Prevalence of retinopathy of prematurity in African preterm infants and evaluation of the accuracy of the WINROP predictive algorithm: an observational clinical study.
Dadjo Amouzou, Dupe S Ademola-Popoola, Tokunbo S Obajolowo, Omotayo O Adesiyun, Lateefat B Olokoba, Victoria A Olatunji
Abstract
Open AccessAIM: To estimate the prevalence of retinopathy of prematurity (ROP) and investigate the prediction and validity of WINROP (Weight, Insulin-Like Growth Factor-1, Neonatal Retinopathy of Prematurity), an online monitoring algorithm in preterm infants at an African low/medium income country (LMIC) based neonatal intensive care unit (NICU). METHODS: Data from the ROP screening proforma for preterm infants between January 2020 and December 31, 2022, with gestational age (GA) from 24 to 32 weeks were entered into the WINROP algorithm system which signals an ROP risk alarm based on weekly weight gain and was compared with ophthalmology examination findings. Sensitivity, specificity, negative and positive predictive values were calculated, and statistical analysis was performed using EPI INFO (version 7.2.2.6). RESULTS: There were 168 preterm infants that met WINROP criteria. The median birth weight was 1380 g (IQR: 1141-1570 g), ranging from 600 g to 2100 g, while the median gestational age at birth was 31 weeks (IQR: 29.0-32.0 weeks), with a range of 24 to 32 weeks The overall prevalence of any ROP was 38.1% (64/168), type 1 ROP and type 2 ROP were 7.1% (12/168) and 8.9% (15/168) respectively. WINROP alarmed in 73 infants born at a gestational age (29.0 weeks, IQR: 28.0-31) and birth weight (1100 g, IQR: 912.5-1200 g). The specificity of WINROP for any ROP was 80.8% (84/104); the sensitivities for type 1 ROP and type 2 ROP were 91.7% (11/12) and 86.7% (13/15) respectively. The negative and positive predictive values were 88.4% (84/95) and 72.6% (53/73) respectively. CONCLUSION: Overall, the WINROP algorithm identified 91.7% of preterm infants who had been diagnosed and treated for type 1 ROP. Though it led to a high number of false alarms, especially in the most vulnerable infants. The algorithm's utility is likely greatest for the larger, more mature preterm infants. This confirms its usefulness to significantly reduce the number of preterm infants who need to go through stressful eye examinations in an African NICU setting where inadequate number of specialists is a major challenge.