Association of type 2 diabetes with liver cirrhosis: a nationwide cohort study

Background The link between the subcategories of liver cirrhosis and type 2 diabetes is not well known. We investigated the association of type 2 diabetes mellitus with alcoholic cirrhosis and cirrhosis without alcohol. Methods This nationwide cohort study used the Taiwan National Health Insurance Research Database. Cirrhotic individuals and their matched controls were identified from 2001-2008. In all, 9 313 cirrhotic patients aged 20 years or older were matched by age, sex, and index date with the non-cirrhotic individuals (n = 37 252). Cirrhosis was categorized into alcoholic cirrhosis and cirrhosis without alcohol. Type 2 diabetes mellitus was identified from January 2001- December 2011. Results The incidence densities (per 1 000 person-months) of type 2 diabetes were as follows: 1.14 (95% CI: 1.09-1.20) in the non-cirrhotic group, 1.88 (CI 1.76-2.01) in patients with cirrhosis, 1.62 (CI 1.48-1.78) in patients with cirrhosis without alcohol, and 2.92 (CI 2.64-3.23) in patients with alcoholic cirrhosis. The adjusted hazards ratio (aHR) for type 2 diabetes mellitus among cirrhotic individuals was 0.774 (CI: 0.715-0.8934). Alcoholic cirrhotic men had a significantly higher risk of type 2 diabetes (aHR 1.182, CI: 1.046-1.335) compared with non-cirrhotic individuals. Increased risks were seen in men (aHR 1.690; CI: 1.455-1.963) and women (aHR 1.715; CI: 1.113-2.645) with alcoholic cirrhosis compared to those with cirrhosis without alcohol. Conclusions This study indicates that alcoholic cirrhosis is a significant risk factor for type 2 diabetes mellitus compared with cirrhosis without alcohol.


INTRODUCTION
Diabetes mellitus is a global health issue. Alcoholic liver diseases have been associated with increased risk of type 2 diabetes [1]. The prevalence of diabetes in cirrhosis has been reported at 12.3-57% [2]. Previous publications have reported a high prevalence of liver diseases in diabetic patients and a high prevalence of diabetes in patients with liver disease [3,4]. Increased risks of diabetes have also been reported in patients with cirrhosis due to hepatitis C (HCV) and alcoholic liver disease but not in patients with cirrhosis due to cholestatic liver disease [1,4]. In Asia, more than half of the liver cirrhosis burden is linked to hepatitis B (HBV) and hepatitis C [5]. The prevalence of cirrhosis among Taiwanese patients with HBV is reported to be 49% [6]. Up to 30% of cirrhotic patients in Taiwan were seropositive for HBeAg while 73% had a serum HBV DNA level >10000 copies/ml [7]. A previous study including Chinese individuals in Taiwan reported that hepatitis B and C virus infection would act independently and synergistically in the development of liver cirrhosis [8]. Huang and colleagues reported that Taiwanese patients with chronic hepatitis B who develop diabetes are at increased risk of liver cirrhosis and its decompensation over time [9]. The exact incidence of liver cirrhosis especially in individual Asian countries is still Clinical Research Paper unknown. From 2011-2013, the incidence rates of primary biliary cirrhosis in Korea were respectively 0.84, 0.92 and 0.87 per 100,000 population [10]. About 33,379 patients from 58 nationwide hospitals in Japan were diagnosed with liver cirrhosis in 2008 [11]. The prevalence rates of non-alcoholic liver disease (NAFLD) in South-Pacific Asia range from 12% to 24% in population subgroups and is about 11.4-41% in Taiwan [12]. About 69.4% of patients with NAFLD have been reported with type 2 diabetes mellitus (T2DM) [13]. In a prospective follow-up study including 8 663 men, heavy drinking was associated with a 2-fold increased risk of type 2 diabetes compared with moderate drinking [14]. It is worth mentioning that similar studies have not shown such relationships in women [15,16].
Liver cirrhosis has been strongly associated with type 2 diabetes [3]. However, the association between T2DM and subcategories of cirrhosis is poorly understood. In this study, liver cirrhosis was categorized into two groups: alcoholic cirrhosis and cirrhosis without alcohol to investigate their association with diabetes mellitus.

