Abstract
Background
Patients with metabolic dysfunction-associated steatotic liver disease (MASLD) are at increased risk of panvascular disease (PD). Noninvasive markers for early risk stratification are needed. We evaluated whether ultrasound-derived fat fraction (UDFF) improves the prediction of PD occurrence in MASLD patients.
Methods
In this prospective cohort, 525 adults with MASLD were enrolled between October 2022 and December 2023 and followed for 6 months for PD occurrence. Baseline UDFF and clinical/metabolic variables were collected. Multivariable logistic regression analyses reported standardized coefficients (per-standard deviation [SD] β) and odds ratios (ORs) per 1-SD (OR). A decision tree model was constructed to explore differences across metabolic subgroups, including obesity, hypertension, diabetes, and multiple metabolic abnormalities. Discriminative performance was compared using the area under the curve (AUC; DeLong test).
Results
Body mass index (BMI), systolic blood pressure (SBP), homeostasis model assessment of insulin resistance (HOMA-IR), and UDFF were each associated with increased PD risk. In the multivariable model, the strongest standardized effects were observed for UDFF (β=3.28; OR, 2.10; 95% confidence interval [CI], 1.45 to 2.75) and BMI (β=1.87; OR, 1.68; 95% CI, 1.25 to 2.34), followed by SBP (β=1.33; OR, 1.15; 95% CI, 1.03 to 1.26). HOMA-IR (β=0.96; OR, 1.65; 95% CI, 1.30 to 2.15), age (β=0.56; OR, 1.06; 95% CI, 1.01 to 1.12), and liver stiffness measurement (β=0.14; OR, 1.03; 95% CI, 1.01 to 1.07) also showed smaller but significant associations. The decision tree model reflected similar patterns and achieved higher discrimination than logistic regression (AUC 0.85 vs. 0.80, P<0.05).
Conclusion
UDFF is a key predictor of PD risk in MASLD patients, underscoring the importance of risk stratification for early intervention and personalized management strategies.
Keywords: Metabolic dysfunction-associated steatotic liver disease, Panvascular disease, Ultrasound-derived fat fraction
GRAPHICAL ABSTRACT
INTRODUCTION
The concept of metabolic dysfunction-associated steatotic liver disease (MASLD) provides a more precise characterization of what was previously termed metabolic associated fatty liver disease, highlighting the pivotal role of metabolic dysfunction in the pathogenesis and progression of fatty liver disorders [1,2]. The designation MASLD underscores its close association with metabolic conditions such as hypertension, diabetes mellitus, and dyslipidemia. Panvascular disease (PD) refers to systemic pathological conditions characterized by atherosclerosis and related vascular abnormalities that affect multiple territories across major vascular beds, including—but not limited to—the cardiovascular, cerebrovascular, peripheral vascular, and valvular arterial systems [3,4]. Emerging studies [5,6] have demonstrated that metabolic abnormalities, which often appear in MASLD patients before hepatic injury develops, are strongly linked to PD. This connection extends beyond hepatic steatosis, reflecting the complex interactions between systemic metabolic dysregulation and vascular pathology, and suggesting a markedly elevated risk of systemic vascular disease in individuals with MASLD.
In managing MASLD, therapeutic strategies should address not only hepatic fibrosis progression but also systemic metabolic dysfunction and its vascular effects [7]. Early detection and intervention targeting metabolic and vascular abnormalities in MASLD patients are crucial to mitigating hepatic damage and preventing major cardiovascular events, thereby improving long-term outcomes. This comprehensive management approach is essential for personalized treatment, particularly in patients with coexisting metabolic and vascular risk factors, and highlights the need for stratified care.
Ultrasound-derived fat fraction (UDFF) is a reliable, noninvasive technique for quantifying hepatic fat content in MASLD patients. It is implemented on the Siemens Acuson Sequoia ultrasound system, with results expressed as a percentage [8–11]. However, MASLD is a multifactorial condition involving multiple metabolic systems, including blood pressure regulation, glucose metabolism, and lipid homeostasis [12,13]. Thus, reliance solely on UDFF may not fully capture the disease’s complexity.
Accurate prediction of PD risk in MASLD patients requires a multidimensional approach that integrates UDFF with metabolic indicators such as body weight, blood pressure, blood glucose levels, and lipid profiles. This comprehensive assessment allows a better understanding of the broader metabolic context and supports more effective clinical decision-making.
In this study, we integrated UDFF with key clinical and metabolic parameters—including blood glucose, lipid profiles, insulin resistance, and body mass index (BMI)—to evaluate their associations with PD risk in MASLD patients. Our objective was to facilitate early detection and intervention before disease progression to more advanced stages. This approach highlights the intricate interplay between MASLD and metabolic dysfunction and provides clinicians with refined decision-support tools for personalized patient management and improved long-term outcomes.
METHODS
Study design
The study enrolled 525 patients with MASLD from the Department of Ultrasound. Baseline data were collected from October 2022 to December 2023, with follow-up conducted through June 2024. Participants were diagnosed with MASLD according to the multi-society Delphi consensus criteria [14], which require the presence of hepatic steatosis and at least one cardiometabolic risk factor. The study adhered to the principles of the Declaration of Helsinki [15], received approval from the institutional ethics committee of the First Affiliated Hospital of the University of Science and Technology of China (approval number: 2024-RE-417), and obtained informed consent from all participants (Fig. 1).
Fig. 1.
Flowchart of patient recruitment. MASLD, metabolic dysfunction-associated steatotic liver disease; UDFF, ultrasound-derived fat fraction; O-MASLD, MASLD with obesity; H-MASLD, MASLD with hypertension; D-MASLD, MASLD with diabetes; M-MASLD, MASLD with multiple comorbidities; PD, panvascular disease.
