Early Prediction Model for Retinopathy of Prematurity Using Placental and Neonatal Risk Factors

Abstract

PURPOSE: Existing prediction models for retinopathy of prematurity (ROP) primarily focus on screening reduction which occur 5 to 7 weeks after birth. We hypothesized that high-risk neonates could be identified much earlier if placental and early postnatal risk factors are incorporated, so that this high risk can be considered during neonatal treatment well before ROP screening begins. METHODS: We included 591 neonates born ≤32 weeks of gestational age (GA) and/or birthweight (BW) ≤1500 grams. Data were retrospectively collected, and placentas were examined for histological abnormalities. The "Placenta as an Additional Predictor for ROP" (PAPROP) model included: GA, BW, mechanical ventilation, postnatal corticosteroids, severe histological chorioamnionitis, and distal villous hypoplasia. This model was internally validated with five-fold cross-validation and compared to a reference model using only GA and BW. RESULTS: The PAPROP model had a discriminatory ability between ROP presence and absence of 0.81 (95% confidence interval [CI] = 0.76-0.86) compared to 0.78 in the reference GA&BW model. This model had a sensitivity of 0.97 and specificity of 0.44 in the test set (threshold 10%). Using clinically relevant thresholds of 10% to 15%, implementing this model in the second postnatal week could lead to a 25% reduction in ROP screenings without missing stage 2 and severe ROP. CONCLUSIONS: The PAPROP model has a high ability to predict ROP development at the end of the second postnatal week, has a potential high clinical utility, and is likely cost-effective. After further external validation, it may aid in creating a personalized neonatal treatment approach for ROP prevention in high-risk neonates.