Predictors of outcome following neonatal encephalopathy in low- and middle-income countries: a systematic review and meta-analysis.

Background: Intrapartum-related neonatal encephalopathy (NE) is a leading cause of neonatal deaths and childhood-onset developmental disabilities worldwide. Accurate prediction of neurodevelopmental outcomes is crucial to support effective neonatal follow-up strategies, guide parental counselling, and inform future neuroprotection research. Whilst NE disproportionately affects those in low- and middle-income countries (LMICs), existing prognostic accuracy research is primarily based in high-income countries. This systematic review and meta-analysis aims to provide a comprehensive synthesis of the predictors of adverse early childhood outcome following NE in LMICs. Methods: Four databases were searched, using terms related to "neonate", "encephalopathy", "predictor", "outcome", and "LMIC". NE was defined as ≥35 weeks' gestation, evidence of intrapartum event, and abnormal neurology on early clinical assessment. Adverse childhood outcome was defined as neurodevelopmental impairment (assessed using a standardised tool) +/- death, at ≥12 months of age. At least two reviewers performed screening of abstracts and full texts, data extraction, and bias assessment (Quality in Prognosis Studies tool). We reported sensitivity and specificity for each predictive tool, stratifying results by therapeutic hypothermia (TH) status. Meta-analyses were performed where possible. The protocol was registered on PROSPERO in January 2024 (CRD42024485734). Results: Of the 7,464 articles screened, 32 were included, totalling 1,538 infants with NE from 14 LMICs. Predictors were categorised into neonatal clinical scores for NE severity (16 studies), neurophysiology (13), neuroimaging (14), biomarkers (10), and post-neonatal neurological clinical assessments (5). Highest prognostic accuracy was demonstrated by MRI (moderate to severe abnormalities; sensitivity 69% and specificity 90%), electroencephalography (early severe background abnormality; sensitivity 87% and specificity 93%), Prechtl's General Movements Assessment (absent fidgety movements; sensitivity 78% and specificity 95%), and Hammersmith Infant Neurological Examination (score <67; sensitivity 88-100% and specificity 88%-100%). Conclusions: A range of standardised tools showed good prognostic accuracy for adverse early childhood outcome following NE. However, this review highlights the paucity of NE research in LMICs using adequate sample sizes and duration of follow-up. Data synthesis and comparability were limited by substantial heterogeneity between study populations, definitions and timing of predictors and outcomes, and variable study quality. Data to evaluate the role of TH on prognostic accuracy were insufficient. Further research to evaluate combinations of the most promising predictors is warranted.