This cross-sectional study, based on the National Basic Public Health Service Program for the elderly and focusing on diseases of glycometabolic and lipid metabolic disorders, recruited 21,025 participants from August 2020 to October 2021 in Minzhong Town and Torch Development Zone, Zhongshan City, Guangdong Province. The recruitment was carried out through local advertisements, invitations, health lectures, or community referrals. All participants had no history of severe infectious diseases, terminal malignancies, recent major surgeries, or trauma and had resided locally for over six months. The questionnaire survey was carried out by qualified, trained investigators. Specifically, it was completed on-site in a one-to-one, face-to-face manner with each question posed individually. The questionnaire included questions related to the name, gender, age, ethnicity, education, marital status, smoking and drinking status, and medical history of the study participants. Physical examination, clinical evaluation, and laboratory testing were performed by expert physicians. The study protocol was approved by the Ethics Committee of Guangdong Pharmaceutical University [Medical Ethics (2019) No.109], and all study participants provided written informed consent. All procedures were conducted by the ethical guidelines of the Declaration of Helsinki. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.

Participants were excluded under the following conditions: if they did not undergo a blood test (n = 4,520), had missing data for calculating the GNRI (n = 7,521), had a history of cancer (n = 153), were aged less than 60 years (n = 1,160), or lacked liver imaging data. Consequently, a total of 7,628 elderly individuals were included in the final analysis (Figure 1).

The sample calculation formula for cross-sectional study is as follows:

N denotes the sample size, (Z _(1−α)/2) is the critical value of the standard normal distribution at the selected confidence level, p represents the expected prevalence, and d is the allowable error. The current prevalence of MASLD in China is 29.6% (10). Thus, the sample size was calculated with α = 0.05, (Z_(1−α)/2) = 1.96, p = 0.296, d = 0.1 × p = 0.0296. The final calculated sample size was 914 cases. Taking into account a 20% questionnaire non-response rate, at least 1,143 cases needed to be included in our study. The number of respondents was 21,025 cases.

The diagnostic criteria used in this study are based on those recommended in the Chinese Guidelines for the Diagnosis and Treatment of Fatty Liver Disease (2024 Edition) (22).The diagnostic criteria for MASLD (Chinese standard) were based on hepatic steatosis identified by imaging or biopsy, exclusion of liver injury from other known causes (presence of genotype 3 HCV infection, drug-induced fatty liver, hepatic leguminous nuclear degeneration, and malnutrition leading to fatty liver), exclusion of excessive alcohol intake (ethanol intake: male: ≥ 210 g/week; female: ≥ 140 g/week), and the presence of at least one metabolic risk factor for cardiovascular disease: (1) Body mass index (BMI) ≥ 24.0 kg/m² or waist circumference (WC) ≥ 90 cm (male) ≥ 85 cm (female) or body fat content and percentage body fat exceeded the standard; (2) Fasting plasma glucose (FPG) ≥ 6.1 mmol/L or 2-h post-load glucose levels ≥ 7.8 mmol/L or glycated hemoglobin ≥ 5.7% or diagnosis of type 2 diabetes or treatment for type 2 diabetes; (3) Blood pressure ≥ 130/85 mmHg or specific antihypertensive drug treatment; (4) Plasma triglycerides (TG) ≥ 1.70 mmol/L or lipid lowering treatment; (5) Plasma high-density lipoprotein cholesterol (HDL-C) ≤ 1.0 mmol/L (male) or ≤ 1.3 mmol/L (female) or lipid lowering treatment. These criteria are similar to the international MASLD diagnostic criteria, except that the thresholds of certain variables have been optimized according to the epidemiological characteristics of the Chinese population.

In addition, to enhance the comparability of the study findings, we also applied the diagnostic criteria from the International Multi-Society Consensus on MASLD (2023) (23).

Ideal body weight is calculated using the Lorentz formula. For males, ideal body weight = height (cm) – 100 – [(height (cm) – 150)/4.0]; for females, ideal body weight = height (cm) – 100 – [(height (cm) – 150)/2.5]. The formula for calculating GNRI: GNRI = [1.489 × serum albumin value (g/L)] + [41.7 × actual body weight (kg)/ideal body weight (kg)]. To ensure the weight of albumin and to mitigate the effect of disease states (oedema, pleural, and abdominal effusions) on body weight, in the GNRI formula, if actual body weight > ideal body weight, actual body weight/ideal body weight = 1; if actual body weight < ideal body weight, the calculation is based on the actual ratio. Grouping was based on GNRI quartiles (Q1–Q4), with GNRI < 104.79 in group Q1, 104.79 ≤ GNRI < 107.51 in group Q2, 107.51 ≤ GNRI < 109.77 in group Q3, and GNRI ≥ 109.77 in group Q4. Furthermore, we also employed four nutrition-related risk levels as defined in previous study: high risk (GNRI < 82), medium risk (82 ≤ GNRI < 92), low risk (92 ≤ GNRI ≤ 98), and no risk (GNRI > 98) (14).

