Metabolic outcomes in non-alcoholic and alcoholic steatotic liver disease among Korean and American adults | BMC Gastroenterology

Baseline characteristics of the participants

The general characteristics of the participants in the KNHANES, NHANES, and KoGES cohorts according to steatotic liver status are summarized in Tables 1 and 2, and Supplementary Table 1. The KNHANES cohort included 22,597 healthy men and 7,017 men with steatotic liver. Among men with steatotic liver, most were categorized under MASLD (n = 3,494), followed by ALD (n = 2,980) and MetALD (n = 543) (Table 1). Meanwhile, there were 29,493 healthy women and 7,042 women with steatotic liver, primarily with MASLD (n = 5,733) (Table 1). Healthy Korean men were generally older, whereas those with steatotic liver were younger; no substantial age differences were observed among the MASLD, MetALD, and ALD groups. For women, no differences were observed among healthy subjects and those with MetALD and ADL. Women with MASLD were generally older than men with MASLD.

There are no established HSI thresholds for determining the presence of steatotic liver in ethnic groups outside Korea. The HSI was analyzed using the threshold of 36, identical to the Korean data, in the NHANES dataset to confirm the prevalence. The prevalence rates were 79%, 14%, 6%, and 2% for KNHANES; 74%, 23%, 1%, and 1% for KoGES; and 25%, 39%, 25%, and 12% for NHANES. Despite the globally reported SLD prevalence of 20–30%, NHANES demonstrated a striking prevalence of 75%, when HSI cutoff value was set at 36 [3]. When comparing the variables used in the HSI calculation between Korea and America, a notable difference was observed in BMI. Therefore, we adopted an HSI threshold of 41 for subsequent analyses for two reasons. First, we approximated the prevalence in the Korean population (Supplementary Fig. 4). Second, we adjusted for the BMI difference, which was approximately 5, between the KNHANES and NHANES cohorts. The NHANES cohort included 3,478 healthy men and 2,863 men with steatotic liver. Among men with steatotic liver, most were categorized under MASLD (n = 1,537), followed by MetALD (n = 707) and ALD (n = 619) (Table 2). Meanwhile, there were 2,794 healthy women and 2,798 women with steatotic liver, primarily with MASLD (n = 1,259) (Table 2). Men with MetALD and ALD were generally younger, similar to those in the KNHANES cohort. Women with MASLD and ALD were generally older and younger, respectively, similar to those in the KNHANES cohort.

The KoGES cohort included 3,235 healthy men, 744 men with MASLD, 98 with MetALD, and 72 with ALD. Meanwhile, there were 3,273 healthy women, 1,291 women with MASLD, 26 with MetALD, and 5 with ALD. In the KNHANES, NHANES, and KoGES cohorts, men with MetALD and ALD were generally younger, whereas those with MASLD varied. Women with MASLD were generally older and had younger ALD patterns.

Across all cohorts, participants with steatotic liver had significantly higher BMI, WC, and TC levels than healthy subjects regardless of sex. Markers of liver injury, such as AST and ALT levels, were elevated in patients in the MASLD, MetALD, and ALD groups compared to healthy participants. The ALT/AST ratio was also consistently increased in patients in the MASLD, MetALD, and ALD groups compared with healthy participants. Blood parameters associated with hypertension, such as SBP and diastolic BP (DBP), were raised in subjects with steatotic liver compared with healthy subjects in all cohorts and regardless of sex. Dyslipidemia parameters, including TC levels, were higher in participants with steatotic liver than in healthy individuals regardless of sex. Conversely, high-density lipoprotein (HDL) levels were lower in participants with steatotic liver than in healthy participants in all cohorts and regardless of sex sexes. Additionally, diabetes parameters, including fasting glucose and HbA1c levels, were elevated in participants with steatotic liver compared to healthy participants in all cohorts and regardless of sex.

Associations between chronic metabolic disease and disease parameters

The prevalence of chronic diseases, such as diabetes, hypertension, and dyslipidemia, increase annually [30,31,32]. Tracking annual prevalence trends using preprocessed cross-sectional data from the NHANES and KNHANES databases revealed an increasing prevalence of diabetes, hypertension, and dyslipidemia over time (Supplementary Fig. 6).

