Observational studies using term-equivalent qMRI have linked preterm maternal milk exposure to improved total and tissue-specific brain growth, white matter development, and functional connectivity in a dose-dependent fashion compared to formula feeding.9,12,28,29,30,31,32,33,34 We previously reported a positive association between donor milk intake and structural brain development in a small subset of human milk-fed very preterm infants as part of an exploratory secondary analysis, but the effects of donor human milk on preterm brain development have not been specifically explored using qMRI.30 In this cohort study of very preterm infants receiving nutritional support with pasteurized donor human milk, mother's own milk, or preterm formula, we aim to investigate the relationship between type of primary enteral nutrition and structural brain development at term-equivalent age. Using mother's own milk as the reference standard, we hypothesize that term-equivalent qMRI will reveal greater differences in brain size and white matter maturation between preterm formula and maternal milk-fed infants than those supported with donor human milk.
Preterm infants admitted to the level IV NICU at Children's National Hospital (Washington, D.C.) from March 2012 through September 2022 were enrolled as part of a prospective, observational study of the antecedents and sequelae of prematurity-related brain injury. As a freestanding children's hospital, all infants admitted to the NICU were out born. Infants born very preterm ( ≤32 weeks gestational age) and/or very low birth weight ( <1500 g) were eligible for inclusion if they were admitted within the first two weeks of life and received at least two weeks of enteral nutrition during NICU hospitalization. Infants with a known or suspected underlying genetic syndrome, metabolic disorder, or perinatal central nervous system infection were excluded from enrollment. Additionally, infants were excluded if either screening head ultrasound or subsequent MRI demonstrated structural brain malformations or significant injury (grade III intraventricular hemorrhage, periventricular hemorrhagic infarction, or cerebellar hemorrhage). This study was approved by the Children's National Hospital Institutional Review Board, and informed, written consent was obtained from the parents of all participants. This study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational cohort studies.
Infant race, ethnicity, and home address were self-reported; the remainder of demographics and clinical data were extracted from the medical record. Home address was used to calculate the overall Social Vulnerability Index (SVI), with indices ranging from 0 to 1 representing low to high vulnerability. Common comorbidities known to affect preterm nutritional advancement and neurodevelopment outcomes were also evaluated, including any diagnosis of infection (positive microbial test or ≥7 days of antibiotic treatment), systemic steroid treatment (hydrocortisone or dexamethasone), bronchopulmonary dysplasia (respiratory support administered at 36 weeks postmenstrual age), surgical necrotizing enterocolitis or spontaneous intestinal perforation (requiring laparotomy or drain placement), patent ductus arteriosus, retinopathy of prematurity (greater than or equal to stage 2), and intraventricular hemorrhage (Papile's grading system). Weekly anthropometric measures (weight, length, head circumference) were also extracted from the medical record, with growth faltering defined as a decrease in weight z-score by >1 standard deviation from birth to term-equivalent age.
Daily nutritional intake (parenteral and enteral) was retrospectively collected from the electronic medical record. The parenteral nutrition protocol initiated protein upon NICU admission with 3 g/kg/day of amino acids and advanced the following day to a goal of 3.5-4 g/kg/day. Intravenous lipids were initiated at 1 g/kg/day within 24 h of NICU admission and advanced by 1 g/kg/day, as tolerated, to a goal of 3 g/kg/day.
A standardized feeding protocol for very preterm infants detailed specific parameters based on birth weight and feeding tolerance. Trophic enteral feeds were initiated at 10-20 ml/kg/day within 24 h of life or once the infant was deemed clinically stable by the medical team, followed by daily advancement by 20 ml/kg/day, as tolerated, to a goal feeding volume of 160 ml/kg/day. Growth faltering was managed by increasing the caloric density of feeds (i.e. via the provision of additional formula supplementation to achieve >24 kcal per ounce) or addition of modular macronutrient supplements (liquid protein, MCT oil), based upon the primary enteral nutrition type and infant's growth velocity. Pasteurized donor human milk was available beginning in June 2014 and was provided with parental consent when maternal milk supply was insufficient until approximately 34 weeks corrected age, at which point infants were gradually transitioned to preterm formula. Maternal (unpasteurized) and donor human milk were both fortified in the same fashion with a liquid bovine-based human milk fortifier to assumed 22 kcal per ounce at a feeding volume of 80 ml/kg/day and 24 kcal per ounce at a feeding volume of 100 ml/kg/day. Daily enteral feeding volume was recorded from admission until discharge or TEA MRI, whichever was achieved sooner. Preterm infants were categorized by primary enteral nutrition type of mother's own milk, donor human milk, or preterm formula ( >50% cumulative enteral feed volume).
