There have been several reports about the clinical association between type 2 diabetes mellitus (DM) and nonalcoholic fatty liver disease (NAFLD). However, most of the studies were about the unilateral effects of type 2 DM on NAFLD, and studies on the reverse relation are rare. Thus, this study was designed to investigate the effect of NAFLD on type 2 DM. We conducted a prospective cohort study on 25,232 Korean men without type 2 DM for 5 years. We serially checked the various metabolic factors including fasting glucose and hemoglobin A1c (HbA1c), and monitored the development of type 2 DM. The incidence rate of type 2 DM was compared according to the degree of NAFLD (normal, mild, and moderate to severe), and a Cox proportional hazards model was used to measure the hazard ratios (HRs) of NAFLD on type 2 DM. The incidence rate of type 2 DM increased according to the degree of NAFLD (normal: 7.0%, mild: 9.8%, moderate to severe: 17.8%, P < 0.001). Even after adjusting for other multiple covariates, the HRs (95% confidence interval [CI]) for type 2 DM development was higher in the mild group (1.09; 0.81-1.48) and moderate to severe group (1.73; 1.00-3.01) compared to the normal group, respectively (P for trend <0.001). Conclusion: The development of type 2 DM is potentially more associated with more progressive NAFLD than a normal or milder state. In addition, NAFLD was an independent risk factor for the future development of type 2 DM. These results suggest the potential availability of NAFLD as an early predictor of type 2 DM. (HEPATOLOGY 2013;57:1378–1383)
Nonalcoholic fatty liver disease (NAFLD) is causing great concern for its clinical association with metabolic diseases such as cardiovascular disease and type 2 diabetes mellitus (DM).1-3 There are animal and human disease models sustaining the hypothesis that a primary hepatic disease could determine the development of type 2 DM.4, 5 In addition, Epidemiologic studies have shown that surrogate markers of NAFLD (e.g., transaminases and γ-glutamyltransferase) and semiquantitative assessment of fatty liver (ultrasound) predict the development of type 2 DM.1, 6
However, so far there have been only limited data for the relationship between the future development of type 2 DM and the baseline degree of NAFLD.1, 7 In addition, there was no definite clinical guideline on the prognosis of NAFLD patients with a nondiabetic and noninsulin intolerance state. For instance, several questions can be raised: “What is the effect of NAFLD on type 2 DM?” or “Can insulin sensitivity get worse in type 2 DM by the effect of NAFLD with the passage of time?” To answer these questions, a prospective cohort study on a large scale should be conducted. Thus, in a large sample of Korean men we performed a prospective cohort study to determine whether ultrasonographically detected NAFLD can predict the future development of type 2 DM with particular respect to the degree of NAFLD.
Subjects and Methods
A prospective cohort study was conducted to examine the association between NAFLD and the future development of type 2 DM. The study cohort population comprised Korean male workers from a medical health check-up program at the health promotion center of Kangbuk Samsung Hospital, Sungkyunkwan University, Seoul, Korea. All employees participated in either annual or biennial health check-ups, as required by Korea's Industrial Safety and Health law. Most of the study population were the employees and family members of various industrial companies from all around the country.
A total of 46,719 men who had been examined with abdominal ultrasonography (US) for a medical check-up in 2005 participated in this study. Among the 46,719 participants, 13,070 men were excluded based on the following exclusion criteria that might influence the type 2 DM or US findings of the liver as a result of other liver disease: 238 had a past history of a malignancy; 323 had a past history of cardiovascular disease; 3,241 were receiving medication for lipid-lowering agents; 6,375 had an alcohol intake of ≥20 g/day; 2,934 had elevated γ-glutamyltransferase (GGT) levels (>100 U/L); 630 had elevated alanine aminotransferase (ALT) levels (>100 U/L); 2,224 had a positive serologic marker for hepatitis B surface antigen (HBsAg); 71 had a positive serologic marker for hepatitis C virus antibody (HCVAb); 523 had abnormal liver US findings of chronic liver disease, liver cirrhosis, and/or current or past history of clonorchiasis in 2005; and 2,524 had a baseline type 2 DM at the initial examinations. Because some participants had more than one exclusion criteria, the total number of men who were eligible for the study was 33,298. We further excluded 8,066 participants who did not attend any follow-up visit between 2006 and 2010. Accordingly, 25,232 participants were included in the final analysis and were observed for the development of type 2 DM. The total follow-up period was 95,170.4 person-years and average follow-up period was 3.77 (standard deviation [SD], 1.38) person-years. Ethics approval for the study protocol and analysis of the data were obtained from the Institutional Review Board of Kangbuk Samsung Hospital.