Sex-specific association patterns between chronic metabolic diseases and clinical parameters were compared between the two cross-sectional cohorts. In the KNHANES cohort (Fig. 1B), hypertension in both sexes showed positive correlations with age, BMI, WC, SBP, DBP, and TG, fasting glucose, HbA1c, and AST levels and negative linear trends with TC and HDL levels. When comparing the sex-specific hypertension-related signatures in the KNHANES and NHANES cohorts, the following variables exhibited correlations independent of sex and ethnicity: age, BMI, WC, SBP, DBP, TG, fasting glucose, and HbA1c levels. In Korea, the pattern of association between liver profile and hypertension risk differed between men and women but was consistently positive in men and women in the United States. TC and HDL levels were negatively related to the prevalence of hypertension in most cases, except for women in the United States.

Dyslipidemia-related signatures largely coincide with those linked to hypertension risk, including TG, fasting glucose, and ALT levels, as well as the ALT/AST ratio, with the exception of DBP and TC. The ALT/AST ratio, which is a predictor of MASLD and IR [16, 33] exhibited a positive relationship with the ratio of dyslipidemia regardless of sex and ethnicity. Furthermore, TC levels showed a positive association with the risk of dyslipidemia in the US dataset and a negative association in the Korean cohort.

Diabetes-related signatures mostly converged with those associated with the risk of diabetes, except for DBP and the ALT/AST ratio, which exhibited a positive relationship with the hypertension ratio depending on sex.

Associations between chronic metabolic disease and hepatic steatosis in the KNHANES cohort

HSI of patients in the KNHANES cohort were transformed into octile percentiles to identify linear trends in hepatic steatosis status with the prevalence of chronic diseases. As HSI increased, all diseases exhibited monotonic escalation patterns (Fig. 2A and B).

Disease prevalence according to hepatic steatosis in the KNHANES cohort. (A, B) Chronic metabolic disease prevalence according to the hepatic steatosis index (HSI) in men (A) and women (B). (C, D) Hazard ratio of chronic metabolic disease according to steatotic liver status in multiple logistic regression analysis in men (C) and women (D). MASLD, metabolic dysfunction-associated steatotic liver disease; MetALD, metabolically associated alcoholic liver disease; ALD, alcoholic liver disease

After an extensive review of existing literature, age and smoking status were identified as important confounding factors that could significantly affect the relationship between steatotic liver status and chronic diseases (Supplementary Table 2). The association between steatotic liver status and chronic diseases was further assessed using a multivariable logistic regression model that adjusted for confounders such as age and smoking status. The Hosmer-Lemeshow test was performed to assess the calibration of the multivariable logistic regression model (Supplement Fig. 7A A). The observed value, represented by the blue bars, indicates the actual number of events, while the expected value, depicted by the red dots, represents the number of events predicted by the logistic regression model. Therefore, the Hosmer-Lemeshow test indicated a close correlation between observed and expected values, validating the robustness calibration of the multivariable logistic regression model in KNHANES. Among men in the KNHANES cohort, the ORs for MASLD, MetALD, and ALD were 4.232 (95% CI, 3.767–4.755), 3.647 (95% CI, 2.764–4.813), and 3.824 (95% CI, 3.356–4.357) for diabetes (Fig. 2C and D), 2.330 (95% CI, 2.108–2.576), 2.312 (95% CI, 1.829–2.921), and 3.301 (95% CI, 2.977–3.660) for hypertension, and 2.055 (95% CI, 1.848–2.285), 2.180 (95% CI, 1.706–2.787), and 2.735 (95% CI, 2.456–3.046) for dyslipidemia, respectively. Among women in the KNHANES cohort, the ORs for diabetes (MASLD > MetALD > ALD), hypertension (ALD > MetALD > MASLD), and dyslipidemia (MASLD > MeALD > ALD) were similar to those in men.

Associations between chronic metabolic disease and hepatic steatosis in the NHANES cohort

HSI in NHANES data were transformed into octile percentiles to identify linear trends in hepatic steatosis status with the prevalence of chronic diseases. The results showed that as HSI increased, all diseases exhibited monotonic escalation patterns (Fig. 3A and B).