Patient grouping
MASLD patients were categorized into four subgroups according to the presence of metabolic comorbidities at enrollment. MASLD with obesity (O-MASLD) was defined as obesity (BMI >25 kg/m2 according to World Health Organization Asia criteria) in the absence of hypertension or diabetes. MASLD with hypertension (H-MASLD) was defined as hypertension (systolic blood pressure [SBP] ≥140 mm Hg or diastolic blood pressure [DBP] ≥90 mm Hg on at least three nonconsecutive measurements), based on the Chinese Guidelines for the Management of Hypertension [16], without obesity or diabetes. MASLD with diabetes (D-MASLD) was defined as diabetes, characterized by fasting plasma glucose (FPG) ≥7.0 mmol/L or hemoglobin A1c (HbA1c) ≥6.5%, according to the American Diabetes Association criteria [17], in the absence of obesity or hypertension. The MASLD with multiple comorbidities (M-MASLD) included patients with two or more metabolic abnormalities—specifically obesity, hypertension, or diabetes—as defined above. Diagnostic thresholds were aligned with internationally recognized guidelines to ensure methodological consistency.
Data collection
Metabolic parameters were collected at study enrollment for all participants. Clinical characteristics included age, sex, alcohol consumption, BMI, SBP, and DBP. Laboratory measurements comprised triglycerides (TG), low-density lipoprotein (LDL), high-density lipoprotein (HDL), FPG, fasting insulin (FINS), HbA1c, homeostasis model assessment of insulin resistance (HOMA-IR), aspartate aminotransferase, alanine aminotransferase, gamma-glutamyl transferase, and total bile acid.
Ultrasound imaging was performed using a Siemens Acuson system equipped with a deep abdominal transducer (DAX) probe, which is a curvilinear broadband piezoceramic array with an operational frequency range of 1.0 to 3.5 MHz. Measurement protocols followed manufacturer-defined specifications. A proprietary calibration marker was aligned with the hepatic capsule to automatically position a fixed-dimension rectangular region of interest (ROI), algorithmically constrained by the system’s quantification software. UDFF measurements were acquired following a standardized protocol: a single certified sonographer obtained three consecutive UDFF readings targeting the central right hepatic lobe (Couinaud segment V). Between acquisitions, participants were asked to assume an upright position to minimize positional bias. During each UDFF acquisition, liver stiffness measurements (LSMs) were obtained from the same ROI using shear wave elastography technology integrated into the ultrasound platform. The system’s multiparametric acquisition mode allowed simultaneous quantification of hepatic fat fraction and viscoelastic properties within identical anatomical coordinates, ensuring both spatial and temporal consistency. Raw UDFF and LSM data were stored in DICOM format and automatically tabulated in system-generated reports. Measurements were performed during apnea, maintaining the interquartile range (IQR)/mean ratio below 30% for both UDFF and LSM to ensure accuracy (Fig. 2).
Fig. 2.
Ultrasound-derived fat fraction (UDFF) and liver stiffness measurement (LSM) measurements in metabolic dysfunction-associated steatotic liver disease (MASLD) patients with different comorbidities. UDFF and LSM data were obtained using a Siemens Sequoia system with a deep abdominal transducer (DAX) transducer. (A) Assessments in MASLD patients with obesity (O-MASLD). (B) Assessments in MASLD patients with hypertension (H-MASLD). (C) Assessments in MASLD patients with diabetes (D-MASLD). (D) Assessments in MASLD patients with multiple comorbidities (M-MASLD).
All enrolled patients underwent systematic longitudinal follow-up for 6 months. During this period, new-onset PD events detected via multi-territorial imaging were recorded. Based on follow-up findings, patients were categorized into two groups: the PD group and the non-PD group.
Statistical analysis
All analyses were performed using SPSS version 26.0 (IBM, Armonk, NY, USA) and R version 4.4.2 (R Foundation, Vienna, Austria). The normality of continuous variables was assessed using the Shapiro–Wilk test. Data are presented as mean±standard deviation (SD) or median (IQR) for continuous variables and as number (%) for categorical variables. Between-group comparisons were conducted using the independent-samples t test or Mann–Whitney U test for continuous data and the chi-square or Fisher exact test for categorical data. A two-sided P<0.05 was considered statistically significant.
Variables that differed significantly between PD and non-PD groups in univariable analyses (P<0.05), along with prespecified clinically relevant covariates, were entered into multivariable logistic regression. To enable comparison of effect sizes across predictors with differing units, continuous variables were z-standardized within each analysis stratum, and standardized coefficients (per-SD β) were reported. For clinical interpretability, odds ratios (ORs) per 1-SD (OR) with 95% confidence intervals (CIs) were also presented.
To identify actionable thresholds and interactions within metabolic subgroups, a decision tree model was constructed using the Quick, Unbiased, Efficient Statistical Tree (QUEST) algorithm with the same candidate predictors. Discriminative performance was quantified by the area under the curve (AUC), and AUCs were compared between the logistic regression and QUEST models using the nonparametric DeLong test.
RESULTS
Clinical characteristics and biochemical indicators of enrolled patients
A total of 525 MASLD patients were included in the study, comprising 64.6% men and 35.4% women. Among all participants, 23.0% (n=121) were classified as O-MASLD, 16.9% (n=89) as H-MASLD, 19.2% (n=101) as D-MASLD, and 20.1% (n=214) as M-MASLD. Additional clinical, biochemical, and imaging characteristics are summarized in Table 1. The classification of PD according to affected vascular territories and diagnostic modalities is provided in Supplemental Table S1. The overall incidence of PD is presented in Supplemental Fig. S1, and the temporal distribution of PD across MASLD subgroups is illustrated in Supplemental Fig. S2. The overall distribution of PD across vascular territories in MASLD patients is shown in Supplemental Fig. S3, while subgroup-specific PD distributions are detailed in Supplemental Table S2.
