Association of a novel Fat-Muscle-BMI Index (FMBI) with NAFLD, liver fibrosis, and atherosclerotic cardiovascular disease risk: a cross-sectional study in 1,592 Chinese adults Authors: Xiaokang Wu, Xinjiang Hou, Jiafeng Yin, Yue Li, Chaoliang Xiong, Hailong Liu, Hao Meng Published: 2026 ABSTRACT AbstractBackgroundNon-alcoholic fatty liver disease (NAFLD) is a global burden with unmet non-invasive biomarker needs; BMI and FMRs have limitations, and their combined index FMBI is unstudied. We aim to explore FMBI’s association with NAFLD, associated liver fibrosis and atherosclerotic cardiovascular disease risk (ASCVD) risk.MethodsThis retrospective study enrolled 1592 eligible participants aged 40–79 years from a Chinese hospital. The Fat-Muscle-BMI Index (FMBI) was calculated as Fat-to-muscle ratio (FMR) × Body mass index (BMI). Multivariable logistic/linear regression with three hierarchical models analyzed FMBI’s associations with NAFLD, hepatic fibrosis and ASCVD; restricted cubic splines (RCS, subgroup and sensitivity/E-value analyses validated the findings.ResultsHigher FMBI quartiles exhibited a progressive increase in NAFLD prevalence (46.0% to 72.4%, P < 0.001). FMBI demonstrated a significant linear positive monotonic relationship with NAFLD (OR = 1.20 per unit increment, P < 0.001), with risk rising markedly as FMBI increased—particularly when FMBI ≥ 10. A significant additive interaction between FMR and smoking was observed (AP = 0.37, P = 0.010). Additionally, FMBI is independently associated with ASCVD risk (β = 0.32, P = 0.001) and hepatic fibrosis (OR = 1.79 for Q4 vs. Q1, P = 0.008) in NAFLD patients. E-value analysis confirmed the robustness of these findings to unmeasured confounding.ConclusionThis study identifies FMBI as a robust NAFLD biomarker exhibiting linear dose–response, additive synergy with smoking, and independent prediction of hepatic fibrosis and ASCVD risk, supporting its use in targeted NAFLD management.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12876-026-04983-3.Keywords: Fat-Muscle-BMI Index (FMBI), Non-alcoholic fatty liver disease (NAFLD), Hepatic fibrosis, Atherosclerotic cardiovascular disease Introduction Non-alcoholic fatty liver disease (NAFLD), defined as hepatic steatosis of ≥ 5% in the absence of excessive alcohol consumption and secondary liver injury causes [1], affects approximately 25% of the global population, with its prevalence projected to rise further by 2030 [2]. As a leading aetiology of cirrhosis and hepatocellular carcinoma, NAFLD is also strongly comorbid with cardiovascular disease, extrahepatic malignancies, metabolic syndrome, type 2 diabetes mellitus (T2DM), and hypertension, imposing a substantial global disease burden [3]. Despite this, validated non-invasive biomarkers for the accurate diagnosis, risk stratification, progression monitoring, and treatment response assessment of NAFLD remain lacking. Liver biopsy, the gold standard for evaluating hepatic fibrosis, is limited in clinical practice by its invasiveness, sampling variability, and poor patient acceptance [4]. In 2023, an international consensus renamed non-alcoholic fatty liver disease (NAFLD) to metabolic dysfunction-associated steatotic liver disease (MASLD) to better reflect the metabolic drivers of the disease and reduce stigmatizing language [5]. Since our data were collected before this change (January 2020 to November 2021), we use the term NAFLD throughout this manuscript, while noting that nearly all NAFLD patients in our cohort would also meet the MASLD criteria [5]. Body mass index (BMI) is widely used for NAFLD screening due to its consistent association with disease risk, yet it has critical limitations in clinical management [6]. First, BMI alone fails to identify 10–20% of "lean" NAFLD cases (BMI < 25 kg/m2 in the general population, < 23 kg/m2 in Asians), leading to underdiagnosis; notably, lean NAFLD patients may face a higher risk of all-cause mortality [7, 8]. Second, BMI does not capture key pathological features driving non-alcoholic steatohepatitis (NASH) progression and disease severity, including myosteatosis, visceral adiposity, and metabolic dysfunction [9–11]. While BMI remains valuable for initial screening and weight management, its standalone diagnostic and predictive value is insufficient [12, 13]. A multidimensional risk stratification approach integrating waist circumference, visceral fat assessment, metabolic markers, and muscle metrics (e.g., grip strength, skeletal muscle index) is therefore essential to improve diagnostic accuracy and guide clinical decision-making [9, 14, 15]. Previous studies have demonstrated the multidimensional value of fat-muscle ratios (FMRs),including the subcutaneous-to-visceral fat ratio (SVR), visceral-to-abdominal fat ratio (VAR), and fat mass-to-fat-free mass ratio, in NAFLD