Predictive Value and Clinical Validation of Serum γ-Glutamyltransferase to Platelet Ratio Index for Type 2 Diabetes Foot Ulcer Risk.

ObjectiveTo investigate the predictive value of the γ-glutamyltransferase (GGT) to platelet (PLT) ratio (GPR) for the risk of diabetic foot ulcer (DFU) in persons living with type 2 diabetes mellitus (T2DM).MethodsWe included 435 persons with T2DM at Zigong Fourth People's Hospital from January 2023 to December 2024. Participants were divided into a DFU group (n = 87) and a non-DFU group (n = 348) based on DFU status. Baseline, clinical, and biochemical data were collected. Using R 4.3.1, 70% of data were randomly assigned as training set and 30% as test set. Univariate and binary logistic regression analyses identified factors influencing DFU. Models included regression, gradient boosting, and random forests. Their predictive ability was evaluated by AUC. A separate validation set of 287 T2DM patients from January to December 2025 was also included.ResultsIn the training set, the DFU group exhibited significantly higher proportions of males, higher body mass index, higher prevalence of smoking history, and higher GPR and low-density lipoprotein (LDL) levels than the non-DFU group (P < .05). Logistic regression analysis identified sex, smoking history, GPR, HbA1c, and LDL as independent risk factors for DFU in T2DM patients. The AUCs for training were 0.797, 0.889, and 0.926 for the nomogram, gradient boosting, and random forest models, respectively, with test set AUCs of 0.825, 0.811, and 0.817. External validation showed the random forest model had the best performance.ConclusionMale sex, smoking, elevated GPR, elevated HbA1c, and elevated LDL are risk factors for DFU in persons living with T2DM. Among the three models built using these factors, the random forest model demonstrated the best performance across the evaluation metrics, followed by the gradient boosting machine model, and then the nomogram model.