Metabolic-Inflammatory Signatures and MASLD Risk: Insights from Composite Biomarkers and Predictive Modelling
A
Xiao-AnLiu1
ChengDing1
Zu-YuWang1
Han-XuanWang¹1
Cheng-RunZhang¹1
Dr.
RenLang1,2✉
Email
1Department of Hepatobiliary, Pancreas & Spleen Surgery, Beijing Chao-Yang HospitalCapital Medical UniversityBeijingChina
2Department of Hepatobiliary, Pancreas & Spleen SurgeryBeijing Chao-Yang HospitalBeijingChina
Xiao-An Liu†¹, Cheng Ding†¹, Zu-Yu Wang†¹, Han-Xuan Wang¹, Cheng-Run Zhang¹, Ren Lang*¹
¹ Department of Hepatobiliary, Pancreas & Spleen Surgery, Beijing Chao-Yang Hospital, Capital Medical University, Beijing, China.
* Correspondence: Dr. Ren Lang, Department of Hepatobiliary, Pancreas & Spleen Surgery, Beijing Chao-Yang Hospital, Beijing, China. Email: dr_langren123@126.com ORCID: https://orcid.org/0000-0001-9759-6725
† These authors contributed equally to this work and should be considered co-first authors.
Abstract
Objective
Metabolic dysfunction-associated fatty liver disease (MASLD) is the most prevalent chronic liver disease worldwide. This study, based on data from the National Health and Nutrition Examination Survey (NHANES), examined associations between composite inflammatory biomarkers—the inflammatory burden index (IBI), systemic inflammation response index (SIRI), aggregate index of systemic inflammation (AISI), and remnant cholesterol inflammatory index (RCII)—and MASLD and developed a predictive model using machine learning.
Methods
Data from 5,112 NHANES participants (1999–2010) were analyzed. Associations of the IBI, SIRI, AISI, and RCII with MASLD were assessed using multivariate logistic regression, restricted cubic splines, and subgroup and sensitivity analyses. Eight machine learning models were constructed: AdaBoost, Decision Tree, Elastic-Net, K-Nearest Neighbors, Multilayer Perceptron, Ridge Regression, Support Vector Machine, and extreme gradient boosting (XGBoost), with a 7:3 ratio for training and validation sets. Model performance was evaluated by receiver operating characteristic (ROC) curves and the area under curve (AUC), and Shapley additive explanation (SHAP) values were used to enhance interpretability.
Results
The prevalence of MASLD was 20.0%. All four inflammatory biomarkers showed significant dose–response and nonlinear positive associations with MASLD (p < 0.001). RCII had the strongest effect (odds ratio [OR] = 5.76, 95% confidence interval [CI]: 4.08–8.14). Subgroup analyses revealed heterogeneity across populations. All biomarkers achieved an area under the curve (AUC) > 0.70, with XGBoost showing the best performance (training AUC = 0.872, test AUC = 0.802). SHAP analysis identified race, gamma-glutamyl transferase, albumin, high-density lipoprotein cholesterol, age, and RCII as major predictors of MASLD.
Conclusion
RCII is a powerful biomarker for predicting MASLD. The XGBoost-based model demonstrated excellent diagnostic value, which may provide a reliable tool for early screening and precision prevention.
Keywords:
Metabolic dysfunction-associated fatty liver disease
National Health and Nutrition Examination Survey
Machine learning
Inflammatory biomarkers
A
A
1. Background
Non-alcoholic fatty liver disease (NAFLD) is one of the most prevalent chronic liver diseases worldwide, affecting approximately 35% of the global adult population. Its prevalence has increased by nearly 50% since the 1990s[1, 2]. NAFLD encompasses a broad spectrum of hepatic conditions, ranging from simple steatosis to more advanced forms such as non-alcoholic steatohepatitis (NASH), cirrhosis, and hepatocellular carcinoma (HCC)[3, 4]. Beyond liver pathology, NAFLD is strongly associated with extrahepatic comorbidities, including cardiovascular disease, type 2 diabetes, chronic kidney disease, and increased risk of premature mortality[5, 6].
Recently, the term NAFLD has been redefined as metabolic dysfunction-associated fatty liver disease (MASLD) to better reflect its integration within systemic metabolic dysfunction[7]. The pathogenesis of MASLD is multifactorial, and inflammation is central to both its initiation and progression. Systemic inflammation exacerbates metabolic dysfunction and accelerates the progression from simple steatosis to more advanced conditions such as NASH and cirrhosis. It is also involved in the development of comorbidities commonly observed in MASLD, including cardiovascular disease and diabetes.
Several composite inflammatory indices have been developed to provide a nuanced assessment of inflammatory activity. These include the Inflammatory Burden Index (IBI), Systemic Inflammation Response Index (SIRI), Aggregate Index of Systemic Inflammation (AISI), and Remnant Cholesterol Inflammatory Index (RCII), each combining multiple biomarkers to provide a comprehensive understanding of systemic inflammation and its impact on disease progression[811].
The IBI incorporates markers such as white blood cell count and C-reactive protein (CRP), which provide an overall measure of inflammatory burden. These markers are consistently elevated across inflammatory states and are useful in assessing MASLD severity[12]. The SIRI, based on neutrophil, monocyte, and lymphocyte counts, reflects the balance between pro- and anti-inflammatory immune cells and has been investigated as a prognostic marker in several diseases, including MASLD, where shifts in immune cell ratios are linked to disease progression[13]. The AISI includes neutrophil, monocyte, lymphocyte, and platelet counts, with particular focus on the inflammatory processes involved in atherosclerosis. Since cardiovascular disease is a major comorbidity in MASLD, the AISI is particularly relevant, as it reflects both the systemic inflammation involved in MASLD and the associated cardiovascular risk[14]. The RCII, a relatively recent index, incorporates remnant cholesterol (RC), which is the difference between total cholesterol (TC) and the cholesterol carried by high-density and low-density lipoproteins. This index reflects systemic inflammation and accounts for lipid metabolism, which plays an essential role in MASLD pathogenesis. Given the prominent role of lipid dysregulation in MASLD, RCII provides a novel perspective on the intersection of lipid metabolism and inflammation in this disease[15].
These indices have demonstrated significant associations with the prognosis of MASLD and other chronic conditions, indicating their potential value for identifying individuals at high risk of progression. However, the mechanisms linking these markers to MASLD are not fully understood. Important questions include the molecular pathways of inflammation-driven hepatic steatosis and fibrosis, the causal role of systemic inflammatory states, and the heterogeneity across different populations. Further investigations are needed to clarify the complex relationships between inflammation and MASLD and to confirm the utility of these indices as biomarkers for MASLD diagnosis and prognosis.
This study aimed to examine the relationships between four composite inflammatory indices (IBI, SIRI, AISI, RCII) and MASLD based on data from the National Health and Nutrition Examination Survey (NHANES, 1999–2010). NHANES, conducted by the Centers for Disease Control and Prevention (CDC), provides comprehensive health and lifestyle data from a representative sample of the U.S. population. Using this dataset, the study examined how these composite inflammatory indices correlate with MASLD and its progression, with the aim of clarifying the role of systemic inflammation and its potential as a target for future prevention and treatment strategies.
2. Methods
2.1 Study Design and Data Source
This study analyzed data from the NHANES conducted between 1999 and 2010. NHANES is a nationally representative, cross-sectional study administered by the CDC. It collects information on health and nutrition, including demographic, behavioral, laboratory, and clinical data, from the U.S. population. The survey employs a multistage, stratified sampling technique to ensure representativeness.
For this analysis, data on the inflammatory indices—IBI, SIRI, AISI, and RCII—were extracted, along with MASLD status and other relevant variables. Covariates included age, sex, ethnicity, body mass index (BMI), smoking status, and alcohol consumption, which were included to account for potential confounding.