Table 1.
Baseline Data for MASLD Patients
| Parameter | Total (n=525) | O-MASLD (n=121) | H-MASLD (n=89) | D-MASLD (n=101) | M-MASLD (n=214) |
|---|---|---|---|---|---|
| Age, yr | 54.38±9.65 | 54.05±8.23 | 54.51±8.11 | 54.14±10.98 | 54.99±8.42 |
| Female sex | 186 (35.4) | 43 (35.5) | 31 (34.8) | 52 (51.5) | 35 (30.2) |
| Alcohol consumption, g/day | 8.52±2.45 | 8.64±2.37 | 8.18±2.05 | 8.69±2.61 | 8.51±2.42 |
| BMI, kg/m2 | 23.19±3.62 | 29.29±2.99 | 19.69±1.50 | 19.86±1.90 | 23.29±2.45 |
| SBP, mm Hg | 122.29±9.55 | 111.66±5.48 | 136.55±3.80 | 119.38±3.48 | 120.29±4.65 |
| DBP, mm Hg | 80.01±6.55 | 73.97±5.39 | 84.70±2.48 | 85.87±2.43 | 75.60±5.67 |
| TG, mmol/L | 1.66 (1.16–1.93) | 1.67 (1.21–1.99) | 1.67 (1.17–1.93) | 1.69 (1.18–1.92) | 1.63 (1.11–1.88) |
| LDL, mmol/L | 3.03±0.63 | 3.15±0.64 | 2.98±0.74 | 2.98±0.57 | 2.88±0.56 |
| HDL, mmol/L | 1.20±0.26 | 1.15±0.25 | 1.21±0.23 | 1.22±0.26 | 1.25±0.28 |
| FPG, mmol/L | 6.98±2.80 | 6.12±2.60 | 4.75±2.53 | 8.75±1.74 | 6.77±2.54 |
| FINS, μU/mL | 13.15 (11.00–17.74) | 10.18 (8.47–15.73) | 10.62 (8.75–16.27) | 16.28 (14.56–17.94) | 12.23 (10.14–18.25) |
| HbA1c, % | 6.78±2.20 | 5.49±1.28 | 4.65±1.07 | 8.99±1.55 | 6.01±1.32 |
| HOMA-IR | 3.67 (2.67–5.26) | 2.12 (1.60–2.53) | 1.89 (1.20–2.16) | 6.09 (5.16–7.64) | 2.70 (1.61–5.47) |
| AST, U/L | 22.75±4.17 | 23.40±4.19 | 21.71±3.93 | 23.37±4.29 | 23.06±3.97 |
| ALT, U/L | 17.15±2.91 | 16.39±2.39 | 15.95±2.48 | 18.50±2.81 | 16.54±2.50 |
| GGT, U/L | 39.0 (24.0–60.46) | 34.55 (21.55–50.82) | 34.17 (20.17–51.17) | 45.67 (28.67–70.01) | 36.39 (20.78–61.39) |
| TBA, μmol/L | 2.40 (1.55–4.17) | 2.39 (1.50–4.45) | 2.32 (1.77–4.22) | 2.5 (1.7–4.0) | 2.36 (1.22–4.14) |
| UDFF, % | 10.16±3.94 | 13.45±3.10 | 7.62±2.32 | 7.23±1.86 | 10.62±4.17 |
| LSM, kPa | 6.1 (5.3–11.4) | 5.6 (4.7–11.2) | 5.9 (5.5–12.4) | 6.1 (5.8–11.6) | 6.4 (5.2–10.8) |
| Proportion of PD cases | 205 (38.9) | 66 (54.5) | 15 (15.3) | 34 (33.7) | 90 (42.1) |
Values are expressed as mean±standard deviation, number (%), or median (interquartile range).
MASLD, metabolic dysfunction-associated steatotic liver disease; O-MASLD, MASLD with obesity; H-MASLD, MASLD with hypertension; D-MASLD, MASLD with diabetes; M-MASLD, MASLD with multiple comorbidities; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TG, triglycerides; LDL, low-density lipoprotein; HDL, high-density lipoprotein; FPG, fasting plasma glucose; FINS, fasting insulin; HbA1c, hemoglobin A1c; HOMA-IR, homeostasis model assessment of insulin resistance; AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, gamma-glutamyl transferase; TBA, total bile acids; UDFF, ultrasound-derived fat fraction; LSM, liver stiffness measurement; PD, panvascular disease.
Follow-up outcomes of MASLD patients and analysis of PD and non-PD parameters
The incidence and proportion of PD in each subgroup are depicted in Supplemental Fig. S4. Within the overall MASLD cohort, the PD group was older and included a higher proportion of women compared with the non-PD group. The PD group also demonstrated significantly higher BMI, SBP, and several metabolic parameters, including TG, LDL, and HOMA-IR (all P<0.01). Moreover, both UDFF (11.77%±4.24% vs. 9.12%±3.36%, P<0.01) and LSM (6.6 kPa [IQR, 5.2 to 12.0] vs. 5.8 kPa [IQR, 4.9 to 10.6], P<0.01) were significantly higher in the PD group than in the non-PD group. No significant differences were observed between groups for DBP, HDL, or hepatic function markers (all P>0.05). Detailed values are provided in Table 2.
Table 2.
