# Lab values in neonates with hypoxic ischemic encephalopathy over time during and after therapeutic hypothermia

## TL;DR

This study tracks lab values in newborns with brain injury over time during cooling treatment, finding that early metabolic changes predict injury severity.

## Contribution

The study introduces machine-learning models using serial lab values to predict MRI-defined injury severity in neonatal HIE.

## Key findings

- Early metabolic derangements like lower pH and base deficit correlate with higher MRI injury scores in neonates with HIE.
- Machine-learning models using lab values at specific timepoints achieve an adjusted R² of 0.47 in predicting injury severity.
- Temporal trends in metabolic, hepatic, and coagulation biomarkers differ between neonates with and without sentinel events.

## Abstract

Hypoxic-ischemic encephalopathy (HIE) remains a leading cause of neonatal neurological injury, and therapeutic hypothermia is the established treatment shown to reduce brain injury in neonates with moderate to severe HIE. The systemic laboratory response to hypoxic-ischemic injury and its relationship to brain injury severity are not fully understood.

This retrospective cohort included 152 neonates born at a gestational age of 35 weeks or greater who met criteria for therapeutic hypothermia for HIE at UF Health Shands Children's Hospital between 2012 and 2024. Laboratory data were collected at seven time intervals from birth through rewarming and analyzed using linear mixed-effects models to characterize temporal trends. Temporal analyses revealed changes across metabolic, hepatic, and coagulation biomarkers during and after therapeutic hypothermia, with several values demonstrating significant variation at specific time points. Neonates were classified by sentinel event status (definite, probable, or none), and temporal trends demonstrated differences between the groups. The laboratory values were correlated with magnetic resonance imaging (MRI) injury severity using the Weeke scoring system.

Early metabolic derangements, including lower pH and more negative base deficit values, were significantly associated with higher MRI injury scores across all regions, including gray matter, white matter, and the cerebellum. Machine-learning models that integrate a combination of early laboratory timepoints improve the prediction of MRI-defined injury, with the best performance achieved using pH at T1, pCO₂ at T1, and lactate at T3 (adjusted R² = 0.47).

These findings demonstrate temporal laboratory trajectories during and after therapeutic hypothermia, supporting the prognostic utility of serial biomarkers and machine-learning-based modeling in neonatal HIE.