Cardiac remodelling in novel cluster-based metabolic (dysfunction)-associated fatty liver disease: a cross-sectional study in China

Abstract

Objectives

To identify novel, data-driven phenotypic clusters of metabolic dysfunction-associated fatty liver disease (MAFLD) and investigate their associations with cardiac remodelling.

Design

Cross-sectional study.

Setting

Secondary care; a single-centre study in China.

Participants

A total of 3233 participants diagnosed with MAFLD were included in the study. The diagnosis was conducted in accordance with the established criteria for MAFLD. The exclusion criteria encompassed a history of significant cardiovascular or hepatic diseases, as well as excessive alcohol consumption.

Primary outcome measures

Echocardiographic parameters of cardiac structure and function.

Results

Four distinct clusters were identified. Cluster 1 (n=1381) comprised men with a normal metabolic state and low Fibrosis-4 Index (FIB-4) levels. Cluster 2 (n=453) included men with the highest body mass index (BMI) and uric acid levels. Cluster 3 (n=474) consisted of men with the most severe glucose and lipid metabolic disturbances. Cluster 4 (n=925) comprised women with the highest FIB-4 levels. Compared with Cluster 1, participants in Clusters 2 and 3 exhibited worse cardiac structure and function, including enlargement of the left atrium (LA), right atrium, left ventricular internal end-diastolic dimension, interventricular septum (IVS), left ventricular posterior wall (LVPW) thickness, right ventricular and reduced left ventricular ejection fraction. Conversely, participants in Cluster 4 had better cardiac structure and function compared with those in Cluster 1. After adjusting for confounders, Cluster 2 showed positive associations with the LA, IVS, LVPW, right atrium and right ventricular (all p<0.001), while Cluster 4 showed negative associations.

Conclusions

The heterogeneity of MAFLD reveals sex-specific patterns of cardiac remodelling: male phenotypes with high BMI and uric acid levels are associated with worse remodelling, whereas a female phenotype characterised by high FIB-4 levels correlates with preserved cardiac function.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The application of an innovative, data-driven cluster analysis methodology provides an unbiased approach to identify distinct metabolic phenotypes within the metabolic (dysfunction)-associated fatty liver disease (MAFLD) population, offering significant methodological innovation.

• The cross-sectional design of the study inherently limits the ability to establish causal relationships between the identified metabolic clusters and outcomes of cardiac remodelling.

• The diagnosis of MAFLD was based on ultrasound imaging rather than the gold-standard liver biopsy, which may have led to misclassification.

• The potential for unmeasured confounding exists due to the lack of comprehensive historical medical data and detailed medication records for all participants.

Introduction

Metabolic (dysfunction)-associated fatty liver disease (MAFLD) is characterised as a liver condition associated with systemic metabolic disorders.¹ With a global prevalence of 25%, MAFLD has emerged as the most prevalent chronic liver disease worldwide.² Various factors, including age, gender, metabolic state, diet, genetic background and gut microbiota, influence both the onset and progression of MAFLD and its associated complications.³

Despite MAFLD having significant associations with cardiovascular risk factors such as obesity, type 2 diabetes mellitus and hyperlipidaemia,³ recent studies indicate that MAFLD independently predicts adverse cardiovascular outcomes.4,7 Therefore, it is crucial to comprehensively assess cardiovascular risk when managing MAFLD. However, research on the correlation between MAFLD and cardiac function and structure remains insufficient. Given that early-stage cardiac abnormalities often lack evident clinical manifestations, early identification of specific MAFLD subtypes is important.

Currently, three phenotypes are used to manage MAFLD: obesity-associated, diabetes-associated and lean MAFLD.¹ This classification aids in reducing the heterogeneity of the disease and more accurately reflects liver steatosis and related metabolic characteristics across various patient groups. However, due to the varying degrees of pathophysiological characteristics and metabolic abnormalities observed in patients with MAFLD, simplifying the condition into three subtypes may not comprehensively address all scenarios. Additionally, some patients may exhibit a combination of characteristics, rendering it challenging for a single classification method to accurately describe their pathological condition. Significantly, circulating metabolites, such as lipids, glucose and uric acid, directly contribute to the pathogenesis of extrahepatic complications in MAFLD, particularly in the development of cardiovascular disease (CVD).⁸ However, the current phenotypes of MAFLD have not fully incorporated these metabolic parameters, thereby limiting clinicians' ability to optimise treatment through early metabolic control.

