Severity of small-for-gestational-age and morbidity and mortality among very preterm neonates

Evaluate the association between small for gestational age (SGA) severity and morbidity and mortality in a contemporary, population of very preterm infants. This secondary analysis of a California statewide database evaluated singleton infants born during 2008–2018 at 24–32 weeks’ gestation, with a birthweight <15th percentile. We analyzed neonatal outcomes in relation to weight for gestational age (WGA) and symmetry of growth restriction. An increase in WGA by one z-score was associated with decreased major morbidity or mortality risk (aRR 0.73, 95% CI 0.68–0.77) and other adverse outcomes. The association was maintained across gestational ages and did not differ by fetal growth restriction diagnosis. Symmetric growth restriction was not associated with neonatal outcomes after standardizing for gestational age at birth. Increasing SGA severity had a significant impact on neonatal outcomes among very preterm infants.

The prenatal diagnosis of fetal growth restriction (FGR), defined as an estimated fetal weight <10th percentile by ultrasound, is intended to identify fetuses who will be born SGA to inform antenatal surveillance, delivery timing and reduce risk of adverse outcomes [12][13][14]. The diagnosis of FGR among infants born SGA has been shown to improve outcomes, yet prior studies have reported detection rates of approximately 20-50% [4,5,[14][15][16][17][18], depending on the severity of SGA, GA and healthcare system [19]. This indicates missed opportunities for detection, enhanced surveillance and risk stratification [13,14]. Increasing FGR detection must be balanced with the risk of overdiagnosis and iatrogenic preterm birth [4,5]. Prenatal diagnosis of FGA and postnatal management of premature SGA have changed over time, necessitating updated evidence.
Historically, compared to symmetric SGA neonates, asymmetric SGA was associated with improved prognosis [20][21][22]. The divergence in prognosis stems from the underlying etiology of SGA. There is limited evidence on how symmetry of SGA may be associated with neonatal outcomes [22].
In this context, we sought to characterize the primary outcome of composite neonatal morbidity and mortality according to a continuum of weight for gestational age (WGA) z-scores and determine if there is a clinically relevant WGA where neonatal risk increases in an early preterm population. We hypothesized that among early preterm infants, <32 weeks' gestation, a lower WGA z-score would be associated with higher neonatal morbidity and mortality. Secondly, we theorized that these associations are attenuated by a prenatal diagnosis of FGR and postnatal asymmetric SGA.

MATERIALS/SUBJECTS AND METHODS Study design
We conducted a secondary analysis of a population-based cohort, the California Perinatal Quality Care Collaborative database (CPQCC), which comprises >90% of neonatal intensive care units (NICUs) in California. The database includes all participating NICU patients who are born 401-1500 grams or 22 0/7 to 31 6/7 weeks gestational age and are admitted on or before 28 days of life [23,24]. The CPQCC links NICU data to birth certificate, death certificate and hospital discharge data. This study was approved by the California Committee for the Protection of Human Subjects (IRB #12-06-0393) and the Stanford University Institutional Review Board (IRB #14746).
We included live, singleton neonates born from 2008 to 2018 at an early preterm gestational age (24 0 -31 6 weeks) and with a birth WGA < 15th percentile [25]. Although SGA is defined as <10th percentile, we included WGA up to the 15th percentile to capture the spectrum of risk associated with lower birthweights. Precedent exists indicating that WGA < 15th percentile is associated with an increased risk of mortality [26,27]. Exclusion criteria comprised severe birth anomalies and missing or implausible birthweight, gestational age, sex, or covariates included in analyses. Implausible was defined as a combination of gestational age and birthweight that were not biologically possible [25].