2.2 Study Population
A
Eligible participants were those with complete data on MASLD status, inflammatory indices (IBI, SIRI, AISI, and RCII), and relevant demographic and health data (age, sex, ethnicity, smoking status, and alcohol consumption). Exclusion criteria were missing data on MASLD, inflammatory indices, or key covariates (e.g., age, sex, BMI, and smoking status), as well as the presence of viral hepatitis or alcohol consumption above defined thresholds (men: >30 g/day, women: >20 g/day). After applying these criteria, 5,112 participants were included in the final analysis (Fig. 1).
Fig. 1
Flowchart of participant selection from the NHANES database (1999–2010)
Click here to Correct
2.3 Variables and Definitions
2.3.1 Exposure Variables
IBI
Defined as the product of CRP (mg/dL) and the neutrophil count (NC, 109/L), divided by the lymphocyte count (LC, 109/L)[16].
SIRI
Defined as the product of the NC and monocyte count (MC, 109/L), divided by the LC[17].
AISI
Defined as the product of NC, platelet count (PC, 109/L), and MC, divided by the LC[18].
RCII
Defined as the product of RC and CRP. RC was calculated as TC minus high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C)[19].
Click here to download actual image
Click here to download actual image
Click here to download actual image
Click here to download actual image
Click here to download actual image
A
2.3.2 Outcome Variable: MASLD
MASLD was defined as the presence of steatotic liver disease (SLD) along with at least one cardiovascular metabolic risk factor. SLD was assessed using the fatty liver index (FLI), calculated as:
Click here to download actual image
where TG = triglycerides, GGT = γ-glutamyl transferase, and WC = waist circumference.
Participants with an FLI ≥ 60 were classified as having SLD. Those with viral hepatitis or alcohol intake exceeding 30 g/day (men) or 20 g/day (women) were excluded.
Cardiometabolic risk factors were defined as the presence of at least one of the following: (1) BMI ≥ 25 kg/m² or WC ≥ 94 cm for men and ≥ 80 cm for women; (2) fasting blood glucose ≥ 100 mg/dL, two-hour post-load blood glucose ≥ 140 mg/dL, physician-diagnosed diabetes mellitus, or current treatment for diabetes mellitus; (3) blood pressure ≥ 130/85 mmHg or current antihypertensive treatment; (4) fasting plasma TG ≥ 150 mg/dL or current lipid-lowering therapy; and (5) plasma HDL-C < 40 mg/dL for men or < 50 mg/dL for women or current lipid-lowering therapy[20].
2.3.3 Covariates
The analysis was adjusted for potential confounders, including demographic factors (age, sex, ethnicity, marital status, and the family income-to-poverty ratio [PIR]); lifestyle factors (smoking status and alcohol consumption); health conditions (hypertension, coronary heart disease, diabetes, and a history of myocardial infarction); and biochemical or clinical factors (albumin, HDL-C, GGT, and LDL-C). To minimize collinearity between covariates, exposures, and outcomes, variables such as BMI, segmented NC, MC, LC, WC, PC, and TC were excluded from the models.
2.4 Statistical Analysis
All statistical analyses were performed using R software (version 4.3.2). Baseline characteristics were summarized as means ± standard deviations for normally distributed continuous variables and compared using the Student’s t-test. Non-normally distributed continuous variables were presented as medians (Q1–Q3) and analyzed using the Wilcoxon rank-sum test. Categorical variables were presented as frequencies and percentages and analyzed using the chi-square test (χ² test) or Fisher’s exact test, as appropriate.
Logistic regression models were used to evaluate associations between the inflammatory indices (IBI, SIRI, AISI, and RCII) and MASLD, with adjusted odds ratios (ORs) and 95% confidence intervals (CIs) reported. Model 1 was unadjusted. Model 2 was adjusted for age, sex, race, marital status, PIR, smoking status, and alcohol consumption. Model 3 was further adjusted for hypertension, coronary heart disease, diabetes, heart disease, albumin, GGT, HDL-C, and LDL-C.
Each inflammatory marker was categorized into four quartiles (Q1–Q4). Restricted cubic spline (RCS) analysis was used to identify any nonlinear relationships. Weighted stratified logistic regression was employed for subgroup analyses by sex, race, marital status, alcohol consumption, hypertension, diabetes, and heart disease. Receiver operating characteristic (ROC) curves were plotted to assess the predictive ability of the exposure factors for MASLD. To ensure the reliability and generalizability of the findings, sensitivity analysis was performed[21].
2.4.1 Machine Learning Model Development
To investigate the predictive potential for MASLD, machine learning models were developed using the study dataset. The data were randomly split into a training set (70%) and a validation set (30%). Eight algorithms were employed to construct the MASLD prediction model: AdaBoost (ADA), Decision Tree (DT), Elastic-Net (ENET), K-Nearest Neighbors (KNN), Multilayer Perceptron (MLP), Ridge Regression (RIDGE), Support Vector Machine (SVM), and extreme gradient boosting (XGBoost). Models were trained on the training set with 5-fold cross-validation and evaluated on the validation set[22].
Model performance was assessed using ROC curves and the area under the curve (AUC) to quantify predictive accuracy[23]. Calibration was assessed using calibration plots and the Brier score, and clinical utility was quantified by decision curve analysis (DCA). All tests were two-tailed, with significance defined as p < 0.05.
We used Shapley additive explanations (SHAP) to interpret the predictions of the optimal machine learning model[24]. SHAP values assign weights to input features, visualizing their contribution to the model output. Using the “shapviz” package in R, we computed and visualized these values to enhance the interpretability of the model’s decision-making process.
2.5 Ethical Considerations
This study used publicly available, de-identified data from the NHANES database.
A
As the dataset did not include personal identifiers, the study was exempt from ethical review by an institutional review board.
A
All procedures were in accordance with the principles of the Declaration of Helsinki.
3. Results
3.1 Baseline Characteristics of Study Participants
The study included 5,112 participants from the NHANES dataset, of whom 20.0% (n = 1,020) were diagnosed with MASLD. A baseline comparison between the MASLD and non-MASLD groups revealed significant differences in demographic, metabolic, and inflammatory markers (p < 0.05).
The MASLD group had a significantly higher proportion of male participants compared to the non-MASLD group (61.0% vs. 56.7%, p = 0.015). Racial distribution also varied significantly, with non-Hispanic Whites comprising a greater proportion of the MASLD group (73.3%) compared with the non-MASLD group (p < 0.001). Marital status showed a significant difference, with the MASLD group comprising more married individuals (57.1% vs. 54.3%, p = 0.001). Economic status, as measured by the PIR, was significantly higher in the MASLD group (median: 2.79 vs. 2.10, p < 0.001), suggesting relatively better economic conditions.
Metabolically, the MASLD group exhibited significantly higher BMI (32.54 vs. 26.23, p < 0.001), consistent with obesity as a major risk factor for MASLD. They also showed a significantly greater prevalence of hypertension (46.9% vs. 33.3%, p < 0.001), diabetes (12.6% vs. 9.2%, p = 0.0001), and prior myocardial infarction (7.8% vs. 5.8%, p = 0.0168), highlighting the strong association of MASLD with cardiovascular and metabolic diseases. Inflammatory and biochemical markers, including TC, GGT, LDL-C, segmented NC, MC, LC, PC, IBI, SIRI, AISI, and RCII, were significantly higher in the MASLD group (p < 0.05), whereas HDL-C and albumin levels were significantly lower in the MASLD group, emphasizing the role of systemic inflammation in MASLD pathogenesis.
The baseline characteristics of the study participants are summarized in Table 1.
Table 1
Baseline characteristics of the study participants
Variable
Overall (n = 5112; 100.0%)
Non-MASLD (n = 4092; 80.0%)
MASLD (n = 1020; 20.0%)
[-value
Age, years
52.00 (37.00–66.00)
52.00 (37.00–66.00)
51.00 (38.00–63.00)
0.1781
Sex: Male
2943 (57.6%)
2321 (56.7%)
622 (61.0%)
0.0152
Sex: Female
2169 (42.4%)
1771 (43.3%)
398 (39.0%)
 