Therefore, this study aims to establish a novel MAFLD classification system and evaluate the differences in cardiac structure and function across various subtypes through Doppler echocardiography. The goal is to identify patients at the highest risk of CVD and to enhance clinical management and individualised treatment strategies.

Methods

Study design and population

This is a cross-sectional single centre study. A data-driven approach based on a comprehensive non-invasive assessment was initially used to identify the subtypes of MAFLD. The model was modified based on the current MAFLD classification, with a particular focus on serological indicators reflecting metabolic status, markers indicating the severity of liver fibrosis and variables possessing significant prognostic value. The characteristics of cardiac remodelling in various MAFLD subtypes were subsequently investigated, and the cardiovascular risk over the following decade was evaluated using the Framingham Risk Score.

This cross-sectional study enrolled participants who underwent both hepatic ultrasonography and echocardiography at the Health Management Center, the Third Affiliated Hospital of Sun Yat-sen University from January 2019 to December 2023. All participants were diagnosed with MAFLD based on standard clinical criteria. The exclusion criteria included: (1) alcohol abusers; (2) patients infected with chronic hepatitis virus; (3) patients with overt heart disease, including those with overt heart failure classified as New York Heart Association functional class III and IV, a left ventricular ejection fraction (LVEF) less than 50%, a history of primary cardiomyopathy, severe valvulopathy and chronic atrial fibrillation and (4) patients with severe hepatic and renal insufficiency. In total, 3233 participants were included for analysis in the present study. The schematic of the study design is shown in figure 1.

Figure 1. Schematic of the study design. ANOVA, analysis of variance; BMI, body mass index; FIB-4, Fibrosis-4 Index; HbA1c, glycated haemoglobin; MAFLD, metabolic dysfunction-associated fatty liver disease; TC, total cholesterol; TG, triglycerides; TyG, Triglyceride-Glucose Index; UA, uric acid.

Patient and public involvement

Formal patient and public involvement was not implemented in this study.

Laboratory measurements

Baseline characteristics were collected from all participants, including age, gender, blood pressure and body mass index (BMI). Blood samples were obtained following an 8-hour fast. The following liver function parameters were assessed: alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, albumin, alkaline phosphatase (ALP), gamma-glutamyl transpeptidase (GGT) and lipid profile. Other biochemical parameters, including fasting blood glucose (FBG), uric acid (UA), serum creatinine, blood urea nitrogen and homocysteine (Hcy), were analysed using the Hitachi 7180 chemical analyzer (Hitachi High-Tech Corp, Japan). The estimated glomerular filtration rate (eGFR) was calculated based on the chronic kidney disease epidemiology collaboration (CKD-EPI) equation. Glycosylated haemoglobin A1c (HbA1c), thyroxine and N-terminal pro-B-type natriuretic peptide levels were determined using a chemiluminescence immunoassay analyser, while haemoglobin, white cell count and platelets were measured with a blood cell analyser. Triglyceride and glucose (TyG) index was calculated as ln(triglyceride×FBG), and remnant cholesterol was estimated as the total cholesterol (TC) minus high-density lipoprotein cholesterol (HDL-C) minus low-density lipoprotein cholesterol.

Four indicators were used to assess inflammatory status: the neutrophil-to-lymphocyte ratio, calculated as neutrophil count/lymphocyte count; Systemic Immune-Inflammation Index, calculated as (platelet count×neutrophil count)/lymphocyte count; Systemic Inflammation Response Index, calculated as (neutrophil count×monocyte count)/lymphocyte count and Systemic Inflammatory and Immune Response Index, calculated as (neutrophil count×monocyte count×platelet count)/lymphocyte count. Additionally, the Fibrosis-4 Index (FIB-4), which reflects the progression of liver fibrosis, was measured as mentioned before.⁹

Definition of MAFLD

The diagnosis of MAFLD is based on the confirmation of liver fat accumulation, which is obtained from imaging studies and supported by evidence from blood biomarkers, along with the presence of at least one of the following three conditions: overweight or obesity, type 2 diabetes or metabolic dysregulation.¹ Fatty liver is evaluated through ultrasonography, characterised by the presence of abnormal liver echogenicity, alterations in liver size and contour and hepatic vascular architecture.¹⁰

Measurements of cardiac remodelling

Cardiac remodelling was evaluated using echocardiographic measurements conducted by trained sonographers at each participating site, following established protocols. The assessment of left heart structure included parameters such as aortic root diameter, left atrial volume, left ventricular end-diastolic diameter (LVDd), interventricular septum (IVS) thickness and left ventricular posterior wall (LVPW) thickness. The right heart structure was primarily evaluated based on the diameters of the right atrium and right ventricle.