Clinical and Laboratory Measurements.
All participants were required to fast for 12 hours before a physical examination by trained staff and physicians using standard protocols.
The body mass index (BMI) was calculated as the weight (kg) divided by the square of the height (m). Height and weight were measured after an overnight fast, with the subjects wearing a lightweight hospital gown and no shoes. Waist circumference (WC) was measured at the midpoint between the lower limit of the ribcage and the iliac crest. Blood pressure (BP) was measured two times in seated subjects after a 5-minute rest using a mercury sphygmomanometer according to the Hypertension Detection and Follow-up Program protocol. The mean of these measurements was used in analyses. The presence and degree of a fatty liver was defined as abnormal hepatic features seen by US (Logic Q700 MR; GE, Milwaukee, WI) which was sustained by standard criteria.8, 9 Abdominal US were carried out by experienced radiologists who were unaware of the aims of the study and were blinded to the laboratory values. Images were captured in a standard fashion, with the patient in the supine position, with the right arm raised above the head. All US images were stored in the image server and were also taken with instant film for later inspection by the radiologists and physicians.
The presence of metabolic syndrome (MetS) was determined according to the joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention.10 Elevated BPs were defined as a systolic or diastolic BP of 130/85 mmHg or higher; elevated fasting serum glucose level was defined as 100 mg/dL or greater; high serum triglyceride levels were defined as 150 mg/dL or more; low high-density lipoprotein (HDL)-cholesterol levels were defined as less than 40 mg/dL in men and elevated WC was defined as more than 90 cm in men.
MetS was also defined according to the definition of the International Diabetes Federation (IDF).11 According to the IDF definition, for participants to be defined as having IDF-defined MetS they must meet three criteria: central obesity defined with ethnicity-specific WC values (≥90 cm in men) plus any two of the following four factors: high serum triglyceride levels (≥150 mg/dL), low HDL-cholesterol levels (<40 mg/dL in men), elevated systolic or diastolic BPs (≥130/85 mm Hg), and high fasting serum glucose levels (≥100 mg/dL). A family history of diabetes was defined as parents, brother, or sister of participants being diagnosed as diabetic by physicians.
Blood samples were collected after more than 12 hours of fasting and were drawn from an antecubital vein. Serum levels of aspartate aminotransferase (AST), ALT, and GGT were measured using Bayer Reagent Packs (Bayer HealthCare, Tarrytown, NY) on an automated chemistry analyzer (ADVIA 1650 Autoanalyzer; Bayer Diagnostics, Leverkusen, Germany). High-sensitivity C-reactive protein (hsCRP) was analyzed by performing particle-enhanced immunonephelometry using the BN System (Dade Behring, Marburg, Germany). Insulin levels were measured with immunoradiometric assays (Biosource, Nivelles, Belgium). Insulin resistance was calculated with the homeostasis model assessment of insulin resistance (HOMA-IR) as described by Matthews et al.12: fasting serum insulin (IU/ dL) × fasting serum glucose (mg/dL)/22.5.
All participants were asked to respond to a questionnaire on health-related behavior and their family history of DM. Questions about alcohol intake included the frequency of alcohol consumption on a weekly basis and the usual amount that was consumed on a daily basis (≥20 g/day). We considered persons reporting that they smoked at that time to be current smokers. In addition, the participants were asked about their weekly frequency of physical activity, such as jogging, bicycling, and swimming that lasted long enough to produce perspiration (≥1 time/week). The development of type 2 DM was assessed from the annual records of all participants and defined as fasting serum glucose ≥126 mg/dL or hemoglobin A1c (HbA1c) ≥6.5%.13 Also, participants who had a history of diabetes, or were currently treated with antidiabetic agents, including insulin-based on the self-report questionnaire at each visit, were considered to have DM. Hypertension was defined as taking antihypertensive medication or having BP ≥140/90 mmHg at the initial examinations.