Non-sedated brain MRI studies were performed at TEA on a 3 Tesla MRI scanner (Discovery MR750; General Electric Medical, Systems, Waukesha, Wisconsin) with an 8-channel receiver head coil approved for safety in neonates. Infants were immobilized using an InfantVacuum Immobilizer (Newmatic Medical, Caledonia, Michigan) and provided with double ear protection. The MRI acquisition protocol included structural imaging with T2 3D-cube and T1 3D-spoiled gradient recalled images (T2: 84 ms echo time, 2500 ms repetition time, field of view 13 cm, 1 mm slice thickness, 160 × 160 acquisition matrix; T1: 3.8 ms echo time, 6.7 ms repetition time, field of view 13 cm, 1 mm slice thickness, 700 ms inversion time, 160 × 160 acquisition matrix). DTI acquisition consisted of a single-shot, echoplanar sequence with 27 or 64 (High Angular Resolution Diffusion Imaging-HARDI) noncollinear direction diffusion gradients with an effective high b-value of 2500 s/mm (3 b = 0 s/mm; HARDI 4 b = 0 s/mm), 80 ms echo time, 8000 ms repetition time, field of view 200 × 200 mm, 3 mm slice thickness and no gap with a 128 × 128 acquisition matrix. Each MRI study was reviewed by an experienced pediatric neuroradiologist (JM) and assigned a Kidokoro score for white matter injury reflecting cystic lesions (score 0-4) or focal signal abnormalities (score 0-3) using T1 and T2 structural imaging.
Volumetric segmentation was performed on coronal T2 Cube 3D images using a validated automated algorithm via the Draw-EM (Developing brain Region Annotation With Expectation-Maximization) package, with subsequent manual inspection and correction as needed by two investigators blinded to enteral nutrition type (KO and KK). Tissue-specific brain volumes were obtained for the cortical and deep gray matter, white matter, amygdala-hippocampus, cerebellum, and brainstem; total brain volume was calculated as the sum of all tissue-specific brain volumes for each infant (Fig. 1). DTI data were preprocessed based on a previously published pipeline, with cubic regions of interest (21-49mm) manually placed by two investigators blinded to enteral nutrition type (KO and KK) using predefined anatomical landmarks in the corpus callosum (genu and splenium), posterior limb of internal capsule, and brainstem (pons). Fractional anisotropy (FA) and mean diffusivity (MD) values, measuring directionality and net diffusion of water molecules, respectively, were calculated for each region of interest. Inter- and intra-rater reliability measures for manually corrected MRI brain volumes and DTI region of interest placement were calculated based on a randomly selected subset of 35 patients using the intraclass correlation coefficient.
Statistical analysis was performed using R Software (R Studio version 2023.12.1.402). Patient characteristics were reported using means and standard deviations (SD) or medians and interquartile ranges (IQRs) for normally and non-normally distributed continuous measures, respectively; categorical data were reported as frequencies (percentages). Univariate analyses using analysis of variance (ANOVA) or Kruskal-Wallis tests for continuous data and chi-square or Fisher's exact tests for categorical data were conducted to identify patient characteristics that differed significantly based on enteral nutrition type. Generalized linear models adjusting for gestational age at birth and postmenstrual age at MRI were used to compare quantitative brain MRI measures (volume and DTI) based on enteral nutrition type. In these models, mother's own milk was used as the reference enteral nutrition type against which donor human milk and preterm formula were compared. Additional models further incorporated infant sex and SVI as potential covariates. P values were two-tailed, with threshold p < 0.025 considered statistically significant to account for multiple comparisons, and confidence intervals set at 95%. Quantitative brain MRI measures (volume and DTI) between donor human milk and formula-fed infants were compared as a post-hoc analysis. A sensitivity analysis to assess nutritional exposure effect was performed between maternal and donor human milk-fed infants using a threshold of >70% of cumulative enteral nutrition.