Left ventricular systolic function was assessed using the LVEF. Diastolic function was evaluated through mitral valve diastolic flow velocities (peak E and peak A), the early diastolic movement velocity of the lateral and interstitial mitral ring walls (e′), and the E/e′ ratio of the mitral valve.

Measurements of Framingham Risk Score

The Framingham Risk Score incorporates several risk factors, including age, gender, TC, HDL-C, systolic blood pressure (SBP), smoking status and the presence of diabetes. By evaluating these factors, the Framingham Risk Score categorises an individual’s cardiovascular risk into low, intermediate or high tiers, assessing the likelihood of developing cardiovascular disease within the next 10 years.¹¹

Statistical analysis

Two-step K-means clustering algorithm was employed for clustering analysis.¹² Clinical indicators used for clustering included gender, BMI, HbA1c, UA, triglycerides (TG), TC, TyG and FIB-4.

The characteristics of patients with MAFLD were presented as mean±SD for quantitative parameters and as percentages for categorical variables. An analysis of variance test was employed for continuous variables with normal distribution, while the χ² test was applied to categorical variables. The Kruskal-Wallis H test was used to compare baseline differences among the different clusters.

Univariate and multivariate linear regression models were constructed to estimate the correlations between echocardiographic parameters of cardiac structure and function among various MAFLD subtypes and the Framingham Risk Score, with results presented as ORs (β) and 95% CIs. A multivariate linear regression model was adjusted for age, heart rate (HR), SBP, diastolic blood pressure, AST, ALT, GGT, ALP, eGFR, Hcy and lymphocyte-to-neutrophil ratio in the second hospitalisation. A two-tailed p value <0.05 was considered statistically significant. All statistical analyses were performed using SPSS.

Results

Cluster analysis of patients with MAFLD

Ultimately, 3233 patients diagnosed with MAFLD were included in the cluster analysis and categorised into four clusters (figure 2). Cluster 1 comprises 1381 individuals (42.7%), characterised by men with a relatively normal metabolic state and low FIB-4 levels. Cluster 2 includes 453 patients (14.5%), primarily distinguished by elevated UA, high BMI and slightly increased cholesterol levels relative to the other clusters. Cluster 3 consists of 474 patients (14.7%), who exhibit relatively high TyG, HbA1c and TG levels, indicating a higher insulin resistance. Cluster 4 encompasses 925 patients (28.6%), marked by a higher proportion of female patients and elevated FIB-4, suggesting a greater degree of hepatic fibrosis in this group (figure 3). Table 1 presents detailed baseline characteristics of the serological indices for these four clusters.

Figure 2. Distribution of four cluster subtypes.

Figure 3. Characteristics of each subtype. BMI, body mass index; FIB-4, Fibrosis-4 Index; HbA1c, haemoglobin A1c; TC, total cholesterol; TG, triglycerides; TyG, triglycerides to glucose ratio; UA, uric acid.

Table 1. Basal characteristics.