Data were expressed as means ± SD or medians (interquartile range) for continuous variables and percentages of the number for categorical variables. One-way analysis of variance (ANOVA) and χ2 tests were used to analyze the statistical differences among the characteristics of the study participants at the time of enrollment in relation to the NAFLD categories. Categories of the NAFLD comprised the following: normal, mild, moderate, and severe. Moderate (n = 1,113, 4.41%) and severe NAFLD (n = 36, 0.14%) were combined into a moderate to severe NAFLD category for analyses, owing to the small number of severe NAFLD. The distributions of continuous variables were evaluated and log transformations were used in the analysis as required. For incident type 2 DM cases, the time of type 2 DM occurrence was assumed to be the midpoint between the visit at which type 2 DM was first diagnosed and the baseline visit (2005). The person-years were calculated as the sum of follow-up times from the baseline until an assumed time of type 2 DM development or until the final examination of each individual. We used Cox proportional hazards models to estimate adjusted hazard ratios (HRs) and 95% confidence intervals (CI) for incident type 2 DM comparing the mild and moderate to severe NAFLD categories versus the normal group. In the multivariate models, we included variables that might confound the relationship between NAFLD and type 2 DM, which include age, WC, triglyceride, HDL-cholesterol, systolic BP, log(hsCRP), log(HOMA-IR), serum creatinine, family history of diabetes, regular exercise, and MetS. For the linear trends of risk, the number of NAFLD categories was used as a continuous variable and tested on each model. To use the Cox proportional hazards models, we checked the validity of the proportional hazards assumption by log-minus-log-survival function and found it to be graphically unviolated. P < 0.05 was considered statistically significant. Statistical analyses were performed with PASW Statistics 18 (SPSS, Chicago, IL).
General Characteristics of the Study Participants.
During the 95,170.4 person-years' follow-up, 2,108 (8.4%) incident cases of type 2 DM developed between 2006 and 2010. Compared with the analytic cohort (n = 25,232), 8,066 participants not included in the analytic cohort were 2.3 years older (44.7 versus 42.4) than the analytic cohort, and had a less favorable baseline metabolic profile in age, BP, triglyceride, fasting serum glucose, recent smoking status, and hypertension (Supporting Table S1).
The baseline characteristics of the study participants in relation to the NAFLD categories are presented in Table 1. At baseline, the mean (SD) age and BMI of study participants were 42.5 (±7.1) years and 24.2 (±2.7) kg/m2, respectively. There were clear dose-response relationships between all of the listed variables and NAFLD categories.
Table 1. Baseline characteristics of participants according to NAFLD categories (N=25,232)
|Person-year (total)||95,170.4||61,936.4||28,942.3||4,291.6|| |
|Person-year (average)||3.77 ± (1.37)||3.78 ± (1.37)||3.75 ± (1.40)||3.73 ± (1.37)|| |
|Age (years)||42.5 ± (7.1)||42.4 ± (7.3)||42.7 ± (6.9)||41.2 ± (6.1)||<0.001|
|BMI (kg/m2)||24.2 ± (2.7)||23.3 ± (2.4)||25.7 ± (2.3)||27.6 ± (2.7)||<0.001|
|WC (cm)||83.6 ± (7.5)||81.2 ± (6.8)||87.8 ± (6.3)||92.6 ± (7.0)||<0.001|
|Systolic BP (mmHg)||114.3 ± (14.0)||112.9 ± (13.6)||116.3 ± (14.0)||120.7 ± (14.9)||<0.001|
|Diastolic BP (mmHg)||77.0 ± (9.4)||75.9 ± (9.1)||78.8 ± (9.6)||80.4 ± (10.3)||<0.001|
|Total cholesterol (mg/dL)||192.5 ± (31.0)||187.4 ± (29.9)||201.2 ± (30.7)||205.9 ± (31.4)||<0.001|
|Triglyceride (mg/dL)||122 (89-171)||107 (80-146)||154 (115-209)||174 (130-235)||<0.001|
|HDL-cholesterol (mg/dL)||49.7 ± (10.0)||51.4 ± (10.4)||46.6 ± (8.3)||45.2 ± (7.4)||<0.001|