Race: Non-Hispanic White
3023 (59.1%)
2275 (55.6%)
748 (73.3%)
< 0.001
Race: Other
2089 (40.9%)
1817 (44.4%)
272 (26.7%)
 
Marriage: Married
2803 (54.8%)
2221 (54.3%)
582 (57.1%)
0.1182
Marriage: Other
2309 (45.2%)
1871 (45.7%)
438 (42.9%)
 
BMI, kg/m²
27.41 (24.07–31.58)
26.23 (23.26–29.74)
32.54 (29.60–36.30)
< 0.001
PIR
2.20 (1.16–4.08)
2.10 (1.13–3.90)
2.79 (1.26–4.77)
< 0.001
Smoking: Not at all
2867 (56.1%)
2272 (55.5%)
595 (58.3%)
0.018
Smoking: Every day
1879 (36.8%)
1507 (36.8%)
372 (36.5%)
 
Smoking: Some days
366 (7.2%)
313 (7.6%)
53 (5.2%)
 
Alcohol intake: No
861 (16.8%)
705 (17.2%)
156 (15.3%)
0.1526
Alcohol intake: Yes
4251 (83.2%)
3387 (82.8%)
864 (84.7%)
 
Hypertension: No
3273 (64.0%)
2731 (66.7%)
542 (53.1%)
< 0.001
Hypertension: Yes
1839 (36.0%)
1361 (33.3%)
478 (46.9%)
 
Diabetes: No
4605 (90.1%)
3714 (90.8%)
891 (87.4%)
0.0014
Diabetes: Yes
507 (9.9%)
378 (9.2%)
129 (12.6%)
 
Coronary heart disease: No
4820 (94.3%)
3870 (94.6%)
950 (93.1%)
0.0902
Coronary heart disease: Yes
292 (5.7%)
222 (5.4%)
70 (6.9%)
 
Heart attack: No
4796 (93.8%)
3856 (94.2%)
940 (92.2%)
0.0168
Heart attack: Yes
316 (6.2%)
236 (5.8%)
80 (7.8%)
 
Segmented neutrophils, ×10⁹/L
4.00 (3.10–5.20)
4.00 (3.00–5.10)
4.40 (3.48–5.50)
< 0.001
Monocytes, ×10⁹/L
0.50 (0.40–0.70)
0.50 (0.40–0.70)
0.60 (0.50–0.70)
< 0.001
Lymphocytes, ×10⁹/L
1.90 (1.60–2.40)
1.90 (1.60–2.40)
2.00 (1.60–2.42)
< 0.001
Waist circumference, cm
98.00
(88.10–108.30)
94.70 (85.60–104.00)
111.00 (104.18–119.43)
< 0.001
Platelet count, ×10⁹/L
251.00
(211.00–296.00)
250.00 (210.00–295.00)
253.00 (214.00–305.00)
0.0369
Total cholesterol, md/dL
196.00
(169.00–224.00)
195.00 (168.00–223.00)
199.50 (172.00–228.00)
0.0034
Albumin, g/dL
42.00 (40.00–45.00)
43.00 (40.00–45.00)
41.00 (39.00–44.00)
< 0.001
GGT, U/L
22.00 (15.00–34.00)
20.00 (15.00–31.00)
28.00 (20.00–42.25)
< 0.001
HDL-C, mg/dL
50.00 (41.00–61.00)
51.00 (42.00–63.00)
45.00 (38.00–54.00)
< 0.001
LDL-C, mg/dL
116.00 (93.00–141.00)
115.00 (93.00–140.00)
120.00 (95.00–143.25)
0.0228
IBI
0.46 (0.18–1.20)
0.40 (0.15–1.05)
0.79 (0.33–1.76)
< 0.001
SIRI
1.09 (0.75–1.62)
1.06 (0.73–1.58)
1.21 (0.84–1.77)
< 0.001
AISI
272.18 (180.00–422.51)
263.31 (174.37–409.73)
302.79 (208.09–466.54)
< 0.001
RCII
5.75 (2.09–14.59)
4.68 (1.75–12.24)
11.66 (5.03–25.06)
< 0.001
Note: Values are presented as n (%) for categorical variables and median (interquartile range) for continuous variables. Abbreviations: GGT, gamma-glutamyl transferase; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; IBI, Inflammatory Burden Index; SIRI, Systemic Inflammation Response Index; AISI, Aggregate Index of Systemic Inflammation; RCII, Remnant Cholesterol Inflammatory Index; TC, total cholesterol.
3.2 Association Between Exposure Factors and MASLD
A
Multicollinearity was evaluated using variance inflation factor (VIF) values, with a prespecified threshold of 5[25]. Only variables meeting this criterion were included in the multivariable models; all VIF values are provided in Additional file 1: Table S1. A multivariable weighted logistic regression analysis was conducted to explore associations between inflammatory markers (IBI, SIRI, AISI, RCII) and MASLD. These markers were grouped into quartiles (Q1–Q4), with Q1 as the reference group. The results demonstrated a dose–response relationship between higher levels of these inflammatory markers and an increased risk of MASLD. For IBI, the OR for Q4 was 3.75 (95% CI: 2.77–5.06, p < 0.001) after adjusting for clinical metabolic factors. Similarly, for SIRI, AISI, and RCII, the ORs for Q4 were 1.41 (95% CI: 1.09–1.82, p = 0.0010), 1.60 (95% CI: 1.26–2.04, p < 0.001), and 5.76 (95% CI: 4.08–8.14, p < 0.001), respectively. These findings indicate that higher levels of inflammatory markers are significantly associated with MASLD and that these indices may serve as potential biomarkers for risk stratification (Table 2).
To further elucidate these associations, generalized additive models (GAMs) were applied to characterize the nonlinear relationships between inflammatory indices and the predicted probability of MASLD. As illustrated in Fig. 2, elevated levels of RCII, IBI, AISI, and SIRI were each positively associated with an increased probability of MASLD in a dose–response fashion, with RCII and IBI exhibiting notably steeper risk gradients. The solid blue curves depict the estimated smooth functions, while the shaded bands denote the corresponding 95% confidence intervals.
Table 2
Multivariate logistic regression analysis of inflammatory indices and MASLD
Indicator
Model 1
Model 2
Model 3
IBI
OR (95% CI)
p-value
OR (95% CI)
p-value
OR (95% CI)
p-value
Continuous
      