Definitions
We calculated a WGA z-score for each infant using sex-specific national reference charts generated with sufficient numbers of early preterm infants [25]. We used continuous z-scores for analysis, and converted them to percentiles to assign severity based on clinically relevant cutoffs: [28] <3rd, 3rd to <5th, 5th to <10th, and 10 to <15th percentile. Fetal growth restriction (FGR) was defined by the diagnosis being coded antenatally. During the timeframe of this cohort, an estimated fetal weight <10th percentile was used to define FGR.
The CPQCC database provides a population-based repository of head circumference measurements from which we calculated z-scores. After calculating head circumference z-scores, we excluded records where head circumference values were ≤ (−4) or ≥ (+4). These outliers were set as missing when performing analyses related to head circumference. We followed the method described by Bocca-Tjeertes et al [22]. to classify the study participants with postnatal asymmetric growth restriction or symmetric growth restriction. Infants with a birthweight >1 standard deviation less than the corresponding head circumference were classified as asymmetric growth restriction. All other infants were classified as symmetric growth restriction. This method characterizes outcomes based on SGA etiology and the phenomenon known as head sparing [22].
Maternal, fetal and delivery factors known to influence neonatal outcomes were analyzed. Hypertension included chronic hypertension, gestational hypertension and/or pre-eclampsia. Diabetes included pre-existing or gestational diabetes. Antenatal magnesium and steroids included receipt of these treatments and does not always indicate a full course. Covariates were selected a priori based on literature review, and included: diabetes [4], hypertension [26], and antenatal steroid administration [4,29]. As described below, gestational age at birth [9,10,30] and FGR diagnosis [4,5,17] were also treated as covariates.

Outcomes
The primary outcome was major neonatal morbidity or death. This composite outcome was previously developed by the CPQCC and is defined as the occurrence of any of the following: infant death during birth hospitalization or within 1 year of birth after continuous hospitalization, chronic lung disease, severe intraventricular hemorrhage, infection, necrotizing enterocolitis, severe retinopathy of prematurity (ROP) or ROP surgery and cystic periventricular leukomalacia (Supplementary Table 1) [31].
Secondary outcomes included Cesarean birth, the need for neonatal endotracheal tube ventilation in the delivery room, an Apgar score <4 at 5 min, neonatal death in the delivery room and infant death. These were selected due to the objective nature, clinical relevance and prognostic value.

Statistical analysis
We calculated maternal, fetal, delivery, and neonatal characteristics according to WGA groups <3rd, 3rd to <5th, 5th to <10th and 10 to <15th percentile. We then conducted multivariable modified Poisson regression models to estimate relative risks (RR) and 95% confidence intervals (CI) for the associations between WGA z-score, treated as a continuous exposure, and the outcomes of interest. Linear associations between WGA z-score and the outcomes were assessed and verified by fitting WGA z-scores with restricted cubic splines. The multivariable regression models adjusted for gestational age at birth (weeks), FGR (diagnosed antenatally), diabetes, hypertension, and antenatal steroid administration. To aid in interpretation of the primary results, we plotted the incidence of major neonatal morbidity or mortality across WGA percentiles for GA groups: 24-26, 27-28, 29-30, and 30-31 weeks. We then replicated these models separately among participants with a diagnosis of FGR and among those without a diagnosis of FGR, based on our hypothesis that an antenatal FGR diagnosis modifies the effect of WGA on neonatal outcomes.
We assessed differences for SGA infants with symmetric growth restriction versus asymmetric growth restriction by conducting multivariable modified Poisson regression models for the outcomes of interest, treating asymmetric growth restriction as the reference group. The models adjusted for the same covariates as above. All analyses were conducted in SAS version 9.4. SAS Code is available from the corresponding author upon request. [2008][2009][2010][2011][2012][2013][2014][2015][2016][2017][2018]196,501 neonates were identified within the California NICU database. After exclusions, 4,689 unique singleton neonates, born in early preterm gestation, without severe anomalies, and with a birthweight <15th percentile remained in the cohort (Supplementary Figure). In comparing the WGA categorical groups, maternal co-morbidities were least common and prior Cesarean birth was most common in those <3rd percentile WGA (Table 1). The rate of Cesarean birth ranged from 90-96% based on severity of WGA percentile. The most common indication for Cesarean birth was fetal distress among those <5th percentile and hypertension for those between the 5th to <15th percentile. Neonates <5th percentile WGA were more likely to be diagnosed antenatally with FGR when compared to 3rd to the <5th or 5th to <10th percentile (82% vs 70% vs 55% respectively). In the 10th to <15th percentiles, 37% had been diagnosed with FGR. The neonates in lower WGA percentile groups were less likely to receive magnesium, but more likely to require intubation, have Apgar scores <4 at 5 min, intraventricular hemorrhage, respiratory distress syndrome, neonatal death in the delivery room, infant death or major morbidity or mortality.