Data source
The data sources were the 2005 and 2010 Longitudinal Health Insurance Databases (LHID 2005 and 2010). These datasets contain de-identified secondary data which have been released to the public for research purposes. This study was exempted from full review by the Institutional Review Board. Informed consents were not applicable.

Study population
The study participants included 1 878 196 enrollees sampled from the LHID 2005 and 2010 ( Figure 1). All participants were aged 20 years and older. Newly diagnosed cirrhotic patients and their matched controls (comparison individuals) were identified from January 2001-December 2008 and matched 1:4 by age, sex and index date. The index date was the first date of cirrhosis detection. The event date was the first date of T2DM diagnosis or prescription of antihyperglycemic agents. Censoring occurred in case of death or withdrawal from the study. Patients were defined as having alcoholic cirrhosis if they had one inpatient and/or two outpatient claims with reported ICD-9 CM Code: 571.2. Cirrhosis without alcohol was identified using the ICD-9 CM Code: 571.5, 571.6. Patients were classified as having type 2 diabetes mellitus if they had one one inpatient and/or two outpatient claims with reported ICD-9 CM codes: 250.xx, 790.21, 790.22, 791.5x, A181. Patients were also classified as having T2DM if they were prescribed antihyperglycemic agents (alpha-glucosidase inhibitor, biguanides, insulin, meglitinides, sulfonylureas and thiazolidinediones) from 2001-2011.

Baseline characteristics and statistical analysis
The demographic variables included sex, age, lowincome, urbanization, and medications (antihypertensives such as ACE inhibitors, β-blocker and diuretics; antihyperlipidemics such as statins and fibrates as well as antiviral drugs). Co-morbidities included hepatitis B (ICD-9: 070. The Chi-square test was used for the nominal variables between cirrhotic and comparison individuals. Pearson Chi-square was used to compare the differences in the distribution of variables between patients with alcoholic cirrhosis and those with cirrhosis without alcohol. A two-tailed t-test was used to compare the mean difference between individuals or groups. The incidence density (per 1 000 person months) of DM and its 95 % confidence interval (95% CI) were calculated. Cox proportional hazard model was used to estimate the hazard ratios (HR) and 95% CI of cirrhosis and other covariates. Because of the high rate of patients with cirrhosis, competing-risk regression was also performed using the Fine and Gray model. Statistical analyses were performed using SAS 9.3 software while a p-value <0.05 was considered as statistically significant.

RESULTS
The final enrollment included 9 313 cirrhotic and 37 252 non-cirrhotic individuals ( Figure 1). Table 1 shows the demographic characteristics of the study population. A significant proportion of cirrhotic patients were lowincome earners (cirrhotic: 1.81% versus non-cirrhotic: 0.75%) and rural inhabitants ( (Table 2). In addition, unspecified viral hepatitis was found to be a significant risk factor for T2DM (aHR 1.233, CI 1.147-1.326). However, the hazard ratio was 0.807 (CI 0.711-0.915) in patients with HBV. Compared with the non-cirrhotic individuals, the hazard ratios for T2DM were 1.182 (CI 1.046-1.335) in men and 0.889 (CI 0.601-1.345) in women with alcoholic cirrhosis ( Table 3). Those of men and women with cirrhosis without alcohol were 0.723 (CI 0.645-0.810) and 0.549 (CI 0.469-0.644), respectively. Table 4 shows the adjusted hazard ratios for T2DM in patients with alcoholic cirrhosis and cirrhosis without alcohol as the reference group. The hazards ratios were 1.929 (CI: 1.742-2.136) and 1.787 (CI 1.339-2.383) for the alcoholic cirrhotic men and women, respectively.