 
1.03 (1.00, 1.07)
0.033
1.04 (1.00, 1.07)
0.028
1.01 (0.99, 1.04)
0.233
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
2.34 (1.76, 3.11)
< 0.001
2.52 (1.90, 3.36)
< 0.001
2.00 (1.49, 2.68)
< 0.001
Q3
3.85 (2.89, 5.13)
< 0.001
4.43 (3.31, 5.93)
< 0.001
3.10 (2.28, 4.21)
< 0.001
Q4
4.99 (3.73, 6.67)
< 0.001
6.11 (4.58, 8.16)
< 0.001
3.75 (2.77, 5.06)
< 0.001
SIRI
      
Continuous
      
 
1.17 (1.09, 1.25)
< 0.001
1.14 (1.06, 1.22)
< 0.001
1.10 (1.02, 1.19)
0.012
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
1.26 (0.96, 1.65)
0.106
1.15 (0.87, 1.52)
0.316
1.07 (0.80, 1.44)
0.629
Q3
1.68 (1.35, 2.10)
< 0.001
1.52 (1.22, 1.90)
< 0.001
1.39 (1.08, 1.78)
0.011
Q4
1.78 (1.42, 2.22)
< 0.001
1.60 (1.27, 2.01)
< 0.001
1.41 (1.09, 1.82)
0.010
AISI
      
Continuous
      
 
1.00 (1.00, 1.00)
< 0.001
1.00 (1.00, 1.00)
0.001
1.00 (1.00, 1.00)
0.028
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
1.50 (1.18, 1.89)
0.001
1.42 (1.11, 1.81)
0.006
1.35 (1.05, 1.74)
0.023
Q3
1.72 (1.37, 2.18)
< 0.001
1.67 (1.31, 2.13)
< 0.001
1.52 (1.16, 2.00)
0.004
Q4
1.87 (1.51, 2.31)
< 0.001
1.82 (1.46, 2.28)
< 0.001
1.60 (1.26, 2.04)
< 0.001
RCII
      
Continuous
      
 
1.02 (1.01, 1.02)
< 0.001
1.02 (1.02, 1.03)
< 0.001
1.01 (1.01, 1.02)
< 0.001
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
3.09 (2.32, 4.10)
< 0.001
3.24 (2.42, 4.34)
< 0.001
2.49 (1.88, 3.32)
< 0.001
Q3
5.96 (4.40, 8.06)
< 0.001
6.95 (5.05, 9.56)
< 0.001
4.53 (3.20, 6.42)
< 0.001
Q4
8.55 (6.44, 11.36)
< 0.001
10.82 (7.94, 14.74)
< 0.001
5.76 (4.08, 8.14)
< 0.001
Note: Model 1: unadjusted model. Model 2: Model 1 with adjustments for age, sex, race, marital status, PIR, and smoking status. Model 3: Model 2 with additional adjustments for hypertension, coronary heart disease, diabetes, heart disease, albumin, alcohol consumption, GGT, HDL-C, and LDL-C. All retained covariates had a variance inflation factor < 5.
Fig. 2
Associations between exposure factors and MASLD
Click here to Correct
A. RCII; B. IBI; C. AISI; D. SIRI.
Note
The blue solid line represents the smoothed curve fit between variables, and the shaded sky-blue area indicates the 95% confidence interval of the fitted results.
3.3 RCS Analysis
The RCS analysis revealed a nonlinear relationship between inflammatory indices and MASLD risk. As IBI, SIRI, AISI, and RCII levels increased, the MASLD risk fluctuated in a nonlinear pattern. This was supported by significant p-overall (p < 0.001) and p-nonlinear values (p < 0.001), confirming a nonlinear effect.
Sex-stratified analysis revealed that IBI and RCII maintained significant associations in both male and female participants (all p-overall < 0.001), whereas AISI was significant only in male participants (p = 0.026). No significant association was observed for the SIRI in either sex (Fig. 3; see also Additional file 1: Figure S1).
Fig. 3
RCS curves for the relationships between RCII, IBI, and MASLD
Click here to Correct
A. RCS curve for RCII; B. sex-stratified RCS curves for RCII; C. RCS curve for IBI; D. sex-stratified RCS curves for IBI.
Note
The shaded area represents the 95% confidence interval.
3.4 Stratified Analysis
Stratified analysis was performed to assess the influence of age, sex, and metabolic disease status on the relationship between inflammatory markers and MASLD. The RCII–MASLD association varied across strata, and the strongest effect was observed in female participants (RCII Q4 vs Q1: OR 19.978, 95% CI 10.677–37.382, p < 0.001). Estimates also differed by age group and among participants with hypertension, diabetes, or heart disease, suggesting heterogeneity of effects across these subpopulations. The association of IBI with MASLD was primarily observed in participants below 65 years (Q4 OR = 6.417, 95% CI: 4.617–8.989, p < 0.001). Similarly, AISI had a strong association with MASLD in participants below 65 years (AISI Q4 OR = 2.032, 95% CI: 1.610–2.564, p < 0.001). The relationship with the SIRI was notably influenced by heart disease status, with a higher OR in participants without a history of heart disease (SIRI Q4 OR = 1.798, 95% CI: 1.422–2.272, p < 0.001) (Figs. 45; see also Additional file 1: Figures S2–3).
Fig. 4
Subgroup analyses of RCII and MASLD
Click here to Correct
Note
Forest plots show odds ratios (ORs) and 95% confidence intervals (CIs) for the association between RCII quartiles and MASLD across subgroups defined by age, sex, race, smoking status, hypertension, diabetes, and history of myocardial infarction. Interaction p-values are presented for each subgroup comparison.
Fig. 5
Subgroup analyses of IBI and MASLD
Click here to Correct
Note
Forest plots show odds ratios (ORs) and 95% confidence intervals (CIs) for the association between IBI quartiles and MASLD across subgroups defined by age, sex, race, smoking status, hypertension, diabetes, and history of myocardial infarction. Interaction p-values are presented for each subgroup comparison.
3.5 ROC Curve and Predictive Value
To evaluate the predictive ability of the inflammatory markers, ROC curve analyses were performed. The RCII and IBI showed good discrimination (AUCs > 0.70) (Fig. 6). Results for SIRI and AISI (Additional file 1: Figure S4) were broadly similar, also meeting the threshold for good diagnostic value (AUC 0.7–0.8). These findings support the potential of these inflammatory markers as screening tools for MASLD. RCII, in particular, has high diagnostic accuracy, suggesting its utility in clinical practice for identifying individuals at risk for MASLD.
Fig. 6
ROC analysis of the RCII and IBI for MASLD prediction
Click here to Correct
Note
Receiver operating characteristic (ROC) curves for (A) RCII_qt (AUC = 0.759) and (B) IBI_qt (AUC = 0.748) in the survey-weighted validation cohort. The diagonal grey line indicates the reference line for a non-informative classifier (AUC = 0.50). ROC curves depict the sensitivity–specificity trade-off across thresholds.
3.6 Sensitivity Analysis
To verify the robustness and generalizability of the findings, unweighted logistic regression was performed as a sensitivity analysis. The results were consistent with those of the primary analyses across all inflammatory markers (Table 3). IBI, SIRI, AISI, and RCII maintained significant effects at higher quartiles (Q3–Q4, all p < 0.05), with particularly strong effect sizes for IBI and RCII (e.g., RCII Q4 OR = 5.16, 95% CI: 3.93–6.79, p < 0.001 in Model 3). While effect magnitudes varied across models, the overall consistent trends confirmed the robustness of the results.
Table 3
Sensitivity analysis results
Indicator
Model 1
Model 2
Model 3
IBI
OR (95% CI)
p-value
OR (95% CI)
p-value
OR (95% CI)
p-value
Continuous
      
