Related Experiment Video
Updated: Jan 6, 2026
Rat Model of Photochemically-Induced Posterior Ischemic Optic Neuropathy
Published on: November 29, 2015
Prediction Models for Recurrence of Retinopathy of Prematurity after Intravitreal Ranibizumab or Laser
Huiqian Kong1, Guanghua Zhou1,2, Yubin Yang1
1Guangdong Eye Institute, Department of Ophthalmology, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Second Zhongshan Road 106#, Guangzhou, 510080, China.
Introduction:
To develop prediction models for retinopathy of prematurity (ROP) recurrence after initial intravitreal injection of ranibizumab (IVR) or laser photocoagulation (LP).
Methods:
This multicenter retrospective cohort study included infants with aggressive posterior ROP (AP-ROP) and type-I ROP. Recurrence was defined as the reappearance of vascular dilation, tortuosity, or new/recurrent neo-vascularization in either eye of infants. The recurrence rates within 6 months after initial treatment were compared. Machine learning (i.e., extreme gradient boost, categorical boost, adaptive boost, and random forest) and multivariable logistic regression were performed to identify risk factors and establish individualized prediction models for ROP recurrence. Area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity were used to evaluate the performance of the prediction models.
Results:
A total of 440 infants were included, with a mean gestational age (standard deviation [SD]) of 28.2 (2.4) weeks and birth weight (SD) of 1.1 (0.4) kg. ROP recurrence occurred in 62 of 344 (18.0%) infants after IVR treatment, while 7 of 96 (7.3%) infants after LP treatment required additional treatment (P < 0.05). Neonatal pneumonia, respiratory distress, septicemia, 5-min Apgar scoring, AP-ROP, and maternal uterine infection were significantly associated with the risk of ROP recurrence (all P < 0.05). The Categorical Boost model achieved the best overall performance. The mean AUC, accuracy, sensitivity, and specificity were 0.96 (95% confidence interval [CI] 0.93-0.99), 85.9% (95% CI 80.3-91.5%), 94.7% (95% CI 89.6-99.8%), and 77.0% (95% CI 67.4-86.6%) on the validation dataset, and 0.94 (95% CI 0.89-1.00), 84.1% (95% CI 77.1-93.2%), 95.5% (95% CI 89.8-100%), and 72.7% (95% CI 59.6-85.5%) on the testing dataset, respectively.
Conclusion:
This study presents a simple, rapid, and reliable predictive strategy for early identifying infants at high risk of ROP recurrence after initial treatment, which is potentially useful in improving the success rate of retreatment and reducing blindness resulting from ROP.