Comparison of All MASLD Patients with and without PD
| Parameter | PD group (n=205) | Non-PD group (n=320) | P value |
|---|---|---|---|
| Age, yr | 55.18±7.16 | 53.77±8.71 | <0.001 |
| Female sex | 84 (41.0) | 102 (31.8) | <0.001 |
| Alcohol consumption, g/day | 8.79±2.25 | 8.34±2.46 | 0.472 |
| BMI, kg/m2 | 25.24±3.53 | 21.88±3.03 | <0.001 |
| SBP, mm Hg | 128.33±9.04 | 117.63±9.82 | <0.001 |
| DBP, mm Hg | 79.25±6.74 | 80.50±6.39 | 0.194 |
| TG, mmol/L | 1.81 (1.24–2.15) | 1.60 (1.13–1.84) | <0.001 |
| LDL, mmol/L | 3.30±0.63 | 2.86±0.57 | <0.001 |
| HDL, mmol/L | 1.19±0.26 | 1.17±0.23 | 0.159 |
| FPG, mmol/L | 6.71±2.86 | 6.54±2.64 | 0.348 |
| FINS, μU/mL | 13.90 (11.92–18.68) | 12.66 (10.41–17.13) | <0.001 |
| HbA1c, % | 7.35±2.18 | 6.42±2.15 | <0.001 |
| HOMA-IR | 4.53 (3.40–6.17) | 3.13 (2.20–4.67) | <0.001 |
| AST, U/L | 22.86±4.15 | 22.48±4.21 | 0.194 |
| ALT, U/L | 17.29±3.03 | 17.07±2.83 | 0.417 |
| GGT, U/L | 40.0 (25.0–61.0) | 39.0 (23.0–60.0) | 0.291 |
| TBA, μmol/L | 2.40 (1.58–4.27) | 2.42 (1.54–4.12) | 0.186 |
| UDFF, % | 11.77±4.24 | 9.12±3.36 | <0.001 |
| LSM, kPa | 6.6 (5.2–12.0) | 5.8 (4.9–10.6) | <0.001 |
Values are expressed as mean±standard deviation, number (%), or median (interquartile range).
MASLD, metabolic dysfunction-associated steatotic liver disease; PD, panvascular disease; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TG, triglycerides; LDL, low-density lipoprotein; HDL, high-density lipoprotein; FPG, fasting plasma glucose; FINS, fasting insulin; HbA1c, hemoglobin A1c; HOMA-IR, homeostasis model assessment of insulin resistance; AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, gamma-glutamyl transferase; TBA, total bile acids; UDFF, ultrasound-derived fat fraction; LSM, liver stiffness measurement.
Significant differences in UDFF and LSM between PD and non-PD groups were also identified across all MASLD subgroups (Supplemental Tables S3–S6). In the O-MASLD subgroup, the PD group showed significantly lower HDL and higher BMI, TG, LDL, and FPG compared with the non-PD group. UDFF (14.31%±3.23% vs. 12.42%±2.62%, P<0.01) and LSM (6.1 kPa [IQR, 4.3 to 11.9] vs. 5.1 kPa [IQR, 4.5 to 10.8], P<0.01) were likewise significantly elevated. In the H-MASLD subgroup, the PD group had significantly higher SBP and DBP, along with increased UDFF (9.42%±2.15% vs. 7.26%±2.19%, P<0.01) and LSM (6.3 kPa [IQR, 5.2 to 13.5] vs. 5.4 kPa [IQR, 4.9 to 11.5], P<0.01). In the D-MASLD subgroup, the PD group exhibited significantly higher FINS, HbA1c, and HOMA-IR compared with the non-PD group, alongside elevated UDFF (7.61%±1.91% vs. 7.03%±1.82%, P<0.01) and LSM (6.9 kPa [IQR, 5.7 to 12.3] vs. 6.1 kPa [IQR, 5.3 to 10.1], P<0.01). Finally, in the M-MASLD subgroup, the PD group demonstrated significantly higher BMI and HOMA-IR. UDFF (11.76%±4.39% vs. 10.11%±2.86%, P<0.01) and LSM (7.2 kPa [IQR, 5.9 to 12.1] vs. 6.4 kPa [IQR, 5.1 to 10.6], P<0.01) were again significantly elevated in the PD group.
Logistic regression analysis of predictors associated with PD
Multivariate logistic regression analyses for each subgroup are presented in Fig. 3. In the overall MASLD cohort, HOMA-IR, UDFF, LSM, BMI, SBP, and age were identified as independent risk factors for PD (all P<0.05). Based on standardized effect sizes, UDFF exhibited the strongest association (β=3.28; OR, 2.10; 95% CI, 1.45 to 2.75), followed by BMI (β=1.87; OR, 1.68; 95% CI, 1.25 to 2.34) and SBP (β=1.33; OR, 1.15; 95% CI, 1.03 to 1.26). HOMA-IR (β=0.96), age (β=0.56), and LSM (β=0.14) were smaller but significant contributors.
Fig. 3.
Multivariable logistic regression analysis of factors associated with panvascular disease (PD) across metabolic dysfunction-associated steatotic liver disease (MASLD) subgroups. (A) Overall MASLD cohort: the largest effects were observed for ultrasound-derived fat fraction (UDFF) (β=3.28), body mass index (BMI) (β=1.87), and systolic blood pressure (SBP) (β=1.33); homeostasis model assessment of insulin resistance (HOMA-IR), age, and liver stiffness measurement (LSM) were also significant (all P<0.05). (B) Obese MASLD subgroup: UDFF showed the highest effect (β=2.48), followed by LSM (β=0.73) and BMI (β=0.59); low-density lipoprotein (LDL) and HOMA-IR were also significant. (C) Hypertensive MASLD subgroup: LSM (β=1.93) and SBP (β=1.90) were dominant; UDFF and age remained significant. (D) Diabetic MASLD subgroup: HOMA-IR (β=2.23) and LSM (β=1.63) were dominant; UDFF (β=1.78), BMI (β=0.78), and hemoglobin A1c (HbA1c) (β=0.18) were also significant. (E) MASLD with multiple comorbidities: UDFF (β=2.91) and HOMA-IR (β=2.11) were largest, with LSM and BMI also significant (all P<0.05). OR, odds ratio per 1-standard deviation (SD); CI, confidence interval; β, standardized logistic coefficient per 1-SD.
