Statistical analyses

All analyses were weighted to account for the complex sampling design using STATA version 14.0 (StataCorp LP, College Station, Texas, USA). The χ² test and one-way analysis of variance were employed to compare the baseline characteristics of participants. A post hoc multiple comparison analysis was performed with Bonferroni correction. The Kaplan–Meier method was used to estimate the overall survival of subjects with MAFLD, stratified by disease subtypes (diabetes, overweight/obesity, and lean metabolic disorder). The log-rank test was performed to compare survival distributions between the groups.

Overall mortality rates were determined and reported as deaths per 1000 person-years at risk. Cox proportional hazard models were used to estimate adjusted hazard ratios (aHRs) for overall mortality. These models were adjusted for age and sex, with further adjustments for education, LTPA, CCI, smoking status, alcohol intake (for models related to metabolic phenotypes), and handgrip strength. The penalized spline smoothing method was used with multivariable adjustment to investigate the association between average daily alcohol consumption and all-cause mortality. P-values for interaction were calculated to assess the relationship between metabolic phenotype and alcohol consumption on all-cause mortality. The combined effect of average alcohol consumption and metabolic phenotype on all-cause mortality was investigated, grouping individuals based on categorized average daily alcohol consumption (abstinence to light, moderate, risk, and heavy drinking) and metabolic phenotypes (overweight/obesity, type 2 diabetes, and lean metabolic disorder). Additionally, we conducted an analysis to assess the risk of all-cause mortality in individuals with MASLD, MetALD, and ALD.

The risk of all-cause mortality between individuals with MASLD, MetALD, and ALD

Among 8635 participants with hepatic steatosis, 7127 individuals (83.0%) were classified as MASLD, 752 subjects (8.8%) as MetALD, and 708 persons (8.3%) as ALD. The remaining 48 subjects exhibited no cardiometabolic risk factors and no excessive alcohol intake. The baseline characteristics of subjects with MASLD, MetALD, and ALD are summarized in Table ​Table4.4. Among the three groups, participants with MASLD were the oldest, predominantly female, had the highest CCI score, the lowest proportion of former and current smokers, the lowest daily alcohol intake, the highest prevalence of low HDL-C, and the lowest mean values of triglyceride, hemoglobin, and handgrip strength. Compared with the MASLD group, the MetALD group comprised younger individuals, with a higher proportion of males, lower mean BMI, higher education levels, higher LTPA, higher daily alcohol consumption, a higher proportion of former or current smokers, and hypertriglyceridemia, lower LDL-C levels, and higher values of hemoglobin and handgrip strength. Individuals with ALD were the youngest, predominantly male, exhibited the highest prevalence of current smokers, the highest daily alcohol intake, higher education levels, higher LTPA, the lowest CCI score, the lowest proportion of low LDL-C, and the highest mean levels of triglyceride, hemoglobin, and handgrip strength.

Table 4

Baseline characteristics of participants with MASLD, MetALD, and ALD.

Characteristics	Steatotic liver disease			P value	Multiple comparison
	MASLD(a)	MetALD(b)	ALD(c)
Number (%)	7127 (83.0)	752 (8.76)	708 (8.3)
Age (years)	49.3±13.7	44.9±9.4	41.6±9.2	<0.001	a≠b≠c
Male gender, n (%)	2185 (30.1)	450 (57.4)	555 (81.0)	<0.001	a≠b≠c
Education, n (%)
Primary	541 (4.6)	24 (2.8)	14 (1.9)	<0.001	a≠b≠c
Secondary	4968 (71.0)	457 (63.3)	355 (47.8)
Higher	1602 (24.3)	270 (34.0)	336 (50.4)
Body mass index (kg/m2)	26.9±4.2	26.3±3.5	26.6±3.5	0.012	a≠b, b≠c
Alcohol use (g/day)	3.4±7.3	38.2±8.9	132.8±107.0	<0.001	a≠b≠c
Smoking status, n (%)
Never	5395 (77.8)	386 (53.8)	259 (34.8)	<0.001	a≠b≠c
Former	922 (9.2)	144 (17.3)	150 (21.3)
Current	810 (13.0)	222 (28.9)	299 (44.0)
High LTPA, n (%)	5216 (80.7)	612 (85.7)	581 (85.2)	0.004	a≠b, a≠c
Charlson Comorbidity index	1.22±0.57	0.12±0.34	0.11±0.31	<0.001	a≠b, a≠c
Metabolic risk factor, n (%)
Hyperglycemia	2404 (26.2)	204 (23.4)	228 (25.2)	0.586
Hypertension	4294 (48.3)	406 (46.4)	404 (49.7)	0.476
Hypertriglyceridemia	4315 (61.6)	554 (76.1)	518 (77.1)	<0.001	a≠b, a≠c
Low HDL-C	4667 (67.2)	390 (54.2)	334 (46.2)	<0.001	a≠b≠c
Central obesity	5713 (79.8)	609 (79.4)	630 (84.8)	0.196
Glucose (mg/dL)	95.5±34.2	91.7±34.9	93.9±25.3	0.194
Total cholesterol (mg/dL)	219.1±46.1	217.3±41.4	217.0±36.4	0.078
LDL-C (mg/dL)	136.4±42.1	127.6±39.0	125.0±34.8	<0.001	a≠b, a≠c
HDL-C (mg/dL)	43.5±10.3	44.1±9.1	43.7±8.7	0.227
Triglycerides (mg/dL)	199.7±119.8	240.4±135.3	256.1±134.6	<0.001	a≠b≠c
Hemoglobin (g/dL)	13.2±1.8	13.8±1.6	14.3±1.5	<0.001	a≠b≠c
Platelet (×109/cumm)	288.8±77.9	286.0±62.9	277.9±62.0	0.183
Creatinine (mg/dL)	0.87±0.31	0.88±0.19	0.92±0.16	0.665
Handgrip strength (kg)	29.1±9.4	34.3±8.3	38.5±8.0	<0.001	a≠b≠c

