Purpose: Postnatal-growth-based predictive models demonstrate strong potential for improving the low specificity of retinopathy of prematurity (ROP) screening. Prior studies are limited by inadequate sample size. We sought to study a sufficiently large cohort of at-risk infants to enable development of a model with highly precise estimates of sensitivity for severe ROP.
Methods: The Postnatal Growth and ROP (G-ROP) Study was a multicenter retrospective cohort study of infants at 30 North American hospitals during 2006-2012. A total of 65 G-ROP-certified abstractors submitted data to a secure, web-based database. Data included ROP examination findings, treatments, complications, daily weight measurements, daily oxygen supplementation, maternal/infant demographics, medical comorbidities, surgical events, and weekly nutrition. Data quality was monitored with system validation rules, data audits, and discrepancy algorithms.
Results: Of 11,261 screened infants, 8334 were enrolled, and 2927 had insufficient data due to transfer, discharge, or death. Of the enrolled infants, 90% (7483) had a known ROP outcome and were included in the study. Median birth weight was 1070 g (range 310-3000g) and mean gestational age 28 weeks (range 22-35 weeks). Severe ROP (Early Treatment of Retinopathy type 1 or 2) developed in 931 infants (12.5%).
Conclusion: Successful incorporation of a predictive model into ROP screening requires confidence that it will capture cases of severe ROP. This dataset provides power to estimate sensitivity with half-confidence interval width of less than 0.5%, determined by the high number of severe ROP cases. The G-ROP Study represents a large, diverse cohort of at-risk infants undergoing ROP screening. It will facilitate evaluation of growth-based algorithms to improve efficiency of ROP screening.
Keywords: Predictive model; prematurity; retinopathy of prematurity; screening.