Predictors of Response to Anti-VEGF Combined with Laser Therapy in Severe Non-Proliferative Diabetic Retinopathy: Development and Validation of a Nomogram Model from Retrospective Data

Chunmei Cui^*Yuehua LiQian Zhang

• Beijing Chaoyang Hospital affiliated to Beijing Capital Medical University, Beijing, China

Background: Anti-vascular endothelial growth factor (anti-VEGF) and laser combination therapy demonstrates variable efficacy in severe non-proliferative diabetic retinopathy, with 30–45% of patients experiencing suboptimal outcomes. This study aimed to develop and validate a clinically deployable nomogram integrating multimodal predictors to quantify individualized treatment response probabilities. Methods: A retrospective cohort study analyzed 280 severe non-proliferative diabetic retinopathy patients (Early Treatment Diabetic Retinopathy Study levels 43–53) receiving combined anti-VEGF (ranibizumab/aflibercept) and laser therapy (2018–2023). The primary outcome was a 12-month composite response (no proliferative diabetic retinopathy progression, ≥2-step Diabetic Retinopathy Severity Scale improvement or ≥30% retinal lesion reduction, and no rescue therapy). Least absolute shrinkage and selection operator regression with the “one standard error” rule selected key predictors from 15 candidate variables. A multivariable logistic regression model was translated into a nomogram, validated temporally (70%/30% split) using area under the curve, calibration curves, and decision curve analysis. Results: Four predictors were retained: glycated hemoglobin variability (adjusted odds ratio 0.63 per 5% increase; 95% confidence interval 0.51–0.78), fluorescein angiography non-perfusion area (adjusted odds ratio 0.68 per 10% increase; 95% confidence interval 0.55–0.84), Diabetic Retinopathy Severity Scale severity (adjusted odds ratio 0.72 per grade; 95% confidence interval 0.55–0.94), and serum albumin (adjusted odds ratio 1.85 per 0.5 g/dL; 95% confidence interval 1.22–2.81). The nomogram achieved robust discrimination (derivation area under the curve 0.821, validation area under the curve 0.754) and calibration (slopes 0.98–0.95; Hosmer-Lemeshow P > 0.60). Decision curve analysis confirmed clinical utility at 15–40% threshold probabilities (net benefit 0.28), outperforming “treat-all” strategies. Conclusions: This validated nomogram—integrating glycemic stability, retinal ischemia, baseline severity, and systemic nutrition—provides individualized response probabilities for anti-VEGF and laser therapy. It enables risk stratification to guide treatment intensification in severe non-proliferative diabetic retinopathy, addressing a critical unmet need in personalized retinopathy management.