If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding Author information. , Division of Gastroenterology, Department of Internal Medicine, Kyung Hee University Hospital at Gangdong, Seoul, Korea, 892 Dongnam-ro, Gangdong-gu, 05278, Seoul, Republic of Korea. Telephone: +82-2-440-6220.
In AFLD, cardiac and hepatic outcomes were higher than in NAFLD regardless of MetR
•
The incidence of cardiac outcomes of NAFLD catches up with AFLD as the number of MetR increases
•
In NAFLD, the incidence of cardiac outcomes increases as the number of MetR increases, but no prominent difference in hepatic outcomes.
•
In AFLD, the number of MetR does not correlate with cardiac and hepatic outcomes
Abstract
Background & Aims
Metabolic risk factors (MetR) are associated with hepatic and cardiac outcomes in patients with fatty liver disease. We aimed to evaluate whether MetR has different effects on alcoholic fatty liver disease (AFLD) and nonalcoholic fatty liver disease (NAFLD).
Methods
We used a standardized common data model from seven university hospitals between 2006 and 2015. MetR included diabetes mellitus, hypertension, dyslipidemia, and obesity. Follow-up data were analyzed for the incidence of hepatic outcomes, cardiac outcomes, and death between AFLD and NAFLD and based on MetR within AFLD and NAFLD.
Results
Out of 3,069 and 17,067 patients with AFLD and NAFLD, respectively, 2,323 (75.7%) and 13,121 (76.9%) had one or more MetR, respectively. Patients with AFLD were at a higher risk of hepatic outcomes (adjusted risk ratio [aRR], 5.81) than those with NAFLD irrespective of MetR. The risk of cardiac outcomes in AFLD and NAFLD became similar with the increasing number of MetR. Patients with NAFLD without MetR demonstrated a lower risk of cardiac outcomes but not hepatic outcomes than those with MetR (aRR, 0.66 and 0.61 for MetR ≥ 1 and MetR ≥ 2, respectively, all P<0.05). In patients with AFLD, hepatic and cardiac outcomes were not associated with MetR.
Conclusions
Clinical impact of MetR in patients with fatty liver disease may differ between patients with AFLD and NAFLD.
Lay summary
With the increasing prevalence of fatty liver disease (FLD) and metabolic syndrome, the increase in associated complications, such as liver and heart diseases, have become important social issues. Particularly, in patients with FLD with excessive alcohol consumption, the incidence of liver and heart disease is pronounced due to the dominant effect of alcohol that exceeds the effects of other factors. Appropriate screening and management of alcohol consumption in patients with FLD are important.
The authors have no conflicts of interest to disclose.
Financial support
None
Authors’ contributions
JL, HS, and HIK were involved in the study concept and design, data analysis, interpretation, drafting of the manuscript, and critical revision of the manuscript. HIK were involved in data acquisition and study supervision.
Data availability statement
CDM data are designed to support a distributed research network. Thus, access to the data is restricted on internal private networks. Therefore, data are not publicly available.
Introduction
Fatty liver disease (FLD) is the most common cause of chronic liver disease, which is responsible for a majority of morbidity and mortality in patients with chronic liver disease worldwide.
Traditionally, FLD can be divided into alcoholic fatty liver disease (AFLD) and non-alcoholic fatty liver disease (NAFLD), which account for almost every patient with FLD.
Progressive forms of FLD eventually result in hepatic fibrosis and primary liver cancer, which are well-known prognostic factors in chronic liver disease.
Additionally, various studies have reported the inseparable relationship between FLD and cardiovascular disease in terms of life-threatening medical events.
With concern that hepatic and cardiac outcomes are important in the clinical course of FLD, recent studies have reported that FLD-related outcomes could be accelerated under the influence of metabolic risk factors (MetR), such as diabetes mellitus, hypertension, dyslipidemia, and obesity.
With the increasing number of patients with FLD worldwide, especially NAFLD, identifying high risk patients is an important issue. In this regard, a panel of international experts has proposed a new classification called metabolic-associated fatty liver disease (MAFLD).
MAFLD is defined as FLD in patients with overweight/obesity, type 2 diabetes mellitus, or evidence of metabolic dysregulation irrespective of any other etiology of liver disease. Although the concept of MAFLD has some advantages in identifying high-risk patients and predicting the outcomes, several debates have been raised.
Most of the patients with MAFLD include those with NAFLD and MetR who are considered to have a relatively benign course than those with AFLD in terms of hepatic and non-hepatic outcomes.
Nevertheless, there have been few attempts to assess the outcomes of AFLD and NAFLD separately, and no study has analyzed FLD-related outcomes according to the presence of MetR.