|LDL-cholesterol (mg/dL)||113.7 ± (26.1)||109.7 ± (25.2)||120.7 ± (26.0)||124.6 ± (27.2)||<0.001|
|Fasting serum glucose (mg/dL)||95.2 ± (8.1)||94.2 ± (7.7)||96.9 ± (8.3)||98.5 ± (8.8)||<0.001|
|HbA1c (%)||5.3 ± (0.3)||5.3 ± (0.3)||5.4 ± (0.3)||5.5 ± (0.3)||<0.001|
|HOMA-IR||1.91 (1.48-2.51)||1.71 (1.36-2.17)||2.32 (1.82-2.93)||2.84 (2.24-3.56)||<0.001|
|Insulin (uU/dL)||8.8 ± (3.3)||7.9 ± (2.7)||10.2 ± (3.4)||12.4 ± (4.6)||<0.001|
|Serum creatinine (mg/dL)||1.13 ± (0.15)||1.12 ± (0.17)||1.14 ± (0.10)||1.14 ± (0.11)||0.002|
|hsCRP (mg/L)||0.05(0.03-0.11)||0.04 (0.02-0.09)||0.07 (0.04-0.14)||0.10 (0.06-0.18)||<0.001|
|AST (U/L)||23 (19-28)||22 (19-26)||25 (21-30)||31 (26-38)||<0.001|
|ALT (U/L)||25 (18-34)||21 (17-27)||32 (25-43)||51 (37-66)||<0.001|
|GGT (U/L)||27 (19-43)||23 (17-35)||36 (25-54)||48 (32-69)||<0.001|
|Current smoker (%)||39.9||35.4||40.7||42.0||0.022|
|Regular exercise (%)||14.8||16.6||12.1||7.7||<0.001|
|Family history of diabetes (%)||30.1||27.8||33.3||37.8||<0.001|
|IDF-defined MetS (%)||9.0||3.4||18.3||37.4||<0.001|
|Development of type2 DM (%)||8.4||7.0||9.8||17.8||<0.001|
In contrast to participants without incident type 2 DM, those with incident type 2 DM were slightly older (42.9 versus 42.4) and more likely to have the hypertension and NAFLD (Supporting Table S2).
Risk of Type 2 DM According to the Severity of NAFLD.
Table 2 shows the HRs and 95% CIs for type 2 DM according to the NAFLD categories. In an unadjusted model, the HRs and 95% CIs for type 2 DM comparing mild to moderate to severe NAFLD versus normal NAFLD were 1.42 (1.30-1.56) and 2.58 (2.22-2.99), respectively (P for trend <0.001). These associations were attenuated, but still remained statistically significant, even after further adjustments for covariates in models 1 and 2. In model 2, the adjusted HRs and 95% CIs for type 2 DM were 1.09 (0.81-1.48) and 1.73 (1.00-3.01), respectively (P for trend <0.001).
Table 2. Hazard ratios (HRs) and 95% confidence intervals (CI) for the incidence of type 2 DM according to NAFLD categories
|NAFLD|| || || || || || |
|Normal||61,936.4||1,146||18.5||1.00 (reference)||1.00 (reference)||1.00 (reference)|
|Mild||28,942.3||758||26.2||1.42 (1.30-1.56)||1.30 (1.04-1.62)||1.09 (0.81-1.48)|
|Moderate to severe||4,291.6||204||47.5||2.58 (2.22-2.99)||1.64 (1.06-2.53)||1.73 (1.00-3.01)|
|P for trend|| || || ||<0.001||<0.001||<0.001|
|Age|| || || || ||1.03 (1.02-1.04)||1.03 (1.01-1.05)|
|WC|| || || || ||1.00 (0.98-1.01)||0.99 (0.97-1.01)|
|Triglyceride|| || || || ||1.00 (1.00-1.01)||1.00 (1.00-1.01)|
|HDL-cholesterol|| || || || ||1.00 (0.99-1.01)||1.01 (0.99-1.02)|
|Systolic BP|| || || || ||1.01 (1.00-1.01)||1.00 (1.00-1.01)|
|Log(hsCRP)|| || || || ||1.10 (1.00-1.21)||1.03 (0.90-1.17)|
|Log(HOMA-IR)|| || || || ||3.26 (2.48-4.30)||2.63 (1.78-3.88)|
|Serum creatinine|| || || || ||1.38 (0.61-3.13)||2.33 (0.75-7.31)|
|Family history of diabetes|| || || || || ||1.76 (1.35-2.29)|
|Regular exercise|| || || || || ||0.88 (0.63-1.23)|
|MetS|| || || || || ||1.79 (1.26-2.54)|
We found a strong association between degree of NAFLD and the subsequent development of type 2 DM in Korean men. This association was independent of age, BMI, total-cholesterol, HOMA-IR, serum creatinine, recent smoking status, regular exercise, hypertension, family history of DM, or metabolic syndrome.
Our study also showed that in a large number cohort of Korean men, 8.4% developed type 2 DM over a 5-year period and the degree of NAFLD predicted the development of subsequent type 2 DM even in the subgroup without type 2 DM at baseline. This was particularly seen in the participants with moderate to severe NAFLD when compared to subjects without NAFLD. Thus, our findings suggest that the degree of NAFLD could independently predict the development of type 2 DM with the elapse of time.
