Consensus Classification of Brain Injury Predicts Long-Term Neurodevelopmental Outcomes in Neonates With Hypoxic-Ischemic Encephalopathy.

BACKGROUND

Accurate early prognostication in neonatal hypoxic-ischemic encephalopathy is vital for family counseling and early interventions. This study evaluated the predictive value of the Canadian standardized magnetic resonance imaging (MRI) classification system for long-term neurodevelopmental outcomes in neonates treated with therapeutic hypothermia.

METHODS

A retrospective cohort study of 172 neonates with moderate-to-severe hypoxic-ischemic encephalopathy who received therapeutic hypothermia (2014-2020) at two tertiary neonatal intensive care units in Calgary, Canada. MRIs (days 4-5) were scored using the Canadian consensus system. The primary outcome was a composite of death or severe neurodevelopmental impairment (sNDI) at 18-24 months. A prespecified sensitivity analysis excluded infants whose follow-up fell during COVID-19 pandemic closures.

RESULTS

Follow-up was completed for 140 infants (81%). Death or sNDI occurred in 37.7% (26/69) of the primary outcome cohort (n = 69, comprising 9 deaths and 60 survivors with complete Bayley Scales of Infant and Toddler Development, 3rd or 4th edition composite scores); grey matter (GM) scores predicted this composite with 80.8% sensitivity, 93.0% specificity, 88.4% accuracy with area under the curve (AUC) of 0.88. For sNDI alone, GM scores yielded the highest specificity (93.0%) and accuracy (88.3%) with an AUC of 0.83. Cerebral palsy (CP) was diagnosed in 7.4% of children; total MRI scores (AUC 0.927) and GM scores (AUC 0.91) demonstrated strong discrimination for motor outcomes. A near-normal MRI yielded a 99.1% negative predictive value for CP, and higher GM scores significantly correlated with functional severity (gross motor function classification system 3 to 5 vs 1 to 2). The inclusion of proton magnetic resonance spectroscopy did not significantly improve predictive performance for any outcome. A prespecified sensitivity analysis excluding infants whose follow-up fell during COVID-19 pandemic closures (n = 102 included) yielded consistent findings, with GM AUC 0.86 for death or sNDI.

CONCLUSIONS

In this single-center external validation of the Weeke scoring framework as adapted in the Canadian consensus, MRI injury scores provide clinically meaningful risk stratification. GM scores offer high specificity for death and sNDI, while Total MRI scores best identify CP risk. Prognostication using structural MRI alone is sufficient in this cohort, consistent with the subsequent decision of the Canadian working group to remove magnetic resonance spectroscopy from the consensus grading form. Conclusions are robust to a sensitivity analysis excluding COVID-era follow-ups.