research. Elevated FMR is positively correlated with NAFLD risk and exhibits superior diagnostic performance in T2DM patients [16–20]. Pathologically, low SVR or high VAR is associated with hepatic steatosis and fibrosis progression, while fat mass-to-fat-free mass ratio identifies body composition imbalance as a core mechanism underlying lean NAFLD; a low liver fat-to-total fat ratio independently predicts hepatic fibrosis risk [21]. However, FMRs have inherent limitations: device-dependent measurements (bioelectrical impedance analysis (BIA), dual-energy X-ray absorptiometry, CT raise comparability concerns, ultrasound-based NAFLD diagnosis leads to methodological heterogeneity, associations are population-specific (e.g., a stronger SVR-NAFLD link in males), and optimal cut-off values remain undefined [22, 23]. Standalone FMRs show limited predictive efficacy, necessitating combination with traditional indices such as BMI for optimized NAFLD risk stratification [21, 24]. Combining FMR with BMI addresses the limitations of each individual metric: BMI reflects overall obesity but neglects body composition distribution, while FMR captures the antagonistic muscle-fat relationship. Their integration enables precise metabolic risk assessment, avoids obesity-centric bias, and facilitates multidimensional evaluation of obesity phenotypes, metabolic risk, muscle function, and clinical prognosis [25–27]. To our knowledge, few study have investigated the association between the product of FMR and BMI-a novel composite index-and NAFLD. This study therefore aimed to explore the association between the Fat-Muscle-BMI Index (FMBI, FMR × BMI) and NAFLD, and further investigate the relationships of FMR and FMBI with hepatic fibrosis and atherosclerotic cardiovascular disease (ASCVD) risk in NAFLD patients. We hypothesized that FMBI would be associated with NAFLD and its complications. Methods Study population This retrospective cross-sectional study utilized data from the Dryad digital repository (10.5061/dryad.7d7wm3809), mainly including individuals who underwent routine health check-ups at Wuhan Union Hospital from January 2020 to November 2021. This was a secondary analysis of a previously published dataset. The study protocol was approved by the Ethics Committee of Wuhan Union Hospital. The detailed methodology and primary findings have been previously reported by Yan et al. [28]. The original study consecutively enrolled 1830 participants aged 40–79 years who voluntarily completed body composition analysis and liver ultrasonography. After applying exclusion criteria (excessive alcohol intake, history of chronic liver disease, acute illness, renal insufficiency, active malignancy, use of corticosteroids, or incomplete data), 1592 participants were included in the final analysis. NAFLD was diagnosed by abdominal ultrasonography based on the presence of hepatic steatosis in the absence of other liver diseases. We used this publicly available dataset to validate the novel FMBI index efficiently, reproducibly, and in line with ethical use of existing clinical data. Definition of NAFLD, hepatic fibrosis, and ASCVD risk NAFLD was diagnosed by conventional abdominal B-mode ultrasonography (Philips IU22, Philips Healthcare, N.V.) performed by certified sonographers; participants with hepatic steatosis and no excluded hepatic comorbidities were classified as having NAFLD. Hepatic fibrosis burden was assessed using the Fibrosis-4 (FIB-4) index, a well-validated non-invasive scoring system. Hepatic fibrosis risk was categorized using FIB-4 index thresholds: low (< 1.30), moderate (1.30–2.67), and high (> 2.67), with moderate and high categories combined for analysis due to limited sample size in the high-risk group. Due to the small sample size of the high-risk group (FIB-4 > 2.67), which would compromise statistical power, moderate and high fibrosis risk groups were combined for subsequent analyses. The 10-year ASCVD risk was calculated using the prediction algorithm recommended by the 2013 American College of Cardiology/American Heart Association (ACC/AHA) guidelines. An ASCVD risk score > 10% was defined as high cardiovascular risk, and a score ≤ 10% as low risk. Anthropometric and laboratory measurements Anthropometric measurements were performed by a trained technician. Body composition was measured using a multi‑frequency bioelectrical impedance analyzer (Tsinghua Tongfan, BCA‑2A, China) according to the manufacturer’s instructions. Height and weight were measured to calculate BMI. Blood pressure was measured after 10 min of rest, and the average of two readings was used. Venous blood samples were collected after an overnight fast of at least 8 h. Routine biochemical tests were performed in the clinical laboratory, including platelet count (PLT), total cholesterol (TC), triglycerides (TG), high‑density