A prediction model for diabetic retinopathy incorporating adipokines and red cell distribution width among patients with type 2 diabetes

Emerging evidence suggests that adipokines such as omentin1 and adiponectin, along with hematological indices like red cell distribution width (RDW), may reflect subclinical inflammation and endothelial dysfunction in diabetes. However, their combined predictive value for diabetic retinopathy (DR) has not been well established. The aim of this study was to test this hypothesis by identifying predictors of DR and constructing a prediction model based on these factors among type 2 diabetes (T2D) patients. A total of 150 patients with T2D, according to the 1999 World Health Organization diagnostic criteria, were retrospectively reviewed at a tertiary hospital between January 2022 and December 2023. Predictors of DR in patients with T2D were identified using a multivariate logistic regression model, and the prediction model was constructed using a receiver operating characteristic (ROC) curve. The predictive value of the constructed model was assessed using the area under the ROC curve (AUC). Among 150 patients, 92 had DR. We observed mean arterial pressure > 120 mmHg, abnormal aspartate aminotransferase, abnormal gamma-glutamyl transpeptidase, high blood urea nitrogen, omentin-1, and apelin were associated with an increased risk of DR in T2D patients. However, low high-density lipoprotein, abnormal alanine aminotransferase, high homeostasis model assessment of insulin resistance, and nesfatin-1 were associated with a reduced risk of DR in patients with T2D. A combination of these factors and omentin-1, adiponectin, or red cell distribution width was used to predict the risk of DR in T2D patients, the apparent AUC of the prediction model was 0.921 (95% CI: 0.880–0.962), with an optimism-corrected AUC of 0.873 (95% CI: 0.821–0.925) after bootstrap internal validation. The Hosmer-Lemeshow test indicated adequate calibration ( p = 0.293), but the calibration plot revealed systematic deviations from perfect calibration, with risk underestimation in the intermediate probability range (0.2–0.6) and overestimation around 0.7. The decision-curve analysis demonstrated positive net benefit across threshold probabilities of 10% to 60%. External validation in independent cohorts is required to confirm these findings. This study identified factors associated with DR in patients with T2D. The prediction model demonstrated good discriminative ability but suboptimal calibration in this cohort. External validation and model recalibration are needed before clinical application.