All the data in this study came from NHANES, which is a research program designed to assess the health and nutrition status of adults and children in the United States administered by the National Center for Health Statistics (NCHS) (32). NHANES has a complex, multistage, probability sampling design that allows it to be generalized to the US population (32). The protocol for the NHANES study was approved by the NCHS Research Ethics Review Board, and its detailed design and data are publicly available at https://www.cdc.gov/nchs/nhanes/. The current study included all participants older than 20 years with complete TyG index, albuminuria, and glomerular filtration rates from NHANES 2011 to 2018 who were in the fasting subsample.

Urinary albumin and creatinine were measured by solid-phase fluorescent immunoassay and modified Jaffe kinetic method, respectively. UACR was calculated by dividing the urinary albumin (mg/dL) concentration by the urinary creatinine concentration (mg/L). UACR was categorized as <30, 30–300, and ≥300mg/g in subgroup analysis. Albuminuria, defined as an UACR>30 mg/g, was considered as an outcome variable in our analysis (33). UACR 30–300 and ≥ 300mg/g correspond to moderately increased albuminuria and severely increased albuminuria, respectively.

TyG index was calculated as ln (fasting triglyceride[mg/dl] × fasting plasma glucose[mg/dl]/2). HOMA-IR was calculated using the following formula: (fasting insulin[µIU/ml] × fasting plasma glucose[mmol/L])/22.5. HOMA-IR was log-transformed for modeling in this study. Metabolic syndrome is defined based on criteria developed jointly by international organizations such as the International Diabetes Federation, the National Heart, Lung and Blood Institute and the American Heart Association (34). Metabolic syndrome includes at least three of the five: elevated waist circumference (WC, cm), hypertension, hyperglycemia, elevated TG and decreased high density lipoprotein cholesterol (HDL-c, mg/dl).

Demographic data (age, gender, and race) was obtained. Some potential covariates were also included in this study, such as smoking status (never/former/current), physical activity (vigorous/moderate/less than moderate), hypertension (yes/no), diabetes (yes/no), cardiovascular disease (yes/no), body mass index (BMI, kg/m²), WC, total cholesterol (TC, mg/dl), HDL-c, low density lipoprotein cholesterol (LDL-c, mg/dl), serum creatinine (Scr, mg/dl), and estimated glomerular filtration rate (eGFR, ml/min/1.73 m²). BMI was categorized as <25, 25-29.9 and ≥30 kg/m², which corresponded to normal weight, overweight, and obese population for participants. The presence of cardiovascular disease was determined based on self-reported history of heart attack, stroke, congestive heart failure, coronary artery disease, or angina. eGFR was calculated according to the CKD Epidemiology Collaboration (CKD-EPI) creatinine equation consisting of age, gender, race, and Scr (35). Detailed measurement procedures for all variables in this study were publicly available in the NHANES database.

All statistical analyses were performed in accordance with Centers for Disease Control and Prevention (CDC) guidelines. A complex multistage cluster survey design was considered, and fasting subsample weights combined with four cycles were applied. Continuous variables were expressed as weighted mean with standard error (SE), and categorical parameters were presented as proportions. Weighted one-way ANOVA and weighted chi-squared tests were used for multiple-group comparisons of continuous and categorical variables respectively. Logistic and linear models were used to test the association between TyG index (continuous/quartile) and albuminuria and UACR in different models. The variance inflation factor (VIF) was employed to assess the level of collinearity among independent variables (36). The VIF is computed using a multivariate linear regression model. The VIF of each independent variable is equal to 1 divided by (1 − R²), where R² represents the coefficient of determination obtained by regressing the independent variables on the remaining independent variables. Independent variables with VIF greater than 5 were excluded. In model 1, no covariates were adjusted. In model 2, age, gender, and race were adjusted. Model 3 was adjusted for age, gender, race, smoking status, physical activity, hypertension, diabetes, cardiovascular disease, BMI, Scr, and eGFR. Weighted Pearson correlation analysis was used to evaluate the correlation between TyG index and other parameters. ROC was used to compare TyG index and TyG-derived indices with log (HOMA-IR). Age (<60/≥60 years), gender (female/male), BMI (normal weight/overweight/obese), diabetes (yes/no), hypertension (yes/no), metabolic syndrome (yes/no), cardiovascular disease (yes/no), and eGFR (<60/60-90/≥90ml/min/1.73 m²) were stratified for subgroup analysis. Empower software (http://www.empowerstats.com) and R version 4.1.0 (http://www.R-project.org) were employed for all analyses. P value <0.05 was considered statistically significant.

