Longitudinal urine metabolomics predicts high-grade brain injury in very preterm infants

Molecular profiling of preterm infants' urine offers a promising noninvasive strategy to identify biomarkers. In this prospective study, urine samples were collected from 64 preterm infants ( < 32 weeks gestation) on postnatal days 1-4, 6, 8, 28 and at term-equivalent age. Forty-three metabolites and neurotransmitters were quantified by LC-MS/MS. Brain injury severity was quantified using the MRI-derived Global Brain Abnormality Score, capturing white matter, cortex, deep grey matter, and cerebellar injury. Distinct temporal patterns of urinary metabolites were associated with global brain injury severity. Early energy-related elevations (lactate, α-hydroxybutyrate) were followed by broader alterations in amino acid, fatty acid, and tryptophan-kynurenine pathways between postnatal days 3-6, defining a promising neurosurveillance window that may implicate mitochondrial dysfunction, oxidative stress, and excitotoxicity as relevant injury mechanisms. At the default decision threshold, multi-block modeling identified 100% of severe injury patients (AUC = 0.849) using day 3-6 data. Exploratory prediction of region-specific MRI anomaly was most robust for cerebellum (AUC = 0.891; full model) and white matter (AUC = 0.724; early model). Urine metabolomics may detect evolving preterm brain injury before MRI at term-equivalent age, with notable results between days 3 and 6. Combining metabolomic insights with clinical and imaging data could allow earlier risk assessment and neuroprotective treatment in preterm infants. IMPACT: Serial urine metabolomics in very preterm infants captures dynamic metabolic disturbances associated with brain injury. Urine patterns show possibly region-specific brain vulnerability. Days 3-6 define a critical window for stratifying risk of severe brain injury at term-equivalent age, shifting focus toward earlier care decisions.