Variables	Total (n=3233)	Cluster 1 (n=1381)	Cluster 2 (n=453)	Cluster 3 (n=474)	Cluster 4 (n=925)	P value
Age, year, mean±SD	48.25±11.82	45.19±11.37	42.41±10.21	52.69±10.35	53.40±11.08	<0.001
Male, n (%)	2308 (71.39)	1381 (100.00)	453 (100.00)	470 (99.16)	4 (0.43)	<0.001
HR, bpm, mean±SD	79.36±12.48	80.02±12.40	78.51±12.58	80.72±12.01	79.60±12.53	0.007
BMI, kg/m2, mean±SD	26.01±3.21	24.82±1.99	30.08±2.05	26.85±3.36	25.37±3.35	<0.001
SBP, mm Hg, mean±SD	128.55±16.81	125.58±15.51	132.10±13.64	132.07±17.19	129.44±18.99	<0.001
DBP, mm Hg, mean±SD	81.36±17.47	81.18±23.69	84.99±9.35	83.80±11.06	78.60±10.30	<0.001
WBC, ×109 /L, mean±SD	6.66±3.34	6.52±1.60	6.85±1.54	7.38±4.36	6.40±4.90	<0.001
Hb, g/L, mean±SD	149.78±14.88	154.92±11.44	158.26±9.93	155.53±12.40	134.99±11.88	<0.001
PLT, ×109 /L, mean±SD	253.36±64.56	247.63±55.56	248.88±55.53	242.97±54.77	269.42±80.94	<0.001
ALT, U/L, mean±SD	37.73±33.21	38.59±27.76	44.53±30.88	40.06±42.06	31.92±35.69	<0.001
AST, U/L, mean±SD	27.92±19.39	27.62±15.39	29.00±14.91	28.62±26.70	27.49±22.07	0.423
GGT, U/L, mean±SD	50.92±54.95	53.24±52.35	57.83±61.05	64.34±78.92	37.36±34.19	<0.001
ALP, U/L, mean±SD	76.13±21.78	74.28±20.09	74.17±18.82	77.25±22.93	79.49±24.58	<0.001
Total bilirubin, µmol/L, mean±SD	11.37±5.05	12.15±4.80	11.81±4.84	11.38±5.47	9.94±5.00	<0.001
Albumin, g/L, mean±SD	46.79±2.89	47.41±2.79	47.27±2.79	46.74±2.95	45.54±2.69	<0.001
Creatinine, µmol/L, mean±SD	70.26±17.14	76.39±13.39	76.27±13.76	74.73±16.26	55.88±15.53	<0.001
BUN, µmol/L, mean±SD	5.22±1.27	5.23±1.16	5.18±1.18	5.39±1.23	5.13±1.46	0.002
UA, µmol/L, mean±SD	431.98±103.07	455.24±89.46	495.45±98.41	441.62±99.09	361.25±87.20	<0.001
eGFR, mL/min/1.73 m2, mean±SD	101.93±15.94	102.44±15.70	104.81±16.12	98.51±16.52	101.58±15.61	<0.001
TC, mmol/L, mean±SD	5.58±1.09	5.50±1.00	5.52±1.01	5.70±1.36	5.66±1.09	<0.001
HDL-C, mmol/L, mean±SD	1.17±0.28	1.16±0.25	1.06±0.20	1.05±0.21	1.31±0.32	<0.001
LDL-C, mmol/L, mean±SD	3.49±0.94	3.54±0.88	3.51±0.87	3.31±1.09	3.50±0.95	<0.001
TG, mmol/L, mean±SD	2.11±1.87	1.83±0.92	2.33±1.38	3.51±3.92	1.72±0.93	<0.001
Apolipoprotein A, g/L, mean±SD	1.45±0.23	1.42±0.22	1.39±0.19	1.39±0.18	1.57±0.25	<0.001
Apolipoprotein B100, g/L, mean±SD	1.09±0.25	1.08±0.24	1.10±0.23	1.12±0.26	1.09±0.26	0.089
Lipoprotein a, mg/L, mean±SD	223.38±235.67	208.38±213.39	203.40±230.77	209.10±239.32	263.67±262.00	<0.001
Remnant cholesterol, mmol/L, mean±SD	0.92±0.70	0.80±0.44	0.96±0.61	1.35±1.32	0.85±0.50	<0.001
FBG, mmol/L, mean±SD	5.65±1.51	5.18±0.57	5.29±0.57	7.31±2.61	5.67±1.43	<0.001
HbA1C, %, mean±SD	5.79±0.95	5.43±0.35	5.48±0.33	6.98±1.47	5.88±0.89	<0.001
TyG, mean±SD	8.95±0.63	8.81±0.49	9.04±0.54	9.53±0.83	8.82±0.55	<0.001
TyG-BMI, mean±SD	233.26±36.08	218.85±23.19	272.10±24.66	255.90±39.40	224.15±35.54	<0.001
Homocysteine, µmol/L, mean±SD	14.19±6.58	15.26±7.42	15.46±6.65	14.43±5.60	11.66±4.64	<0.001
TSH, µIU/mL, mean±SD	1.96±2.56	1.83±2.30	1.82±1.01	1.80±1.17	2.31±3.78	<0.001
Free thyroxine, pmol/L, mean±SD	12.95±1.89	13.08±1.76	12.96±1.40	13.02±1.83	12.67±2.28	<0.001
Free triiodothyronine, pmol/L, mean±SD	4.54±0.94	4.65±0.93	4.66±0.54	4.52±0.75	4.33±1.15	<0.001
FIB-4, mean±SD	0.19±0.14	0.17±0.10	0.14±0.09	0.20±0.11	0.23±0.19	<0.001
APRI, mean±SD	0.30±0.30	0.30±0.26	0.31±0.26	0.32±0.35	0.29±0.33	0.264
NLR, mean±SD	1.75±0.73	1.75±0.77	1.73±0.57	1.77±0.71	1.75±0.73	0.835
SII, mean±SD	448.19±237.21	436.25±220.46	435.32±193.56	435.28±219.13	478.94±282.85	<0.001
SIRI, mean±SD	0.81±0.54	0.83±0.59	0.84±0.41	0.92±0.56	0.70±0.49	<0.001
SIAII, mean±SD	208.99±168.86	208.68±159.15	215.89±135.51	229.96±171.43	195.33±193.50	0.003
LNSII, mean±SD	5.99±0.47	5.98±0.44	5.99±0.41	5.97±0.47	6.04±0.54	0.015
NT-proBNP, pg/mL, mean±SD	43.78±55.88	35.94±37.25	89.49±105.09	18.54±9.28	18.50±NA	0.710