According to the inclusion criteria, a total of 9872 subjects were included in the study, with an average age of 46.81 ± 0.35 years, including 48.94% male. We compared general information and clinical indicators for subjects in the non-albuminuria, moderately increased albuminuria and severely increased albuminuria groups ( Table 1 ). Compared with non-albuminuria group, age, male, smoking proportion, less than moderate physical activity, hypertension prevalence, diabetes prevalence, cardiovascular disease prevalence, metabolic syndrome prevalence, BMI, WC, FPG, FIns, TG, Scr, urinary albumin, and HOMA-IR were significantly increased in moderately increased albuminuria and severely increased albuminuria groups ( P <0.001), and LDL-c, HDL-c, urinary creatinine, and eGFR were decreased ( P <0.05). There were also differences in race distribution among groups ( P <0.001). Note that TyG levels were higher in the moderately increased albuminuria and severely increased albuminuria groups than in the non-albuminuria group ( P <0.001).

According to the TyG level of all subjects, they were divided into four groups from low to high: quartile I (TyG<8.07), quartile II (8.07<TyG ≤ 8.50), quartile III (8.50<TyG ≤ 8.96), and quartile IV (TyG>8.96) ( Table 2 ). Compared with quartile I-TyG group, age, male, smoking proportion, hypertension prevalence, diabetes prevalence, cardiovascular disease prevalence, metabolic syndrome prevalence, BMI, WC, FPG, FIns, TG, TC, Scr, urinary albumin, and HOMA-IR were significantly increased in quartile II-TyG, quartile III-TyG, and quartile IV-TyG groups increased significantly ( P <0.001), and LDL-c, HDL-c, and eGFR decreased significantly ( P <0.001). Most importantly, with the gradual increase of TyG level, the UACR level (19.34 ± 2.44 vs. 21.88 ± 2.94 vs. 25.75 ± 2.59 vs. 72.46 ± 10.90, P <0.001) and the number of people with albuminuria (7.13% vs. 7.66% vs. 8.27% vs. 16.58%, P < 0.001) gradually increased.

Our results showed that higher TyG index was associated with increased likelihood of increased albuminuria. This association was significant in model 1 [OR (95%CI): 1.871 (1.725-2.030), P <0.001], model 2 [OR (95%CI): 1.772 (1.622-1.936), P <0.001], model 3 [OR (95%CI): 1.391 (1.258-1.539), P <0.001], and the model with additional adjustment for log (HOMA-IR) [OR (95%CI): 1.346 (1.204-1.506), P <0.001] or metabolic syndrome [OR (95%CI): 1.316 (1.178-1.470), P <0.001]. We transformed TyG index from a continuous variable to a categorical variable. Compared to the lowest TyG index quartile, the highest TyG index quartile had 32.0% and 27.3% increased risk of albuminuria in the model with additional adjusted for log (HOMA-IR) or metabolic syndrome, respectively ( Table 3 ). Using the UACR as a dependent variable for the linear regression analysis, we can also find that TyG index is closely related to UACR ( Table 4 ).

Weighted Pearson correlation analysis suggested that the TyG index was consistently positively correlated with age, gender, race, smoking status, hypertension, diabetes, cardiovascular disease, WC, FPG, TG, TC, LDL-c, Scr, urinary albumin, UACR, albuminuria, and negatively correlated with HDL-c, urinary creatinine, and eGFR, whether unadjusted, or adjusted for log (HOMA-IR), metabolic syndrome ( P <0.001) ( Table 5 ).

​After removing covariates with VIF greater than 5, we performed a multivariate logistic regression analysis ( Table 6 ). The independent variables included the TyG index, log (HOMA-IR), metabolic syndrome, age, gender, race, smoking status, physical activity, hypertension, diabetes, cardiovascular disease, BMI, Scr, and eGFR. The dependent variable was albuminuria. The results showed that TyG index [OR (95%CI): 1.283 (1.138-1.446), P <0.001] was more strongly associated with albuminuria than log (HOMA-IR) [OR (95%CI): 1.138 (0.908-1.426), P =0.263] and metabolic syndrome [OR (95%CI): 1.213 (1.024-1.438), P <0.05].

As Figure 1 shows the performance for evaluating the endpoint among TyG index, TyG-derived indices, and log (HOMA-IR) for albuminuria risk, the AUC of the marker is as follows: BMI 0.557, WC 0.589, log (HOMA-IR) 0.597, TyG-BMI (TyG × BMI) 0.587, TyG-WC (TyG ×WC) 0.616, and TyG 0.616.

Our subgroup analysis showed that the degree of association between TyG index levels and albuminuria was inconsistent across populations. The interaction test suggested that the relationship between TyG index and albuminuria was influenced by age (<60/≥60), BMI (normal weight/overweight/obese), diabetes (yes/no), and metabolic syndrome (yes/no) stratification ( P <0.05). Age < 60, overweight/obesity, diabetes, and metabolic syndrome may be effector modifiers ( Figure 2 ).