Open in a separate window

ALD, alcohol-associated liver disease; BMI, body mass index; BP, blood pressure; LDL-C, low-density lipoprotein cholesterol; LTPA; leisure time physical activity; HDL-C, high-density lipoprotein cholesterol; MASLD, metabolic dysfunction-associated steatotic liver disease; MetALD, metabolic dysfunction- and alcohol-associated liver disease.

Data are presented as the mean±standard deviation or number (percentage).

The χ² test and one-way analysis of variance were used to compare the characteristics of the study participants. Post hoc multiple comparison analysis was performed with Bonferroni correction.

a≠b≠c: There is statistical difference among MASLD (a), MetALD (b), and ALD (c).

a≠b, a≠c: There is statistical difference between MASLD (a) and MetALD (b) and MASLD (a) and ALD (c), while there is no statistical difference between MetALD (b) and ALD (c).

b≠c: There is statistical difference between MetALD (b) and ALD (c).

a≠b, b≠c: There is statistical difference between MASLD (a) and MetALD (b) and between MetALD (b) and ALD (c), while there is no statistical difference between MASLD (a) and ALD (c).

Participants with MASLD, MetALD, and ALD exhibited overall mortality rates of 13.86, 7.91, and 7.94 per 1000 person-years, respectively. After adjusting for factors such as age, sex, education, smoking status, LTPA, CCI, and handgrip strength, it was observed that MetALD was significantly associated with an increased risk of all-cause mortality (aHRs 1.57, 95% CI 1.04–2.35, P=0.032) compared with MASLD. ALD displayed a higher, albeit not statistically significant, risk of all-cause mortality (aHR 1.49, 95% CI 0.83–2.71, P=0.174) compared with MASLD.

Interpretation

MAFLD was initially introduced as a fatty liver disease with concomitant manifestations of metabolic dysregulation¹. It is now recognized as a heterogeneous disease with multiple metabolic phenotypes. Few studies have investigated mortality outcomes of MAFLD based on its diagnostic criteria. Hence, we categorized MAFLD participants into three metabolic phenotypes and assessed whether long-term mortality risks differed. Our analysis, focusing on the impact of metabolic phenotype at the MAFLD population level, uses patients with the overweight/obese phenotype as a reference group to assess the prognostic value of metabolic phenotypes. Intriguingly, our results indicate that the prognosis of MAFLD patients varies based on their metabolic phenotypes. Specifically, diabetes (aHR 1.59, 95% CI 1.18–2.13) and lean metabolic phenotypes (aHR 1.28, 95% CI 1.01–1.64) were independent predictors for all-cause mortality among patients with MAFLD. These findings support the idea that MAFLD patients have heterogeneous risks of overall mortality that vary according to accompanying metabolic dysregulation. Thus, classifying MAFLD subtypes based on the metabolic phenotype can potentially be used for risk stratification to promote early clinical interventions.