Another critical issue is the effect of concomitant comorbidities (except for MetR) on AFLD and NAFLD. In fact, the patients with AFLD and NAFLD include different number of comorbidities, and it is possible that the effects of comorbidities on the outcomes can vary.
Therefore, the effects of MetR on FLD outcomes should be analyzed separately from the influence of comorbidities; however, no prior studies have investigated this issue.
Therefore, we investigated the effects of MetR on the incidence of hepatic and cardiac outcomes using a common data model, including seven university hospital databases, to identify clinical information for risk stratification in patients with FLD. To compare the outcomes between the AFLD and NAFLD groups of patients, we analyzed 1) intergroup differences between AFLD and NAFLD, and 2) intragroup differences according to MetR within groups of patients with AFLD and NAFLD.
Materials and methods
Data source
This multicenter observational study used data from seven university hospitals in six provinces of South Korea. It is difficult to combine and analyze data from different hospitals because different datasets are built using different data models and often local terminologies. Furthermore, integrating data from multiple hospitals entails not only problems of data-standardization but also pseudonymization issues.
In order to overcome these limitations and secure the basis for integrated data analysis, we used the data previously converted to the standardized Observational Medical Outcome Partnership (OMOP)-Common Data Model (CDM) v5.3.
Since 2018, OMOP-CDM version data from various hospitals have been converted and utilized through the platform called FEEDER-NET (https://feedernet.com), a health big-data platform supported by the National CDM projects in Korea. The OMOP-CDM version of hospital data has been validated in several observational health data science and informatics (OHDSI) studies.
Conversion and Data Quality Assessment of Electronic Health Record Data at a Korean Tertiary Teaching Hospital to a Common Data Model for Distributed Network Research.
Association between proton pump inhibitor use and gastric cancer: a population-based cohort study using two different types of nationwide databases in Korea.
Non-standardized medical records and terms were mapped to the Systemized Nomenclature for Medicine-Clinical Terms (SNOMED-CT) code, and Logical Observation Identifiers, Names, and Codes (LOINC) were also used for numeric data extraction.
McDonald CJ, Huff Sm, Fau - Suico JG, Suico Jg Fau - Hill G, Hill G Fau - Leavelle D, Leavelle D Fau - Aller R, Aller R, Fau - Forrey A, et al. LOINC, a universal standard for identifying laboratory observations: a 5-year update. 2003(0009-9147 (Print)).
The analysis in this study included data from various hospitals (Kyung Hee University Hospital at Gangdong, Hallym University Kangdong Sacred Heart hospital, Kangwon National University Hospital, Daegu Catholic University Medical Center, Pusan National University Hospital, Wonkwang University Hospital, and Ajou University Hospital) corresponding to the Research Border-free Zone (RFZ), a research-free zone for multi-institution distributed big data research. Institutions affiliated with the RFZ provide researchers from other institutions with the same level of CDM research rights as those permitted to in-house researchers.
All data were obtained from the hospital based CDM database. This study protocol was approved by the Institutional Review Board (IRB) of Kyung Hee University in Korea, and all study methods were performed in accordance with the appropriate guidelines and regulations. The study protocol also conforms to the ethical guidelines of the Declaration of Helsinki as reflected in prior approval by the institution’s IRB (KHNMC IRB 2020-08-003).
Study subjects and definitions
In this retrospective, multicenter, observational, comparative cohort study, we included patients aged over 20 years and cohort entries from July 2006 to December 2015 with the end of the observational period of December 2019. All study variables were defined by standard concept identification and the codes for the concept sets as summarized in Supplementary table 1. The medical histories of 23,884 patients diagnosed with FLD for the first time were initially extracted from a database of the seven hospitals. The patients included only those with confirmed AFLD or NAFLD according to the standard code. If available, CDM data regarding drinking were used to confirm the diagnostic accuracy of AFLD and NAFLD. Patients with AFLD were defined by diagnostic codes (50325005) with or without current drinker code (228279004, 86933000, and 219006) confirmed within 1 year of diagnosis. Patients with NAFLD were defined by the presence of both diagnostic codes (197321007) with or without non-drinker code (105542008) confirmed within 1 year at the time of diagnosis. Patients with both AFLD and NAFLD codes were excluded.
For the study analysis, we excluded patients with any time before to 1 year after medical history of cirrhosis and primary liver cancer (n=617). We also excluded patients with chronic viral hepatitis (hepatitis B and C virus infection, n=1,316) and minor etiology of chronic liver disease including primary biliary cholangitis, primary sclerosing cholangitis, Wilson’s disease, hemochromatosis, and autoimmune hepatitis that was diagnosed at any time point during the observational period (n=64). Additionally, to achieve the research purpose to study only the effect of metabolic risk factors on AFLD and NAFLD, the patients with any time before to 1 year after medical history of severe comorbidities as organic manifestations (heart failure, ischemic heart disease, atrial fibrillation, chronic obstructive lung disease, cerebral infarction, and chronic kidney disease, n=1,751) were excluded.
