Validation of the TWO-ROP Algorithm in the Shenzhen Prematurity Eye Study
Related Experiment Video
Updated: Oct 30, 2026
An Alternative and Validated Injection Method for Accessing the Subretinal Space via a Transcleral Posterior Approach
Published on: December 7, 2016
Xinyu Zhao1, Zhenquan Wu1, Zhen Yu1
1Shenzhen Eye Hospital, Shenzhen Eye Medical Center, Southern Medical University, Shenzhen, China; Jinan University, Guangzhou, China.
Purpose:
To evaluate the performance of the TWO-ROP algorithm in a large Chinese cohort under alternative United States (US) and Chinese retinopathy of prematurity (ROP) screening criteria.
Design:
Multicenter retrospective cohort study.
Participants:
A total of 14,198 infants screened for ROP between 2011 and 2025 across 56 centers in China.
Methods:
Infants were identified through the Eyecare-Cloud platform as part of the Shenzhen Prematurity Eye Study. The TWO-ROP algorithm aims to reduce screening examinations by recommending a single examination at 36 weeks or later postmenstrual age (PMA) for lower-risk infants. Its performance was evaluated under the US and Chinese ROP screening criteria within the same cohort, with alternative 35- and 38-week rules also explored.
Main Outcome Measures:
Rates of any ROP and TR-ROP, potentially missed TR-ROP cases, and sensitivity for detecting TR-ROP.
Results:
Overall, 10,024 infants were eligible for TWO-ROP under the US screening criteria and 6,750 under the Chinese criteria. The rates of TR-ROP were 2.94% and 1.32%, respectively. Under the US criteria, the 36-week screening rule identified 275 of 295 TR-ROP cases with sensitivity of 93.2% (95% confidence interval [CI], 89.7%-95.8%). Among 20 potentially missed cases, 14 (70%) had aggressive ROP. Advancing the screening time point to 35 weeks increased sensitivity to 98.6% (95% CI, 96.6%-99.6%). Under the Chinese criteria, the 36-week rule identified all 89 TR-ROP cases with sensitivity of 100% (95% CI, 95.9%-100%). Among infants born from 2021 to 2025, the exploratory 38-week rule identified 22 of 25 TR-ROP cases, with a sensitivity of 88.0% (95% CI, 68.8%-97.5%); all three potentially missed cases occurred in local hospitals.
Conclusion:
The performance of the TWO-ROP algorithm varied under the US and Chinese ROP screening criteria. These findings highlight the importance of adapting screening strategies to local criteria and individualizing screening timing according to clinical settings and levels of care.