lipoprotein cholesterol (HDL‑C), low‑density lipoprotein cholesterol (LDL‑C), alanine aminotransferase (ALT), aspartate aminotransferase (AST), uric acid (UA), and fasting blood glucose (FBG). PLT was determined using a Beckman Coulter analyzer; biochemical parameters were measured on a Beckman AU5800 automated analyzer. The fat‑to‑muscle ratio (FMR) was calculated as total fat mass divided by total muscle mass. Type 2 diabetes mellitus (T2DM) and hypertension were defined based on self‑reported medical history or current use of relevant medications. Calculation of FMBI A novel composite metric, the Fat-Muscle-BMI Index (FMBI), was constructed as the product of FMR and BMI, with the formula: FMBI = Fat-to-muscle ratio (FMR) × Body mass index (BMI). Statistical analysis Descriptive statistics Participant baseline characteristics were stratified by FMBI quartiles (Q1–Q4). Continuous variables were presented as mean ± standard deviation (normally distributed) or median (interquartile range) (non-normally distributed), with normality assessed by the Shapiro–Wilk test. Categorical variables were expressed as frequencies and proportions. One-way ANOVA or the Kruskal–Wallis test was used for comparisons of continuous variables across groups, and the χ2 test for categorical variables. Covariate selection (Boruta algorithm) Boruta was a machine-learning feature selection method that identifies important variables. The Boruta algorithm (a random forest-based method) was used for covariate selection: variable importance was compared with permuted "shadow" features, and features that consistently outperformed shadow variables across 500 iterations were retained as confirmed predictors. Logistic regression for NAFLD, hepatic fibrosis and ASCVD risk score Multivariable logistic regression analyses were used to assess the association between FMBI and risks of NAFLD and NAFLD-associated hepatic fibrosis. Three hierarchical models were constructed for each outcome: Model 1 (unadjusted); Model 2 (adjusted for sex and tobacco use); Model 3 (fully adjusted for sex, tobacco use, ALT, AST, UA, FBG, TG, HDL-C, systolic blood pressure [SBP], hypertension, and T2DM). Continuous FMBI was analysed per unit increment; categorical FMBI was analysed by quartiles (Q1 as reference) for NAFLD and fibrosis, and by tertiles (T1 as reference) for additional NAFLD validation, with trend significance evaluated by treating quartiles/tertiles as ordinal variables. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. Multivariable linear regression models were used to assess the association between FMBI (continuous and quartiles) and ASCVD risk score, with the same three hierarchical adjustment models as above; trend tests treated FMBI quartiles as ordinal variables, and regression coefficients (β) and 95% CIs were reported. Restricted cubic splines (RCS) with four knots (placed at the 5th, 35th, 65th, and 95th percentiles of FMBI) were used to model the incremental risk pattern between FMBI and NAFLD risk, with adjustment for Model 3 covariates. A reference point of FMBI = 9.832 was set to calculate ORs, and nonlinearity was tested by likelihood ratio tests comparing linear and spline models. Subgroup and interaction analyses Subgroup analyses were performed using multivariable logistic regression stratified by sex, age, BMI, tobacco use, hypertension, and T2DM status. Interaction terms (FMBI × subgroup variable) were included to test for effect modification, with continuous FMBI analysed per unit increment; results were visualized by forest plots, and interaction was considered significant at P < 0.05. All subgroup models retained Model 3 covariates. To quantify incremental predictive performance for NAFLD, we added BMI, FRM, and the combined FRM–BMI measure separately to the prespecified basic multivariable model and calculated continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI) with 95% confidence intervals. Confidence intervals were derived via nonparametric bootstrapping and two-sided p-values were used to assess statistical significance. The basic model served as the reference. The interaction between smoking status and FMBI on NAFLD risk was assessed on both multiplicative and additive scales. Multiplicative interaction was tested by including product terms in logistic regression models and using likelihood ratio tests. Additive interaction was quantified using three established indices: Relative Excess Risk due to Interaction, Attributable Proportion, and Synergy Index, with their 95% CIs. Sensitivity analysis (E-value and others) E-value analysis was used to assess robustness to unmeasured confounding. E-value analysis was performed to evaluate the robustness of the FMBI-NAFLD association to unmeasured confounding. Sensitivity analyses were conducted to verify the consistency of the