Age was classified into two groups: age < 75 and age ≥ 75. Height, weight, and WC were measured by investigators who had received uniform training, using standard measurement methods. BMI was calculated as weight in kilograms divided by height in meters squared, was categorized as two groups: BMI < 24.0 kg/m² and BMI ≥ 24.0 kg/m². Ethnic group was categorized as Han ethnicity and ethnic minorities. Educational attainment was classified as primary school and below, junior high school, high school and undergraduate and above. Marital status was categorized into four groups: unmarried, married or cohabiting, divorced or separated, and widowed. Participants were classified as never smokers (individuals who had never smoked in their lives), former smokers (those who had smoked ≥ 100 cigarettes but had quit), and current smokers (those who had smoked ≥ 100 cigarettes and were still smoking). Regular drinkers were defined as individuals who consumed alcoholic beverages at least once a week for a continuous period of six months. Those who did not meet this criterion were defined as non-drinkers. After fasting for at least 8 h, fasting venous blood samples were collected from each study participant by professional medical workers. The collected blood samples were stored in a low-temperature transport box and sent to the hospital's clinical laboratory within 2 h for biochemical parameter tests. The biochemical parameters included fasting plasma glucose, total cholesterol (TC), TG, low-density lipoprotein cholesterol (LDL-C), HDL-C, alanine aminotransferase (ALT), aspartate aminotransferase (AST), and serum uric acid (UA). Hypertensive patients were defined as those who met at least one of the following criteria: self-reported hypertension, systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, hypertension documented in medical records, or receiving antihypertensive treatment. Glycemic status was defined as a three-category variable (normal, pre-diabetes, diabetes). Diabetic patients were defined as those who met at least one of the following criteria: self-reported diabetes, FPG ≥ 7.0 mmol/L, glycated hemoglobin (HbA1c) ≥ 6.5%, diabetes documented in medical records, or receiving antidiabetic treatment. Pre-diabetes was defined as meeting laboratory criteria (FPG 5.6–6.9 mmol/L or HbA1c 5.7–6.4%) in the absence of diabetes diagnosis or treatment. Normal glycemic status was defined as FPG < 5.6 mmol/L and HbA1c < 5.7% with no evidence of diabetes. This study categorized patients with MASLD into three subtypes. Individuals diagnosed with diabetes are classified as MASLD (diabetes subtype), while among those without diabetes, participants with a BMI ≥ 24 kg/m² [according to Chinese BMI standards (24)] are considered to have MASLD (overweight/obesity subtype). Finally, in the non-diabetic population, individuals with a BMI < 24 kg/m² but with metabolic risk abnormalities are classified as MASLD (lean metabolic disorder subtype), while the remaining individuals are classified as non-MASLD (9). In parallel, an international classification was applied using the Asian-specific BMI cut-off of ≥ 23 kg/m², as specified in the consensus MASLD diagnostic criteria (2023) (23). Within this framework, non-diabetic patients with BMI ≥ 23 kg/m² were categorized into the overweight/obesity subtype, and those with BMI < 23 kg/m² but with accompanying cardio metabolic abnormalities were assigned to the lean metabolic disorder subtype.

