The NHANES is a nationally representative, cross-sectional study designed to evaluate the health and nutritional status of the US population. It employs a standardized protocol integrating detailed household interviews (capturing demographic, dietary, and health-related data), comprehensive physical examinations in mobile examination centers (MECs), and advanced laboratory testing of biological specimens. The analysis utilized data collected across nine survey cycles between 2001 and 2018. We excluded specific groups from our analysis to ensure a focused study population: (1) Age <40 years; (2) individuals with liver cancer; (3) individuals with high alcohol intake, defined as more than three alcoholic beverages per day for men and more than two per day for women; (4) individuals diagnosed with hepatitis B or C, identified by positive hepatitis B surface antigen or hepatitis C antibody/HCV RNA tests; and (5) individuals with ferritin saturation levels exceeding 50%. (6) Lack of information to evaluate MASLD; (7) individuals with UACR >30. Figure 1 provides a visual representation of the patient selection process.

The diagnosis of MASLD requires both evidence of hepatic steatosis and at least one metabolic risk factor (e.g., obesity, diabetes, or metabolic syndrome components), while excluding other liver diseases and excessive alcohol consumption. Liver biopsy is the gold standard for diagnosing hepatic steatosis, but due to its limitations such as invasiveness, non-invasive methods are preferred in clinical practice (15, 16). Currently, Controlled attenuation parameter techniques based on vibration controlled transient elastography have become an important alternative diagnostic tool due to their high accuracy and ability to simultaneously assess fibrosis. However, this study was limited by the availability of liver ultrasound data in most survey cycles of NHANES, so the US Fatty Liver Index (USFLI) was used to define MASLD. This methodological choice was based on previous validation studies confirming that a USFLI score ≥30 as a cutoff has reasonable diagnostic efficacy (AUROC of 0.80–0.85 compared to MRI-confirmed steatosis) (17, 18).

Urinary albumin and creatinine levels were used to calculate the UACR. To further investigate potential differences within the normal range of UACR levels, we categorized participants with UACR <30 mg/g into three subgroups based on previous studies (12): low (<4.67 mg/g), intermediate (4.67–7.67 mg/g), and high (7.68–30 mg/g) level groups.

The primary endpoints of this study were all-cause and cardiovascular disease (CVD)-specific mortality, as determined by records from the National Death Index (NDI) up to December 31, 2019. All-cause mortality includes deaths from heart disease, malignant neoplasms, nephritis, nephrotic syndrome and nephrosis, Alzheimer's disease, diabetes mellitus, chronic lower respiratory diseases, influenza and pneumonia, cerebrovascular diseases, accidents (unintentional injuries), and other causes. CVD-cause mortality as deaths due to cardiovascular disease. Further details on these classifications are provided in a previous study (19).

In this study, we adjusted for potential confounders that could influence the outcomes. Demographic characteristics were collected through standardized questionnaires during household interviews, including age, gender, race (Mexican American, White, Black, and other races), poverty status, education level (below high school, high school, and above high school), smoking status and alcohol assumption. Relevant laboratory measures, including liver function tests (ALT, AST) blood albumin, blood urea nitrogen, creatinine, and estimated glomerular filtration rate (eGFR), were obtained through mobile testing centers. Additionally, we collected information on the presence of hypertension (yes/no), diabetes (yes/no), and chronic kidney disease (yes/no), and CVD (yes/no; The above detailed definitions are attached in the Supplementary file).