 
1.02 (1.01, 1.03)
0.004
1.02 (1.01, 1.03)
0.002
1.00 (0.99, 1.02)
0.69
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
2.00 (1.58, 2.54)
< 0.001
2.26 (1.77, 2.88)
< 0.001
1.87 (1.46, 2.40)
< 0.001
Q3
2.92 (1.58, 2.54)
< 0.001
3.45 (2.72, 4.37)
< 0.001
2.54 (1.99, 3.24)
< 0.001
Q4
4.13 (3.30, 5.16)
< 0.001
5.11 (4.05, 6.46)
< 0.001
3.28 (2.55, 4.20)
< 0.001
SIRI
      
Continuous
      
 
1.14 (1.07, 1.22)
< 0.001
1.10 (1.03, 1.18)
0.006
1.06 (0.98, 1.14)
0.156
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
1.26 (1.02, 1.55)
0.029
1.12 (0.90, 1.38)
0.303
1.07 (0.86, 1.34)
0.527
Q3
1.61 (1.32, 1.97)
< 0.001
1.41 (1.14, 1.73)
0.001
1.30 (1.05, 1.62)
0.016
Q4
1.78 (1.46, 2.18)
< 0.001
1.51 (1.23, 1.86)
< 0.001
1.33 (1.07, 1.65)
0.011
AISI
      
Continuous
      
 
1.00 (1.00, 1.00)
0.002
1.00 (1.00, 1.00)
0.012
1.00 (1.00, 1.00)
0.134
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
1.45 (1.17, 1.78)
0.001
1.31 (1.06, 1.62)
0.013
1.27 (1.02, 1.58)
0.036
Q3
1.70 (1.39, 2.09)
< 0.001
1.55 (1.25, 1.91)
< 0.001
1.44 (1.16, 1.79)
< 0.001
Q4
1.88 (1.53, 2.30)
< 0.001
1.66 (1.35, 2.05)
< 0.001
1.47 (1.18, 1.78)
0.001
RCII
      
Continuous
      
 
1.01 (1.01, 1.01)
< 0.001
1.01 (1.01, 1.02)
< 0.001
1.01 (1.00, 1.01)
< 0.001
Categorical
      