In O-MASLD, all five predictors were significant, with UDFF showing the largest standardized effect (β=2.48; OR, 2.22; 95% CI, 1.25 to 2.34), followed by LSM (β=0.73), BMI (β=0.59), LDL (β=0.32), and HOMA-IR (β=0.31). In H-MASLD, LSM (β=1.93; OR, 1.45; 95% CI, 1.23 to 1.67) and SBP (β=1.90; OR, 1.68; 95% CI, 1.25 to 2.34) were dominant predictors, while UDFF (β=0.54) and age (β=0.34) had smaller effects. In D-MASLD, HOMA-IR (β=2.23; OR, 2.75; 95% CI, 1.45 to 3.05) and LSM (β=1.63; OR, 1.75; 95% CI, 1.30 to 2.35) were the principal contributors, with HbA1c (β=0.18), UDFF (β=1.78), and BMI (β=0.78) also reaching significance. In M-MASLD, UDFF (β=2.91; OR, 2.21; 95% CI, 1.45 to 2.75) and HOMA-IR (β=2.11; OR, 2.03; 95% CI, 1.66 to 2.21) exerted the strongest effects, while LSM (β=0.73) and BMI (β=0.68) had modest yet significant associations (all P<0.05).
Predicting the risk of PD in MASLD patients based on decision tree and cutoff value
We further developed a decision tree model to predict PD risk in MASLD patients and to determine cutoff values for key risk factors. Variables associated with PD occurrence were incorporated into the model to identify major predictors and their respective threshold points.
In the overall MASLD cohort, the decision tree analysis revealed a stratified model based on gender, HOMA-IR, UDFF, and LSM. The incidence of PD was significantly higher among women than men (45.2% vs. 35.7%; node 1). Among women, when HOMA-IR exceeded 6.9 (node 4), the PD incidence was 52.2%, and when UDFF exceeded 12% (node 8), the incidence increased to 75.0%. Among men (node 2), when UDFF exceeded 11% (node 6) and LSM exceeded 6.5 kPa (node 10), the PD incidence reached 80.0% (Fig. 4). In the O-MASLD subgroup, BMI, HOMA-IR, and UDFF were key stratification variables. When BMI exceeded 27 kg/m2, HOMA-IR exceeded 2.4, and UDFF exceeded 9%, the probability of PD occurrence was approximately 72.2% (node 0–node 2–node 6–node 8). In the H-MASLD subgroup, SBP and LSM were the dominant predictive factors. When SBP was ≤140 mm Hg, the probability of PD occurrence was only 8.5% (node 0). However, when SBP exceeded 140 mm Hg and LSM exceeded 6.5 kPa, the probability rose to 66.7% (node 0–node 2–node 6). In the D-MASLD subgroup, HOMA-IR and LSM served as the primary stratification factors. When HOMA-IR exceeded 6.5 and LSM exceeded 6.2 kPa, the probability of PD occurrence reached 70.6% (node 0–node 2–node 6). In the M-MASLD subgroup, BMI, HOMA-IR, UDFF, and LSM formed sequential stratification nodes, indicating that multiple metabolic parameters jointly influence PD incidence. When BMI exceeded 26 kg/m2, UDFF exceeded 10%, and LSM exceeded 7.5 kPa, the probability of PD occurrence was approximately 76.0% (node 0–node 2–node 6–node 10). When SBP exceeded 145 mm Hg and HOMA-IR exceeded 5.9, the probability increased to 84.0% (node 0–node 1–node 4–node 8) (Fig. 5).
Fig. 4.
Decision tree model predicting panvascular disease (PD) occurrence in all metabolic dysfunction-associated steatotic liver disease (MASLD) patients. Gender, homeostasis model assessment of insulin resistance (HOMA-IR) >6.9, and ultrasound-derived fat fraction (UDFF) >11% were primary variables stratifying PD risk. LSM, liver stiffness measurement.
Fig. 5.
Decision tree models predicting panvascular disease (PD) occurrence across metabolic dysfunction-associated steatotic liver disease (MASLD) subgroups. (A) Obese MASLD patients: body mass index (BMI) >27 kg/m2 was the initial node, followed by liver stiffness measurement (LSM) >6.5 kPa, homeostasis model assessment of insulin resistance (HOMA-IR) >2.4, and ultrasound-derived fat fraction (UDFF) >9%. (B) Hypertensive MASLD patients: systolic blood pressure (SBP) >140 mm Hg was the primary variable, with UDFF >11% and LSM >6.5 kPa further stratifying PD risk. (C) Diabetic MASLD patients: HOMA-IR >6.5 and UDFF >11% were critical thresholds for PD risk prediction, followed by LSM >6.2 kPa. (D) MASLD patients with multiple comorbidities: BMI >26 kg/m2 was the initial node, with SBP >145 mm Hg, HOMA-IR >5.9, and LSM >7.5 kPa as subsequent predictors.
Diagnostic efficacy of logistic regression and decision tree prediction for PD occurrence in MASLD patients
We next evaluated the predictive performance of logistic regression and decision tree models for estimating PD risk in MASLD patients. Model discrimination was assessed using the AUC and compared with the DeLong test. The logistic regression model achieved an AUC of 0.80 in the overall MASLD cohort, with the highest performance observed in the M-MASLD subgroup (AUC=0.85). In contrast, the D-MASLD and H-MASLD subgroups had AUCs of 0.82 and 0.80, respectively, while the O-MASLD subgroup showed the lowest AUC of 0.72.