MetR were defined as standardized diagnosis codes and/or numeric data within any time before to 1 year after and nearest to the index date for the following:
dyslipidemia, defined as a serum total cholesterol level ≥ 200 mg/dL, low-density lipoprotein (LDL)-cholesterol level ≥ 100 mg/dL, triglyceride level ≥ 150 mg/dL, high-density lipoprotein (HDL)-cholesterol < 40 mg/dL, or a diagnosis of dyslipidemia; and
Direct data extraction from the electronic medical records for detailed information on alcohol consumption
The CDM code for the amount of drinking varied from 12.8% to 52.6% according to the conversion rate of each hospital data. To evaluate whether the standard code-based classification sufficiently classified actual patients with AFLD and NAFLD, we conducted a manual review of electronic medical records from accessible data using the same inclusion criteria as the CDM code-based extraction. A total of 3,253 patients (approximately 16.2% of patients in the entire study group) were extracted from Kyung Hee University Hospital at Gangdong using the same inclusion criteria. Data related to alcohol consumption were manually reviewed except for cases where the patient refused to disclose their level of alcohol consumption or when it was difficult to obtain complete information due to loss of data (Supplementary table 2). The results of the additional analysis showed that the agreement between CDM code and clinical diagnosis was 97.9% for AFLD (heavy or very heavy drinker with FLD) and 98.8% for NAFLD (equal to or less than moderate drinking with FLD). These results suggest that AFLD and NAFLD defined by standard codes could be useful in distinguishing FLD based on whether there was a history of heavy drinking.
Study outcomes
This study was conducted to address the effects of MetR on the incidence of cirrhosis, primary liver cancer, cardiac outcomes, and death in patients with AFLD and NAFLD. The hepatic outcomes included the incidence of cirrhosis and primary liver cancer. The cardiac outcomes included the incidence of ischemic heart disease, heart failure, and atrial fibrillation.
Subsequently, intergroup comparisons (AFLD vs. NAFLD, depending on the presence of the same level of MetR [all patients, patients without MetR, patients with ≥ 1 MetR, and patients with ≥ 2 MetR]), and intragroup comparisons (AFLD without MetR vs. AFLD with 1 [or 2] or more MetR; NAFLD without MetR vs. NAFLD with 1 [or 2] or more MetR) were performed.
Statistical analyses
OHDSI analysis tools were embedded in the interactive analysis platform, ATLAS. ATLAS version 2.7.6 was used, and we analyzed the platform on FEEDER-NET, a health big-data platform based on OHDSI-CDM and supported by the Korean National Project.
Conversion and Data Quality Assessment of Electronic Health Record Data at a Korean Tertiary Teaching Hospital to a Common Data Model for Distributed Network Research.
In this study, OHDSI analysis tools were used for primary data extraction, annual and cumulative incidence analysis, and regression analysis for each hospital. The data extraction of the study population and the incidence-based analysis were performed from the overall data; normally distributed continuous variables are presented as mean ± standard deviation, and data for categorical variables are presented as numbers (percentages). The chi-square test was used to examine the relationships between categorical variables, and Student’s t-test was used to compare the mean values of continuous variables across the groups. The Kaplan–Meier method with the log-rank test was used to compare the cumulative incidence of the outcome in terms of intergroup and intragroup analyses. The hazard ratios (HRs) and 95% confidence intervals (CIs) for the incidence (event per person-year) of outcomes based on Cox proportional hazard models (COX-PH) were used only when the analysis was possible in each hospital (all the target and comparator groups had data available including six of seven hospitals). COX-PH analysis was performed for hepatic and cardiac outcomes in terms of intergroup and intragroup comparisons using unadjusted and age-sex adjusted regression analysis. Once the data were extracted from each hospital, the risk of outcome data was combined and analyzed using meta-analyses, which were performed to estimate the incidence (by person-years) using a random-effects model. Statistical analyses for pooled data were performed using R v4.1.3 (R Foundation Inc.; http://cran.r-project.org), and meta-analysis was performed using Review Manager (RevMan, version 5.4.1; Cochrane Collaboration). A p-value <0.05 was considered as significant.
