We included DFU patients diagnosed with diabetes from the NHANES database between 1999 and 2004, who had a history of non-healing ulcers for more than four weeks according to the diabetes questionnaire data. Mortality follow-up data for this population was obtained from the National Death Index (NDI). Patients under the age of 40 and pregnant women were excluded based on the scope of the questionnaire.

ACR, eGFR and cystatin C were used as indicators to assess CKD, and the calculations required serum creatinine, urinary creatinine, and urinary protein, where eGFR was calculated using the CKD-EPI formula (8). The CKD-EPI formula for creatinine values using mg/dL is specified as follows: eGFR = 141 × min(Scr/κ, 1)^α × max(Scr/κ, 1)^-1.209 × 0.993^Age × 1.018 [if female] _ 1.159 [if black], where Scr is serum creatinine, κ is 0.7 for females and 0.9 for males, α is -0.329 for females and -0.411 for males, min indicates the minimum of Scr/κ or 1, and max indicates the maximum of Scr/κ or 1.

Covariates potentially associated with DFU and likely to influence the association of DFU with CKD were included in this study. The multiple estimation of chained equations (MICE) method was used to estimate the missing covariate data. Diabetic patients included in the survey should contain any of the following conditions: (1) HbA1c ≥6.5%; (2) fasting blood glucose ≥7.0 mmol/L (≥ 126 mg/dL); (3) random blood glucose ≥11.1 mmol/L(≥ 200 mg/dL); and (4) had been diagnosed with diabetes by a physician. Biochemical indicators include total cholesterol (TC), serum albumin. Hypertension was defined as a diagnosis confirmed by a physician. Peripheral neuropathy (PN) was defined as having at least one insensate site in both feet. Peripheral arterial disease (PAD) was defined as a left ankle or right ankle brachial index (ABI) < 0.9. Body mass index (BMI) was obtained directly from body measurements. The current smokers were defined as participants who were reported smoking occasionally or daily during the past 7 days, or over 100 cigarettes.

NHANES data meeting the inclusion criteria were extracted, merged, filtered, and analyzed using weighted data in R Studio (version 4.4.1). For baseline characteristics, means and standard deviations were calculated for continuous variables in the deceased and non-deceased groups using the ‘svyby’ function from the ‘survey’ package, while percentages and 95% confidence intervals were calculated for categorical variables. We obtained p-values by fitting a generalized linear model (GLM) using the ‘svyglm’ function and calculating t-tests for the regression coefficients to assess whether there was a significant between-group difference between the surviving and dying groups.

We transformed eGFR and ACR into categorical variables using clinical classification: eGFR was divided into four categories (≥90, 60-89, 30-59, <30 ml/min/1.73 m²), and ACR into three categories (<30, 30-300, >300 mg/g). The optimal cutoff value for cystatin C was determined using the cutoff in the R package, divided into two groups of ≤0.99 and >0.99 mg/L for analysis ( Figure 1 ).

The Kaplan-Meier (KM) survival curves visually demonstrates the relationship between renal function indices at various clinical grades and the survival probability of DFU patients over time.

Using the ‘svyglm’ function and adjusting for confounding variables, a multivariable logistic regression analysis was conducted with a generalized linear model (GLM) to investigate the relationship between renal function indices at different clinical stages and the mortality risk in DFU patients. The odds ratio (OR) represents the change in mortality for each unit increase in renal function indices.

A Cox proportional hazards regression analysis was conducted to explore the relationship between CKD and the mortality risk in DFU patients. The hazard ratio (HR) was used to measure the impact of renal function on the mortality risk of DFU patients over time. The proportional risk assumption was tested using the Scheinfeld residual test. The proportional risk assumption was tested using the Scheinfeld residual test.

Using the ‘survivalROC’ package and based on survival data including survival time and status, the true positive rate (TPR) and false positive rate (FPR) are calculated using the Kaplan-Meier method to generate time-dependent ROC curves, while also computing the area under the curve (AUC) to quantify predictive performance. By traversing multiple time points, an AUC-time curve is plotted to analyze the trend of predictive accuracy over time, thus evaluating the long-term prognostic value of renal function indicators in the prognosis of DFU.

A total of 31,126 people participated in the NHANES survey during the four cycles from 1999-2004. In our study, 10930 participants who were under 40 years of age and pregnant, 13263 with missing renal function data, 6790 non-DFU patients, and those who did not meet the diagnosis after data interpolation were excluded. Finally, 112 participants were included in our study, 29 in the surviving group and 83 in the mortality group ( Figure 2 ).

The mean age of the overall sample was 65 years, of which 43 (38.39%) were females and 69 (61.61%) were males. Among the 112 DFU patients, the median follow-up duration was 116 months, with a mortality rate of 74.1%. The average age at the time of death was 63 years old. In our study, DFU mortality is significantly influenced by age, PAD, eGFR, and cystatin C levels. Older age (P < 0.01) and the presence of PAD (P < 0.001) are linked to higher mortality rates. Additionally, both eGFR (P < 0.05) and cystatinC (P < 0.01) indicate that patients with DFU and CKD face an elevated risk of death ( Table 1 ).

We converted cystatin C to a categorical variable using the optimal cutoff value, while eGFR and ACR were converted based on clinical guidelines. DFU patients with cystatin C levels > 0.99 mg/L had significantly shorter survival times compared to those with levels ≤ 0.99 mg/L. Similarly, patients with ACR < 30 mg/g had longer median survival times than those with ACR > 30 mg/g. However, no significant difference in survival was observed between patients with ACR > 300 mg/g and those with ACR between 30 and 300 mg/g. Patients with eGFR < 30 ml/min/1.73m² also exhibited significantly shorter survival times. These results were confirmed by the log-rank test ( Figure 3 ).

In the multivariable logistic regression analysis, we constructed three models to gradually control for confounders in order to explore the relationship between CKD and the mortality risk of DFU. Model 1 did not adjust for any confounders, while Model 2 adjusted for basic demographic variables such as age, gender, and ethnicity. In these two models, CKD was significantly associated with the mortality risk of DFU patients, suggesting that CKD may have a direct impact on mortality risk. However, in Model 3, after further adjusting for potential confounders such as hypertension, peripheral artery disease, and smoking, the association between CKD and mortality risk was no longer significant. This indicates that the effect of CKD on mortality risk may be indirect, potentially mediated through these related comorbidities (e.g., hypertension, PAD) rather than CKD itself directly increasing the risk of death ( Table 2 ).

In the weighted Cox proportional hazards regression analysis, age (HR 1.05 [95% CI 1.01,1.09], P = 0.021), serum albumin [0.37(0.18,0.76), P = 0.007], smoking [3.08(1.29,7.38), P = 0.012], and cystatin C [1.69(1.14,2.50), P = 0.012] were significantly associated with mortality in patients with DFU. This suggests that these factors contribute to increased mortality risk due to their long-term cumulative effects on the body, which are more accurately captured in a time-dependent model like cox regression, compared to a static linear analysis. Schoenfeld residuals were used to test the proportional hazards assumption, and no violation was observed ( Figure 4 ).

The time-dependent ROC curve analysis indicates that the correlation between cystatin C and mortality outcomes in patients with DFU increased steadily over time, whereas the correlation between ACR and eGFR decreased after 10 years. We plotted AUC-time curves in months at 12-month intervals to further show that the close correlation between cystatin C and the risk of death in patients with DFU remained stable after 5 years of follow-up (AUC>0.70), whereas the correlation between ACR and eGFR and the risk of death from DFU declined after 7 years of follow-up, with the AUC value dropping below 0.70 after 10 years.( Figure 5 ).