Longitudinal Urine Metabolomics Predicts High-Grade Brain Injury in Very Preterm Infants

Preterm birth leaves many infants vulnerable to brain injury, but detecting damage early is notoriously difficult. A new viral science news report highlights research that uses longitudinal urine metabolomics to forecast high-grade brain injury before magnetic resonance imaging (MRI) can reveal it.

In a prospective study, researchers collected urine from 64 very preterm infants (<32 weeks gestation) on multiple postnatal days: days 1–4, 6, 8, 28, and again at term-equivalent age. The goal was to map how small-molecule signatures in urine change over time in relation to later neurologic outcomes.

Across the samples, investigators quantified 43 metabolites and neurotransmitters using LC-MS/MS, a technique that separates and identifies compounds based on mass and chromatographic behavior. Instead of relying on a single time point, the team analyzed temporal trajectories to capture evolving biochemical processes during brain development.

Brain injury severity was measured with an MRI-derived Global Brain Abnormality Score. This composite metric summarizes injury across major regions, including white matter, cortex, deep grey matter, and cerebellum—providing a global view of damage rather than a single localized endpoint.

The key finding was that urinary metabolites displayed distinct time-linked patterns that correlated with global injury severity. Notably, early elevations in energy-related molecules—such as lactate and α-hydroxybutyrate—appeared first, suggesting metabolic stress in the earliest window after birth.

Between postnatal days 3 and 6, these early signals were followed by broader biochemical disruptions spanning amino acids, fatty acids, and the tryptophan–kynurenine pathway. Together, these pathway shifts point to mechanisms that can include mitochondrial dysfunction, oxidative stress, and excitotoxicity.

For severe injury cases, multi-block modeling achieved striking performance: at the default decision threshold, it identified 100% of severe injury patients, with an area under the curve (AUC) of 0.849, using only day 3–6 data. This suggests a potentially actionable neurosurveillance window for risk stratification.

Region-specific MRI prediction was even more varied. The model was strongest for cerebellar anomaly (AUC 0.891 in the full model) and showed more moderate but still meaningful performance for white matter in an early model (AUC 0.724).

Overall, the study suggests that urine metabolomics may detect brain injury dynamics well before term-equivalent MRI, with days 3–6 emerging as a particularly informative period. If validated, this noninvasive biochemical early-warning system could enable earlier monitoring and potentially guide neuroprotective interventions for very preterm infants.