Continuous variables that conformed to normal distribution were described by mean ± standard deviation and were compared between groups using the independent samples t-test; continuous variables that did not conform to normal distribution were expressed as median and upper and lower quartiles [M (P₂₅, P₇₅)], and comparisons between groups were made using a nonparametric test (Mann-Whitney test). Categorical variables were described by frequency and percentage, while the differences among groups were compared using the chi-square test. To explore the independent association between GNRI and the risk of MASLD in the elderly population, and to estimate the odds ratios (ORs) and 95% confidence intervals (CIs), multivariate binary logistic regression models were employed. Three models were incorporated to control for confounders. The included covariates are based on previous study covering MASLD, aging, and nutrition (20, 21, 25–31). Crude model was non-adjusted. Model 1 was adjusted for gender, age group, ethnic group, marital status and educational attainment, model 2 was further adjusted for smoking status, drinking status, BMI group and WC, and model 3 was further adjusted for ALT, AST, UA, TC, TG, HDL-C, hypertension and glycemic status. All the values of variance inflation factors in the final model were below five, confirming the absence of multicollinearity. To investigate whether there was a nonlinear association between GNRI and the risk of MASLD in the elderly population, restricted cubic splines (RCS) (3 knots: at the 10th, 50th, and 90th percentiles) and likelihood ratio test were conducted. The selection of knots for the RCS curves was guided by the minimization of Akaike's Information Criterion. If the association was nonlinear, the inflection point was determined by identifying the GNRI value at which the predicted OR from the model equaled 1. The receiver operator characteristic curve (ROC) and the area under the curve (AUC) were employed to access the predictive accuracy of multivariate-adjusted GNRI for the risk of MASLD. Mediation analyses were used to investigate whether the relevance of GNRI to MASLD could be explained by BMI, WC, ALT, AST, TC, TG, LDL-C, and HDL-C after adjusting for covariates in the model 3. Stratified analyses were conducted based on gender, age, BMI, smoking status, drinking status, hypertension, and glycemic status. Additionally, we used multinomial logistic regression with multivariable adjustment to further explore whether an increase in GNRI (or higher quantiles) would increase the risk of a particular MASLD subtype compared to individuals without MASLD. Three models were included in the study to control for confounding factors. The crude model was unadjusted. Model 1 adjusted for gender, age group, ethnic group, marital status, and educational attainment. Model 2 further adjusted for smoking status and drinking status. Model 3 further adjusted for ALT, AST, UA, TC, TG, HDL-C, and hypertension. All logistic regression models were developed in adherence to the widely accepted criterion of a minimum of 10 outcome events per predictor variable. In this study, several variables (ethnic group, marital status, and educational attainment) exhibited some degree of missingness (11.55%, 11.06%, and 10.45%, respectively). For all other analytical variables, the missing rates were below 1% or the data were complete. All missing data were imputed with multiple imputation methods.

We conducted additional analyses to examine whether GNRI is associated with the risk of MASLD (as defined by the 2023 International Multi-Society Consensus) and its subtypes in Chinese elderly individuals. We also added a full analysis of the international MASLD criteria, and the results of this part are presented in the supplementary material.

All statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC) and R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value < 0.05 was regarded as statistically significant.

Among the 7,628 participants, 40.4% were male and 59.6% were female, and the prevalence of MASLD was 38.8% (2,956 cases). The median age of the participants was 67 years, and the median BMI was 23.82 kg/m². The characteristics of individuals with MASLD or non-MASLD are summarized in Table 1. According to the GNRI nutritional risk classification, 96.6% of all participants in this study were classified as no risk, including 99.6% of those with MASLD and 94.7% of those without MASLD. Individuals with MASLD were more likely to be aged < 75 years, female, married or cohabiting, never smokers, non-drinkers, BMI > 24.0 kg/m². Compared with those who were without MASLD, participants with MASLD tended to have a higher GNRI, BMI, WC, FPG, ALT, AST, UA, TC, TG, LDL-C, and a lower HDL-C. Additionally, those with MASLD had a higher prevalence of hypertension and diabetes.

Table 2 presents the multivariate binary logistic regression analysis of GNRI and the risk of MASLD. In Model 3, after adjusting for potential confounding factors, each 1-unit increase in GNRI was associated with a 12% higher risk of MASLD (OR = 1.12, 95% CI: 1.10–1.13). Similarly, when GNRI was categorized into quartiles, compared with the first quartile (Q1), the higher the GNRI level, the higher the prevalence of MASLD in the elderly population. After full adjustment, the ORs and 95% CIs for GNRI quartiles and the risk of MASLD were as follows: Q2: OR = 1.56, 95% CI: 1.31–1.84; Q3: OR = 1.93, 95% CI: 1.64–2.28; Q4: OR = 2.70, 95% CI: 2.28–3.20; P for trend < 0.001. Collectively, these results indicate a dose-response relationship, where both per-unit increases and higher quartiles of GNRI consistently elevate MASLD risk.

