Contemporary Definitions of Infant Growth Failure and Neurodevelopmental and Behavioral Outcomes in Extremely Premature Infants at Two Years of Age

Introduction: The association of 2-year neurodevelopmental and behavioral outcomes with in-hospital or post-discharge growth failure (GF) using contemporary definitions for preterm infants is unknown. Methods In a secondary analysis of a preterm cohort, changes in anthropometric z-scores were examined between birth and hospital discharge, and from discharge to 2 years. The 2-year evaluation included Bayley Scales of Infant Development (BSID-III) and Child Behavior Checklist (CBCL). Results Among 629 infants, accelerated linear growth from birth to discharge was associated with higher BSID-III cognitive scores (+ 3.2 points [IQR 0.02, 6.4]) while in-hospital GF was not associated with any outcomes. Infants with weight GF after discharge had lower BSID-III motor scores (−3.1 points [−5.9, −0.2]). Infants with accelerated weight growth after discharge had increased odds of behavioral problems on the CBCL (aOR 1.9 [1.03, 3.5]). Discussion In-hospital and post-hospitalization growth metrics are modestly associated with neurodevelopmental outcomes with length gains apparently most beneficial.


INTRODUCTION
Providing adequate nutrition to the premature infant for the rst two years is essential for neurodevelopment, the gut microbiome, bone health, and metabolism (1)(2)(3)(4). The most common markers of adequate nutritional status are serial measurements of anthropometrics (weight, length, and head circumference). However, the de nition of inadequate growth, or "growth failure," during the neonatal intensive care unit (NICU) admission is controversial (5). Furthermore, there is limited information on what is optimal growth once the infant is discharged from the NICU.
Extremely preterm infants are at particularly high risk for both growth failure and neurodevelopmental impairment. Early studies examining growth failure in extremely low birth weight (ELBW) infants de ned growth failure based on changes in growth velocity and small for gestational age status at 36 weeks postmenstrual age (PMA) (6-8). Greater weight gain was associated with improved Bayley Scales of Infant Development (BSID)-II mental and psychomotor subscales at 18 months (6, 8) and at 22 months (6). Similar ndings were found for head circumference growth (6, 8). However, increased weight gain or "catch up growth," has also been associated with lower lean body mass and increased total body fat in adolescence (9). Catch up growth in infancy has additionally been associated with later obesity and increased risk for metabolic diseases (10). Studies utilizing growth metrics such as growth velocity and small for gestational age status at 36 weeks are limited because they did not take into account the infant's initial growth percentiles.
To address this concern, many groups now consider changes in weight or length z-score as a better measure for adequate growth (2,(11)(12)(13). Studies of the relationship between changes in z-score during the initial hospitalization of ELBWs and neurodevelopmental outcomes show mixed results (8, 14,15), and there has been minimal research evaluating the association of z-score changes after discharge with neurodevelopment (16).
The objective of this study was to examine changes in weight, length, and head circumference z-scores from birth to discharge and from discharge to two years and their associations with neurodevelopmental and behavioral outcomes. We hypothesized that poor linear growth would be associated with adverse neurodevelopmental and behavioral outcomes at two years of age.

Patient Population:
This study is a secondary analysis of the Preterm Erythropoietin Neuroprotection (PENUT) Trial (NCT #01378273) (17). PENUT was a randomized, double-blinded, placebo-controlled trial of erythropoietin for neuroprotection in infants born 24 through 27 completed weeks of gestation. The study was conducted at 19 sites and 30 neonatal intensive care units in the United States between December 2013 and September 2016 (17). All infants enrolled in the PENUT trial who survived and were assessed for longterm developmental outcomes were eligible for this study excepting those requiring invasive mechanical ventilation at discharge, as their growth may be particularly disrupted (N = 17) (17). The PENUT Trial was approved by an institutional review board at each site. Parental consent was obtained prior to infant enrollment. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) reporting guidelines.
We collected data about maternal characteristics, pregnancy, and delivery, as well as infant characteristics including anthropometric measurements, time to regain birth weight, exposure to medications, and comorbidities during their NICU stay (17). At 20 to 33 months corrected age (CA), infants were evaluated by certi ed examiners who assessed cognitive, motor, and language development with the BSID-lll. All BSID-III subscales were scaled based on the age at which the assessment was performed. Child Behavior Checklist (CBCL) for ages 1-5 years was administered, and T scores for emotionally reactive, anxious/depressed, somatic complaints, withdrawn, sleep problems, attention problems, aggressive behavior problems, depressive problems, anxiety problems, autism spectrum problems, attention de cit/hyperactivity problems, oppositional de ant problems, and total problem score were collected. The Modi ed Checklist for Autism in Toddlers (M-CHAT-R) was also given, and the total score was collected.

