Related Experiment Video
Updated: Jul 14, 2026
Monitoring Dynamic Growth of Retinal Vessels in Oxygen-Induced Retinopathy Mouse Model
Published on: April 2, 2021
Prognostic models for predicting sight-threatening retinopathy of prematurity in preterm infants: A systematic review
Stella Moutzouri1,2, Anna-Bettina Haidich2, Aikaterini K Seliniotaki1
12nd Department of Ophthalmology, School of Medicine, Faculty of Health Sciences, Aristotle University of Thessaloniki, Thessaloniki, Greece.
Insights
Numerous retinopathy of prematurity (ROP) prediction models exist, but methodological flaws and insufficient validation hinder clinical use. Future research needs robust validation in diverse populations to improve ROP screening.
Area of Science:
- Neonatal Medicine
- Ophthalmology
- Biostatistics
Background:
- Retinopathy of prematurity (ROP) is a leading cause of childhood blindness in premature infants.
- Accurate prediction models are crucial for optimizing screening protocols in neonatal intensive care units (NICUs).
Purpose of the Study:
- To systematically identify, describe, and critically appraise studies developing and validating prognostic models for sight-threatening ROP.
- To assess the methodological quality, risk of bias, and applicability of existing ROP prediction models.
Main Methods:
- Comprehensive literature search of PubMed, Embase, trial registers, and grey literature up to October 2025.
- Data extraction using CHARMS checklist; reporting adherence to TRIPOD-SRMA and PRISMA guidelines.
- Quality assessment using PROBAST+AI for model development, validation, and risk of bias.
Main Results:
- Identified 35 unique prognostic models for sight-threatening ROP from 30 development studies.
- Common predictors included gestational age, birth weight, and postnatal weight gain.
- Most models suffered from methodological limitations, inadequate reporting, and limited external validation, impacting generalizability and clinical applicability.
Conclusions:
- Despite many ROP prediction models, significant methodological weaknesses and lack of robust validation impede widespread clinical adoption.
- Future research must focus on prospective, multicenter validation in diverse cohorts using standardized reporting and readily available predictors.
- Improved model development and validation are essential for reliable ROP screening tools in NICUs.
Abstract:
To identify, describe and critically appraise studies developing and/or validating models for predicting sight-threatening retinopathy of prematurity (ROP) in preterm infants undergoing screening in neonatal intensive care units. PubMed, Embase via Ovid, trial registers, and grey literature were searched from inception to 21 October 2025. Eligible studies included those that developed and/or validated prognostic models for sight-threatening ROP in screened preterm infants. Data extraction followed the CHARMS checklist. Reporting adhered to TRIPOD-SRMA and PRISMA guidelines. Model quality, risk of bias, and applicability were assessed using PROBAST+AI. The protocol is available on Open Science Framework (https://osf.io/dgc4y). Thirty-five unique prognostic models for sight-threatening ROP were identified from 30 development studies. Gestational age, birth weight, and postnatal weight gain were the most frequently reported predictors, followed by sex. Twenty-three models underwent at least one external validation (155 validations in total). Most development studies showed substantial methodological concerns related to study design, handling of missing data, lack of external validation, and inadequate reporting of performance measures. Studies assessing model performance (apparent, internal, or external) were also at high overall risk of bias due to design, analytical, and reporting shortcomings. Collectively, these limitations may reduce the generalizability and clinical applicability of existing models. Despite numerous ROP prediction models, methodological limitations and limited robust validation preclude widespread clinical adoption. Future research should prioritize development and prospective, multicentre validation across large, diverse cohorts, with emphasis on readily available and objectively measured predictors, alongside strict adherence to established methodological and reporting standards (TRIPOD-AI, PROBAST+AI) to ensure reliable, generalizable tools for optimizing ROP screening.