By comparison, the decision tree model demonstrated superior discrimination, with an overall AUC of 0.85 (DeLong test P<0.05). Notably, the model achieved particularly high predictive accuracy in the M-MASLD (AUC=0.95) and D-MASLD (AUC=0.87) subgroups. In the H-MASLD and O-MASLD subgroups, the AUCs were 0.78 and 0.74, respectively (Fig. 6).
Fig. 6.
Receiver operating characteristic (ROC) curves comparing logistic regression and decision tree models for predicting panvascular disease risk in metabolic dysfunction-associated steatotic liver disease (MASLD) patients. (A) Logistic regression model: total MASLD (area under the curve [AUC]=0.80), MASLD with obesity (O-MASLD; AUC=0.72), MASLD with hypertension (H-MASLD; AUC=0.80), MASLD with diabetes (D-MASLD; AUC=0.82), and MASLD with multiple comorbidities (M-MASLD; AUC=0.85). (B) Decision tree model: total MASLD (AUC=0.85), O-MASLD (AUC=0.74), H-MASLD (AUC=0.78), D-MASLD (AUC=0.87), and M-MASLD (AUC=0.95).
DISCUSSION
This study is the first to systematically evaluate clinical indicators, UDFF, and LSM as predictors of PD in patients with MASLD. Both logistic regression and decision tree models were employed to analyze risk factors for PD occurrence across different MASLD subtypes, underscoring the central role of UDFF and LSM in PD risk prediction. The highest incidence of PD was observed in the O-MASLD subgroup (54.5%), followed by the M-MASLD subgroup (42.1%), with the lowest incidence in the H-MASLD subgroup (15.3%). Notably, significant differences in UDFF, LSM, and HOMA-IR across subgroups directly influenced PD risk.
In the overall cohort, the proportion of women was higher in the PD group (41.0%) than in the non-PD group (31.8%). This sex difference is plausibly attributable to sex-specific metabolic regulation. Cardiovascular pathophysiology demonstrates marked sex-based distinctions influenced by hormonal milieu [18]. Women with MASLD tend to exhibit greater hepatic steatosis burden [19], and hormonal effects on insulin resistance may contribute to atherosclerosis, consistent with the established association between insulin resistance and carotid plaque formation [20]. PD patients were also slightly older than non-PD patients, consistent with the greater representation of older individuals and women with diabetes within the hypertensive subgroup.
Across the entire MASLD cohort, both multivariable logistic regression and the decision tree consistently identified UDFF as the strongest predictor of PD. In men, UDFF >11% constituted the primary split and, together with LSM >6.5 kPa, provided clear stratification. In women, UDFF >12% combined with HOMA-IR >6.9 delineated a high-risk subgroup. Subgroup-specific patterns were biologically coherent: UDFF dominated in the O-MASLD subgroup, reflecting the central role of hepatic fat accumulation and adiposity in driving panvascular injury [21]. LSM and SBP were the main determinants in the H-MASLD subgroup. HOMA-IR and LSM predominated in the D-MASLD subgroup, potentially due to increased hepatic glycogen deposition in patients with diabetes [22]. Hyperglycemia and insulin resistance are known to elevate LSM [23], supporting the importance of early insulin resistance control and dynamic LSM monitoring. In the M-MASLD subgroup, UDFF and HOMA-IR, with LSM as an adjunct, jointly influenced PD risk. By DeLong testing, the decision tree achieved a significantly higher AUC than logistic regression for PD prediction in the overall cohort (0.85 vs. 0.80, P<0.05), indicating that key risk relationships are nonlinear and interaction-driven. The decision tree model effectively captured these complex interactions, particularly within composite phenotypes such as M-MASLD (AUC=0.95).
Collectively, these findings support a phenotype-tailored framework for PD risk stratification in MASLD. In the obesity-dominant phenotype, longitudinal UDFF monitoring should be prioritized to track steatosis progression. In the hypertension-dominant phenotype, LSM combined with SBP provides a reliable index of fibrotic and hemodynamic risk. In the diabetes-dominant phenotype, monitoring HOMA-IR and LSM enables early detection of key pathological transitions. In multimorbidity phenotypes, integrating BMI, HOMA-IR, UDFF, and LSM allows quantification of synergistic metabolic risk. Aligning monitoring priorities and therapeutic intensity with these phenotypic profiles is expected to enhance diagnostic precision, guide treatment sequencing, optimize resource utilization, and support dynamic, phenotype-driven care pathways in MASLD.
This prospective study has several limitations. First, it was conducted at a single center with recruitment primarily from the ultrasound department, which may introduce selection bias. Second, most PD events were peripheral; while this distribution is biologically plausible given that systemic atherosclerosis underlies PD, it emphasizes the importance of continued longitudinal surveillance. Third, the 6-month follow-up period limits conclusions regarding long-term cardiovascular outcomes. Early findings are reported because clinically meaningful signals were observed within this timeframe; however, larger multicenter studies with extended follow-up are needed to validate these results.
In conclusion, through logistic regression and decision tree analysis, this study clearly demonstrates the importance of UDFF and LSM in predicting PD risk in MASLD patients, underscoring the profound impact of metabolic dysfunction on disease progression. The results indicate that logistic regression effectively identifies key risk factors, whereas the decision tree provides an intuitive representation of complex, nonlinear relationships. The complementary use of these two analytical approaches offers a robust framework for stratified risk assessment and personalized management of patients with MASLD.
Footnotes
CONFLICTS OF INTEREST
No potential conflict of interest relevant to this article was reported.
AUTHOR CONTRIBUTIONS
Conception or design: X.Q., L.H. Acquisition, analysis, or interpretation of data: X.Q., L.H., X.R., M.T., T.Y., L.X. Drafting the work or revising: X.Q., L.H., H.J., C.P., L.H. Final approval of the manuscript: X.Q., L.H., H.J., X.R., M.T., T.Y., L.X., C.P., L.H.
