EEG-Based Decision-Support Algorithm for Hypoxic-Ischemic Encephalopathy in Term Newborns

N/ANot Yet RecruitingNCT07764315

Assistance Publique - Hôpitaux de Paris310 enrolled

Overview

This retrospective, multicenter, observational study evaluates how well a decision-support algorithm can tell apart mild forms of hypoxic-ischemic encephalopathy (HIE) from moderate or severe forms in full-term newborns born after a lack of oxygen around birth (perinatal asphyxia). The algorithm reads the raw (non-compressed) EEG signal. Its output is compared with the reference reading of the full conventional EEG made by a panel of pediatric neurophysiologists together with the baby's clinical information. The study uses only medical data that already exists and asks nothing of the babies or their families.

The algorithm under investigation was developed in a previous retrospective clinical study (registered as NCT05114070; French CNIL/ethics ref CER-2021-023; approved by the Sorbonne University ethics committee) and is based on background-rhythm analysis of EEG. Newborns are analyzed in two subgroups: those who received a treatment liable to modify the EEG tracing, and those who did not. Conventional EEG recordings are collected from the neurophysiology departments of participating level-3 maternity hospitals over the period 01 January 2016 to 30 April 2026. No follow-up or sampling is performed.

Study Type

OBSERVATIONAL

Enrollment

310

Conditions

Hypoxic-Ischemic Encephalopathy Perinatal Asphyxia Neonatal Brain Injury

Eligibility

Medical Language ↔ Plain English

Inclusion Criteria: * No parental opposition to the use of the child's medical data within 1 month of the mailed information notice * Gestational age ≥ 36 weeks of amenorrhea * Birth weight ≥ 1800 g * Born in a context of perinatal asphyxia (any cause) leading to suspected HIE, with EITHER biological signs of metabolic acidosis (pH ≤ 7 or base deficit ≥ 16 mmol/L or lactate ≥ 11 mmol/L within the first hour of life, any blood sample) OR a history of asphyxia with Apgar ≤ 5 at 10 minutes or need for ventilatory resuscitation continued at 10 minutes of life * Non-compressed conventional EEG (EEGc) recorded before the 6th hour of life * EEG including the signal of active electrodes Fp1, Fp2, T3 and T4 Exclusion Criteria: * Participation in a therapeutic biomedical research liable to modify the EEG tracing * Fewer than 20 minutes of artifact-free EEG recording * Signal from active electrodes Fp1, Fp2, T3 and T4 not exploitable

Outcomes

Primary Outcomes

Sensitivity and specificity of the algorithm versus the gold standard for discriminating mild vs moderate/severe HIE

Diagnostic performance computed against the reference EEG reading by pediatric neurophysiologists complemented by clinical data; 95% confidence intervals by the exact (Clopper-Pearson) method.

Time frame: Within the first 6 hours of life

Secondary Outcomes

PPV, NPV, false-alarm ratio, accuracy, balanced accuracy and F1-score of the algorithm vs gold standard (overall and by subgroup)

Secondary diagnostic-performance metrics computed overall and within the treated / untreated subgroups.

Time frame: Within the first 6 hours of life

Proportion of transfers not leading to therapeutic hypothermia

Share of included newborns not treated by therapeutic hypothermia at the time, including those transferred from a level 1/2a/2b maternity or managed within the same facility.

Time frame: During the initial hospitalization (up to 28 days of life)

Evaluate the inter-rater agreement between the first two expert neurophysiologists

Inter-rater agreement will be assessed using Cohen's kappa coefficient (κ), with a 95% confidence interval.

Time frame: EEG recordings from the first 6 hours of life