There is growing evidence showing that MAFLD and type 2 diabetes mellitus frequently coexist, potentiating adverse outcomes and individual progression of both disorders¹⁶. MAFLD patients with diabetes often exhibit higher severity and progression of MAFLD, leading to an increased risk of liver-related mortality^17–19. Some studies with biopsy-proven MAFLD have identified diabetes as an independent risk factor for significant fibrosis²⁰. Another histologic study revealed that non-alcoholic fatty liver disease, recently renamed MASLD, and type 2 diabetes had an additive effect on the development of aggressive outcomes such as cirrhosis and death²¹. Our study found that type 2 diabetes is an independent risk factor for mortality in MAFLD patients, increasing all-cause mortality by nearly 60% compared to non-diabetic patients with overweight or obesity. The connection between diabetes and mortality in MAFLD patients is intricate and seems to be related to insulin resistance. Our analysis revealed that MAFLD individuals with the diabetes phenotype were older and had more underlying comorbidities and metabolic risk factors such as hyperglycemia, hypertension, and low HDL-C, as well as higher serum creatinine levels compared with their overweight/obese counterparts. Considering these cardiometabolic risk factors in MAFLD individuals with diabetes, devastating microvascular and macrovascular complications of diabetes, such as cardiovascular disease and nephropathy, could be attributed to the increased risk of all-cause mortality in MAFLD patients with the diabetes phenotype.

The lean metabolic phenotype is one of the three defining criteria for MAFLD, and it plays a critical role in adverse clinical outcomes. Our study showed that 11.8% of the MAFLD cohort exhibited the lean metabolic phenotype using a conservative lower BMI for Asian populations. Our lean participants can be characterized as metabolically obese individuals with a normal weight, given their notably high prevalence of hypertriglyceridemia and low HDL-C. We found that MAFLD subjects with the lean metabolic phenotype were associated with a nearly 30% increased risk of all-cause mortality compared with those with overweight/obese MAFLD, independent of physiological, lifestyle, and comorbid factors. The prognosis of lean individuals with fatty liver disease remains subject to debate. Although NAFLD individuals with a lean body habitus have less severe liver disease than their obese counterparts, lean individuals still experience both hepatic and extrahepatic complications²². An updated meta-analysis of 14 observational studies revealed that lean MASLD was associated with a 1.6-fold increased long-term risk of all-cause mortality, independent of age, sex, and cardiometabolic risk factors²³. The mechanisms underlying this heightened mortality risk among lean MAFLD patients remain unclear, but evidence suggests a higher risk of liver-related events and mortality^24,25, potentially linked to genetic factors like the patatin-like phospholipase domain-containing 3 gene variant, particularly the GG genotype. Moreover, hypertriglyceridemia has been associated with MASLD progression in lean individuals, highlighting the importance of metabolic dysregulations in disease prognosis²⁶. The relationship between hypertriglyceridemia and MASLD progression requires further investigation. Additionally, sarcopenia has emerged as a prognostic factor for all-cause mortality in MASLD patients, particularly in lean persons with metabolic dysregulations²⁷. Handgrip strength, indicative of sarcopenia, appears to be more pronounced in lean individuals with metabolic dysregulations and may further contribute to the mortality risk in MAFLD patients with the lean metabolic phenotype. These findings underscore the complex interplay between metabolic factors, genetic predisposition, and sarcopenia in shaping the prognosis of lean individuals with MAFLD.

While regular alcohol drinking is common in the general population, its impact on overall health appears to intensify notably in individuals with metabolic risk factors²⁸. However, existing epidemiological evidence regarding the relationship between alcohol intake and mortality risk among patients with MAFLD is limited. Some studies suggest that alcohol consumption may influence the prognosis of MAFLD^29,30. Our comprehensive investigation aimed to elucidate the potential adverse effects of alcohol consumption in MAFLD patients, revealing a J-shaped association between alcohol consumption and overall mortality. Using Cox models with penalized splines, we observed that harmful alcohol consumption manifested after surpassing 50 g/day, while moderate drinking exhibited a favorable impact on metabolic traits such as hyperglycemia and hypertension, resulting in a reduced risk of overall mortality. These findings may be explained by the complex effects of alcohol on cardiovascular and liver health. Moderate alcohol drinking may confer cardiovascular benefits, but excessive alcohol intake can contribute to cardiovascular diseases and related mortality^31,32, especially in patients with metabolic dysregulations like those seen in MAFLD. Additionally, a population-based cohort study in Finland demonstrates a J‐shaped association between alcohol consumption and all‐cause mortality, with a dose-dependent risk increase with alcohol intake for incident advanced liver disease and cancers³³. However, when stratifying mortality risk according to alcohol intake categories, we found that subjects in higher drinking categories did not significantly differ in mortality risk from abstainers or light drinkers. This outcome may be due to the limited number of Thai participants consuming excessive alcohol, hindering our ability to estimate potential risks in these categories. Conflicting results from the US National Health and Nutrition Examination Survey III cohort regarding the detrimental effects of excessive alcohol consumption on MAFLD mortality risk may stem from differences in study design, particularly the use of participants without MAFLD as the reference and the oversight in categorizing daily alcohol consumption, thereby neglecting its nonlinear association with mortality risk⁷.

