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

Samantha Sadoo^1,2*Phillip Wanduru^3,4Eva Loucaides⁵Carol Nanyunja^6,7Haitao Hu⁸Emily Webb⁷Hannah Blencowe⁷Frances Cowan⁹Eric O Ohuma⁷Kirsty Le Doare⁵Cally J Tann^2,6,7

• ¹London School of Hygiene and Tropical Medicine, University of London, London, United Kingdom
• ²University College Hospital Neonatal Unit, London, United Kingdom
• ³Makerere University, Kampala, Uganda
• ⁴Karolinska Institutet, Stockholm, Sweden
• ⁵City St George's University of London, London, United Kingdom
• ⁶MRC/UVRI and LSHTM Uganda Research Unit, Entebbe, Uganda
• ⁷London School of Hygiene & Tropical Medicine Department of Infectious Disease Epidemiology, London, United Kingdom
• ⁸The Chinese University of Hong Kong, Hong Kong, Hong Kong, SAR China
• ⁹Imperial College London, London, United Kingdom

Background: Intrapartum-related neonatal encephalopathy (NE) is a leading cause of neonatal deaths and childhood-onset developmental disabilities globally. 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 predictors of adverse early childhood outcome after 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 standardised tool) +/- death, at ≥12 months of age. At least two reviewers performed screening of abstracts and full texts, data extraction, and bias assessment (QUIPS tool). We reported sensitivity and specificity for each predictive tool, stratifying results by therapeutic hypothermia status. Meta-analyses were performed where possible. The protocol was registered on PROSPERO in January 2024 (CRD42024485734). Results: Of 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%, specificity 90%), EEG (early severe background abnormality; sensitivity 87%, specificity 93%), Prechtl's General Movements Assessment (absent fidgety movements; sensitivity 78%, specificity 95%) and Hammersmith Infant Neonatal Examination (score <67; sensitivity 88-100%, specificity 88-100%). Conclusions: A range of standardised tools showed good prognostic accuracy for adverse early childhood outcome after NE. However, this review highlights the paucity of NE research in LMICs with 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. There was insufficient data to evaluate the role of TH on prognostic accuracy. Further research to evaluate combinations of the most promising predictors is warranted.