Yue-Feng He, Yue Zhou, Jian-Gang Lu
CAR is independently associated with prevalent DFU in hospitalized patients with T2DM, but its association is largely driven by acute infection-related inflammation. The integrated risk model demonstrated good internal discrimination and modest external performance, and may assist in cross-sectional risk stratification for hospitalized patients; however, local recalibration is recommended before clinical deployment, and prospective studies in non-infected populations are needed before broader implementation.
OBJECTIVE: To compare the screening performance of the neutrophil-to-lymphocyte ratio (NLR), systemic immune-inflammation index (SII), and C-reactive protein-to-albumin ratio (CAR) for risk stratification of diabetic foot ulcer (DFU) in patients with type 2 diabetes mellitus (T2DM), and to develop an integrated risk screening model with clinically applicable cutoff values.
METHODS: This multicenter retrospective cross-sectional study included hospitalized patients with T2DM from Dujiangyan People's Hospital and Chongqing University Fuling Hospital between January 2024 and December 2025. Demographic characteristics, clinical variables, and laboratory parameters were collected, and NLR, SII, and CAR were calculated. Univariate and multivariate logistic regression analyses were performed to identify risk factors associated with DFU, and these results were combined with random forest analysis for variable selection. Four risk screening models were subsequently constructed. Model discrimination was evaluated using receiver operating characteristic (ROC) curves, area under the curve (AUC), and DeLong's test. Optimal cutoff values were determined using the Youden index. Model stability was assessed using bootstrap resampling (1000 iterations). Calibration performance was evaluated using the Hosmer-Lemeshow test and calibration curves, while clinical utility was assessed by decision curve analysis (DCA). A prespecified sensitivity analysis excluding patients with moderate-to-severe DFU infection was performed. The Fuling cohort served as an independent external validation cohort, with the development-derived threshold fixed and applied without recalibration.
RESULTS: A total of 962 patients from the Dujiangyan cohort (161 with DFU, 16.7%) and 600 patients from the Fuling cohort (204 with DFU, 34.0%) were included. Multivariate logistic regression analysis demonstrated that CAR was significantly associated with DFU in both cohorts [Dujiangyan cohort: odds ratio (OR) per 0.1-unit increase = 1.029, 95% confidence interval (CI): 1.017-1.041, P < 0.001; Fuling cohort: OR = 1.013, 95% CI: 1.001-1.025, P = 0.031]. The integrated risk model (Model 4) achieved an AUC of 0.84 (95% CI: 0.80-0.87) in internal validation, with a bias-corrected C-index of 0.83 (95% CI: 0.80-0.87) and good calibration (Hosmer-Lemeshow P = 0.866). In the external validation cohort, Model 4 achieved an AUC of 0.61 (95% CI: 0.56-0.66) using the fixed development-derived threshold (0.136). However, a prespecified sensitivity analysis excluding patients with moderate-to-severe DFU infection substantially attenuated the CAR association in both cohorts, indicating that CAR's performance is largely driven by acute infection-related inflammation.
CONCLUSION: CAR is independently associated with prevalent DFU in hospitalized patients with T2DM, but its association is largely driven by acute infection-related inflammation. The integrated risk model demonstrated good internal discrimination and modest external performance, and may assist in cross-sectional risk stratification for hospitalized patients; however, local recalibration is recommended before clinical deployment, and prospective studies in non-infected populations are needed before broader implementation.