Growth Parameters Analyzed:
Birth to Discharge. Weight, length, and head circumference were collected at birth, 14 days, and discharge. Fenton growth curves were utilized to calculate weight, length, and head circumference zscores for growth parameters during the NICU hospitalization (18). Weight, length, and head circumference growth failure were de ned as: z-score discharge -z-score birth ≤-0.8, as de ned by the Academy of Nutrition and Dietetics (13). Weight, length, or head circumference accelerated growth was de ned as: z-score discharge -z-score birth ≥0.8 for weight or length. Normal growth was de ned as a zscore change of -0.799 to 0.799. If z-score changes were > 3 or <-3, they were truncated to 3 or -3 to avoid bias in likely input errors.
Discharge to two years. Weight, length, and head circumference were measured at two years corrected age (CA), and body mass index (BMI) z-scores were calculated using Centers for Disease Control and Prevention (CDC) growth curves (19). Changes in weight, length, and head circumference z-score were assessed from discharge to two years CA. Weight, length, or head circumference growth failure were de ned as: z-score 2 years -z-score discharge ≤-0.8 (13). Weight or length accelerated growth was de ned as: z-score 2 years -z-score discharge ≥0.8. Normal growth was de ned as a z-score change of -0.799 to 0.799. If z-score changes were > 3 or <-3, they were truncated to 3 or -3 to avoid bias in likely input errors.

Statistical Analyses:
Summary demographic variables are presented as mean and standard deviation (SD) except for days to regain birthweight which is presented as median with interquartile range (IQR). Z-score calculations were completed using the peditools library in R, and the cohorts were divided into growth failure, normal growth, and accelerated growth for weight, length, and head circumference growth. For all inferential analyses, generalized estimating equations (GEE) with robust standard errors were used to appropriately account for potential correlation of outcomes for same-birth siblings (20). Baseline and demographic factors were compared across growth trajectory groups using a multivariate Wald test.
Linear GEE models adjusted for gestational age, maternal education, > 14 days of dexamethasone, length z-score at birth, severe intracranial hemorrhage (grade III or IV) (ICH), erythropoetin use, and feeding status at discharge were used to compare growth categories with respect to neurodevelopmental scores (BSID-III and M-CHAT-R). GEE logistic regression models were used to determine adjusted odds ratios (aOR) for having a borderline/clinical score in CBCL sub-parameters based on post-discharge growth trajectory, adjusting for gestational age, maternal education, > 14 days of dexamethasone, length z-score at birth, severe ICH, erythropoetin use, and feeding status at discharge. Model outputs are presented as adjusted mean difference or aOR with 95% con dence intervals (CI). All analyses were conducted using R statistical package (Version 4.1.2, Foundation for Statistical Computing, Vienna, Austria). A p-value < 0.05 was considered statistically signi cant.

Growth from Birth to NICU Discharge
Six hundred and twenty-nine infants met criteria for this secondary analysis. Maternal and child characteristics for infants with length growth failure, normal growth, and accelerated growth are shown in Table 1. Maternal and child characteristics for infants with weight growth failure, normal growth, and accelerated growth are shown in Supplementary Table 1. Common co-morbidities of prematurity including necrotizing enterocolitis (stage 2b or 3), severe ICH, severe bronchopulmonary dysplasia (BPD), symptomatic culture positive sepsis, and length of stay are included. The linear growth failure group had more mothers with a high school education or less and infants born at a lower gestational age ( Table 1). The weight growth failure group were less likely to be small for gestational age at birth (Supplementary Table 1). Table 1. Cohort subject characteristics during the timeframe of birth to hospital discharge. Growth failure and accelerated growth is de ned as having length growth failure or accelerated length growth at discharge. Categorical variables are represented by n (%). Continuous variables are represented with mean (SD) or median (IQR). P-values were calculated with Chi-squared tests for categorical and ANOVA for continuous variables. Severe NEC was de ned as Bell's stage 2b to 3. Severe ICH was de ned as grade III or IV either unilateral or bilateral, according to Papile staging. BPD was de ned as requiring nasal canula or higher levels of respiratory support at 36 weeks PMA. Severe sepsis was de ned as cultureproven bacterial or fungal sepsis resulting in blood-pressure support or substantive new respiratory support. Growth from Discharge to Two Years Cohort characteristics are described for weight in Table 2 and length in Supplementary Table 2. Mothers of infants with accelerated weight and linear growth after discharge were more likely to have a high school education or less (Table 2 and Supplementary Table 2). Infants with weight growth failure after discharge were more likely to have had severe ICH and severe BPD ( Table 2). Those with accelerated linear growth after discharge were likely to be lower gestational age at birth and small for gestational age (Supplmentary Table 2). After adjusting for gestational age, maternal education, >14 days of dexamethasone, erythropoietin use, length z-score at birth, PIH, severe ICH, and feeding status at discharge, weight growth failure at follow-up was associated with decreased motor (-3.1 [-5.9, -0.2], p=0.035) scores, but not language (-2.0 [-5.0, 1.0] or cognitive scores (-1.1 [-4.0, 1.6]), Figure 2. There were no differences in BSID-III cognitive, motor, and language scores with changes in length or head circumference z-score.
When adjusting for gestational age, maternal education, >14 days of dexamethasone, erythropoietin use, length z-score at birth, PIH, severe BPD, feeding status at discharge, and severe ICH, infants with accelerated weight growth at two years had increased odds of total behavioral problems compared to those with normal growth and growth failure (OR 1.88 [1.03,3.45], p=0.041, Figure 3). Behavioral outcomes on the CBCL did not differ between linear or head circumference growth cohorts. There were no signi cant differences in M-CHAT-R scores.