association between FMBI and NAFLD, hepatic fibrosis, and ASCVD risk across different analytical approaches. All statistical analyses were performed using R Statistical Software (Version 4.2.2, The R Foundation) and the Free Statistics Analysis Platform (Version 2.2, Beijing, China), with a two-tailed P < 0.05 considered statistically significant. Results Baseline characteristics of the study population Baseline characteristics of the 1592 participants stratified by FMBI quartiles are presented in Table 1, with significant differences observed for most variables across quartiles (P < 0.05). Sex distribution varied markedly, with the highest proportion of females in Q4 versus Q1. Age showed a progressive increase across FMBI quartiles. Metabolic parameters including BMI,FMR, FBG, TG, and HDL-C exhibited significant ascending trends across quartiles. The prevalence of hypertension, T2DM and NAFLD increased significantly with higher FMBI quartiles. No significant interquartile differences were observed for PLT and AST, indicating substantial metabolic heterogeneity across FMBI strata. Notably, UA levels decreased in Q4, likely due to the higher proportion of women in this group. Table 1Baseline characteristics of participantVariablesTotal (n = 1592)Q1 (n = 398)Q2 (n = 398)Q3 (n = 398)Q4 (n = 398)pSex, n (%) < 0.001 Female444 (27.9)46 (11.6)57 (14.3)106 (26.6)235 (59) Male1148 (72.1)352 (88.4)341 (85.7)292 (73.4)163 (41)Age (year) [n (%)] < 0.001 40–49354 (22.2)113 (28.4)110 (27.6)75 (18.8)56 (14.1) 50–59709 (44.5)200 (50.3)185 (46.5)180 (45.2)144 (36.2) 60–69360 (22.6)64 (16.1)59 (14.8)103 (25.9)134 (33.7) 70–79169 (10.6)21 (5.3)44 (11.1)40 (10.1)64 (16.1)BMI (kg/m2)25.4 ± 2.923.1 ± 1.824.8 ± 1.726.0 ± 2.227.8 ± 3.3 < 0.001FMR0.4 ± 0.10.3 ± 0.00.3 ± 0.00.4 ± 0.00.5 ± 0.1 < 0.001PLT(109/L)211.1 ± 53.3209.1 ± 54.3208.7 ± 53.0210.4 ± 51.5216.3 ± 54.30.152ALT(U/L)21.0 (15.0, 30.0)20.0 (14.0, 26.0)21.0 (15.0, 30.0)23.5 (16.0, 33.8)22.0 (16.0, 32.0) < 0.001AST(U/L)23.1 ± 11.022.6 ± 14.022.4 ± 9.123.6 ± 10.023.9 ± 10.10.19UA(umol/L)366.8 ± 95.8361.7 ± 89.2372.2 ± 91.7376.8 ± 98.7356.3 ± 102.00.009FBG(mmol/L)5.5 ± 1.75.3 ± 1.75.6 ± 1.75.4 ± 1.45.9 ± 1.8 < 0.001TC(mmol/L)4.5 ± 1.14.4 ± 1.04.5 ± 1.14.5 ± 1.14.6 ± 1.20.08TG(mmol/L)1.4 (1.0, 2.2)1.2 (0.9, 1.8)1.5 (1.0, 2.4)1.6 (1.1, 2.3)1.5 (1.0, 2.2) < 0.001HDL.C(mmol/L)1.1 ± 0.31.2 ± 0.31.1 ± 0.31.1 ± 0.31.2 ± 0.3 < 0.001LDL.C(mmol/L)2.7 ± 0.92.6 ± 0.82.7 ± 0.92.7 ± 0.92.7 ± 0.90.767SBP(mm Hg)130.9 ± 15.8127.3 ± 14.5131.3 ± 16.3130.2 ± 16.1134.9 ± 15.4 < 0.001DBP(mm Hg)81.6 ± 11.280.6 ± 10.982.6 ± 11.681.1 ± 11.582.1 ± 10.60.054Tobacco use, n (%)538 (33.8)158 (39.7)153 (38.4)145 (36.4)82 (20.6) < 0.001Alcohol use, n (%)520 (32.7)143 (35.9)163 (41)137 (34.4)77 (19.3) < 0.001FMBI10.1 ± 3.76.3 ± 1.08.5 ± 0.610.6 ± 0.715.1 ± 3.4 < 0.001Hypertension, n (%)943 (59.2)177 (44.5)227 (57)249 (62.6)290 (72.9) < 0.001Diabetes, n (%)498 (31.3)100 (25.1)127 (31.9)123 (30.9)148 (37.2)0.004NAFLD, n (%)973 (61.1)183 (46)238 (59.8)264 (66.3)288 (72.4) < 0.001Normality test was assessed by the Shapiro‑Wilk test; variables with normal distribution are shown as mean ± SD, and those with non‑normal distribution as median (IQR)ALT Alanine aminotransferase, AST Aspartate aminotransferase, BMI Body mass index, DBP Diastolic blood pressure, FBG Fasting blood glucose, FMR Fat-to-muscle ratio, HDL-C High-density lipoprotein cholesterol, LDL-C Low-density lipoprotein cholesterol, NAFLD Non-alcoholic fatty liver disease, PLT Platelets, SBP Systolic blood pressure, TC Total cholesterol, TG Triglycerides, UA Uric acid Baseline characteristics of participant Normality test was assessed by the Shapiro‑Wilk test; variables with normal distribution are shown as mean ± SD, and those with non‑normal distribution as median (IQR) ALT Alanine aminotransferase, AST Aspartate aminotransferase, BMI Body mass index, DBP Diastolic blood pressure, FBG Fasting blood glucose, FMR Fat-to-muscle ratio, HDL-C High-density lipoprotein cholesterol, LDL-C Low-density lipoprotein cholesterol, NAFLD Non-alcoholic fatty liver disease, PLT Platelets, SBP Systolic blood pressure, TC Total cholesterol, TG Triglycerides, UA Uric acid Covariate selection Univariate logistic regression identified multiple variables with significant unadjusted associations with NAFLD (Table S1), most notably FMBI,age categories, and key metabolic parameters (all P < 0.001). Boruta analysis (Fig. 1) confirmed the following NAFLD-associated predictors: TG, ALT, UA, FBG, T2DM, hypertension, HDL-C, SBP, tobacco use, AST, and sex. Alcohol use, TC, and age were deemed statistically insignificant, while diastolic blood pressure (DBP) showed tentative importance. This machine learning approach validated the manually selected covariates and identified TG as the highest-impact metabolic predictor. Based on these results, sex, tobacco