Study design and population
The NHANES is an ongoing survey with a multistage and probabilistic sampling design that aims to survey the nutritional and health conditions of US individuals. The survey protocol was authorized by the National Center for Health Statistics’ ethical review board, and written consent was obtained. Details of NHANES are provided on its official website. We analyzed NHANES data from 2005–2018. Of the initial 70,190 participants enrolled, those younger than 18 years (n = 28,047) were excluded. Participants with diabetes (n = 8,130), impaired kidney function (n = 1), and those with missing GAR data (n = 19,774) were excluded. Individuals with missing covariate data (n = 5,868) or missing death data (n = 42) were excluded. Finally, 8,328 eligible participants were included in the study (Fig. 1).
2.2 Exposure definition
NHANES measured the blood glucose and albumin levels through venous blood which were collected in the mobile examination center (MEC) and stored at -20°C before analyzing26. The measurement of blood glucose and albumin can be accessed from the official website. (https://wwwn.cdc.gov/nchs/nhanes/AnalyticGuidelines.aspx). Based on previous studies, GAR was calculated accordingly: \(\frac{\text{S}\text{e}\text{r}\text{u}\text{m} \text{G}\text{l}\text{u}\text{c}\text{o}\text{s}\text{e}(\text{m}\text{m}\text{o}\text{l}/\text{L})}{\text{S}\text{e}\text{r}\text{u}\text{m} \text{A}\text{l}\text{b}\text{u}\text{m}\text{i}\text{n}(\text{g}/\text{L})}\times {10}^{2}\) 11.
2.3 Outcome definition
In our study, kidney injury was defined as eGFR ≤ 60 ml/min per 1.73m2. The eGFR was estimated by CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) creatinine-based Eq. 27. The creatine data was obtained through linkage of NHANES data28–30.
2.4 Ascertainment of covariates
In our study, several variables were analyzed as potential confounders, based on previous studies31–33. These variables include age, sex (male or female), race/ethnicity (non-Hispanic black, non-Hispanic white, other race, Mexican American and other Hispanic), marital status (married, divorced, or living alone), PIR (< 1, 1–2, 2–4, ≥ 4), education level (less than high school, high school diploma, or higher), smoking, alcohol consumption, physical activity, coronary heart disease (yes or no), hyperlipidemia, BMI, ALT, AST, ALP, and ALB. PIR was divided into three categories:<1, 1–2, 2–4 and > 434. Smoking was categorized as three states: never smokers (< 100 cigarettes in their lifetime), former smokers (close to 100 cigarettes in their life time) and current smokers (currently smoked some days and more than 100 cigarettes)34. Alcohol consumption was divided as former (quit drinking when s/he answered the questionnaire), moderate (less than two drinks per day), never, mild (two drinks per day), heavy (three drinks or more per day)34. We calculated BMI as weight (kg) divided by height in meters squared and categorized it as underweight (BMI < 18.5 kg/m2), normal (18.5 < BMI < 25.0 kg/m2), overweight (25.0 kg/m2 < BMI < 30.0 kg/m2) and obesity (BMI > 30.0 kg/m2)35. We classified hyperlipidemia as the fasting plasma concentration of triglyceride surpasses a threshold value (> 150 mg/dL)36.
2.5 Statistical analysis
All statistical analyses were conducted in accordance with recommendations from the Centers for Disease Control (CDC). The Anderson-Darling test was performed to check the distribution of continuous variables. Continuous variables with Gaussian distribution were presented as means with standard errors (SE). Continuous variables with non-Gaussian distribution were presented as medians with interquartile ranges (IQR). Categorical parameters were presented as proportions with SE. Differences between four groups were calculated using the Kruskal–Wallis H test (continuous variables with non-Gaussian distribution) and ANOVA tests (continuous variables with Gaussian distribution). Differences between groups were calculated using the chi-squared test for categorical variables. Univariate and multivariate weighted logistics regressions were conducted to identify the relationship between GAR and kidney injury. Model 1 was established after adjusting for gender, race, marital status, PIR, and education level. Model 2 was adjusted for gender, race, marital status, PIR, education level, BMI, smoking, alcohol consumption, and physical activity. Model 3 was constructed after adjusting for covariates such as gender, race, marital status, PIR, education level, BMI, smoking, alcohol consumption, physical activity, coronary heart disease, hyperlipidemia, ALT, AST and ALP. After adjusting all other covariates in different subgroups, we tested the interaction. Age, race, sex, marital status, PIR, education level, alcohol, smoke, coronary heart disease and hyperlipidemia were stratified for subgroup analyses. Additionally, RCS analysis was performed to investigate the nonlinear relationship between GAR and kidney injury. Finally, Cox proportional hazard model and Kaplan-Meier curve were conducted to identify the correlation between GAR and all-cause death and cardiovascular death among participants with kidney injury. All statistical analyses were performed in R 4.3.2. Statistical significance was defined as a bilateral P value < 0.05.