Results
Demographic characteristics and trends of outcomes in AFLD and NAFLD
Table 1 summarizes the baseline characteristics of the patients with FLD. The study population included 3,069 patients with AFLD and 17,067 patients with NAFLD; those with AFLD included a higher proportion of males (91.2% in AFLD and 63.2% in NAFLD) and older patients (mean age, 49.12 years in AFLD and 47.33 years in NAFLD) than those with NAFLD. Those without MetR included 746 (24.3%) patients with AFLD and 3,946 (23.1%) patients with NAFLD. In patients without MetR, a similar trend was observed; AFLD included a higher proportion of males (88.5% in AFLD and 61.8% in NAFLD) and older age (mean age, 48.20 in AFLD and 43.84 years in NAFLD) than those with NAFLD. According to the presence of each MetR, males were more likely to have AFLD than NAFLD (range, 91.5%–92.8% in AFLD and 59.8%–66.5% in NAFLD). As for age, there were no statistically significant differences between the AFLD and NAFLD groups in terms of diabetes mellitus, hypertension, and dyslipidemia, while those with obesity and AFLD were older (mean age, 48.68 years in AFLD and 46.098 years in NAFLD) than those with obesity and NAFLD. The AFLD group had a higher proportion of males (MetR ≥ 1, 92.0% in AFLD, and 63.5% in NAFLD; MetR ≥ 2, 92.4% in AFLD, and 64.1% in NAFLD), and older age (mean age: MetR ≥ 1, 49.99 years in AFLD and 48.81 years in NAFLD; MetR ≥ 2, 50.01 years in AFLD, and 48.83 years in NAFLD) in patients with both ≥ 1 MetR and ≥ 2 MetR. Generally, the presence of MetR tended to increase age in patients with FLD. Based on the data extracted from each hospital, there were differences in the absolute values or ratios according to region or hospital; however, the trend was maintained overall in terms of the proportion of males and the mean age (Supplementary table 3).
Table 1Baseline Characteristics of Patients with fatty liver disease.
Obesity data from three centers (Kangwon National University Hospital, Daegu Catholic University Medical Center, and Pusan National University Hospital) are not available.
306
2,245
Male, n (%)
284 (92.8%)
1,493 (66.5%)
<0.001
Age, mean (SD)
48.68 (9.98)
46.09 (12.75)
0.002
Presence of 1 or more metabolic risk factors
2,323
13,121
Male, n (%)
2,137 (92.0%)
8,329 (63.5%)
<0.001
Age, mean (SD)
49.99 (10.10)
48.41 (12.72)
<0.001
Presence of 2 or more metabolic risk factors
1,556
9,288
Male, n (%)
1,437 (92.4%)
5,956 (64.1%)
<0.001
Age, mean (SD)
50.01 (10.12)
48.83 (12.61)
0.005
Values are presented as mean ± standard deviation or number (%).
The chi-square test was used to examine the relationships between categorical variables, and Student’s t-test was used to compare the mean values of continuous variables across the groups. Level of significance, p value <0.05.
∗ Obesity data from three centers (Kangwon National University Hospital, Daegu Catholic University Medical Center, and Pusan National University Hospital) are not available.
The incidences of four outcomes (cirrhosis, liver cancer, heart disease, and death) were estimated in the entire patient group (Figure 1). The mean follow-up duration was 5.52 years in the AFLD group and 5.56 years in the NAFLD group. In all patients, the 5-year and 10-year cumulative incidence of cirrhosis was 0.88% and 2.93% in the AFLD and 0.13% and 0.53% in NAFLD groups, respectively. The 5-year and 10-year liver cancer rate was 0.07% and 0.30% in the AFLD group, respectively, and 0.02% and 0.09% in the NAFLD group, respectively. For heart disease, the 5-year and 10-year incidence was 0.37% and 1.96% in the AFLD group and 0.25% and 1.48% in the NAFLD group, respectively. The 5-year and 10-year cumulative mortality rate was 1.35% and 2.02% in the AFLD group and 0.54% and 0.91% in the NAFLD group, respectively. Overall, the incidence of cirrhosis, liver cancer, and death was higher in the AFLD group than those in the NAFLD group. In contrast, the proportional difference in heart disease was less prominent than those of other outcomes. The risk of incidence of hepatic and cardiac outcomes between the AFLD and NAFLD groups (Supplementary table 4) and the adjusted risk ratio (aRR) revealed that hepatic outcomes were commoner in the AFLD group than those in the NAFLD group (aRR [95% confidence interval, CI]: 5.81 [4.14–8.15], P <0.001) but not cardiac outcomes (aRR [95% CI]: 1.19 [0.89–1.58], P>0.05).
Figure 1Cumulative incidence of cirrhosis, liver cancer, heart disease, and death in patients with alcoholic fatty liver disease and non-alcoholic fatty liver disease.