These findings are the most remarkable points of this study corresponding to our study aim. We initially focused on investigating the effect of NAFD on type 2 DM because these studies must be helpful to research the prognosis of NAFLD. In the clinical field, clinicians sometimes encounter NAFLD patients with nondiabetic and noninsulin intolerance states. Nevertheless, it is not easy to suggest a definite clinical guideline, because there was no specific guideline for NAFLD patients without metabolic diseases like DM or insulin intolerance. However, if these clinical outcomes would be addressed in consecutive studies, it might contribute to NAFLD care in terms of preventing type 2 DM. For example, if NAFLD could be confirmed as the greater predictor of type 2 DM, we could manage the risk of type 2 DM more efficiently. On the basis of these contexts, we think that our study has important clinical implications.
There have been several studies showing findings in accordance with our study. Bae et al.1 showed the combined effect of NAFLD and impaired fasting glucose on the development of type 2 DM. However, their study is not enough to suggest the independent role of NAFLD on the development of type 2 DM, because the higher risk group of type 2 DM only existed in the impaired fasting glucose state. Although Shibata et al.7 also demonstrated that NAFLD is a risk factor for type 2 DM in middle-aged Japanese men, it has a limitation as a case-control study that is not enough to show a causative relation. However, our study was designed as a prospective cohort study with a large sample size that would be better able to reveal the significant association between NAFLD and type 2 DM.
As the mechanism of these results, we can suggest theories concerning the effects of NAFLD on glucose metabolism. First, based on the “lipotoxic” hypothesis, the influx of free fatty acid from the excessive adipose tissue to the peripheral tissues would induce the development of insulin resistance especially.14 Second is the well-established relationship between fatty liver and insulin resistance. Kotronen et al.15 reported that in a population of 271 nondiabetic individuals with varying degrees of intrahepatic fat content, serum C-peptide was the strongest correlate of liver fat. Gastaldelli et al.16 also suggested that excessive fatty liver had an important systemic consequence to adversely affect insulin sensitivity in humans. In a rodent model, increasing or decreasing intrahepatic fat content had an opposite effect on muscle insulin sensitivity, which suggests that fat accumulation in the liver may be the primary event leading to peripheral insulin resistance.17 Thus, peripheral insulin resistance may be the consequence of intrahepatic fat accumulation, resulting in the lower release of humoral factors that impair insulin sensitivity in peripheral tissue.18 In summary, NAFLD can be a cause of insulin resistance as well as an outcome.
There are several limitations in our study. First, the presence of NAFLD was assessed by US instead of a pathologic method, as it is inappropriate to perform invasive tests in a population-based epidemiological study.19 In addition, although US is regarded as reasonably accurate, it cannot identify fatty infiltration of the liver below the threshold of 30%.20 Particularly, we did not have information on the intra- or interobserver reliability of US examinations. However, all examinations were carried out and interpreted by experienced radiologists evaluating live echo images using widely established methods and criteria. Second, alcohol intake, a common cause of fatty liver disease in Korean men, was self-reported. Therefore, it might have been underestimated. However, we initially excluded the participants with serum GGT level >100 U/L, a commonly used marker of alcohol consumption.
Third, bias from follow-up loss may have affected our results. Participants not included in the analysis (n = 8,066) were older and had less favorable metabolic profiles at baseline than those in the analytic cohort. It is likely that follow-up loss will be unexpected, especially in those who are in poor health. This follow-up loss of high-risk people would probably lead to a conservative bias and subsequent underestimation of risk. Fourth, participants were self-selected, so this study may show participant selection bias.
In conclusion, our findings, which were obtained from a large cohort, indicated the clinical utility of NAFLD as an early predictor of type 2 DM. These findings support the concept that NAFLD is an early feature of type 2 DM, and may be helpful to set preventive strategies for type 2 DM. Accordingly, we have to keep in mind the significance of NAFLD as a possible predictor of type 2 DM as well as a surrogate marker of fatty infiltration.
J-H.R. coordinated the study, analyzed the data, and wrote the article. S.K.P. and M.H.S. collected and interpreted the data, contributed to discussion, and reviewed and edited the article. S.K.P. and M.H.S. contributed equally to the work reported here and should be considered as first authors. H.C.S. contributed to review of the article. J-H.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.