The NHANES is an ongoing program designed to assess the health and nutritional status of the U.S. population. Ethical oversight for all surveys is provided by the Ethics Review Committee of the National Center for Health Statistics, with participants giving written informed consent. The survey employs multistage, stratified sampling to ensure representativeness.

This study utilized data from the NHANES database, covering the years 1999 to 2018. Patients were selectively excluded based on predefined criteria. Exclusions included individuals without DKD (n = 97,651), those under the age of 20 (n = 28), individuals with a baseline cancer diagnosis (n = 596), and those lacking data on HbA1c (n = 1), FPG (n = 1,240), mortality status (n = 2), fasting weights (n = 194), or covariates (n = 283). Additionally, individuals treated for anemia within the previous three months were excluded (n = 264) due to potential influences on erythrocyte longevity and HbA1c measurements. After applying these criteria, the analysis included 1,057 individuals with DKD (Fig. 1).

The diagnostic criteria for DKD were established based on diagnostic standards for diabetes and aligned with the CKD guidelines set forth by the Kidney Disease Improving Global Outcomes (KDIGO) working group [20]. These criteria encompassed either an elevated urinary albumin-to-creatinine ratio (ACR) exceeding 30 mg/g or a reduced estimated glomerular filtration rate (eGFR) falling below 60 ml/min/1.73 m² [3].

For the diagnosis of diabetes, several parameters were considered: a clinical diagnosis of diabetes by a healthcare professional, a HbA1c level exceeding 6.5%, a fasting blood glucose level of ≥ 7.0 mmol/L, a random blood glucose level of ≥ 11.1 mmol/L, or a random blood glucose level exceeding 11.1 mmol/L after a two-hour oral glucose tolerance test (OGTT). Meeting any of these criteria confirmed the diagnosis of diabetes [21].

We employed the standardized HGI formula, developed by Hempe et al., a pioneering figure in HGI research [14]. This formula is grounded in data extracted from the 2005–2016 NHANES. The derivation of this formula involved a cohort of 18,675 adults who were either untreated for diabetes or did not self-report diabetes.

The projected HbA1c value (projected HbA1c = 0.024 FPG + 3.1) is determined by entering the FPG value into a regression equation describing the linear correlation between FPG and HbA1c in the reference population. HGI is computed as the disparity between the measured HbA1c and the predicted HbA1c.

All-cause mortality, representing death resulting from any cause, was designated as the primary outcome. CVD mortality was delineated utilizing International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes I00 to I09, I11, I13, I20 to I51, and I60 to I69. Mortality data, acquired through linkage of the cohort database with the National Death Index up to December 31, 2019, underwent thorough analysis.

Several covariates potentially influencing the outcomes were incorporated, covering age (years), sex (male and female), race/ethnicity (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, or other races including multi-racial), education level (< 9 years, 9–13 years, or ≥ 13 years), household poverty-to-income ratio (PIR), smoking status (never, current, and former), alcohol intake (never, moderate, and heavy), use of diabetes medications (other, insulin, oral medications, and unknown), diabetes duration (< 10 years, ≥ 10 years, and unknown). Hypertension was characterized by systolic and diastolic blood pressure readings of at least 140 mmHg and 90 mmHg, respectively, self-reported hypertension, physician-diagnosed hypertension, or the requirement for medication to regulate elevated blood pressure [22]. CVD was self-reported as a diagnosis of congestive heart failure, angina pectoris, myocardial infarction, or coronary heart disease. Cancer diagnoses were self-reported by a physician. Hyperlipidemia was delineated by triglyceride (TG) levels ≥ 150 mg/dL (1.7 mmol/L) or total cholesterol (TC) levels ≥ 200 mg/dL (5.18 mmol/L), low-density lipoprotein (LDL) levels ≥ 130 mg/dL (3.37 mmol/L), or, in men, high-density lipoprotein (HDL) levels < 40 mg/dL (1.04 mmol/L), or, in women, HDL levels < 50 mg/dL (1.30 mmol/L) [23]. The healthy eating index (HEI) score was computed in accordance with HEI-2015 guidelines [24]. Physical activity was defined as engaging in moderate to vigorous exercise, fitness programs, or recreational activities for more than 10 min per week; otherwise, participants were classified as inactive. The eGFR was calculated using the CKD-EPI 2009 (Chronic Kidney Disease Epidemiology Collaboration) creatinine equation.

This study rigorously adhered to NHANES guidelines, meticulously adjusting for complex sampling designs and weights. Weighted averages for continuous variables and percentages for categorical ones were reported, ensuring accuracy and reliability. Statistical comparisons were conducted using appropriate methods: continuous variables were analyzed using ANOVA or the Kruskal-Wallis test, while categorical data differences were assessed using the chi-square test. To assess the impact of HGI on both all-cause mortality and CVD mortality, survey-weighted Cox regression analysis was employed. HGI was categorized into three tertiles, with the second tertile serving as the reference for exploratory analysis. To mitigate potential confounding factors, four multivariate models were constructed: Model I adjusted for age and sex. Model II expanded adjustments to include race/ethnicity, education level, PIR, body mass index (BMI), smoking status, comorbidities (hypertension, CVD, hyperlipidemia), physical activity, and alcohol intake. Model III further included variables such as diabetes duration, use of diabetes medications, hemoglobin, eGFR, and HEI. Model IV introduced HbA1C. To explore the nonlinear relationship between HGI and mortality outcomes, restricted cubic spline analysis with 4 knots was conducted. The likelihood ratio test was utilized to assess nonlinearity, ensuring robustness in the analysis. Additionally, analyses were stratified by various factors including age (< 60 or ≥ 60 years), sex (male or female), race/ethnicity (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, or other races including multi-racial), PIR (< 1, 1–3, ≥ 3), smoking status (never, ever, or current), BMI (< 30 or ≥ 30), HbA1c (< 7% or ≥ 7%), hypertension (yes or no), hyperlipidemia (yes or no), CVD (yes or no), GFR category (G1, G2, G3a + G3b, G4 + G5), use of diabetes medications (other, insulin, oral medications, and unknown), and diabetes duration (< 10 years, ≥ 10 years, and unknown). Interactions between HGI and these factors were rigorously assessed, enhancing the depth of analysis and interpretation.

Sensitivity analyses were also conducted to ensure the robustness of our findings: (1) Variation in HGI calculation: The existing literature shows slight variations in the calculation of HGI, primarily due to differences in the populations used to derive the formula for predicting HbA1c. To address this, we developed distinct predictive formulas for HbA1c using both the entire cohort and a subset specifically focused on DKD. These formulas were then used to calculate HGI, validating its association with mortality outcomes. (2) Mitigating Reverse Causality Bias: To mitigate potential reverse causality bias, we excluded participants who experienced mortality within a 2-year follow-up period from the analysis.