Bold values indicate statistical significance.

ALP, alkaline phosphatase; ALT, alanine aminotransferase; APRI, AST to Platelet Ratio Index; AST, aspartate aminotransferase; BMI, body mass index; BUN, blood urea nitrogen; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; FBG, fasting blood glucose; FIB-4, Fibrosis-4 Index; GGT, gamma-glutamyl transferase; Hb, haemoglobin; HbA1c, glycated haemoglobin A1c; HDL-C, high-density lipoprotein cholesterol; HR, heart rate; LDL-C, low-density lipoprotein cholesterol; LNSII, lymphocyte-to-neutrophil ratio in the second hospitalisation; NLR, neutrophil-to-lymphocyte ratio; NT-proBNP, N-terminal pro-B-type natriuretic peptide; PLT, platelet count; SBP, systolic blood pressure; SIAII, Systemic Inflammatory and Immune Response Index; SII, Systemic Immune-Inflammation Index; SIRI, Systemic Immune Response Index; TC, total cholesterol; TG, triglycerides; TSH, thyroid stimulating hormone; TyG, triglyceride and glucose index; UA, uric acid; WBC, white cell count.

Cardiac remodelling and the Framingham Risk Score in MAFLD subtypes

Table 2 presents the differences among the four clusters in terms of cardiac structure, functional parameters and Framingham Risk Scores. Compared with the other groups, Cluster 2 exhibited significant cardiac structural remodelling, characterised by larger atrial and ventricular volumes or diameters, as well as IVS and LVPW (all p<0.001). Cluster 4 showed a lower degree of atrial remodelling (all p<0.001) with the exception of the IVS, whereas Cluster 3 showed a significantly reduced IVS. In terms of cardiac function, Cluster 4 had the highest LVEF. Furthermore, compared with the other two clusters, both Cluster 3 and Cluster 4 had lower E/A ratios and higher E/e′ ratios, indicating impaired diastolic function in these groups. Regarding Framingham Risk Scores, Cluster 2 contained a higher proportion of low-risk individuals, whereas high-risk individuals were predominantly concentrated in Cluster 3.

Table 2. Cardiac structure and function.

Variables	Total (n=3233)	Cluster 1 (n=1381)	Cluster 2 (n=453)	Cluster 3 (n=474)	Cluster 4 (n=925)	P value
Cardiac structure
Aortic root, mm, mean±SD	26.75±2.92	26.83±2.87	27.80±2.73	27.97±2.87	25.48±2.58	<0.001
Left atrium, mm, mean±SD	30.55±3.42	30.19±3.27	32.34±3.08	31.57±3.56	29.67±3.27	<0.001
LVDd, mm, mean±SD	44.80±4.15	45.09±4.06	46.74±4.12	45.32±4.07	43.15±3.74	<0.001
IVS, mm, mean±SD	9.90±1.36	9.86±1.30	10.38±1.21	0.42±1.22	9.45±1.42	<0.001
LVPW, mm, mean±SD	9.21±1.31	9.19±1.47	9.61±1.07	9.60±1.13	8.83±1.12	<0.001
Right atrium upper and lower diameter, mm, mean±SD	41.73±4.51	41.69±4.46	43.56±4.39	42.40±4.33	40.56±4.37	<0.001
Right atrium left and right diameter, mm, mean±SD	32.11±6.63	32.43±3.98	33.35±4.21	32.31±4.06	30.93±10.49	<0.001
Right ventricular anteroposterior diameter, mm, mean±SD	21.84±3.04	21.90±3.06	22.98±2.87	22.22±2.88	20.99±2.93	<0.001
Cardiac function
LVEF, %, mean±SD	67.24±4.62	67.17±4.50	66.84±4.61	66.54±4.99	67.90±4.53	<0.001
E/A ratio, mean±SD	1.04±0.40	1.11±0.37	1.09±0.52	0.92±0.40	0.97±0.34	<0.001
E/E′ ratio, mean±SD	9.07±3.16	8.39±3.17	8.94±2.37	9.59±3.00	9.88±3.32	<0.001
Framingham classification, n (%)						<0.001
Low risk	2610 (81.18)	1230 (89.52)	419 (92.70)	221 (47.02)	740 (80.52)
Medium risk	448 (13.93)	119 (8.66)	28 (6.19)	162 (34.47)	139 (15.13)
High risk	157 (4.88)	25 (1.82)	5 (1.11)	87 (18.51)	40 (4.35)
Framingham score, mean±SD	6.48±6.51	4.89±4.92	4.46±3.76	12.68±9.29	6.68±5.98	<0.001