use, ALT, AST, UA, FBG, TG, HDL-C, SBP, hypertension, and T2DM were included in all fully adjusted models (Model 3). Boruta algorithm-based feature selection for NAFLD-associated predictors. The Boruta algorithm, a random forest-based method, was used to identify important predictors of NAFLD. Variable importance (Z-score) is plotted for each candidate variable. TG, triglycerides; ALT, alanine aminotransferase; UA, uric acid; FBG, fasting blood glucose; T2DM, type 2 diabetes mellitus; HDL-C, high-density lipoprotein cholesterol; SBP, systolic blood pressure; AST, aspartate aminotransferase; DBP, diastolic blood pressure; TC, total cholesterol Association between FMBI and NAFLD A significant positive association between FMBI and NAFLD was observed across all analytical approaches (Table 2). Per unit increment in continuous FMBI was associated with a significantly increased NAFLD risk in the fully adjusted model (OR = 1.20, 95% CI: 1.14–1.25, P < 0.001). The FMBI ranged approximately from 6 to 15 in this study; each one-unit increase in continuous FMBI reflected a small incremental elevation in fat-muscle metabolic burden. Per unit increment in continuous FMBI was associated with a significantly increased NAFLD risk in the fully adjusted model (OR = 1.20, 95% CI: 1.14–1.25, P < 0.001). Further clinical meaningful analysis showed that per 5-unit increment in FMBI was linked to a substantially higher NAFLD risk in the fully adjusted model. Quartile analysis revealed a clear dose-dependent relationship: after full adjustment, Q4 had a 3.95-fold higher odds of NAFLD compared with Q1, with a significant trend persisting in the fully adjusted model (P < 0.001). The association remained reliable across sequentially adjusted models, confirming FMBI’s independent relationship with NAFLD after accounting for demographic, metabolic, and comorbidity confounders. Table 2Multivariate logistic regression analysis between FMBI and NAFLDVariableModel1Model2Model3OR (95%CI)P valueOR (95%CI)P valueOR (95%CI)P valueFMBI Continuous1.15 (1.11 ~ 1.18) < 0.0011.28 (1.22 ~ 1.33) < 0.0011.2 0(1.14 ~ 1.25) < 0.001FMBI CategoriesQ11(Ref)1(Ref)1(Ref)Q21.75 (1.32 ~ 2.31) < 0.0011.88 (1.41 ~ 2.51) < 0.0011.46 (1.07 ~ 2.00)0.019Q32.31 (1.74 ~ 3.08) < 0.0013.09 (2.28 ~ 4.20) < 0.0012.02 (1.45 ~ 2.81) < 0.001Q43.08 (2.29 ~ 4.13) < 0.0017.12 (4.96 ~ 10.22) < 0.0013.95 (2.64 ~ 5.90) < 0.001Trend test1.45 (1.32 ~ 1.59) < 0.0011.88 (1.68 ~ 2.10) < 0.0011.54 (1.36 ~ 1.74) < 0.001Model 1: no adjustedModel 2: adjusted sex, tobaccoModel 3: adjusted sex, tobacco use, ALT, AST, UA, FBG, TG, HDL.C, SBP, Hypertension, DiabetesQ Quartiles FMR, fat-to-muscle ratio, NAFLD Non-alcoholic fatty liver disease, ALT Alanine aminotransferase, AST Aspartate aminotransferase, BMI Body mass index, FBG Fasting blood glucose; HDL-C High-density lipoprotein cholesterol, SBP Systolic blood pressure; TG Triglycerides; UA Uric acid Multivariate logistic regression analysis between FMBI and NAFLD Model 1: no adjusted Model 2: adjusted sex, tobacco Model 3: adjusted sex, tobacco use, ALT, AST, UA, FBG, TG, HDL.C, SBP, Hypertension, Diabetes Q Quartiles FMR, fat-to-muscle ratio, NAFLD Non-alcoholic fatty liver disease, ALT Alanine aminotransferase, AST Aspartate aminotransferase, BMI Body mass index, FBG Fasting blood glucose; HDL-C High-density lipoprotein cholesterol, SBP Systolic blood pressure; TG Triglycerides; UA Uric acid RCS analysis (Fig S1) showed that NAFLD risk rose steadily with increasing FMBI without an obvious threshold. The nonlinearity test yielded P = 0.245, supporting a linear graded association between FMBI and NAFLD risk across the clinically relevant range of FMBI values. RCS analysis (Fig S1) demonstrated a significant linear graded association between FMBI and NAFLD risk. The risk rose steadily without a clear threshold, and this linear relationship persisted across the clinically relevant range of FMBI values. Subgroup and interaction analysis Forest plot analysis (Fig. 2) showed a consistent positive association between FMBI and NAFLD across most pre-defined subgroups. A significant interaction was observed for tobacco use: the effect of FMBI on NAFLD risk was weaker in tobacco users compared with non-users. Subgroup analysis of the association between FMBI and NAFLD risk. Forest plot showing the odds ratios (ORs) and 95% confidence intervals (CIs) for NAFLD per unit increment in continuous FMBI across predefined subgroups Both FMBI and smoking status were independently associated with increased NAFLD risk (ORs > 1.9, P < 0.001). The multiplicative interaction term between FMBI and smoking was not statistically significant. However, additive interaction analysis yielded significant results (Table 3, Fig. 3): AP due to interaction indicated that 37% of the excess NAFLD risk in individuals with both high FMBI and smoking