framework has two potential applications. First, better understanding of the different mechanisms leading to preterm birth may allow for specific interventions to prevent or delay delivery. For instance, medically indicated deliveries for maternal or fetal health may be driven by different disease pathways compared to preterm birth occurring due to spontaneous onset of preterm labor, and necessitate different approaches to prevention.10 Second, such a framework may add an explanatory factor for why preterm infants have varied outcomes.8,9,11 Beyond individual genetics, variability in health outcomes within the same gestational age category is multifactorial in origin. Potential reasons include deleterious adaptation to early-life stressors, organ injury, exposure to pro-inflammatory stimuli, and the aftermath of aberrant repair.12 Understanding the contribution of early-life factors towards variability in later health outcomes is especially attractive, as targeted interventions at this stage could prevent a series of maladaptations that may be more difficult to rectify later in life.

In this issue of Pediatric Research, Letouzey et al.13 use this framework of reason for preterm birth, to report an association between being delivered for maternal hypertensive diseases or isolated fetal growth restriction (FGR) and late onset bacteremia among very preterm infants (<32 weeks gestation).13 Using a large population-based cohort of 2052 very preterm infants who survived >72 h, the authors report an incremental hazard of 1.7 (95% CI 1.2–2.5) for infants delivered for maternal hypertensive disorders without associated FGR (n = 278), 2.6 (95% CI 1.9–3.6) for infants born to mothers with hypertensive disorders and associated FGR (n = 295), and 2.9 (95% CI 1.9–4.4) among infants delivered for isolated FGR (n = 111), compared to infants delivered after spontaneous onset of preterm labor (n = 824) when adjusting for gestational age, infant sex, maternal age, maternal body mass index, mother’s place of birth, and maternal smoking during pregnancy. The risk among infants delivered after preterm premature membrane rupture or abruption was not significantly different.

While the authors do not report the degree of overlap between infants with FGR and postnatal categorization of small for gestation age (SGA), increased risk of late onset sepsis (LOS) in SGA infants has been previously reported.14,15 Hypothesized reasons for their susceptibility include neutropenia and low IgG levels. However, interventions to rectify these deficiencies using granulocyte-stimulating factors and immunoglobulin replacement have not resulted in LOS reduction.16,17 The authors of the current study use additional analysis to exclude central line duration as an intermediary variable that may account for the association. In addition to making a strong case for why the differences are unlikely to be explained by differential need for central lines, the authors discuss other hypotheses including epigenetic changes and difference in delivery processes among the groups to explain the observed association.

The study demonstrates an interesting association that invites further inquiries. One approach may be to explore the pathogens involved. The authors do not describe the microbiology of the late onset bacteremia across the birth groups but note that 66.4% of cases were from Coagulase-negative staphylococci (CoNS), an organism whose role as a contaminant versus true pathogen that can difficult to discern in clinical care, especially if only one culture is obtained. This in conjunction with SGA infants being a higher-risk group18 and potentially being subjected to more frequent sepsis evaluations introduces the risk of ascertainment bias—more frequent evaluations increase chances of isolating contaminants. Work by others have shown that CoNS is the main driver of the difference in rates of late onset bacteremia between SGA and non SGA infants.14 However, it is unlikely that all CoNS cases are contaminants. Are there specific CoNS virulence factors that give this pathobiont an advantage in preterm infants with antenatal stress? Does in utero stress alter host innate immunity in ways that favor CoNS over other common intestinal microbes? If there is a specific predilection to CoNS that is unexplained by use of medical devices, then revealing the mechanisms involved could inform prevention.

A second line of investigation may involve defining preterm birth categories differently. While the heterogeneity of preterm birth is recognized, the optimal grouping to reveal meaningful categories that predict neonatal outcomes is less established. Approaches to create categories have ranged from classification rules made by experts to use of machine learning and clustering techniques.2,5,6 Most of these approaches categorize preterm birth by maternal presentation, intentionally avoiding consideration of delivery mode or other delivery management that could be practice based. This approach is robust when aiming to understand mechanisms leading to preterm birth in order to target prevention. However, when aiming to predict the complications and outcome of an infant delivered prematurely, such an approach loses critical information. Obstetric practice and delivery management decisions clearly influence the initial clinical presentation of the infant that then impact all subsequent outcomes. For instance, administration of antenatal steroid impacts the severity of illness of a preterm infant at birth and survival thereafter. What is practice in one instance becomes “biology” in the next.

Letouzey et al.’s approach uses only maternal presentation and reveals an association with late onset bacteremia in the neonate. In contrast, a strategy that uses both patient characteristics, such as onset of spontaneous preterm labor, and practice characteristics, such as delivery mode, is used to explain differential risk of early onset sepsis in preterm infants.19 Preterm infants compared to term infants have a manifold higher risk of early onset sepsis, acquired by vertical transmission of pathogens during delivery. As a result, most preterm infant were routinely started on empiric antibiotics at birth.19,20 However, infants born to mothers via cesarean section prior to onset of labor or membrane rupture have a minimal risk of pathogen transmission from mother to child, and comprise over 30% of the very preterm population. This low association between delivery criteria and risk of early onset sepsis was confirmed in cohort studies and the current national guidelines recommend that infants meeting these criteria be managed without routine initiation of empiric antibiotics at birth.20,21,22 Taking this concept further, it is possible that vertical transmission of commensal/mutualistic microbes may be as important to neonatal outcomes as vertical transmission of pathogens. Given that what colonizes is also that which may infect, infants that differ in the trajectory and composition of microbiome formation would differ in prevalence and microbiology of late onset infection. Stepping back and querying the reason for preterm delivery of such infants, i.e., delivered in this “sterile” manner, one will find the vast majority are delivered for maternal hypertensive disorders and for FGR. Thus, the reason for preterm birth drives obstetric practice that together result in specific patterns of clinical presentation in the neonate. Pursuing a comprehensive categorization of birth characteristics that combines reason for birth and the management it triggers may be a next logical step to understand if and how these factors effect outcomes.

Lastly, evidence is strong that fetal physiology has both short- and long-term impact. Biomarkers in umbilical cord blood of preterm infants are associated with increased risk of bronchopulmonary dysplasia and pulmonary hypertension suggesting this progression may begin in utero.23,24 Are there fetal epigenetic changes or other fetal biochemical processes that alter susceptibility to LOS, as has been reported for metabolic syndrome and cardiovascular disease?25

How does one parse out the multiple presentations, and subsequently triggered practices, into cohesive, clinically useful groups that reveal why a baby may have a specific predilection? Can we alter aberrant patterns of fetal or early life inflammation in a targeted fashion (as is common in the adult inflammatory diseases so convincingly portrayed in advertisements) to prevent some of the complications of prematurity? Can we change the iatrogenic intestinal dysbiosis that is common in preterm infants with probiotics and if so what combination of microbes? In the last decade large trials of both probiotics and bovine lactoferrin failed to significantly reduce LOS rates. Are we randomizing the right babies, or are we diluting our effects by not recognizing key target groups? Letouzey et al.’s paper provides an interesting association, but we need to dig so much further into the start of the preterm infant’s journey to improve outcomes. In the words of James R. Sherman, “You can’t go back and make a new start, but you can start right now and make a brand new ending”. We need to understand the beggining of preterm birth better to ensure a better outcomes for preterm infants.