Bold values indicate statistical significance.

E/A ratio, early/atrial filling ratio; E/e′ ratio, ejection fraction E′ ratio; IVS, interventricular septum; LVDd, left ventricular diastolic diameter; LVEF, left ventricular ejection fraction; LVPW, left ventricular posterior wall.

Linear regression analyses between MAFLD subtypes and cardiac remodelling and Framingham Risk Score

As shown in table 3, in the univariate linear regression analysis, compared with Cluster 1, Cluster 2 and Cluster 3 exhibited a significant increase in the structural remodelling of the left atrium, IVS and LVPW, while Cluster 4 showed a negative change in cardiac parameters (all p<0.001). In the multivariate model, the conclusions remained consistent, except for Cluster 3, which showed no significant correlation with the IVS (β (95% CI): 0.18 (−0.01 to 0.36); p=0.062), LVPW (β (95% CI): 0.15 (−0.01 to 0.31); p=0.073). In addition, Cluster 2 exhibited an increase in right atrial diameter while Cluster 4 showed decreased correlation. As for other cardiac function assessment, although Cluster 3 and Cluster 4 displayed significant correlations with LVEF and E/A ratio in the univariate linear regression, these associations were not observed after adjustment. Regarding the Framingham Risk Score, Cluster 3 was associated with a higher risk compared with Cluster 1 (β (95% CI): 5.06 (4.49 to 5.62); p<0.001), while Cluster 4 showed a lower risk (β (95% CI): −1.33 (−1.81 to −0.84); p<0.001).

Table 3. Linear regression analysis assessing the relationships of MAFLD subtype with the cardiac structural and functional parameters.

Variables	Univariate		Multivariate
	β (95% CI)	P value	β (95% CI)	P value
Left cardiac structure
Left atrium, mm
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	2.14 (1.79 to 2.49)	<0.001	2.07 (1.61 to 2.52)	<0.001
Cluster 3	1.37 (1.03 to 1.71)	<0.001	0.85 (0.38 to 1.32)	<0.001
Cluster 4	−0.52 (−0.80 to −0.25)	<0.001	−1.09 (−1.50 to −0.69)	<0.001
IVS, mm
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	0.52 (0.38 to 0.66)	<0.001	0.48 (0.30 to 0.65)	<0.001
Cluster 3	0.56 (0.42 to 0.69)	<0.001	0.18 (−0.01 to 0.36)	0.062
Cluster 4	−0.41 (−0.52 to −0.30)	<0.001	−0.65 (−0.80 to −0.49)	<0.001
LVPW, mm
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	0.44 (0.32 to 0.56)	<0.001	0.42 (0.27 to 0.57)	<0.001
Cluster 3	0.43 (0.32 to 0.55)	<0.001	0.15 (−0.01 to 0.31)	0.073
Cluster 4	−0.33 (−0.42 to −0.24)	<0.001	−0.46 (−0.60 to −0.33)	<0.001
Right cardiac structure
Right atrium upper and lower diameter, mm
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	1.87 (1.40 to 2.34)	<0.001	2.05 (1.44 to 2.65)	<0.001
Cluster 3	0.71 (0.25 to 1.17)	0.003	0.29 (−0.34 to 0.92)	0.366
Cluster 4	−1.13 (−1.49 to −0.76)	<0.001	−1.39 (−1.93 to −0.85)	<0.001
Right atrium left and right diameter, mm
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	0.93 (0.23 to 1.63)	0.010	1.28 (0.73 to 1.83)	<0.001
Cluster 3	−0.11 (−0.80 to 0.58)	0.745	0.29 (−0.28 to 0.87)	0.319
Cluster 4	−1.50 (−2.05 to −0.95)	<0.001	−1.82 (−2.32 to −1.33)	<0.001
Right ventricular, mm
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	1.09 (0.77 to 1.40)	<0.001	1.07 (0.66 to 1.48)	<0.001
Cluster 3	0.33 (0.02 to 0.64)	0.040	0.23 (−0.20 to 0.66)	0.295
Cluster 4	−0.91 (−1.16 to −0.66)	<0.001	−1.23 (−1.59 to −0.86)	<0.001
Cardiac function
LVEF, %
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	−0.33 (−0.82 to 0.16)	0.187	−0.37 (−1.03 to 0.29)	0.276
Cluster 3	−0.63 (−1.11 to −0.15)	0.010	−0.29 (−0.98 to 0.40)	0.407
Cluster 4	0.74 (0.35 to 1.12)	<0.001	0.51 (−0.08 to 1.11)	0.090
E/A ratio
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	−0.02 (−0.06 to 0.02)	0.396	−0.00 (−0.05 to 0.05)	0.878
Cluster 3	−0.19 (−0.23 to −0.15)	<0.001	−0.05 (−0.10 to 0.01)	0.082
Cluster 4	−0.13 (−0.17 to −0.10)	<0.001	0.02 (−0.03 to 0.06)	0.424
Framingham scores
Cluster 1	0.00 (Reference)		0.00 (Reference)
Cluster 2	−0.43 (−1.06 to 0.20)	0.185	−0.40 (−0.94 to 0.14)	0.150
Cluster 3	7.79 (7.16 to 8.41)	<0.001	5.06 (4.49 to 5.62)	<0.001
Cluster 4	1.79 (1.30 to 2.29)	<0.001	−1.33 (−1.81 to −0.84)	<0.001

