The Pooled Diagnostic Accuracy of Neuroimaging, General Movements, and Neurological Examination for Diagnosing Cerebral Palsy Early in High-Risk Infants: A Case Control Study.

INTRODUCTION: Clinical guidelines recommend using neuroimaging, Prechtls' General Movements Assessment (GMA), and Hammersmith Infant Neurological Examination (HINE) to diagnose cerebral palsy (CP) in infancy. Previous studies provided excellent sensitivity and specificity for each test in isolation, but no study has examined the pooled predictive power for early diagnosis. METHODS: We performed a retrospective case-control study of 441 high-risk infants born between 2003 and 2014, from three Italian hospitals. Infants with either a normal outcome, mild disability, or CP at two years, were matched for birth year, gender, and gestational age. Three-month HINE, GMA, and neuroimaging were retrieved from medical records. Logistic regression was conducted with log-likelihood and used to determine the model fit and Area Under the Curve (AUC) for accuracy. RESULTS: Sensitivity and specificity for detecting CP were 88% and 62% for three-month HINE, 95% and 97% for absent fidgety GMs, and 79% and 99% for neuroimaging. The combined predictive power of all three assessments gave sensitivity and specificity values of 97.86% and 99.22% (PPV 98.56%, NPV 98.84%). CONCLUSION: CP can be accurately detected in high-risk infants when these test findings triangulate. Clinical implementation of these tools is likely to reduce the average age when CP is diagnosed, and intervention is started.