Several epidemiological studies have demonstrated that the combined effects of harmful drinking and metabolic dysfunction on liver disease are supra-additive^34,35. A population-based study conducted in Finland revealed that weekly binge drinking (≥60 g per occasion) combined with metabolic syndrome had significant supra-additive effects on liver-related outcomes³⁶. A study conducted in the US on adults with fatty liver disease showed that there is a combined effect of excessive drinking and metabolic syndrome on all-cause mortality¹⁰. Consistent with both studies, our investigation found that daily heavy drinking (>50 g for women and>60 g for men) was associated with a significantly higher risk of all-cause mortality among lean individuals with metabolic dysregulation. The risk is up to 3.39 times higher compared to abstainers or light drinkers with overweight/obesity. This finding suggests that heavy drinking and metabolic risk factors have a synergistic effect on mortality in lean patients with MAFLD. However, our analysis did not establish a synergistic relationship between excessive alcohol consumption and type 2 diabetes in increasing the risk of all-cause mortality. The physiological mechanisms that cause this phenomenon are intricate and not yet entirely comprehended, especially concerning how it affects only lean individuals. It is possible that the combination of metabolic dysfunction, excessive alcohol consumption, and other unhealthy lifestyle habits common in lean individuals with high-risk alcohol intake may result in worse clinical outcomes.

The Delphi consensus panel recently introduced a new nomenclature for fatty liver disease, aiming to eliminate stigmatizing terms and defining excessive liver fat accumulation as "steatotic liver disease (SLD)", subclassified into MASLD, MetALD, and ALD¹³. However, questions remain regarding the alcohol intake thresholds defined in this nomenclature. Our study explored the intricate relationship between alcohol consumption, metabolic risk factors, and all-cause mortality among individuals with different classifications of SLD. We observed distinct lifestyle and metabolic profiles among those classified into the MASLD, MetALD, and ALD groups. Notably, participants with MetALD exhibited a higher daily alcohol consumption compared to those with MASLD, while individuals with ALD displayed the highest daily alcohol intake. Our findings revealed a concerning trend in mortality rates among individuals with different classifications of SLD. Specifically, after adjusting for physiological, lifestyle, and comorbid factors, we found that individuals with MetALD had a significantly increased risk of all-cause mortality compared to those with MASLD. Although ALD subjects had a higher mortality risk, it did not reach statistical significance. This finding provides valuable insights into the complex interplay between alcohol intake and metabolic risk factors, which aligns with prior evidence of interactions between these factors on the risk of liver disease progression in patients with SLD^8,28,35. Future research should focus on understanding mechanisms driving mortality outcomes, especially cardiovascular disease, cancers, and cirrhosis, to inform targeted interventions for this population.

The strengths of this study included the large nationally representative sample and comprehensive analysis of the effects of metabolic phenotypes and alcohol intake on mortality while controlling for demographic factors, education level (as a rough indicator of socioeconomic status), lifestyle behaviors, comorbidities, and a parameter of sarcopenia. Reporting of alcohol consumption at screening intended that there was no recall bias. However, certain limitations should be acknowledged. First, the diagnosis of fatty liver disease was based on the LAP score, which is suitable for large population-based studies when acquiring liver biopsy or imaging is not feasible. The LAP score has been validated across various ethnicities, including populations in the US, Europe, and Asia^37–42. Second, this is an observational study, and daily alcohol consumption was self-reported via questionnaires administered by interviewers, potentially leading to underreporting of consumption levels. Nonetheless, this is unlikely, as we observed the association between higher categories of alcohol intake and lower proportions of low HDL-C, a biological marker of alcohol use⁴³, thus supporting the reliability of self-reported alcohol intake. Third, all covariates were available only at baseline, so we were unable to capture changes in potential confounding factors over time, such as glycemic control, weight change, and drinking habits during follow-up. Finally, due to the restricted nature of the data, we could not evaluate cause-specific mortality as there was no follow-up data on disease-specific adverse events.