17 - Comparing MRI-based Scoring Systems for Neuroprognostication in Neonatal Encephalopathy

Publication Number: 1015.17

Michal Yaar, The Hospital for Sick Children, Toronto, ON, Canada; Vann Chau, The Hospital for Sick Children, Toronto, ON, Canada; Jessy Parokaran Varghese, The Hospital for Sick Children, Toronto, ON, Canada; Daphne Kamino, The Hospital for Sick Children, Toronto, ON, Canada; Linh G. Ly, The Hospital for Sick Children, Toronto, ON, Canada; Eva Mamak, Hospital for Sick Children, Toronto, ON, Canada; Elysa Widjaja, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, United States; Emily WY. Tam, The Hospital for Sick Children, Toronto, ON, Canada

Background: Although many MRI-based scoring systems are well established, it is unclear which one has the best prognostic ability. Accurate neuroprognostication helps identify infants at risk for neurodevelopmental impairments (NDI), enabling early interventions that can improve long-term outcomes.

Objective: To determine which scoring system has the best prognostic ability for developmental outcomes at 36 months of age.

Design/Methods: A prospective cohort of newborns ≥36 weeks with neonatal encephalopathy had an early postnatal brain MRI. Each MRI was independently scored using five established scoring systems: Barkovich, Weeke, Trivedi, Rutherford, and NICHD-NRN. Neurodevelopmental outcomes at 36 months of age were obtained through standardized developmental assessments - Bayley Scales of Infant and Toddler Development 3rd Edition (Bayley-3) and the Child Behavior Checklist (CBCL). Logistic regression analyses and receiver operating characteristic (ROC) curves were used to evaluate prognostic performance of MRI scores for developmental outcomes. To find the best predictive model from the Weeke scoring components, we used pairwise correlation to reduce highly correlated variables, and then backward selection to optimize the linear regression model.

Results: Of 102 subjects enrolled with an MRI, 88 (86%) completed developmental assessments (7 died, 1 withdrew, 5 lost to follow-up, and 1 unable to attend due to pandemic). Brain MRIs were performed at a median age of day 4 of life (interquartile range 3-4). The Weeke total score showed the highest predictive value for Bayley-3 motor outcomes (AUC 0.7647). For CBCL internalizing scores, the Weeke gray matter subscore had the strongest association (AUC 0.7735), followed by the Barkovich basal ganglia score (AUC 0.7674) and the Weeke total score (AUC 0.7640). None of the systems reliably predicted cognitive and language outcomes, or CBCL externalizing outcomes (AUC < 0.7). The most accurate predictive model for 36-month Bayley-3 motor scores included 5 variables from Weeke: basal ganglia, posterior limb of internal capsule, and perirolandic cortex severity, cerebellar hemorrhage location, and basal ganglia lactate levels.

Conclusion(s): MRI based scoring systems have limited reliability in predicting developmental outcomes other than motor. Weeke total score demonstrates the highest reliability for predicting motor outcomes at 36 months. Only 5 parameters from the Weeke scoring system are required to accurately predict motor outcomes.