From
The incidence of major neonatal morbidity or mortality was plotted based on WGA percentile and stratified by gestational age. The rate of major morbidity or mortality decreased with increasing WGA percentile and increasing gestational age (Fig. 1). Among neonates born 29-32 weeks' gestation, there appeared to be an inflection point at <3rd percentile of WGA, below which major morbidity and mortality increased. In contrast, visually, there was an overall gradual continuum of increasing risk with decreasing WGA percentile among neonates born 24-28 weeks' gestation.
Increasing WGA z-score was associated with improved neonatal outcomes (Table 2), before and after adjusting for FGR, maternal diabetes, maternal hypertension, and antenatal steroids. These relationships persisted when adjusting for gestational age. The relative risk of the primary outcome, major morbidity or mortality, decreased with each unit increase in WGA z-score (aRR 0.73, 95% CI 0.68-0.77, p < 0.0001). This would be interpreted clinically as a neonate in the 14th percentile WGA having a 27% lower risk of major morbidity or mortality compared to a neonate in the 2nd percentile, independent of confounders. Each increase in WGA z-score also demonstrated a reduction in the relative risk of secondary outcomes including: cesarean birth, need for neonatal endotracheal tube ventilation, Apgar score <4 at 5 min, neonatal delivery room death and infant death. The estimated effect of  (7) 101 (7) 90 (6) Necrotizing enterocolitis d 55 (7) 43 (7) 80 (5) 59 (4) Infant death 231 (26) 85 (14) 145 (9) 105 (7) Severe ROP or surgery for ROP 97 (14) 44 (8) 101 (8) 83 (6) increasing WGA z-score was most pronounced for preventing delivery room death (aRR 0.17, 95% CI 0.12-0.24). Indicating that for one increase in z-score, or from the 2nd to the 14th percentile, there is an 83% reduction in delivery room death. Stratifying outcomes by the prenatal diagnosis with FGR (n = 2585) versus without (n = 2104) did not alter the relationship of increasing WGA z-score on neonatal outcomes (Table 3). More neonates had asymmetric growth restriction (n = 3,257) than symmetric growth restriction (n = 1,324), as previously defined by Bocca-Tjeertes et al [22]. (Table 4). Symmetric growth restricted neonates were born at earlier gestational ages compared with asymmetric growth restricted neonates (mean 26.0 [SD 1.4] weeks versus 29.7 [SD 1.3] weeks, p < 0.0001). In the crude regression models, neonates with symmetric growth restriction had an increased risk of major morbidity or mortality, intubation, Apgar <4 at 5 min, delivery room and infant death compared with neonates with asymmetric growth restriction.
These associations persisted after adjusting for maternal comorbidities, FGR, and antenatal steroid use. Specifically, the most pronounced findings among symmetric growth restriction were an increased risk of delivery room death (aRR 24.6, 95% CI 8.9-68.3) and infant death (aRR 5.8 95% CI 4.8-6.9). Given the differential in gestational age of delivery between symmetric and asymmetric neonates, gestational age was added to the adjustment model. When gestational age was added to the adjustment model, there were no significant associations between symmetric growth restriction and any of the outcomes. Thus, observed differences in outcomes may be attributed to gestational age differences.