Q1
Ref
 
Ref
 
Ref
 
Q2
2.56 (1.97, 3.33)
< 0.001
2.91 (2.23, 3.80)
< 0.001
2.37 (1.81, 3.11)
< 0.001
Q3
4.06 (3.16, 5.22)
< 0.001
5.19 (4.01, 6.72)
< 0.001
3.64 (2.78, 4.77)
< 0.001
Q4
6.55 (5.13, 8.36)
< 0.001
8.85 (6.86, 11.42)
< 0.001
5.16 (3.93, 6.79)
< 0.001
3.7 Machine Learning Model Development
Our study applied the Boruta algorithm with shadow features to identify 16 potentially effective predictive variables (Additional file 1: Figure S5, green module). The shadow feature variables included race, GGT, RCII qt, age, HDL-C, IBI qt, hypertension, PIR, albumin, diabetes, SIRI qt, AISI qt, smoking, heart attack, coronary heart disease, and sex.
This study systematically compared the predictive performance of eight machine learning models using a 70/30 training–validation split. Figure 7 depicts the ROC curves for each model. The AUC values in the test set were as follows: ADA 0.799, DT 0.717, ENET 0.758, KNN 0.713, MLP 0.777, RIDGE 0.758, SVM 0.766, and XGB 0.802. Among these, the XGB algorithm demonstrated the best predictive performance (AUC 0.802), indicating excellent prediction accuracy.
A
Calibration curves indicated that XGB achieved the closest alignment with the 45° reference line, with the lowest Brier score (0.127), followed by MLP (0.132) and ENET/RIDGE (0.136). Meanwhile, DT and KNN showed noticeable miscalibration at higher predicted risks. In addition, DCA revealed that XGB consistently provided the greatest net clinical benefit across a wide range of threshold probabilities, outperforming the “treat-all” and “treat-none” strategies. ADA and MLP offered moderate benefit, and DT/KNN showed limited utility (Additional file 1: Figure S6). Based on these findings, XGB was identified as the optimal model owing to its high specificity, low false-positive rate, and stable overall performance.
Fig. 7
ROC curves of the eight machine learning models in the training and test sets
Click here to Correct
3.8 Interpretability Analysis
We used SHAP values to examine individual predictions of MASLD risk from the optimal XGB model. Key predictive factors included race, GGT, albumin, HDL-C, age, and RCII qt. The SHAP summary plot and SHAP variable importance ranking (Fig. 8) highlighted the significance of these variables, with the ranking indicating their influence. These results emphasized the critical role of inflammatory markers and demographic characteristics in MASLD risk assessment.
Fig. 8
SHAP interpretation of the XGB model for MASLD prediction.
Click here to Correct
A. SHAP summary plot of the XGB model; B. SHAP variable importance ranking
We analyzed the model’s ability to predict MASLD risk using SHAP. In the waterfall plot (Fig. 9A), for a given sample, the model’s baseline (expected) prediction was 0.0418; after sequentially adding the SHAP contributions of the relevant predictors, the final predicted probability increased to 0.205. The feature “Race = Others” significantly decreased the prediction value (–0.0852), while the feature “GGT = 44” increased the prediction value (+ 0.0447). Specifically, RCII contributed positively to the prediction value (+ 0.0278). The SHAP force plot (Fig. 9B) further illustrated the overall impact of these features on the prediction. The analysis indicated that race, GGT, albumin, and the RCII are key factors influencing the prediction.
Fig. 9
Interpretation of individual predictions
Click here to Correct
A. SHAP waterfall plot of individual prediction results. B. Force plot showing SHAP values for each feature.
4. Discussion
This study examined the relationships between four composite inflammatory indices (IBI, SIRI, AISI, and RCII) and MASLD using 1999–2010 NHANES data. The findings revealed the significant role of systemic inflammation in the pathogenesis of MASLD and the robust associations between these inflammatory markers and MASLD risk. These results provide a deeper understanding of the molecular mechanisms underlying MASLD and its implications for public health and clinical applications.
MASLD is closely tied to systemic metabolic dysfunction, including obesity, insulin resistance, and dyslipidemia. At the molecular level, these disruptions trigger inflammatory responses that contribute to liver damage and fibrosis progression[26, 27]. Recent research supports the role of chronic low-grade inflammation in the development and progression of MASLD. Consistent with previous reports, our results indicate that elevated levels of inflammatory indices—IBI, SIRI, AISI, and RCII—are linked to increased MASLD risk[13, 28].
The inflammatory response in MASLD is driven by the activation of several molecular pathways, including the nuclear factor kappa B (NF-κB) pathway; inflammasome signaling; and the upregulation of proinflammatory cytokines such as tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), and CRP[29, 30]. These mediators can exacerbate insulin resistance, adipocyte dysfunction, and hepatic lipogenesis, thereby accelerating the progression from simple steatosis to NASH and, eventually, cirrhosis[31]. Our findings show that markers such as the IBI, which combines the CRP level and the neutrophil-to-lymphocyte ratio, reflect the extent of this inflammatory burden. This aligns with the findings of molecular evidence linking these markers to different stages of MASLD progression[3].
In addition, our analysis revealed sex-specific differences. Both stratified analyses and RCS curves indicated that the association between RCII and MASLD was stronger in female participants than in male participants. This observation is consistent with previous reports that sex hormones, body fat distribution, and lipid metabolism differ by sex, thereby modifying the influence of dyslipidemia and systemic inflammation on hepatic steatosis[32]. Women may experience distinct metabolic and hormonal effects, including the protective role of estrogen on lipid regulation and inflammatory pathways before menopause, whereas men are more likely to exhibit visceral adiposity and higher remnant cholesterol levels[33]. These differences may help explain the stronger RCII–MASLD association observed in women in our study.
Moreover, indices that combine lipoprotein metabolism markers, such as the AISI and RCII, highlight the importance of lipid dysregulation in MASLD. Elevated LDL-C and TG, in addition to decreased HDL-C, are characteristic of metabolic syndrome and are closely associated with increased hepatic fat accumulation[34]. The role of RC, captured by the RCII, further highlights the intricate relationship between lipid metabolism and inflammation, both of which contribute to liver injury and an increased risk of cardiovascular comorbidities in MASLD[35].
MASLD is now recognized as the most common liver disease worldwide, affecting an estimated 25–30% of the global population, with prevalence rising steadily, particularly among individuals with obesity and type 2 diabetes[36, 37]. This growing burden has considerable public health implications, given its association with cardiovascular diseases, diabetes, and chronic kidney disease[38]. Our findings reinforce the importance of identifying early inflammatory markers to predict the onset and progression of MASLD, thereby supporting the development of targeted preventive and therapeutic strategies.
Given the systemic nature of MASLD, public health interventions should address broader metabolic dysfunctions that contribute to disease development. Lifestyle modifications, including weight reduction, dietary changes, and regular physical activity, are the cornerstone of MASLD prevention and management[39, 40]. Moreover, our findings suggest that the IBI, SIRI, AISI, and RCII may be valuable for early detection and risk stratification, enabling healthcare providers to identify high-risk individuals before the onset of advanced hepatic injury. Integrating these markers into routine clinical practice could substantially enhance population-level management of MASLD and reduce the burden of liver-related morbidity and mortality[41].
In clinical settings, these composite inflammatory indices may serve as non-invasive biomarkers for monitoring disease progression in individuals with MASLD. Although liver biopsy remains the gold standard for diagnosing NASH and staging fibrosis, it is invasive, costly, and prone to sampling errors[42]. Non-invasive tests such as liver elastography and serum biomarkers are increasingly used, but further validation of inflammatory indices as diagnostic tools is necessary. Our study suggests that IBI, SIRI, AISI, and RCII could complement or, in some contexts, substitute for existing tests, thereby offering a more comprehensive assessment of the systemic inflammation underlying MASLD[43].