Supplementary Material
The Classification of Panvascular Disease by Affected Vascular Territories and Corresponding Diagnostic Approaches
The Distribution of Panvascular Disease among Different MASLD Subgroups
Comparison of O-MASLD Patients with and without PD
Comparison of H-MASLD Patients with and without PD
Comparison of D-MASLD Patients with and without PD
Comparison of M-MASLD Patients with and without PD
Proportion of panvascular disease (PD) in each group of metabolic dysfunction-associated steatotic liver disease (MASLD) patients. Non-PD group accounts for 62%; MASLD with obesity (O-MASLD), 13%; MASLD with hypertension (H-MASLD), 3%; MASLD with diabetes (D-MASLD), 6%; MASLD with multiple comorbidities (M-MASLD), 16%.
The proportion of panvascular disease (PD) over time in different group of metabolic dysfunction-associated steatotic liver disease (MASLD) patients. MASLD with obesity (O-MASLD) patients exhibited the most rapid increase, with incidence exceeding 50% by 6 months. MASLD with multiple comorbidities (M-MASLD) patients showed a similarly steep rise, surpassing 40% by 6 months. MASLD with diabetes (D-MASLD) and MASLD with hypertension (H-MASLD) patients demonstrated a more moderate increase in PD incidence. The overall PD trend represents the combined data from all subgroups, illustrating a progressive rise in incidence over time.
overall distribution of panvascular disease (PD) across vascular territories in patients with metabolic dysfunction-associated steatotic liver disease (MASLD). Peripheral arterial disease (49%) and cerebrovascular disease (41%) were the most prevalent, followed by cardiovascular disease (6%), aortic disease (2%), and microvascular complications (2%).
The number and proportion of panvascular disease (PD) in each metabolic dysfunction-associated steatotic liver disease (MASLD) subgroup. At 6 months, MASLD with obesity (O-MASLD) patients had the highest PD proportion (54.5%), whereas MASLD with hypertension (H-MASLD) patients had the lowest (15.3%). MASLD with multiple comorbidities (M-MASLD) patients demonstrated a marked increase, reaching 42.1% at 6 months. The overall PD proportion across all subgroups at 6 months was 39.0%. D-MASLD, MASLD with diabetes.
REFERENCES
- 1. Targher G, Byrne CD, Tilg H. MASLD: a systemic metabolic disorder with cardiovascular and malignant complications. Gut. 2024;73:691–702. doi: 10.1136/gutjnl-2023-330595. [DOI] [PubMed] [Google Scholar]
- 2. Chan WK, Chuah KH, Rajaram RB, Lim LL, Ratnasingam J, Vethakkan SR. Metabolic dysfunction-associated steatotic liver disease (MASLD): a state-of-the-art review. J Obes Metab Syndr. 2023;32:197–213. doi: 10.7570/jomes23052. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Zhou X, Yu L, Zhao Y, Ge J. Panvascular medicine: an emerging discipline focusing on atherosclerotic diseases. Eur Heart J. 2022;43:4528–31. doi: 10.1093/eurheartj/ehac448. [DOI] [PubMed] [Google Scholar]
- 4. Musiałek P, Montauk L, Saugnet A, Micari A, Hopkins LN. The cardio-vascular future of panvascular medicine: the basics. Kardiol Pol. 2019;77:899–901. doi: 10.33963/KP.15034. [DOI] [PubMed] [Google Scholar]
- 5. Gries JJ, Lazarus JV, Brennan PN, Siddiqui MS, Targher G, Lang CC, et al. Interdisciplinary perspectives on the co-management of metabolic dysfunction-associated steatotic liver disease and coronary artery disease. Lancet Gastroenterol Hepatol. 2025;10:82–94. doi: 10.1016/S2468-1253(24)00310-8. [DOI] [PubMed] [Google Scholar]
- 6. Miao L, Targher G, Byrne CD, Cao YY, Zheng MH. Current status and future trends of the global burden of MASLD. Trends Endocrinol Metab. 2024;35:697–707. doi: 10.1016/j.tem.2024.02.007. [DOI] [PubMed] [Google Scholar]
- 7. Yanai H, Adachi H, Hakoshima M, Iida S, Katsuyama H. Metabolic-dysfunction-associated steatotic liver disease-its pathophysiology, association with atherosclerosis and cardiovascular disease, and treatments. Int J Mol Sci. 2023;24:15473. doi: 10.3390/ijms242015473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Verdan S, Torri GB, Marcos VN, Moreira MHS, Defante MLR, Fagundes MDC, et al. Ultrasound-derived fat fraction for diagnosing hepatic steatosis: a systematic review and meta-analysis. Eur Radiol. 2025;35:7421–30. doi: 10.1007/s00330-025-11652-8. [DOI] [PubMed] [Google Scholar]
- 9. Dillman JR, Thapaliya S, Tkach JA, Trout AT. Quantification of hepatic steatosis by ultrasound: prospective comparison with MRI proton density fat fraction as reference standard. AJR Am J Roentgenol. 2022;219:784–91. doi: 10.2214/AJR.22.27878. [DOI] [PubMed] [Google Scholar]
- 10. De Robertis R, Spoto F, Autelitano D, Guagenti D, Olivieri A, Zanutto P, et al. Ultrasound-derived fat fraction for detection of hepatic steatosis and quantification of liver fat content. Radiol Med. 2023;128:1174–80. doi: 10.1007/s11547-023-01693-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Labyed Y, Milkowski A. Novel method for ultrasound-derived fat fraction using an integrated phantom. J Ultrasound Med. 2020;39:2427–38. doi: 10.1002/jum.15364. [DOI] [PubMed] [Google Scholar]