Comparisons of intergroup and intragroup outcomes according to MetR
Intergroup analysis was performed in patients without MetR, MetR ≥ 1, and MetR ≥2 for the incidence of cirrhosis, liver cancer, heart disease, and death (Figure 2). In patients without MetR, AFLD had a higher cumulative 5-year and 10-year incidence of all four outcomes than NAFLD (cirrhosis, 0.58% and 2.88% in AFLD group and 0.17% and 0.58% in NAFLD group; liver cancer, 0.14% and 0.43% in AFLD group and 0% and 0.03% in NAFLD group; heart disease, 0.43% and 1.87% in AFLD group and 0.11% and 0.83% in NAFLD group; death, 1.15% and 1.87% in AFLD group and 0.44% and 0.72% in NAFLD group; all P values <0.05, log-rank test). In patients with MetR ≥ 1, AFLD and NAFLD revealed no statistical difference in the incidence of heart disease (5- and 10-year incidences, 0.26% and 2.03% in AFLD group and 0.28% and 1.68% in NAFLD group, all P > 0.05). There was no statistical difference in the incidence of liver cancer and heart disease between the AFLD and NAFLD groups in patients with MetR ≥ 2 (5- and 10-year incidence: liver cancer, 0.07% and 0.26% in AFLD group and 0.04% and 0.12% in NAFLD group; heart disease, 0.20% and 1.97% in AFLD group and 0.30% and 1.90% in NAFLD group; all P values > 0.05). In contrast, the incidence of cirrhosis and death was higher in AFLD group than those in NAFLD group irrespective of MetR (all P values < 0.05).
Figure 2Intergroup comparison of cumulative incidence of cirrhosis, liver cancer, heart disease, and death in (A) patients without metabolic risk factors (MetR) (B) MetR ≥ 1, and (C) MetR ≥ 2. Log-rank tests used to examine the differences in cumulative risks, level of significance, p <0.05
In terms of intragroup comparative analysis (Figure 3), there was no statistical difference in the incidence rates of cirrhosis, liver cancer, heart disease, and death among the patients without MetR, MetR ≥ 1, and MetR ≥ 2 in AFLD group (5- and 10-year incidence: cirrhosis, 0.58% and 2.88%, 0.88% and 3.04%, 0.72% and 2.82%, respectively; liver cancer, 0.14% and 0.43%, 0.04% and 0.26%, 0.07 and 0.26%, respectively; heart disease, 0.43% and 1.87%, 0.26% and 2.03%, 0.20%, and 1.97%, respectively; and death, 1.15% and 1.87%, 1.41% and 2.07%, 1.31%, and 1.83%, respectively; all P values > 0.05). In the NAFLD group, there was no statistical difference between the incidence of cirrhosis and liver cancer among patients without MetR, MetR ≥ 1, and MetR ≥ 2 (5- and 10-year incidence: cirrhosis and liver cancer, 0% and 0.03%, 0.02% and 0.10%, 0.04% and 0.12%, respectively; all P values > 0.05). In contrast, the incidence of heart disease was significantly higher in patients with MetR ≥ 1 and MetR ≥ 2 than in those without MetR (5- and 10-year incidence: no MetR vs. MetR ≥ 1, 0.11% and 0.83% vs. 0.28% and 1.68%, respectively; no MetR vs. MetR ≥ 2, 0.11% and 0.83% vs. 0.30% and 1.90%, respectively; all P values <0.05). The incidence of mortality was higher in those with MetR ≥ 1 and MetR ≥ 2 than that in those without MetR; however, there was no statistical difference between the groups (5- and 10-year incidence: no MetR vs. MetR ≥ 1, 0.44% and 0.72% vs. 0.57% and 0.96%, respectively; no MetR vs. MetR ≥ 2, 0.44% and 0.72% vs. 0.52% and 1.03%, respectively; all P values > 0.05). The annualized incidence and number of patients at risk are summarized in Supplementary table 5.
Figure 3Intragroup comparison of cumulative incidence of cirrhosis, liver cancer, heart disease, and death according to metabolic risk factors in (A) patients with alcoholic fatty liver disease (AFLD) and (B) with non-AFLD. Log-rank tests used to examine the differences in cumulative risks, level of significance, p <0.05
Intergroup and intragroup risk of hepatic and cardiac outcomes according to MetR
Intergroup comparisons (AFLD vs. NAFLD) were performed in patients without MetR, MetR ≥ 1, and MetR ≥ 2 to compare the risk of incidence of hepatic and cardiac outcomes (Table 2). The aRR in patients without MetR, MetR ≥ 1, and MetR ≥ 2 in terms of cardiac outcomes were not statistically different across the groups (aRR [95% CI], 1.46 [0.77–2.75] in patients without MetR; 1.16 [0.84–1.61] in patients with MetR ≥ 1; 0.94 [0.63–1.39] in patients with MetR ≥ 2, all P values > 0.05). In contrast to cardiac outcomes, the hepatic outcomes consistently favored the NAFLD group in patients without MetR, MetR ≥ 1, and MetR ≥ 2 (aRR [95%CI]: 2.94 [1.47–5.87] in patients without MetR; 6.74 [3.69–12.29] in patients with MetR ≥ 1; 5.56 [2.68–11.51] in patients with MetR ≥ 2, all P value < 0.05).