DISCUSSION
This large, multi-center study examined the association between growth parameter z-score changes to two year neurodevelopmental and behavioral outcomes among extremely premature infants. Utilizing contemporary de nitions of growth failure currently endorsed by Academy of Nutrition and Dietetics (13), we did not nd signi cant differences in BSID-III scores based on growth parameters. However, we found weight growth failure during initial hospitalization was associated with an increased M-CHAT-R score, more concerning for autism. In contrast, we found accelerated linear growth was associated with a modest increase in BSID-III cognitive score and decreased M-CHAT-R. Weight growth failure from discharge to two year follow-up was associated with modestly decreased motor score, and accelerated weight z-score gain was associated with increased total behavioral problems. These ndings suggest the association of in-hospital growth failure with neurodevelopmental outcomes is not as robust as once believed, and potentially needs to be re-evaluated.
Studies examining growth during the NICU hospitalization and subsequent neurodevelopment are heterogenous and con icting. Different metrics (velocities vs. changes in z-scores), differing time points of evaluation (birth vs. nadir, 36 weeks corrected age vs. discharge), and different growth curves (Fenton vs. INTERGROWTH) have been used. In this study, we utilized changes in z-scores utilizing the growth failure de nition applied by Goldberg et al. using the Fenton growth curve from birth to discharge (13,18). One study examining ELBW and very preterm infants utilizing these de nitions did not nd differences in neurodevelopmental outcomes on the BSID-II at 12 months (1). In contrast, one study found a decline in weight z-score > 1.2 to be associated with increased days to oral feeding (21), a known strong predictor of neurodevelopmental outcomes, and another found a weight z-score change > 1 to be associated with BSID-III scores (22). Other studies are similar to ours, identifying poor linear growth as detrimental to cognitive and language outcomes (8, 15,23). These results call for standardized guidelines to be established prior to applying these de nitions for clinical and quality improvement initiatives.
This study found weight growth failure from discharge to two years was associated with lower BSID-III motor scores. There is less research on growth following NICU discharge in extremely preterm infants.
One study found that increase in head circumference z-score from discharge to follow-up was associated with improved mental processing composite score (24). Another large cohort study found weight velocity following discharge decreased the odds of neurodevelopmental impairment in a non-linear fashion (25). A study examining growth from discharge to two years found gains in weight/length z-score was associated with decreased odds of cognitive impairment at 10 years old (16). These studies suggest that growth following NICU discharge should be closely monitored by the pediatrician.
Although there are multiple studies showing preterm infants are at risk for behavioral and psychiatric problems, this is one of the few studies examining behavioral outcomes and growth trajectories during and after hospitalization (26-28). We found there were increased total behavioral problems with accelerated weight growth after discharge. One study found poor fetal growth in utero was associated with increased risk for attention-de cit hyperactivity disorder (29). Another recent study in extremely preterm infants did not nd differences with different growth trajectories during NICU hospitalization and behavioral/psychiatric outcomes (16). Our study also found modest (< 1 point) differences in M-CHAT-R scores during NICU hospitalization. Accelerated linear growth during NICU hospitalization was associated with a decreased M-CHAT-R score and those with weight or head circumference growth failure during NICU hospitalization had an increased M-CHAT-R score. Of note, we used the M-CHAT-R. Previous research utilizing the original M-CHAT in ELBW infants found poor sensitivity and positive predictive value for autism (30). However, a recent study examining the M-CHAT-R and comparing that to the CBCL and the BSID in infants less than 30 weeks' gestation found they were associated with neurobehavioral and CBCL outcomes, suggesting the revised M-CHAT may have utility as a developmental screen (31).
The previously mentioned study and our study call for further research regarding extremely premature infants and autism screening.
One potential reason we did not appreciate substantial differences in neurodevelopmental outcomes in infants with weight growth failure during NICU hospitalization is due to the methodology of examining birth weight z-score rather than weight nadir z-score. The use of weight at birth versus nadir weight is currently controversial. Weight nadir z-score may better re ect expected physiologic weight loss (32). A study examining weight gain velocity calculations recommended the utilization of the weight nadir or day seven weight rather than birth weight (33). Day seven was deemed to not be signi cantly different than nadir and less laborious for clinicians and researchers (33). Our study chose to use birth weight z-score as this described in Goldberg et al. as this information is standardly available at majority of NICUs and thus is more generalizable.