To investigate whether there was a nonlinear association between GNRI and the prevalence of MASLD, RCS, and likelihood ratio test were conducted (Figure 2). The results demonstrated a nonlinear J-shaped association between GNRI and the risk of MASLD (P for overall < 0.001; P for nonlinear = 0.008). Taking GNRI = 107.59 as the reference point (OR = 1), we observed that: when GNRI < 107.59, the OR for MASLD was consistently < 1, suggesting a protective effect. When GNRI ≥ 107.59, the OR increased progressively, indicating an increased risk.

The ROC and AUC were employed to assess the predictive accuracy of multivariate-adjusted GNRI for the risk of MASLD (Figure 3). The GNRI demonstrated predictive capability for the prevalence of MASLD in the elderly population (AUC: 0.802, 95% CI: 0.792–0.812).

Mediation analyses indicated that BMI, WC, TC, TG, LDL-C, HDL-C, ALT, and AST significantly mediated the association between GNRI and MASLD (Figure 4). Among the examined mediators, BMI accounted for the largest proportion of the total effect of GNRI on MASLD (21.1%, 95% CI: 6.4%−29.2%), followed by ALT (5.6%), TG (2.4%), AST (2.1%), HDL-C (1.7%), LDL-C (1.5%), TC (1.1%), and WC (0.7%), all of which exhibited statistically significant mediation effects (P < 0.05).

After adjusting for potential confounders, stratified analyses based on gender, age, BMI, smoking status, drinking status, hypertension, and glycemic status demonstrated that the risk of MASLD increased with increasing levels of the GNRI quartile for people with different characteristics (Table 3). Among them, male with high GNRI levels, those aged ≥ 75 years, those with a BMI < 24.0 kg/m², current smokers, regular drinkers, people with hypertension and normal glycemic status have a relatively higher risk of MASLD.

The ORs and 95% CIs for the association between GNRI and MASLD subtypes, estimated using multinomial logistic regression with multivariable adjustment and the non-MASLD group as the reference, are presented in Table 4. Regardless of the MASLD subtype, higher GNRI levels (or higher quantiles) were significantly associated with an increased risk of that subtype compared with individuals without MASLD. When GNRI was categorized into quartiles, the risk of belonging to the overweight/obese subtype was the highest, followed by the diabetic subtype and then the lean metabolic disorder subtype. Similar trends were observed when GNRI was analyzed as a continuous variable. After adjusting for covariates in model 3, when the GNRI Q1 group was set as the reference, the ORs and 95% CIs were as follows: MASLD (diabetes): Q2: OR = 1.54, 95% CI: 1.20–1.97; Q3: OR = 2.08, 95% CI: 1.64–2.64; Q4: OR = 3.36, 95% CI: 2.66–4.25; MASLD (overweight/obesity): Q2: OR = 2.21, 95% CI: 1.76–2.78; Q3: OR = 3.04, 95% CI: 2.43–3.79; Q4: OR = 4.90, 95% CI: 3.93–6.11; MASLD (lean metabolic disorder): Q2: OR = 1.82, 95% CI: 1.39–2.37; Q3: OR = 2.23, 95% CI: 1.72–2.89; Q4: OR = 3.04, 95% CI: 2.34–3.95; all P_−trend < 0.001.

This is a comprehensive study exploring the associations between GNRI and the risk of MASLD and its subtypes. We found that GNRI could be a strong predictor of MASLD risk in Chinese older adults using ROC analysis and RCS modeling—which revealed a significant non-linear relationship and identified GNRI ≥ 107.59 as a critical risk threshold. We also investigated the effect of mediating variables on the outcome using mediation analysis. In addition, we used the latest Chinese diagnostic criteria for MASLD and utilized ultrasonographic findings of hepatic steatosis to aid in the diagnosis. In order to enhance comparability, a parallel analysis using international diagnostic criteria for MASLD (2023) was conducted, yielding largely consistent findings. Furthermore, we rigorously accounted for potential confounders through multivariate adjustment across all analyses. Despite these strengths, several limitations warrant consideration. First, this is a cross-sectional study, so we cannot infer causal associations. Second, these results are based on older adults in China, which may limit replication and generalization to other populations. Third, although we attempted to control for confounding variables through multifactorial adjustment and stratified analyses, there may be other residual confounders. Finally, while GNRI was originally developed for hospitalized older adults, numerous studies have extended its application to community-dwelling elderly populations and those with specific chronic conditions, aligning with our approach. Nevertheless, further validation of GNRI for quantifying over nutrition-related risks (as opposed to its initial undernutrition focus) remains valuable.