Table 2Risk of adverse cardiac and hepatic outcomes on intergroup comparison between those with AFLD and those with NAFLD according to MetR.
Presence of MetR
Outcome incidence
Outcome incidence (Events/Follow-up time [person-year])
Model adjusted for age and sex in each hospital data included, and meta-analyses were undertaken to obtain cumulative incidence estimates (by person-year) using a random-effects model.
Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
∗ Model adjusted for age and sex in each hospital data included, and meta-analyses were undertaken to obtain cumulative incidence estimates (by person-year) using a random-effects model.
† Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
Intragroup comparisons between AFLD and NAFLD patients are summarized in Table 3. In patients with AFLD, there was no statistically significant difference between patients without MetR vs. MetR ≥ 1 and MetR ≥ 2 in terms of cardiac outcomes (no MetR vs. MetR ≥ 1, 0.98 [0.56–1.71]; no MetR vs. MetR ≥ 2, 1.04 [0.58–1.85], all P values > 0.05) and hepatic outcomes (no MetR vs. MetR ≥ 1, 1.09 [0.37–3.19]; no MetR vs. MetR ≥ 2, 1.08 [0.33–3.52], all P values > 0.05). In patients with NAFLD, however, cardiac outcomes were significantly more favorable in patients without MetR than those in patients with MetR ≥ 1 and MetR ≥ 2 (0.66 [0.46–0.94] and 0.61 [0.43–0.87]; respectively, all P values < 0.05). There was no significant risk difference in the hepatic outcomes in patients with NAFLD with any MetR than those without MetR (1.35 [0.52–3.52] and 1.59 [0.65–3.92], respectively, all P values > 0.05).
Table 3Risk of adverse cardiac and hepatic outcomes on intragroup comparison in those with AFLD and those with NAFLD according to MetR.
Outcome incidence
Outcome incidence (Events/Follow-up time [person-year])
Model adjusted for age and sex in each hospital data included, and meta-analyses were undertaken to obtain cumulative incidence estimates (by person-year) using a random-effects model.
Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
∗ Model adjusted for age and sex in each hospital data included, and meta-analyses were undertaken to obtain cumulative incidence estimates (by person-year) using a random-effects model.
† Data of hazard ratios and 95% confidence intervals for the incidence (event per person-year) of outcomes by Cox proportional hazard models were used in six hospitals, except Wonkwang University Hospital.
Different effects of newly developed MetR on the outcomes in AFLD and NAFLD
Subgroup analysis was performed to identify the incidence of outcomes of one or more newly developed MetR (nMetR) within 5 years in the patients with no MetR (Figure 4). Although the absolute increment in the incidence of outcomes was consistently higher in the AFLD group than that in the NAFLD group, the rate of increase was different across the four outcomes. In patients with AFLD, 5- and 10-year cumulative outcome incidences were higher in nMetR patients than in patients without MetR (cirrhosis, 0.58% and 2.88% vs. 0.92% and 5.50%; liver cancer, 0.14% and 0% vs. 0% and 0%; heart disease, 0.43% and 1.87% vs. 1.87% and 5.50%; and death, 1.15% and 1.87% vs. 2.75% and 2.75%, respectively). In patients with NAFLD, the 5- and 10-year cumulative outcome incidences of nMetR patients compared with those in patients without MetR were higher in terms of the incidence of cirrhosis (0.17% and 0.58% vs. 0.67% and 1.72%), heart disease (0.11% and 0.83% vs. 0.34% and 1.89%), death (0.44% and 0.72% vs. 0.52% and 1.55%, respectively), and non-liver cancer (0% and 0.03% vs. 0% and 0.17%, respectively). Overall, both the AFLD and NAFLD groups, where MetR begain within 5 years, demonstrated a more prominent incidence of outcomes than patients without MetR at baseline in terms of cirrhosis, heart disease, and death (Supplementary Figure 2).
Figure 4Comparison of the incidence of cirrhosis, liver cancer, heart disease, and death between patients without metabolic risk factors (MetR) and those with newly developed MetR within 5 years of cohort entry in (A) patients with alcoholic fatty liver disease (AFLD) and (B) with non-AFLD.