We anticipate that linear growth affected cognitive and autism screening due to differences in lean body mass (bone, muscle, water). Compared to weight and head circumference, length is understood to be the anthropometric measure best associated with lean body mass/fat-free mass (34). Fat-free mass has been associated with improved neurodevelopmental outcomes. One study examining preterm infants found higher four month to four year fat-free mass gains were associated with improved full-scale IQ and speed of processing performance, whereas term to four month fat mass gains were associated with lower working memory (35). In infants with congenital gastrointestinal anomalies, higher fat free mass was associated with improved vocabulary, cognitive exibility, and general cognitive function at preschool age (36). Another study examining extremely preterm infants found higher fat-free mass z-score was associated with larger brain size at term (37). Currently, body composition tools are deemed costly, and cannot be utilized in unstable infants. Further research involving use of low-cost tools such as ultrasound and bioimpedance are necessary to apply them in the clinical arena.
The strength of this secondary analysis is the utilization of a large, contemporary cohort of extremely premature infants across 30 NICUs in the United States (17). This study included a battery of developmental tests, including not only the use of BSID-III scores, but also CBCL and MCHAT-R scores (17). This allows for a more comprehensive developmental assessment. We also applied the use of anthropometric z-score assessments which are being adapted by many quality improvement networks in the United States; thus, our results are relevant to current clinical practice. Accounting for all of these factors, we believe these results are generalizable to other US-based NICUs.
This study also has limitations. It is a retrospective chart review, and nutritional management continues to evolve. Because the study was a multi-center randomized control trial for a different purpose, there is a lack of serial growth measurements during the hospital stay and post-discharge. Lengths were also performed per-unit protocol, and it is unclear if all sites were using length boards, which have been shown to be more accurate measures of an infant's length (38). Because of the multi-center nature, there is not one standard parenteral and enteral nutritional protocol. We also do not have information on dietary intake after discharge.
In conclusion, extremely preterm infant growth trajectories during and after the NICU hospitalization showed modest associations with neurodevelopmental and behavioral outcomes. In-hospital linear growth appeared to be the most neuroprotective. These ndings suggest that our current nutritional guidelines de ning growth failure during NICU hospitalization may require further evaluation. Furthermore, nutritional studies exploring how to improve linear growth and how to measure lean body mass are imperative to improve neonatal nutrition in preterm infants. Also, further research is required on optimal growth trajectories after NICU hospitalization. Our study elicits further questions to be raised regarding the optimal metrics of adequate nutrition during hospitalization and following discharge in the extremely preterm infant.

Declarations
Clinical Trial Registration: This study is a secondary analysis of pre-existing data from the PENUT Trial Registration: NCT01378273 Con icts of Interest: None of the authors report any con icts of interest. infants who experienced weight growth failure from discharge to two year follow-up to those that did not, adjusting for gestational age, maternal education, >14 days of dexamethasone, erythropoietin use, length z-score at birth, pregnancy induced hypertension, severe bronchopulmonary dysplasia, and grade 3-4 intracranial hemorrhage. Estimated mean difference in the accelerate growth group with 95% CI is presented above the boxplot, with signi cant differences in bold.

Figure 3
Total Behavioral Problems score and post-discharge weight growth. Odds of borderline/clinical Child Behavioral Checklist total problem score in infants with accelerated weight growth, weight growth failure, and normal weight growth from discharge to two years. Number within each group as well as proportion and percentage with borderline/clinical scores within each group is shown. Adjusted odds ratio (aOR) for borderline/clinical total problem score compared to those with normal growth is shown adjusted for gestational age, maternal education, >14 days of dexamethasone, erythropoietin use, length z-score at birth, pregnancy induced hypertension, severe bronchopulmonary dysplasia, and grade 3-4 intracranial hemorrhage.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryTables.docx