exposure was attributable to their interaction. Table 3Assessment of additive and multiplicative interactions between FMBI and tobacco use in relation to the outcomeMeasuresEstimatesCI.95.lowCI.95.upP. valueOR001NANANAOR011.981.532.570OR101.931.52.50OR114.62.867.40OR(FMBI on outcome [Tobacco.use = = 0]1.981.532.570OR(FMBI on outcome [Tobacco.use = = 1]2.381.453.90OR(Tobacco.use on outcome [FMBI = = 0]1.931.52.50OR(Tobacco.use on outcome [FMBI = = 1]2.321.413.810Multiplicative scale1.20.692.10.52Relative Excess Risk due to Interaction (RERI)1.68−0.473.830.06Attributable Proportion (AP)0.370.050.680.01Synergy Index (SI)1.880.973.630 Assessment of additive and multiplicative interactions between FMBI and tobacco use in relation to the outcome Additive interaction between FMBI and smoking on NAFLD risk. The four groups represent combinations of FMBI (B: 0 and 1) and smoking status (A: 0, non-smoker; 1, smoker). The odds ratios (ORs) for NAFLD are presented for each group Incremental predictive analysis Relative to the basic model, inclusion of BMI produced certain improvement in reclassification and discrimination. Addition of FRM yielded a smaller but statistically significant gain. The combined FRM–BMI metric provided intermediate improvement (Table 4). Table 4Incremental predictive value of FRM, BMI, FRM-BMI for NAFLDModelNRI(Continuous) (95% CI)P valueIDI (95% CI)P valueBasic modelRefRefRefRef + BMI0.4787[0.3821–0.5754]00.0593[0.0478–0.0709]0 + FRM0.2684[0.1688–0.3681]00.0245[0.0166–0.0325]0 + FRM-BMI0.3216[0.223–0.4202]00.0378[0.0283–0.0473]0 Incremental predictive value of FRM, BMI, FRM-BMI for NAFLD Sensitivity analyses FMBI and ASCVD risk in NAFLD In the fully adjusted model (Model 3), per unit increment in FMBI was significantly associated with higher ASCVD risk scores in NAFLD patients (β = 0.32, 95% CI: 0.13–0.51, P = 0.001). Given that FMBI values ranged from approximately 6 to 15 in our cohort, this corresponds to an approximate 2.9 percentage‑point absolute increase in 10‑year ASCVD risk when comparing the highest versus the lowest FMBI quartile – a difference that is clinically meaningful for cardiovascular risk stratification. Quartile analysis revealed a significant dose‑response relationship (P for trend < 0.001), with Q4 having a 3.24‑point higher ASCVD score than Q1 (95% CI: 1.32–5.16, P = 0.001) (Table S2). The strength of the association increased after adjustment for metabolic confounders (Model 2 vs. Model 1), confirming FMBI’s independent contribution to ASCVD risk in NAFLD patients. FMBI and NAFLD-associated hepatic fibrosis Fully adjusted models demonstrated a significant dose-dependent relationship between FMBI quartiles and NAFLD-associated fibrosis risk (P for trend = 0.023), with Q4 having a 1.79-fold higher odds of fibrosis compared with Q1. Continuous FMBI showed a modest but significant association with fibrosis risk (Table S3). FMBI tertile analysis for NAFLD Tertile analysis of FMBI confirmed a substantiated dose-dependent relationship with NAFLD risk (Table S4): after full adjustment, T3 had a 3.01-fold higher odds of NAFLD than T1 (95% CI: 2.14–4.24, P < 0.001), with a significant trend across all models (P for trend < 0.001). Although adjustment for metabolic factors reduced the magnitude of ORs, statistical significance was preserved, further validating FMBI’s independent association with NAFLD. E-value analysis for unmeasured confounding E-value analysis confirmed the robustness of the FMBI-NAFLD association to unmeasured confounding (Table S5, Fig S2). For Q4 versus Q1, an unmeasured confounder would need to have an OR ≥ 3.39 with both FMBI and NAFLD to explain away the observed association. Continuous FMBI showed moderate robustness. These magnitude-dependent E-values reinforce that the FMBI-NAFLD association is unlikely to be explained by unmeasured confounding. Discussion FMBI shows a significant linear dose‑response association with NAFLD risk, interacts additively with smoking to further increase risk, and remains independently associated with hepatic fibrosis and ASCVD risk beyond traditional metabolic factors. Among these, the additive interaction between FMBI and smoking is particularly noteworthy, as it quantifies a synergistic effect that explains over one‑third of excess NAFLD risk in co‑exposed individuals. These findings support FMBI as an effective tool for identifying high‑risk NAFLD populations and guiding targeted clinical management. Sarcopenic obesity (SO), is a well-established predictor of severe hepatic pathological outcomes in NAFLD and an independent risk factor for significant hepatic fibrosis [29]. Sarcopenia-induced muscle mass loss exacerbates insulin resistance, promotes hepatic fat accumulation and chronic inflammation, and accelerates the progression from simple steatosis to NASH and fibrosis [30–32]; one study