- 12. Kwak M, Kim HS, Jiang ZG, Yeo YH, Trivedi HD, Noureddin M, et al. MASLD/MetALD and mortality in individuals with any cardio-metabolic risk factor: a population-based study with 26.7 years of follow-up. Hepatology. 2025;81:228–37. doi: 10.1097/HEP.0000000000000925. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Li Y, Yang P, Ye J, Xu Q, Wu J, Wang Y. Updated mechanisms of MASLD pathogenesis. Lipids Health Dis. 2024;23:117. doi: 10.1186/s12944-024-02108-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Rinella ME, Lazarus JV, Ratziu V, Francque SM, Sanyal AJ, Kanwal F, et al. A multisociety Delphi consensus statement on new fatty liver disease nomenclature. Hepatology. 2023;78:1966–86. doi: 10.1097/HEP.0000000000000520. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. World Medical Association World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA. 2013;310:2191–4. doi: 10.1001/jama.2013.281053. [DOI] [PubMed] [Google Scholar]
- 16. Chinese Society of Cardiology, Chinese Medical Association, Hypertension Committee of Cross-Straits Medicine Exchange Association, Cardiovascular Disease Prevention and Rehabilitation Committee, Chinese Association of Rehabilitation Medicine Clinical practice guidelines for the management of hypertension in China. Zhonghua Xin Xue Guan Bing Za Zhi. 2024;52:985–1032. doi: 10.3760/cma.j.cn112148-20240709-00377. [DOI] [PubMed] [Google Scholar]
- 17. American Diabetes Association Professional Practice Committee 4. Comprehensive medical evaluation and assessment of comorbidities: standards of care in diabetes-2025. Diabetes Care. 2025;48:S59–85. doi: 10.2337/dc25-S004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Regitz-Zagrosek V, Kararigas G. Mechanistic pathways of sex differences in cardiovascular disease. Physiol Rev. 2017;97:1–37. doi: 10.1152/physrev.00021.2015. [DOI] [PubMed] [Google Scholar]
- 19. Meng L, Wang J, Yang H, Hu Y, Yang Z. Ultrasound-derived fat fraction to assess liver steatosis in obese patients with polycystic ovary syndrome. Clin Exp Med. 2025;25:130. doi: 10.1007/s10238-025-01635-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Ishizaka N, Ishizaka Y, Takahashi E, Unuma T, Tooda E, Nagai R, et al. Association between insulin resistance and carotid arteriosclerosis in subjects with normal fasting glucose and normal glucose tolerance. Arterioscler Thromb Vasc Biol. 2003;23:295–301. doi: 10.1161/01.atv.0000050142.09911.0b. [DOI] [PubMed] [Google Scholar]
- 21. Xu R, Wang Z, Dong J, Yu M, Zhou Y. Lipoprotein(a) and panvascular disease. Lipids Health Dis. 2025;24:186. doi: 10.1186/s12944-025-02600-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Scoditti E, Sabatini S, Carli F, Gastaldelli A. Hepatic glucose metabolism in the steatotic liver. Nat Rev Gastroenterol Hepatol. 2024;21:319–34. doi: 10.1038/s41575-023-00888-8. [DOI] [PubMed] [Google Scholar]
- 23. Han K, Tan K, Shen J, Gu Y, Wang Z, He J, et al. Machine learning models including insulin resistance indexes for predicting liver stiffness in United States population: data from NHANES. Front Public Health. 2022;10:1008794. doi: 10.3389/fpubh.2022.1008794. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
The Classification of Panvascular Disease by Affected Vascular Territories and Corresponding Diagnostic Approaches
The Distribution of Panvascular Disease among Different MASLD Subgroups
Comparison of O-MASLD Patients with and without PD
Comparison of H-MASLD Patients with and without PD
Comparison of D-MASLD Patients with and without PD
Comparison of M-MASLD Patients with and without PD
Proportion of panvascular disease (PD) in each group of metabolic dysfunction-associated steatotic liver disease (MASLD) patients. Non-PD group accounts for 62%; MASLD with obesity (O-MASLD), 13%; MASLD with hypertension (H-MASLD), 3%; MASLD with diabetes (D-MASLD), 6%; MASLD with multiple comorbidities (M-MASLD), 16%.
The proportion of panvascular disease (PD) over time in different group of metabolic dysfunction-associated steatotic liver disease (MASLD) patients. MASLD with obesity (O-MASLD) patients exhibited the most rapid increase, with incidence exceeding 50% by 6 months. MASLD with multiple comorbidities (M-MASLD) patients showed a similarly steep rise, surpassing 40% by 6 months. MASLD with diabetes (D-MASLD) and MASLD with hypertension (H-MASLD) patients demonstrated a more moderate increase in PD incidence. The overall PD trend represents the combined data from all subgroups, illustrating a progressive rise in incidence over time.
overall distribution of panvascular disease (PD) across vascular territories in patients with metabolic dysfunction-associated steatotic liver disease (MASLD). Peripheral arterial disease (49%) and cerebrovascular disease (41%) were the most prevalent, followed by cardiovascular disease (6%), aortic disease (2%), and microvascular complications (2%).
The number and proportion of panvascular disease (PD) in each metabolic dysfunction-associated steatotic liver disease (MASLD) subgroup. At 6 months, MASLD with obesity (O-MASLD) patients had the highest PD proportion (54.5%), whereas MASLD with hypertension (H-MASLD) patients had the lowest (15.3%). MASLD with multiple comorbidities (M-MASLD) patients demonstrated a marked increase, reaching 42.1% at 6 months. The overall PD proportion across all subgroups at 6 months was 39.0%. D-MASLD, MASLD with diabetes.