DISCUSSION
Our study demonstrated several important findings. Among preterm infants born <32 weeks' gestation and <15th percentile Fig. 1 Major neonatal morbidity or mortality as a function of weight for gestational age percentile. Incidence of major neonatal morbidity or mortality by weight for gestational age percentile among 4,771 early preterm (<32 weeks' gestation) infants after stratifying by gestational age. WGA, increasing WGA z-score was associated with a decrease in neonatal morbidity and mortality. We demonstrated this relationship across gestational age groups. Additionally, the relationship between WGA z-score and major morbidity or mortality was not modified by the diagnosis of FGR. Lastly, symmetrically growth restricted neonates had worse outcomes, which was explained by earlier gestational age at delivery. The degree of protection conferred by an increasing WGA percentile was notable for immediate outcomes, longer-term outcomes, and the composite outcome of major morbidity or mortality. Our results parallel conclusions of a large single-center cohort from 1988-1996, which demonstrated an increase in neonatal morbidity and mortality for term neonates ≤3rd percentile, but preterm infants had no inflection point and instead experienced a gradual increase in adverse outcomes with increasing severity of growth restruction [10]. Our findings demonstrate a visual inflection point at <3rd percentile of WGA in neonates 29-32 weeks' gestation, but not <29 weeks. This likely reflects an improvement in outcomes of prematurity since the 1980s and 1990s, particularly for those born very preterm, but not extremely preterm [32].
Preterm SGA neonates are affected by both severity of growth restriction and prematurity [10]. Prematurity is likely acting as an effect modifier, therefore isolating a relationship between growth restriction and neonatal outcomes in a premature population is challenging [2,10]. A recent prospective study combining results from nine Maternal-Fetal Medicine Unit Network studies, found no difference in neonatal morbidity between SGA and appropriate for  gestational age (AGA) neonates who were born <32 weeks' gestation [2]. This study [2] as well as others [30,33] have hypothesized that gestational age dictates neonatal outcomes more than growth restriction, in a premature population. Specifically, when <27 weeks, neonatal survival improves by 2% for each day in utero [33]. This relationship is also evident in our data, where >80% of SGA neonates <27 weeks experienced major morbidity or mortality, but with one additional week of gestation, outcomes improve and the severity of SGA begins to exhibit an effect (Fig. 1). We notably isolated the effect of SGA severity on neonatal outcomes by reporting WGA z-scores and adjusting models for gestational age and maternal comorbidities [30]. Additionally, we restricted the population to <32 weeks, concentrating on pathologically small neonates, as opposed to constitutional growth restriction from biologic variability [34]. Among our high risk, preterm, SGA cohort, we demonstrate a Cesarean birth rate of 90-96% (Table 1), which decreased with each increase in WGA z-score ( Table 2). The intrauterine environment of a growth restricted fetus is hypothesized to activate maladaptive physiologic fetal changes [35]. When uterine contractions are superimposed onto a fetal framework predisposed to hypoxemia, the fetus may be unable to compensate [35,36]. Clinically, this translates into the perception that preterm fetuses with growth restriction will not tolerate labor [37]. A study of preterm births with SGA from 2011-2014 noted a rate as high as 88% [37]. A rise in the national Cesarean birth rate since this publication and the gestational ages included, likely contributes to our higher observed rate.
Prior studies have found FGR to be more specific for extreme SGA percentiles [5,14,18,38,39]. Similarly, our results demonstrate neonates categorized as <5th percentile have a higher likelihood of being diagnosed with FGR compared to the 5 to <10th percentile. In our study, 55-82% of SGA ( <10th percentile) neonates were diagnosed with FGR antenatally, which is higher than prior studies [4,5,14,16,17,19]. Compared with neonates not diagnosed with FGR, the FGR cohort was more severely growth restricted and had a higher risk of infant death. The challenge of FGR identification in a preterm population must be understood in context of the United States prenatal diagnosis system. Growth ultrasounds are performed at the discretion of the obstetrician, driven by maternal or fetal comorbidities or an abnormal screening fundal height. After an initial anatomy scan, not all are screened with growth ultrasounds, and not all who are screened are diagnosed correctly. Based on the cadence of maternal visits in the 2nd and early 3rd trimesters, it is possible to miss an evolving process. Therefore, some selection bias exists: those that receive growth ultrasounds and have the opportunity to be diagnosed with FGR are already deemed higher risk. This may account, in part, for the lack of detection in some SGA infants, and the higher likelihood of worse outcomes in those noted as FGR.
The FGR paradigm of prenatal diagnosis is that detection of pathologically small fetuses allows for surveillance, expedites intervention and leads to improved outcomes [33]. However, studies are conflicting as to whether the diagnosis of FGR affects outcomes [4,5,16,17]. Our study showed antenatal FGR diagnosis did not impact the relationship between SGA severity and neonatal outcomes. This finding may stem from FGR not being sensitive or specific for SGA [15,18], FGR having a higher sensitivity for severe FGR [14], and the heterogeneity in surveillance protocols for detection and management of FGR.