The predictive models developed in this study using machine learning algorithms, particularly XGBoost, further demonstrated the potential of these inflammatory markers in assessing MASLD risk. Integrating clinical, demographic, and inflammatory data into robust predictive models could improve diagnostic accuracy and inform personalized therapeutic interventions. Nevertheless, prospective studies in larger and more diverse populations are needed to validate these findings and optimize the models for clinical use[23].
Taken together, the RCII emerged as the most informative of the four inflammatory indices (IBI, SIRI, AISI, and RCII). In conventional analyses, RCII demonstrated the strongest and most consistent association with MASLD, with effect estimates remaining stable across sensitivity analyses. Likewise, machine learning comparisons consistently ranked RCII among the top predictors, and SHAP analyses highlighted its substantial global contribution alongside its localized effects on individual risk predictions. Biologically, this is plausible because the RCII integrates RC (a TG-rich, ApoB-containing fraction linked to hepatic lipid accumulation) with CRP (a systemic inflammatory marker). This combination allows RCII to capture the metabolic–inflammatory burden more directly than leukocyte-based composites[44]. Clinically, these findings suggest that the RCII may serve as a practical, integrative biomarker to refine MASLD risk stratification beyond traditional lipid parameters and cell-count indices.
While our study provides valuable insights, several limitations should be addressed in future research. First, the cross-sectional nature of the NHANES data precluded the establishment of causal relationships between inflammation and MASLD. Longitudinal studies are needed to determine whether elevated inflammatory indices precede the onset of MASLD or occur as a consequence of the disease[45]. Second, while our analysis adjusted for several potential confounders, unmeasured factors may still have influenced the relationship between inflammation and MASLD. For instance, genetic predisposition, gut microbiota composition, and environmental exposures may all contribute to the disease’s development and progression[46]. Future studies should incorporate these variables to provide a more comprehensive understanding of the molecular mechanisms underlying MASLD.
Furthermore, while we focused on four inflammatory indices, other biomarkers of immune dysregulation and metabolic dysfunction, such as adipokines and oxidative stress markers, were not included. Investigating the interplay between these additional markers and the indices studied here could reveal novel insights into MASLD pathogenesis and help enhance risk prediction models[47]. Finally, the external validity of our findings would be improved by including more diverse populations, particularly from regions outside the U.S., where dietary habits, lifestyle factors, and genetic predispositions may differ.
5. Conclusion
This study highlighted the significant role of systemic inflammation in the development and progression of MASLD and demonstrated the potential of inflammatory indices as biomarkers for early diagnosis and risk stratification. The mechanisms underlying these associations suggested that therapeutic strategies targeting inflammation and lipid dysregulation may hold promise in MASLD management. At the population level, public health efforts should prioritize lifestyle interventions to mitigate the rising prevalence of MASLD, while in clinical settings, incorporation of these indices could enhance patient stratification and outcomes. Despite some limitations, our findings provide a foundation for future research into non-invasive diagnostic tools and personalized treatment strategies for MASLD, ultimately contributing to the broader goal of reducing the global burden of metabolic liver disease.
List of abbreviations
AISI
Aggregate index of systemic inflammation
AUC
Area under the curve
CRP
C-reactive protein
FLI
Fatty liver index
GGT
Gamma-glutamyl transferase
HDL-C
High-density lipoprotein cholesterol
IBI
Inflammatory burden index
LDL-C
Low-density lipoprotein cholesterol
MASLD
Metabolic dysfunction-associated fatty liver disease
MC
Monocyte count
NC
Neutrophil count
NHANES
National Health and Nutrition Examination Survey
OR
Odds ratio
PC
Platelet count
PIR
Poverty income ratio
RC
Remnant cholesterol
RCII
Remnant cholesterol inflammatory index
RCS
Restricted cubic spline
ROC
Receiver operating characteristic
SD
Standard deviation
SHAP
Shapley additive explanation
SLD
Steatotic liver disease
SIRI
Systemic inflammation response index
TG
Triglyceride
TC
Total cholesterol
WC
Waist circumference
Supplementary Material
File format: .docx
Title of data: Variance inflation factor between variables, RCS curves for the relationship between AISI, SIRI, and MASLD, Subgroup analysis chart for AISI and SIRI, ROC analysis of AISI, SIRI, and MASLD, Feature selection based on the Boruta algorithm, calibration and decision curve plots
Description of data: Contains the complete variance inflation factor results (Table S1), restricted cubic spline curve (Figure S1), subgroup analysis figures (Figures S2–S3), ROC curves (Figure S4), Boruta feature selection figure (Figure S5), and calibration and decision curve analyses (Figure S6).
Declarations
Ethics approval and consent to participate
A
A
This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research participants were approved by the Ethics Review Board of the National Center for Health Statistics (NCHS).
A
Written informed consent was obtained from all participants.
Consent for publication
A
All authors consent to publication of this work.
A
Data Availability
The data that support the findings of this study are available in National Health and Nutrition Examination Survey (NHANES) at https://www.cdc.gov/nchs/nhanes/index.html. These data were derived from the following resources available in the public domain:- NHANES survey cycles 1999–2010 (used in this analysis), https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=1999
Competing interests
The authors declare that they have no competing interests
A
Funding
None.
A
Author Contribution
XAL, CD, and ZW completed manuscript drafting, data analysis, interpretation, and visualization. HXW and CRZ contributed to methodology, conceptualization, and review of the manuscript. RL supervised the study, critically revised the manuscript, and is responsible for the overall integrity of the work. All authors read and approved the final manuscript and agree to be accountable for its accuracy and integrity.
Acknowledgements
We acknowledge the NHANES database for providing access to their valuable datasets and thank all contributors involved in data collection and management.
Electronic Supplementary Material
Below is the link to the electronic supplementary material
Reference
1.
Pouwels S, Sakran N, Graham Y, et al. Non-alcoholic fatty liver disease (NAFLD): a review of pathophysiology, clinical management and effects of weight loss[J]. BMC Endocrine Disorders, 2022, 22(1): 63.
2.
Younossi Z M. Non-alcoholic fatty liver disease – A global public health perspective[J]. Journal of Hepatology, 2019, 70(3): 531–544.
3.
Cobbina E, Akhlaghi F. Non-alcoholic fatty liver disease (NAFLD) – pathogenesis, classification, and effect on drug metabolizing enzymes and transporters[J]. Drug Metabolism Reviews, 2017, 49(2): 197–211.
4.
Eslam M, Sanyal A J, George J, et al. MAFLD: A Consensus-Driven Proposed Nomenclature for Metabolic Associated Fatty Liver Disease[J]. Gastroenterology, 2020, 158(7): 1999–2014.e1.
5.
Duell P B, Welty F K, Miller M, et al. Nonalcoholic Fatty Liver Disease and Cardiovascular Risk: A Scientific Statement From the American Heart Association[J]. Arteriosclerosis, Thrombosis, and Vascular Biology, 2022, 42(6): e168-e185.
6.
Guo T, Yan W, Cui X, et al. Liraglutide attenuates type 2 diabetes mellitus-associated non-alcoholic fatty liver disease by activating AMPK/ACC signaling and inhibiting ferroptosis[J]. Molecular Medicine (Cambridge, Mass.), 2023, 29(1): 132.
7.