has further confirmed a positive correlation between SO and the severity of NAFLD-related fibrosis [33]. While high BMI is a major driver of NAFLD, pure obese phenotypes (without sarcopenia) are associated with milder hepatic histological injury compared with SO phenotypes [34]. Notably, lean NAFLD (normal BMI with hepatic steatosis) may present with milder metabolic abnormalities at diagnosis but has a long-term prognosis (e.g., cirrhosis, hepatocellular carcinoma risk) that is not superior to obese NAFLD [35]. These observations align with our findings of a dose-dependent association between FMBI and NAFLD-related fibrosis risk, as FMBI integrates both adiposity and muscle mass to capture the SO phenotype that is missed by BMI alone.Patients with concurrent SO and NAFLD have a significantly higher risk of significant fibrosis and ASCVD events than those with pure obesity [36, 37]. Lean NAFLD is highly prevalent in Asian populations, and despite mild initial metabolic parameters, it may closely linked to SO (with lean NAFLD potentially acting as a risk factor for SO development), and further highlighting the limitations of BMI-only risk assessment for obesity-related NAFLD [38]. Our study demonstrates that per unit increment in FMBI is associated with a 20% higher risk of NAFLD and an approximately 33% higher risk of NAFLD-related ASCVD, validating FMBI’s consistent predictive value for both NAFLD and its cardiovascular comorbidities. The insidious progression of lean NAFLD and its link to SO further reinforce the inadequacy of BMI-based risk stratification [39–41]. Supported by rigorous covariate screening and unmeasured confounding evaluation, our results confirm a strong linear association between FMBI and NAFLD, with no apparent safe threshold for elevated FMBI: FMBI ≥ 10 is associated with increased NAFLD risk, which rises continuously and nearly linearly with further increases in FMBI. By integrating muscle and adiposity indicators, FMBI effectively captures both muscular and non-muscular NAFLD subtypes, underscoring the clinical importance and necessity of incorporating muscle parameters into NAFLD risk assessment. Sarcopenia promotes NAFLD development and progression through multi-system interactive mechanisms: (1) Insulin resistance: Reduced muscle mass impairs peripheral glucose disposal, exacerbating systemic insulin resistance and hepatic lipid accumulation [42–44]; (2) Chronic inflammation: Sarcopenia is associated with increased pro-inflammatory cytokines (TNF-α, IL-6) and decreased adiponectin, which activate hepatic stellate cells and drive fibrogenesis [44–46]; (3) Myokine dysregulation: Decreased irisin and elevated myostatin directly enhance hepatic steatosis by regulating lipid metabolism [45, 46]; (4) Nutritional and behavioural factors: Protein deficiency and sedentary lifestyles worsen metabolic disturbances, creating a vicious cycle between sarcopenia and NAFLD [47]. Emerging evidence confirms smoking as an independent NAFLD risk factor, despite historical debate [48–51]. Cigarette smoke contains over 4000 bioactive compounds that induce cellular damage; nicotine directly promotes NAFLD development and exacerbates obesity-induced hepatic steatosis, suggesting a biological synergy between smoking and obesity in metabolic liver injury [51]. The "obesity plus smoking" phenotype represents a high-risk combination with significantly elevated complication risks [48].This study reveals a significant additive interaction between smoking and FMBI in NAFLD for the first time. The synergistic effect, where FMBI's impact is stronger in non-smokers, explains over one-third of the excess risk in co-exposed individuals. This finding has direct clinical implications: individuals with both high FMBI (e.g., ≥ 10) and a smoking habit should be prioritized for NAFLD screening and lifestyle intervention. These novel insights underscore the urgency of targeted interventions for high-FMBI smokers to reduce NAFLD and its comorbidities. We acknowledged the important prior work by Yan et al., who first reported the association between FMR and NAFLD, liver fibrosis, and ASCVD risk using the same cohort^28^. Our study extends their findings in three key aspects: we propose a novel composite index (FMBI) that integrates FMR and BMI; we systematically evaluated its incremental predictive value over FMR and BMI alone using NRI and IDI analyses; and we identified a significant additive interaction between FMR and smoking, not previously reported. A very recent study by Song et al. also performed a secondary analysis of the same Wuhan Union Hospital cohort and evaluated several non-conventional lipid parameters, including TyG-BMI, for NAFLD detection and ASCVD risk prediction [52]. Our study differs in that we focus on body composition-derived indices (FMR and FMBI), whereas Song et