Work

Risk factors for light perception or no light perception after vitrectomy for proliferative diabetic retinopathy: a retrospective cohort study

Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background This study aimed to identify preoperative and perioperative risk factors for light perception (LP) or no light perception (NLP) after vitrectomy for proliferative diabetic retinopathy (PDR) and, as an exploratory step, to develop and internally validate preliminary prediction models. Methods This study included 238 eyes of 238 patients undergoing vitrectomy for PDR; 14 candidate predictors were evaluated using Firth penalised logistic regression, least absolute shrinkage and selection operator (LASSO) regression, and random forest. Performance was assessed by the area under the receiver operating characteristic curve (AUC), calibration slope, Brier score and decision curve analysis, with bootstrap optimism correction and repeated five-fold cross-validation. Results Nine eyes (3.8%) reached the endpoint of LP or NLP. Preoperative iris neovascularisation (odds ratio 39.88, 95% confidence interval 6.29-405.45), higher glycated haemoglobin (odds ratio 1.92 per %, 1.05–4.35) and diabetic nephropathy (odds ratio 15.50, 1.21-2397.50) were independently associated with LP or NLP, whereas preoperative intraocular pressure and visual acuity were not. Optimism-corrected AUCs were 0.971, 0.975 and 0.988; with only nine events, these estimates are unstable and best regarded as unreliable upper bounds, and the Firth model was the best calibrated. Conclusions Iris neovascularisation, poor glycaemic control and diabetic nephropathy were the strongest predictors of LP or NLP. The best-calibrated Firth model, translated into a nomogram, is the most suitable candidate for risk stratification, pending external validation.
Language: English
Subfield: Ophthalmology
Field: Medicine
Open Access status: green
OpenAlex ID: W7235893578
Cites: 29
Cited by:
Related to: 10