重症脳損傷のない早産児における神経発達予後予測バイオマーカーとしての機能的結合性

Despite a decline in severe neonatal brain injury in preterm infants, neurodevelopmental impairment remains prevalent. Identifying early biomarkers for neurodevelopmental impairment, particularly in infants without severe neonatal brain injury, is crucial for intervention. This study explores whether brain dysmaturation, indicated by functional connectivity alterations at term-equivalent age, predicts neurodevelopmental impairment severity at 24 months corrected age in preterm infants without severe neonatal brain injury. In this observational cohort study, preterm infants born < 31 weeks' gestation without severe neonatal brain injury underwent resting-state functional MRI at term-equivalent age. Neurodevelopmental outcomes at corrected age 24 months were assessed using Bayley-III cognitive and motor composite scores, cerebral palsy severity, and neurosensory impairments. Functional connectivity alterations were analyzed in relation to cognitive, language, and motor outcomes. Machine learning models were applied to assess the predictive value of functional connectivity features alongside neonatal exposures for neurodevelopmental impairment severity. Among the 122 preterm infants, 89 (73%) infants had no/mild neurodevelopmental impairment, 27 (22%) had moderate neurodevelopmental impairment, and 6 (5%) showed severe neurodevelopmental impairment. Compared with the no/mild neurodevelopmental impairment group, the moderate/severe neurodevelopmental impairment group was significantly lower in gestational age, and required longer durations of invasive mechanical ventilation, oxygen therapy, vasopressors, and total parenteral nutrition during admission. Compared with term-born controls, a clear trend emerged across neurodevelopmental impairment severity levels: as impairment increased from the no/mild group to the moderate and severe groups, the clustering coefficient increased, whereas the global efficiency decreased. Statistical comparisons between the no/mild and moderate/severe groups, relative to term-born controls, confirmed these patterns (clustering coefficient: t = -4.38, P < 0.001; global efficiency: t = 3.44, P < 0.001). Infants with no/mild neurodevelopmental impairment exhibited enhanced connectivity in the limbic system (t = -5.21, P < 0.001) and between the thalamus and basal ganglia (t = -5.9, P < 0.001), but this compensatory connectivity weakened with increasing neurodevelopmental impairment severity. The thalamo-cortical (frontal lobe, limbic system), thalamo-basal ganglia, and thalamo-cerebellar connectivity were strongly associated with cognitive, language, and motor performance at follow-up. A predictive model incorporating these functional connectivity features and neonatal adverse exposure parameters achieved 82% accuracy. Distinct disruptions in functional connectivity at term-equivalent age in very preterm infants without severe neonatal brain injury may predict the severity of later neurodevelopmental impairment. Early functional connectivity assessment holds promise as a biomarker for identifying high-risk infants who may benefit from timely neurodevelopmental interventions.