DISCUSSION
In this large nationwide cohort study, a positive association was found between alcoholic cirrhosis as a risk factor for the development of type 2 diabetes. Previous studies have reported significant proportions of cirrhotic patients with diabetes [17][18][19][20][21]. Nonetheless, cirrhosis was not analyzed by subgroups. In a case-control study, 47% of patients with cryptogenic cirrhosis had T2DM compared to 22% of the controls [22]. Another study showed that the risks of T2DM were 7.6 and 5.3 times higher in hepatitis C and alcoholic cirrhosis, respectively [1]. However, only a few variables were considered during analysis.
In this study, when categorized by sex, a higher risk was found particularly in men with alcoholic cirrhosis. However, after including cirrhosis subgroups in the competing risk analysis, T2DM risk was found to be significantly higher in both men and women. Heavy amounts of alcohol have shown direct diabetogenic effects [23]. One study has reported a significant association between alcohol intake and T2DM with a relative risk of 1.5 per 137.8 g of alcohol intake in the past week [15]. Other studies found that moderate alcohol consumption was negatively correlated with T2DM even with an intake beyond 48 g / day [24]. Alcohol consumption has been inversely associated with fasting and post-load insulin levels [25,26]. It has also been reported that excess alcohol may reduce insulin-mediated glucose uptake [21] and can cause injury to pancreatic islet β-cells resulting in type 2 diabetes mellitus [27]. In addition, alcohol can also cause fatty liver disease leading to dysfunction of the mitochondria, thereby increasing the risk of T2DM [28,29]. Included alcoholic cirrhosis and cirrhosis without alcohol ﹡Significantly different between cirrhotic and comparison patients, p < 0.05 ﹟Significantly different between patients with alcoholic cirrhosis and cirrhosis without alcohol p < 0.05 Abbreviation: CI, confidence interval, COPD, chronic obstructive pulmonary disease. ﹡ aHR represented adjusted hazard ratios: adjusted for low-income, residents, comorbidity, antihypertensive, antihyperlipidemics and antiviral drugs Abbreviation: CI, confidence interval, COPD, chronic obstructive pulmonary disease. ﹡ aHR represented adjusted hazard ratios: adjusted for low-income, residents, comorbidity, antihypertensive, antihyperlipidemics and antiviral drugs Participants who used higher doses of antihypertensive and antihyperlipidemic agents were also found with higher risks of T2DM when compared with non-users. Past studies have associated diuretics and β-blocker with increased risk of diabetes [26]. When analyzed by sex, we found that long-term use of ACEI, β-blocker, diuretic and fibrates women were positively associated with T2DM in both men and women. Therefore, it was important that such confounders be adjusted.
We also found a lower risk of T2DM among patients suffering from HBV. The HR was lower but not significant for HCV. However, unspecified hepatitis was significantly associated with increased risk of T2DM. Previous studies have reported diverging conclusions on the association between HCV and T2DM [30]. Results from a metaanalysis indicate that HBV itself may not be pro-diabetic [31]. Among the cirrhotic individuals in this study, about 22.33 % had HBV. However, whether all the cases of cirrhosis were HBV-derived cannot be fully explained.
Hepatogenous diabetes is a common complication of liver cirrhosis [32,33]. As noted earlier, recent studies have reported increased risk of diabetes in patients with liver cirrhosis due to hepatitis C and alcoholic liver disease. However, no increased risk was found in patients with liver cirrhosis due to cholestatic liver disease [1]. Nonalcoholic fatty liver disease [34], chronic viral hepatitis [35,36], hemochromatosis [37], alcoholic liver disease [1,15] and cirrhosis [27] have been associated with T2DM. However, their relationships may be independent of the lifestyle risk factors and other metabolic diseases [15].
Our study made use of a large sample size with a longer period of follow-up. However, it was not without limitations. First, the severity of cirrhosis could not be determined. Second, there was a dearth of information about lipid and blood glucose levels. Lastly, information on alcohol intake was not available. Alcoholic cirrhosis was defined based on the ICD-9 CM codes, and hence information bias cannot be ruled out.

CONCLUSIONS
In summary, patients with alcoholic cirrhosis were found with a higher risk of type 2 diabetes mellitus compared to those with cirrhosis without alcohol. Monitoring of blood glucose levels is recommended for patients with cirrhosis.

Author contributions
PHH and YPL designed the study, managed the literature searches and summarized previous related work. JYH and ONN analyzed and interpreted the data. PHH, CCL and CCH wrote the manuscript. All authors critically reviewed and approved the final version of the manuscript. Abbreviation: CI, confidence interval, COPD, chronic obstructive pulmonary disease. ﹡ aHR represented adjusted hazard ratios: adjusted for low-income, residents, comorbidity, antihypertensive, antihyperlipidemics and antiviral drugs