Controversy exists as to the clinical utility of asymmetric growth restriction versus symmetric growth restriction. This dichotomy Table 4. Association between symmetric versus asymmetric growth restriction and outcomes at birth among early preterm infants born <15th percentile WGA, California (N = 4,581, Symmetric growth restriction N = 1324, Asymmetric growth restriction N = 3257). was intended to stratify risk based on the etiology: symmetric growth restriction is correlated with an early insult (infectious or genetic etiologies) and portends a poorer prognosis, whereas asymmetric growth restriction is an adaptive response to a recent event, often utero-placental insufficiency [20,22,40]. This is consistent with our results: compared to asymmetric growth restriction, symmetric growth restriction was associated with an increase in major morbidity and mortality, which was driven by earlier gestational age. Physiologic adaptation to the intrauterine milieu, a lack of reserve and earlier gestation in symmetric growth restriction may contribute to the challenge with immediate transition to extra-uterine life. Similarly, a study of 1,268 SGA neonates, <32 weeks, concluded that asymmetric growth restriction neonates had lower odds of mortality and respiratory complications compared to symmetric growth restriction neonates. Excess mortality in symmetric growth restriction was attributed to respiratory distress [20]. Antepartum studies of symmetry of growth using ultrasound did not demonstrate predictive accuracy [41]. Guidelines recommended against using symmetric growth restriction for prognostication and our results support this in a contemporary, very preterm cohort [28]. The continuum of risk associated with SGA neonates is vital to how physicians counsel patients, delivery timing and to the allocation of population-based resources [33]. While the birthweight is not known to obstetricians at the time of management decisions, outcomes from SGA neonates can be extrapolated to pregnant patients with suspected FGR for counseling purposes [2,30]. Prior studies have challenged whether the categorization of SGA < 10th percentile is still a meaningful threshold for defining risk. Defining SGA using a high percentile jeopardizes applying interventions without benefit, and possible harm. Conversely, using a lower percentile to define SGA may cause missed surveilling an at-risk population [1,17,18]. Some propose different thresholds demarcating interventions, individualized to gestational age [1,42]. According to our data, while neonates in the 10-15th percentile are not without risk, defining SGA as <10th percentile is appropriate, with <3rd percentile being the most atrisk cohort. Understanding this spectrum is vital for both the obstetrician and neonatologist.
Clinically, patients and providers cannot control severity of SGA, but management decisions often dictate gestational age at delivery. Each week gained in gestation during the very preterm period improves outcomes. For example, our data suggests that an infant born at the 3rd percentile WGA at 29 weeks versus 27 weeks' gestation could decrease risk of neonatal morbidity and mortality by approximately 40%. Therefore, a higher threshold for delivery is warranted for growth restricted fetuses <29 weeks' gestation, as composite morbidity and mortality is >50% for all neonates with a WGA <10th percentile.
The detection of FGR in the 5th to <10th percentile of birth weight remains low. Future studies are needed to determine the optimal detection and surveillance protocols that can reduce mortality and morbidity [15,17,43]. While we present data on singleton births, twin gestations are disproportionally responsible for the burden of morbidity in a preterm population [44][45][46][47]. Studies are needed to elucidate how the severity of SGA impacts outcomes by twin status and birth order in a contemporary, very preterm population.
Major strengths of our study include the prospectively collected data and the large, population based, sample size. This afforded the statistical power to detect differences in rare outcomes, such as neonatal death, and prevent selection bias. The level of detail in this contemporary cohort has not previously been demonstrated and allows for precision in counseling amongst a unique population of very preterm SGA neonates.
A neonatal database has limitations, primarily a lack of depth into antenatal management and maternal characteristics. FGR was based on neonatal chart abstraction and therefore misclassification was possible. It is not recorded in the database when FGR was diagnosed, the nomogram used, antenatal surveillance or delivery decisions. Thus, we cannot elucidate what aspect of care is responsible for our observations. Our database also lacks reliable cord pH, which is an important variable to understanding the maternal-fetal interaction and intrauterine milieu at delivery. While the granularity of this database allowed us to investigate asymmetric versus symmetric growth through the use of head circumference, we do recognize the inherent subjectivity in this measurement. Inaccuracies were reduced through a database logic check, including head circumference measured on the day of birth or the following day only and excluding extreme values in our analysis. Lastly, stillbirth is a major adverse outcome related to severe FGR [1,6,43] that our database could not capture. Therefore, we are not reporting the full affliction of FGR.
Across decreasing WGA percentiles, neonatal morbidity and mortality increased in this contemporary, population-based cohort, of very preterm infants across gestational ages. Symmetric growth restriction was less common in preterm gestation and when adjusting for gestational age, was not associated with poorer outcomes.

DATA AVAILABILITY
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.