Bold values indicate statistical significance.

Multivariate linear regression Model was adjusted for age, HR, SBP, DBP, AST, ALT, GGT, ALP, eGFR, Hcy, and LnSII.

ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; DBP, diastolic blood pressure; E/A ratio, early/atrial filling ratio; eGFR, estimated glomerular filtration rate; GGT, gamma-glutamyl transferase; Hcy, homocysteine; HR, heart rate; IVS, interventricular septum; LnSII, (log-transformed) systemic immune-inflammation index; LVEF, left ventricular ejection fraction; LVPW, left ventricular posterior wall; MFALD, metabolic (dysfunction)-associated fatty liver disease; SBP, systolic blood pressure; SE, standard error.

Discussion

This study investigated the relationship between data-driven MAFLD typing and cardiac remodelling. Ultimately, we identified four distinct subtypes of MAFLD and revealed that men with higher BMI and UA levels were associated with more pronounced cardiac remodelling. Conversely, a cluster predominantly consisting of women with a higher FIB-4 index exhibited better heart structure and function. These findings contribute not only to our understanding of the link between MAFLD and cardiac health but also offer new perspectives and directions for future clinical diagnosis and treatment.

Inflammation, oxidative stress and metabolic disorders in MAFLD are proposed to promote changes in cardiac structure by activating specific signalling pathways,¹³ thereby potentially affecting cardiac function¹⁴ and the calcium handling capacity of cardiomyocytes.^{15 16} Consequently, patients with MAFLD characterised by elevated lipid levels and insulin resistance in our study exhibited an increased risk of cardiovascular diseases, as indicated by a higher Framingham Risk Score. However, our current findings suggest that the correlation between MAFLD subtypes and cardiac structure does not necessarily translate into immediate cardiac functional impairment. This could be attributed to the early compensatory mechanisms of the cardiac structure, suggesting the possibility that persistent metabolic stress could lead to a progressive deterioration of both cardiac structure and function over time.^{17 18} Therefore, long-term cardiac monitoring for patients with MAFLD, as well as the adoption of individualised treatment strategies, including lifestyle improvements and pharmacological interventions, could be crucial for preventing the occurrence and progression of cardiovascular diseases.