Recent studies on FLD within the concept of MAFLD have attempted to analyze whether there are differences in the morbidity or mortality between MAFLD and NAFLD.
However, the integrated approach of FLD with MetR has some problems in that it is difficult to clearly evaluate the effects of MetR on FLD of different etiologies. Particularly, since a majority of patients with FLD consist not only of NAFLD but also AFLD, it is important to separate the analyses to identify the effects of MetR on the incidence of hepatic and cardiac outcomes.
Furthermore, it is important to exclude well-known clinical factors, such as chronic viral hepatitis and severe underlying comorbidities, related to the outcomes
to analyze the effects of MetR on FLD per se. Nevertheless, no study has evaluated the effects of MetR on the incidence of outcomes in two different categories of patients with FLD after excluding confounding factors. Our study has demonstrated that both hepatic and cardiac outcomes were much more frequent in patients with AFLD than in those with NAFLD. However, when the incidence analysis was limited to patients with AFLD, the incidence of outcomes did not depend on the severity of MetR. In contrast, the incidence of heart disease in patients with NAFLD increased by MetR, the results became similar to those of AFLD. In patients with NAFLD, the incidence of liver cancer and mortality increased but there was no statistical difference between the cumulative incidence and adjusted regression analysis. These results suggest that the presence of MetR affects the incidence of hepatic and cardiac outcomes differently in patients with AFLD and those with NAFLD, thus indicating that careful risk stratification is required under the concept of MAFLD.
No previous studies have directly compared AFLD and NAFLD groups of patients according to the presence of MetR. One hospital cohort study evaluated the incidence of liver and cardiovascular disease in patients with fatty liver disease.
The risk of hepatic outcomes in the AFLD group compared to the NAFLD group was similar to that in our study in age- and sex-adjusted regression analysis. Our intergroup analysis revealed that the incidence of hepatic outcomes was at least 3–6 times higher in the AFLD group than that in the NAFLD group and was dependent on MetR. Although a gradual increase in liver cancer was observed, most of the hepatic outcomes were cirrhosis, which demonstrated a consistently higher incidence in the AFLD group than the NAFLD group. As the incidence of liver cancer before advanced fibrosis or cirrhosis is very low, the majority proportion is considered to be the occurrence of cirrhosis as our study results.
Interestingly, the results of previous studies did not demonstrate a statistically significant difference in the cardiovascular risk between patients with AFLD and NAFLD even after adjusting the regression model for MetR, such as diabetes, hypertension, and dyslipidemia. Our intergroup analysis revealed that the difference in the incidence of heart disease between the AFLD and NAFLD groups disappeared with an increase in the number of MetR. Overall, the AFLD group had consistently more frequent hepatic and cardiac outcomes than the NAFLD group irrespective of the number of MetR. In contrast, in NAFLD, there was a pattern of catching up with cardiac outcomes, as in AFLD, as the severity of MetR increased.
In the intragroup analysis, the AFLD group did not demonstrate clinical effects according to an increase in MetR for all outcomes. However, both cardiac outcomes and mortality increased in the NAFLD group. In terms of AFLD, the effects of MetR may not be clinically significant on the incidence of hepatic outcomes, such as cirrhosis or liver cancer. In NAFLD, increase in MetR affects the cardiac outcomes and mortality as reported previously.
These results suggest that AFLD and NAFLD may be differently affected depending on the severity of MetR. The result of the patients with AFLD can be explained by the following two possibilities. First, alcohol consumption itself could have a dominant effect on outcomes in the AFLD group irrespective of MetR. Previous studies targeting FLD and alcohol consumption have reported MetR as an independent risk factor for overall survival and the incidence of severe forms of liver disease.
Considering the results of previous studies and those of our study on the distribution of alcohol consumption suggest that AFLD has a dominant effect on the incidence of hepatic outcomes over cardiac outcomes.
Furthermore, it is possible that alcohol consumption itself had a greater influence than MetR on the occurrence of clinically significant outcomes. Second, it is possible that not only the number of MetR at cohort entry but also the newly developed MetR with time has different effects on the outcomes in the AFLD and NAFLD groups. To identify the difference in the outcomes according to the number of MetR, the outcome incidence was compared between the groups in which MetR occurred within 5 years (nMetR) and the group without MetR at cohort entry. As a result, in the cumulative incidence of cirrhosis and heart disease more than doubled in the nMetR group for both the AFLD and NAFLD groups, and this trend was far more prominent in the AFLD group. Particularly, the group in which nMetR developed demonstrated a higher incidence of outcomes than the group that already had MetR at cohort entry. This result suggests that more careful monitoring and management are required in patients with nMetR, particularly in the AFLD group. Therefore, further studies need to consider the following: 1) when screening the risk groups according to MetR in patients with FLD, different risk stratification is needed between the AFLD and NAFLD groups; and 2) it is necessary to investigate in patients with nMetR whether we can effectively predict the high-risk group of hepatic and cardiac outcomes in patients with FLD.