Disease prevalence according to hepatic steatosis in the NHANES cohort. (A, B) Chronic metabolic disease prevalence according to the hepatic steatosis index (HSI) in men (A) and women (B). (C, D) Hazard ratio of chronic metabolic disease according to steatotic liver status in multiple logistic regression analysis in men (C) and women (D). MASLD, metabolic dysfunction-associated steatotic liver disease; MetALD, metabolically associated alcoholic liver disease; ALD, alcoholic liver disease

The association between steatotic liver and chronic diseases was assessed using a multivariate logistic regression model. The Hosmer-Lemeshow test indicated a close correlation between observed and expected values, validating the robustness calibration of the multivariable logistic regression model in NHANES(Supplement Figs. 7B and 8B). Among men in the NHANES cohort, the ORs for MASLD, MetALD, and ALD were 1.524 (95% CI: 1.269–1.830), 1.328 (95% CI: 1.009–1.748), and 1.893 (95% CI: 1.438–2.492) for diabetes (Fig. 3C and D), 1.960 (95% CI, 1.713–2.242), 2.215 (95% CI, 1.848–2.656), and 2.166 (95% CI, 1.783–2.631) for hypertension, and 1.621 (95% CI, 1.423–1.846), 1.770 (95% CI, 1.484–2.111), and 1.778 (95% CI, 1.472–2.147) for dyslipidemia, respectively. Among women in the NHANES cohort, the ORs for diabetes, hypertension (ALD > MASLD > MetALD), and dyslipidemia (MetALD > MASLD > ALD) were similar to those in men.

We also analyzed the association between alcohol consumption and the prevalence of chronic metabolic diseases and found no significant increase in the KNHANES or NHANES cohorts. However, in the KNHANES cohort, an inverse relationship was observed among women, wherein higher alcohol consumption was associated with a lower prevalence of chronic metabolic diseases (Supplementary Fig. 8).

Risks of chronic metabolic diseases in SLD

To examine the causal relationship between chronic diseases and steatotic liver status, we used longitudinal data from the KoGES cohort and performed Cox regression analysis. The Schoenfeld analysis was utilized to verify the proportional hazards assumption, the basic tenet of the Cox proportional hazards model. A Schoenfeld p-value exceeding 0.05 confirms that the variable’s effect does not vary with time, thereby satisfying the proportional hazards assumption(Supplementary Fig. 10). Figure 4 shows the causal effects of the steatotic liver status on the development of chronic metabolic diseases. To predict the development of chronic metabolic diseases in patients with steatotic liver disease, we used data from 2 years.

Causal relationship between chronic metabolic disease and steatotic liver status in the KoGES cohort. (A) Hazard ratio of chronic metabolic disease according to steatotic liver status in Cox regression analysis in men. (B) Hazard ratio of chronic metabolic disease according to steatotic liver status in Cox regression analysis in women. Causality: Cox regression analysis was performed to assess the relationship between baseline steatotic liver status and the incidence of chronic metabolic diseases over the 2-year follow-up period

Based on the 2-year data, we calculated the risk of early development of chronic metabolic diseases. Both MASLD and MetALD were significant predictors of diabetes in men (HR, 4.076; 95% CI, 2.172–7.647 and HR, 8.012; 95% CI, 2.365–27.139, respectively) and women (HR, 6.450; 95% CI, 5.072–8.202 and HR, 10.218; 95% CI, 4.130–25.285, respectively). For hypertension, MetALD was a significant predictor in men (HR, 3.637; 95% CI, 1.111–11.900), whereas MASLD was a significant predictor in women (HR, 2.662; 95% CI, 1.803–3.930). For dyslipidemia, the 2-year data showed no significant predictive value based on steatotic liver status in either sex (Fig. 4A and B). Regarding overall causality, among men (Fig. 4A), the HRs for MASLD, MetALD, and ALD were 4.354 (95% CI, 3.371–5.624), 6.037 (95% CI, 3.654–9.975), and 5.831 (95% CI, 3.343–10.171) for diabetes, 1.517 (95% CI, 1.257–1.830), 2.069 (95% CI, 1.319–3.245), and 2.050 (95% CI, 1.242–3.383) for hypertension, and 1.398 (95% CI, 1.120–1.743), 2.109 (95% CI, 1.338–3.325), and 2.867 (95% CI, 1.818–4.522) for dyslipidemia, respectively. Meanwhile, the number of women with ALD was insufficient to provide reliable HR estimates; however, their HRs for diabetes and dyslipidemia were similar to those in men.