Our research has revealed a significant correlation in patients with MAFLD where men with the highest BMI and UA levels exhibit a strong association with adverse cardiac remodelling. This finding suggests that the impact of BMI on the liver-heart axis is not solely mediated by disruptions in lipid metabolism; rather, it appears to be a consequence of direct fat deposition. As we know, an elevated BMI likely points to a pronounced expansion of body adipose tissue,¹⁹ which can lead to excessive accumulation of fat within hepatocytes.²⁰ The fat deposition in the liver may disrupt the energy metabolism of myocardium by releasing adipokines and proinflammatory cytokines. Moreover, hepatic steatosis may indicate ectopic fat deposition in the myocardium and pericardium.²¹ Studies have demonstrated that an increase in epicardial adipose tissue, which is associated with obesity or fatty liver disease, can trigger a transformation in its biological properties.²² The latter promotes the production of inflammatory cytokines, which initiate inflammatory responses and fibrotic processes, thereby inducing coronary artery diseases, chronic heart failure and atrial fibrillation.^{23 24 25}

Furthermore, visceral adipose tissue supplies free fatty acids to the liver and induces UA production.²⁶ Hyperuricaemia impacts the cardiovascular system primarily through the deposition of UA on vascular walls, potentially resulting in damage to the vascular intima and endothelial function. The activation of UA crystals initially enhances platelet aggregation, which could increase the risk of arterial wall hardening and thrombosis.^{27 28} As the levels of UA rise, they contribute to vasoconstriction, intensify inflammatory responses and increase the production of reactive oxygen species.^{29 30} These combined effects further accelerate the proliferation of vascular smooth muscle cells and promote the progression of atherosclerosis, thereby significantly elevating the risk of cardiovascular diseases.³¹

Interestingly, our observations have highlighted a notable disparity in the subtypes of MAFLD among different genders. Previous studies have indicated that lipid and insulin-related pathways and inflammatory processes in adipose and liver tissues seem to play a more prominent role in nonalcoholic fatty liver disease (NAFLD) among men,^{32 33} which is consistent with our observation of more pronounced metabolic dysregulations associated with MAFLD in the male population. However, in contrast to previous studies,34,38 our results indicate that female patients with MAFLD present with more severe liver fibrosis but demonstrate improved cardiac structure and function, as well as a lower risk of cardiovascular disease. Therefore, the elevated FIB-4 index observed in Cluster 4 is likely principally driven by the older age of this subgroup, rather than by clinically significant hepatic inflammation or damage. Moreover, this cluster exhibited multiple metabolically protective features, including the highest levels of HDL-C, the lowest TG levels, optimal glycaemic control and minimal inflammation. These factors together may explain the comparatively maintained cardiac function observed in these individuals. Besides, this cluster may suggest that oestrogen could exert a protective effect on the cardiovascular system by influencing vascular endothelial function,³⁹ lowering blood pressure⁴⁰ and improving lipid metabolism.⁴¹ Additionally, compared with men, women may accumulate less visceral fat even in the presence of liver fibrosis,⁴² which could result in a reduced impact on the heart. Furthermore, women often exhibit a greater awareness of self-health management,⁴³ which enables them to achieve early detection and intervention. Finally, it is also important to acknowledge that our derived clusters demonstrated considerable overlap with sex distribution, which could raise concerns regarding potential confounding. However, the clustering captured significant metabolic heterogeneity beyond these demographic influences. Specifically, even within the sex-homogeneous subgroups—particularly among men—clusters exhibited clinically meaningful distinctions. The results suggest that while sex influences the overall structure of metabolic variation, the clustering method has identified intrinsic subgroups that may indicate distinct metabolic risk trajectories, independent of sex.

There exist some limitations in our study. Primarily, MAFLD was primarily diagnosed using ultrasound, which is less sensitive than liver biopsy or elastography; this lower sensitivity may have resulted in underdiagnosis. Second, the lack of granular data on comorbidities and medications precluded a full adjustment for confounding. For instance, antihypertensive medications (eg, ACEI, angiotensin receptor blocker (ARB)) that promote reverse remodelling, potentially confounding LVDd and left ventricular mass index (LVMI) interpretations, and antidiabetic agents (eg, sodium-glucose cotransporter inhibitors) with known cardioprotective effects that may bias diastolic function parameters. Therefore, it remains possible that some of the observed phenotypic differences between clusters are attributable to these unmeasured treatment variables rather than to the MAFLD subtypes themselves. Third, although we used the Framingham Risk Score to estimate participants’ cardiovascular risk, the accuracy of these predictions must be confirmed in large, long-term prospective studies. Finally, the generalisability of the clustering results in this study requires further validation in an external cohort.

Conclusion

The novel classification method developed in the study offers a more refined approach to distinguishing the characteristics of patients with MAFLD and predicting the risk of cardiac remodelling across various subtypes. Our results demonstrate a correlation between elevated BMI and UA levels in men and adverse cardiac remodelling, while in women, the highest FIB-4 index is associated with superior cardiac structure and function. These insights have the potential to inform the development of personalised management strategies in patients with MAFLD.

Data availability statement

No data are available.