Previous studies on FLD have evaluated the incidence of cardiovascular events, which aimed to assess complications of vascular origin.
These studies reflect the underlying mechanism mainly associated with NAFLD. However, when atrial fibrillation occurs in patients with excessive alcohol consumption, ischemic stroke has been reported to increase following atrial fibrillation.
That is, many stroke events may be caused by embolic events that occur after atrial fibrillation in patients with AFLD (The effect of non-vascular heart disease) This pattern is different from that of NAFLD. Although the association between NAFLD and atrial fibrillation has been frequently reported,
a recent mendelian randomized study showed that NAFLD has a high correlation with small/large vessels and stroke; however, there is a possibility that it is not related to embolic events.
Interestingly, by the additional CDM-based analysis, the incidence of stroke showed the hepatic outcome patterns in the intergroup analysis and the cardiac outcome in the intragroup analysis (Supplementary figure 3–6 and Supplementary table 6–7). Future research is needed to evaluate the different incidence patterns of heart disease and stroke in patients with AFLD and NAFLD according to MetR.
This study has several advantages. With the recent increase in the prevalence of FLD, it is necessary to separately analyze the effects of MetR on FLD to facilitate clinical risk stratification. Our study aimed to provide basic information for future research by excluding confounders, such as chronic viral hepatitis and severe comorbidities. Additionally, this multicenter observational study, using standardized data extraction and CDM-based analysis, provides methodological examples of conducting studies that require long-term observational studies across multiple regions.
However, there are some problems that could not be addressed by this study. First, due to the limitations of data acquisition, MetR included only four representative diseases: diabetes mellitus, hypertension, dyslipidemia, and obesity. Therefore, compared to the MAFLD concept, it was impossible to completely separate the patients with MetR corresponding to the gray zone according to the differences in the definition.
Additionally, it was not possible to analyze how different components of each MetR could affect the AFLD and NAFLD groups differently. Second, the difference in baseline comorbidities or the risk of outcomes change according to treatment history could not be analyzed separately. Third, the degree of alcohol consumption could not analyzed separately. However, in our study results from a single-center EMR review, the CDM code for AFLD and NAFLD could divide the patients with FLD based on certain amounts of alcohol consumption (i.e., heavy drinking) Although the incidence of outcomes was likely to vary depending on the regional drinking patterns and the possibility of abstinence in the AFLD group, the rate of alcohol cessation is very low in Korea, it is believed that the study cohort reflects the general patient population.
Fourth, it was not possible to analyze the degree of steatosis, inflammation, and fibrosis because this study was analyzed based on clinical diagnosis and indicators and not based on biopsy results, which is the gold standard test for the diagnosis of FLD.
In conclusion, compared with the NAFLD group, the AFLD group showed more hepatic and cardiac outcomes irrespective of the number of MetR. In the NAFLD group, the proportion of cardiac outcomes increased with an increase in the number of MetR and reached that of cardiac outcomes in the AFLD group. In contrast, in the AFLD group, both hepatic and cardiac outcomes were not associated with the number of MetR. Overall, the clinical impact of MetR in patients with FLD may differ between AFLD and NAFLD.
Acknowledgements
Data extraction and statistical analyses of each of the 7 medical centers were performed using FEEDER-NET (EvidNet Inc.,https://feedernet.com), a healthcare big-data platform.
Appendix A. Supplementary data
The following is/are the supplementary data to this article:
Conversion and Data Quality Assessment of Electronic Health Record Data at a Korean Tertiary Teaching Hospital to a Common Data Model for Distributed Network Research.
Association between proton pump inhibitor use and gastric cancer: a population-based cohort study using two different types of nationwide databases in Korea.
McDonald CJ, Huff Sm, Fau - Suico JG, Suico Jg Fau - Hill G, Hill G Fau - Leavelle D, Leavelle D Fau - Aller R, Aller R, Fau - Forrey A, et al. LOINC, a universal standard for identifying laboratory observations: a 5-year update. 2003(0009-9147 (Print)).
00052-6&id=ga1.jpg){kind=link}
00052-6&id=gr1.jpg){kind=link}
00052-6&id=gr2.jpg){kind=link}
00052-6&id=gr3.jpg){kind=link}
00052-6&id=gr4.jpg){kind=link}