al. focused on metabolic parameters. These two approaches are complementary rather than competing, as they reflect different physiological pathways, including insulin resistance versus fat-muscle imbalance. Additionally, the additive interaction between FMR and smoking was not examined in Song et al. We also evaluated the predictive value of FMBI over these metabolic parameters using NRI and IDI analyses (Table S6). The recent nomenclature change from NAFLD to MASLD does not affect the validity of our findings, as the two definitions identify largely overlapping populations, and the metabolic risk factors central to MASLD are already well-represented in our NAFLD cohort 5. This study possesses several key strengths, including the development of FMBI along with the application of rigorous covariate selection methods (univariate regression and Boruta algorithm) to minimize confounding. Furthermore, the linear progressive increase between FMBI and NAFLD was validated using RCS analysis, and interaction effects were assessed on both multiplicative and additive scales. Finally, we used E-value analysis to evaluate the robustness of our findings to potential unmeasured confounding.However, the study has inevitable limitations that should be acknowledged: (1) the retrospective cross-sectional design only confirms a correlation, not a causal relationship, between FMBI and NAFLD and its complications; (2) objective biases in indicator detection (e.g., BIA-based body composition analysis, which is less accurate than reference methods such as dual-energy X-ray absorptiometry or CT, and ultrasound-based NAFLD diagnosis) may affect the accuracy of results; (3) the single-center study design introduces potential selection bias; (4) the study population is exclusively Chinese, limiting the generalizability of findings to other ethnic groups. Conclusion In this retrospective cross-sectional study of 1,592 Chinese adults, we demonstrated that FMBI, a novel composite index of FMR and BMI, is significantly associated with NAFLD risk in a linear dose–response manner, exhibits an additive interaction with smoking, and independently predicts NAFLD-related hepatic fibrosis and ASCVD risk. For clinical practice, an FMBI ≥ 10, especially with smoking, should raise suspicion of NAFLD and prompt further evaluation for fibrosis and heart risk. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Acknowledgements The authors would like to thank the investigators at Tongji Medical College of Huazhong University of Science and Technology for sharing their data. Authors’ contributions Xiaokang Wu: Conceptualization, Methodology, Investigation, Writing—Original Draft, Writing—Review & Editing. Jiafeng Yin and Yue Li: Software, Formal Analysis and Data Curation. Chaoliang Xiong: Investigation, Resources. Hailong Liu and Hao Meng: Investigation. Xinjiang Hou: Visualization and manuscript revision. All authors have read and agreed to the published version of the manuscript. Funding This project was funded by Key Research and Development Plan Projects in Shaanxi Province (2021SF-133). Data availability Data are available in a public, open access repository. The datasets analyzed in this study are openly available in the Dryad Digital Repository ([https://doi.org/10.5061/dryad.7d7wm3809]) 28. The following supporting information can be downloaded at: Supplementary Figures and Tables. Declarations Ethics approval and consent to participate The study involved the Publicly open-data from the Dryad digital repository including raw available clinical information data collected by other research teams from various countries and hospitals, has been used and analyzed by some studies. This data was from individuals who underwent routine health check-ups at Wuhan Union Hospital from January 2020 to November 2021, and the study protocol was approved by the Ethics Committee of Wuhan Tongji Medical College (S155), and all procedures were performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki. Not applicable. This study is a secondary analysis of publicly available, anonymized data from the Dryad digital repository. The original study was approved by the Ethics Committee of Wuhan Tongji Medical College (S155), which waived the requirement for individual informed consent. Consent for publication Not applicable. This manuscript does not contain any individual person’s data in any form (including images, videos, or clinical details) that would compromise anonymity. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Data Availability Statement Data are available in a public, open access repository. The datasets analyzed in this study are openly available in the Dryad Digital Repository ([https://doi.org/10.5061/dryad.7d7wm3809]) 28. The following supporting information can be downloaded at: Supplementary Figures and Tables.