This study included 3,412 participants, who were categorized into three groups based on UACR. For variables with missing values, we conducted multiple imputations based on “mice” package. Descriptive statistics were presented for both categorical and continuous variables. For continuous variables, the one-way analysis of variance was used to compare multiple samples. Categorical data were expressed as frequencies (n, %) and analyzed using the Chi-square test, with P < 0.05 considered statistically significant. Multiple Cox proportional hazards regression models were used to examine the association between UACR and the risk of all-cause mortality and CVD-cause mortality across three analytical models. Crude Model was adjusted for age, gender, ethnicity, poverty, Model 1 adjusted for, blood urea nitrogen, albumin eGFR, SII Alt, and Ast on the basis of Crude Model., and Model 2 included additional variables, including smoking status, history of hypertension, and diabetes, and CVD. Hazard ratios (HR) and corresponding 95% confidence intervals (CI) were calculated. Kaplan-Meier (KM) survival analysis was employed to compare all-cause and cardiovascular mortality risks across the three UACR. Additionally, a restricted cubic spline (RCS) regression model was employed to assess the linear relationship between continuous UACR levels and risk of death. Subgroup analyses were conducted, and interactions within subgroups were evaluated using the likelihood ratio test. Analyses were conducted using NHANES-recommended sampling weights to account for the complex survey design. Statistical analysis was performed using R Statistical Software (Version 4.3.3), with statistical significance defined as a P-value < 0.05. The study was approved by the National Center for Health Statistics Ethics Review Board, and all participants provided informed consent.

The baseline characteristics of the cohort, classified according to UACR levels, are shown in Table 1. The average age of the participants was 59.58 years, with 57.44% being male and 76.08% identifying as White. The prevalence of hypertension and diabetes was 64.29% and 37.69%, respectively. Compared to the low level of UACR, participants with the high level of UACR were older, had higher SII levels, and were more likely to have a history of hypertension, diabetes, and CVD. In contrast, the high level of UACR group had lower educational levels, higher poverty rates. There were no significant differences in BMI, Ast, blood urea nitrogen, blood albumin and smoking among the three groups. A total of 720 (15.93%) deaths occurred during the follow-up period, 238 of which were attributed to CVD. Participants in the group with high UACR levels had a higher mortality rate than those in the group with low UACR levels.

Table 2 presents the relationship between UACR and Risk of death in patients with MASLD. In fully adjusted multivariable Cox models, each 1 mg/g increment in UACR was associated with a statistically significant 4% increase in all-cause mortality risk (HR 1.04, 95% CI 1.03–1.05). When categorized by UACR risk levels and in the fully adjusted model (Model 3), participants with intermediate level of UACR showed 81% higher mortality risk (HR 1.81, 95% CI 1.32,2.48; P < 0.001), and participants with high level of UACR showed 169% higher mortality risk (HR 2.69, 95% CI 2.07,3.50; P < 0.001), compared with the low level of UACR group. These findings were consistent across the three models (Model 1, Model 2, and Model 3). The KM curve further indicated that participants with low level of UACR had a survival advantage over those with higher UACR levels (Figure 2A). The RCS regression demonstrated a non-linear association between UACR and all-cause mortality, displaying a distinct inverted L-shaped pattern (Figure 3A). When the level of UACR >7.467, the risk of all-cause death was significantly increased.

For CVD-cause mortality, a similar pattern was observed for relationship between UACR and CVD-cause death (Table 2). In multivariable-adjusted models, each 1 mg/g increase in UACR was associated with a 5% higher risk of CVD-cause death (HR 1.05, 95% CI 1.02–1.08). When grouped by UACR risk levels in the fully adjusted model (Model 3), compared with the low level of UACR group, both the intermediate level of UACR group [HR, 2.02 (95% CI, 1.14, 3.57), P < 0.05] and high level of UACR group [HR, 2.97 (95% CI, 1.76, 4.99), P < 0.0001] had a significantly higher risk of CVD-cause death. The KM curve further indicated that participants with low level of UACR had a survival advantage over those with higher UACR levels (Figure 2B). The relationship between UACR and CVD-cause mortality was also non-linear and inverted L-shaped based on the RCS model (Figure 3B). When the level of UACR >7.195, the risk of CVD-cause death was significantly increased.

Subgroup analyses assessing the stability of the association between UACR and mortality (including both all-cause and CVD-cause deaths) are presented in Tables 3, 4. The trends in the effects of UACR on mortality were consistent with those observed in the overall cohort. No significant interactions were found between UACR and the risk of death across the stratified variables (all p-interactions > 0.05).