Lee C H, Lui D T, Lam K S. Non-alcoholic fatty liver disease and type 2 diabetes: An update[J]. Journal of Diabetes Investigation, 2022, 13(6): 930–940.
8.
Xiong Z, Xu W, Wang Y, et al. Inflammatory burden index: associations between osteoarthritis and all-cause mortality among individuals with osteoarthritis[J]. BMC public health, 2024, 24(1): 2203.
9.
Ma R, Cui L, Cai J, et al. Association between systemic immune inflammation index, systemic inflammation response index and adult psoriasis: evidence from NHANES[J]. Frontiers in Immunology, 2024, 15: 1323174.
10.
A combined analysis of TyG index, SII index, and SIRI index: positive association with CHD risk and coronary atherosclerosis severity in patients with NAFLD - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38260163/.
11.
Yu Y, Zhang Y, Zhu C, et al. Remnant cholesterol inflammatory index, calculated from residual cholesterol to C-reactive protein ratio, and stroke outcomes: a retrospective study using the National institutes of health stroke scale and modified Rankin scale[J]. Lipids in Health and Disease, 2025, 24(1): 228.
12.
Bay B, Tanner R, Gao M, et al. Residual cholesterol and inflammatory risk in statin-treated patients undergoing percutaneous coronary intervention†[J]. European Heart Journal, 2025, 46(32): 3167–3177.
13.
Wang Y, Chen S, Tian C, et al. Association of systemic immune biomarkers with metabolic dysfunction-associated steatotic liver disease: a cross-sectional study of NHANES 2007–2018[J]. Frontiers in Nutrition, 2024, 11: 1415484.
14.
Kawanaka M, Nishino K, Kawada M, et al. Predictors of Atherosclerotic Cardiovascular Disease Events in Patients with Metabolic Dysfunction-associated Steatotic Liver Disease[J]. Internal Medicine (Tokyo, Japan), 2025, 64(12): 1791–1798.
15.
Carli F, Della Pepa G, Sabatini S, et al. Lipid metabolism in MASLD and MASH: From mechanism to the clinic[J]. JHEP reports: innovation in hepatology, 2024, 6(12): 101185.
16.
The inflammatory burden index is a superior systemic inflammation biomarker for the prognosis of non-small cell lung cancer - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/36852672/.
17.
Zhao S, Dong S, Qin Y, et al. Inflammation index SIRI is associated with increased all-cause and cardiovascular mortality among patients with hypertension[J]. Frontiers in Cardiovascular Medicine, 2023, 9: 1066219.
18.
Zinellu A, Collu C, Nasser M, et al. The Aggregate Index of Systemic Inflammation (AISI): A Novel Prognostic Biomarker in Idiopathic Pulmonary Fibrosis[J]. Journal of Clinical Medicine, 2021, 10(18): 4134.
19.
Wang Y, Bi L, Li Q, et al. Remnant cholesterol inflammatory index and its association with all-cause and cause-specific mortality in middle-aged and elderly populations: evidence from US and Chinese national population surveys[J]. Lipids in Health and Disease, 2025, 24(1): 155.
20.
Tian T, Zhang J, Xie W, et al. Dietary Quality and Relationships with Metabolic Dysfunction-Associated Fatty Liver Disease (MAFLD) among United States Adults, Results from NHANES 2017–2018[J]. Nutrients, 2022, 14(21): 4505.
21.
Association between dietary inflammatory index and Stroke in the US population: evidence from NHANES 1999–2018 - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38166986/.
22.
Jin W, Xu L, Yue C, et al. Development and validation of explainable machine learning models for female hip osteoporosis using electronic health records[J]. International Journal of Medical Informatics, 2025, 199: 105889.
23.
Machine learning models for predicting metabolic dysfunction-associated steatotic liver disease prevalence using basic demographic and clinical characteristics - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/40155991/.
24.
Fu Q, Wu Y, Zhu M, et al. Identifying cardiovascular disease risk in the U.S. population using environmental volatile organic compounds exposure: A machine learning predictive model based on the SHAP methodology[J]. Ecotoxicology and Environmental Safety, 2024, 286: 117210.
25.
Kim J H. Multicollinearity and misleading statistical results[J]. Korean Journal of Anesthesiology, 2019, 72(6): 558–569.
26.
Natural history and progression of metabolic dysfunction-associated steatotic liver disease - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/39243773/.
27.
A new definition for metabolic dysfunction-associated fatty liver disease: An international expert consensus statement - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/32278004/.
28.
The Potential Role of C-Reactive Protein in Metabolic-Dysfunction-Associated Fatty Liver Disease and Aging - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/37893085/.
29.
Friedman S L, Neuschwander-Tetri B A, Rinella M, et al. Mechanisms of NAFLD development and therapeutic strategies[J]. Nature medicine, 2018, 24(7): 908–922.
30.
Wan X, Xu C, Yu C, et al. Role of NLRP3 Inflammasome in the Progression of NAFLD to NASH[J]. Canadian Journal of Gastroenterology & Hepatology, 2016, 2016: 6489012.
31.
A TNFα/Miz1-positive feedback loop inhibits mitophagy in hepatocytes and propagates non-alcoholic steatohepatitis - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/37040844/.
32.
Sex Differences in Nonalcoholic Fatty Liver Disease: Estrogen Influence on the Liver–Adipose Tissue Crosstalk | Antioxidants & Redox Signaling[EB/OL]. [2025-08-29]. https://www.liebertpub.com/doi/abs/10.1089/ars.2021.0044?doi=10.1089%2Fars.2021.0044&utm_source=chatgpt.com&journalCode=ars.
33.
Palmisano B T, Zhu L, Stafford J M. Estrogens in the Regulation of Liver Lipid Metabolism[J]. Advances in experimental medicine and biology, 2017, 1043: 227.
34.
Ko J, Sequeira I R, Skudder-Hill L, et al. Metabolic traits affecting the relationship between liver fat and intrapancreatic fat: a mediation analysis[J]. Diabetologia, 2023, 66(1): 190–200.
35.
MASLD: a systemic metabolic disorder with cardiovascular and malignant complications - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38228377/.
36.
Obesity and MASLD: Is weight loss the (only) key to treat metabolic liver disease? - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38782182/.
37.
The Global Epidemiology of Nonalcoholic Fatty Liver Disease and Nonalcoholic Steatohepatitis Among Patients With Type 2 Diabetes - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38521116/.
38.
Steatotic liver disease, MASLD and risk of chronic kidney disease - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38141808/.
39.
EASL-EASD-EASO Clinical Practice Guidelines on the Management of Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38852583/.
40.
Dietary and pharmacological treatment in patients with metabolic-dysfunction associated steatotic liver disease - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38262842/.
41.
Breaking the barriers: the role of gut homeostasis in Metabolic-Associated Steatotic Liver Disease (MASLD) - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38512763/.
42.
Abdelhameed F, Kite C, Lagojda L, et al. Non-invasive Scores and Serum Biomarkers for Fatty Liver in the Era of Metabolic Dysfunction-associated Steatotic Liver Disease (MASLD): A Comprehensive Review From NAFLD to MAFLD and MASLD[J]. Current Obesity Reports, 2024, 13(3): 510–531.
43.
Ilagan-Ying Y C, Banini B A, Do A, et al. Screening, Diagnosis, and Staging of Non-Alcoholic Fatty Liver Disease (NAFLD): Application of Society Guidelines to Clinical Practice[J]. Current Gastroenterology Reports, 2023, 25(10): 213–224.
44.
Cholesterol Remnants, Triglyceride-Rich Lipoproteins and Cardiovascular Risk - PMC[EB/OL]. [2025-08-29]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10002331/?utm_source=chatgpt.com.
45.
Zhang Y, Qiu X, Wu Z, et al. Comprehensive analysis of the association between volatile organic compound pollutants and chronic kidney disease in hypertensive populations: insights from multi-omics approaches and identification of potential therapeutic targets[J]. Environmental Research, 2025, 282: 121966.
46.
Clinical and genetic risk factors for progressive fibrosis in metabolic dysfunction-associated steatotic liver disease - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38967582/.
47.
Molecular mechanisms in MASLD/MASH-related HCC - PubMed[EB/OL]. [2025-08-29]. https://pubmed.ncbi.nlm.nih.gov/38349726/.
Total words in MS: 5856
Total words in Title: 12
Total words in Abstract: 263
Total Keyword count: 4
Total Images in MS: 9
Total Tables in MS: 3
Total Reference count: 47