This prospective observational study was conducted at Government Medical College, Kozhikode, Kerala, India. Patients with diagnosis of cirrhosis liver and satisfying the inclusion criteria attending the Department of Gastroenterology from January 2009 to January 2011 were enrolled in the study and were regularly followed since enrolment until death or January 2013 whichever was earlier. Patients aged between 18 to 75 years with diagnosis of cirrhosis with or without antecedent DM were included in the study group.

The diagnosis of cirrhosis was based on a combination of history, physical examination, laboratory and radiological investigations and upper GI endoscopy. Diagnosis of DM, impaired fasting glucose and impaired glucose tolerance were based on the expert committee follow up report on the diagnosis of diabetes mellitus; 2003 [17]. Patients with secondary diabetes, chronic pancreatitis, Hepatocellular Carcinoma (HCC) or other neoplasms at the time of enrolment, chronic renal failure, chronic cardio-respiratory illnesses and pregnant females were excluded from the study. Institutional Research and Ethical Committee approval were obtained to conduct the study. Informed written consent was obtained from patients before enrolment into the study.

All patients underwent complete haemogram, renal and liver biochemistries, serum electrolytes, α fetoprotein and abdominal sonogram at baseline. Child-Turcotte-Pugh (CTP) and Model for End stage Liver Disease (MELD) scoring were calculated using presence of ascites or encephalopathy, albumin, creatinine, bilirubin and International Normalised Ratio (INR) [18]. At each visit the patients were enquired about any major health related events or hospitalizations. Investigations were repeated six monthly in all patients or when hospitalised. Patients who failed to follow up regularly were contacted over telephone. Diagnostic paracentesis was performed at baseline in patients with ascites or at development of ascites while on follow up and or whenever there was clinical suspicion of Spontaneous Bacterial Peritonitis (SBP). Development of SBP, other infections/sepsis, hepatic encephalopathy, renal failure, hepatocellular carcinoma as well as death were noted. Patients were assigned into two groups; patients with cirrhosis and diabetes (Group A), and patients with cirrhosis without DM (Group B). Patients with cirrhosis and diabetes were further sub grouped as those with diagnosis of diabetes prior to the diagnosis of cirrhosis {Antecedent Diabetes (AD)} and those having diagnosis of diabetes after detection of cirrhosis (hepatogenous diabetes or HD). Demography, aetiology, laboratory data and the disease course over the follow up period were compared between Group A and B as well as between AD and HD.

Quantitative variables were expressed as mean ± standard deviation and compared using t-test or Mann Whitney U test. Categorical variables were compared using Chi-square test. Predictors of mortality were analysed initially using univariate analysis and those found statistically significant were put into a multivariate forward conditional logistic regression to obtain the best model with adjusted odds ratios and 95% confidence intervals. A p-value <0.05 was considered statistically significant.

A total of 291 patients were found to be eligible for the study. Out of this, 69 patients lost to follow up and hence excluded from analysis and the rest 222 patients were included in the final analysis. A total of 116 patients had DM and cirrhosis, of which 59 were diagnosed to have HD while 48 had AD and nine patients had a simultaneous diagnosis of cirrhosis and diabetes. Alcohol and cryptogenic cirrhosis were the most common cause in both the groups, followed by hepatitis B and C. During the course of study, majority of patients were euglycaemic (n=106/222).

Patients were divided into cirrhosis with diabetes (Group A) and cirrhosis without diabetes (Group B) and baseline characteristics are depicted in [Table/Fig-1].

Mean age of patients were 54.90 years in Group A and 54.05 years in Group B. In the study, 96 (82.76%) of those with diabetes and 88 (83.02%) of those without diabetes were males. The mean duration of diabetes was 35.04 months for the entire diabetic cohort. Also, 55 (47.4%) of those with diabetes and 63 (59.4%) of non-diabetics gave history of alcohol consumption in cirrhogenic doses. Duration of cirrhosis in months was 20.05 and 19.63 respectively for Group A and B. The CTP and MELD scores were 9.96 and 16.21 in Group A versus 9.90 and 16.02 in Group B. The total serum bilirubin, albumin and Prothrombin Time (PT) prolongation were 3.80 mg/dl, 2.81 gm/dl and 6.1 seconds for Group A while were 3.76 mg/dl, 2.97 gm/dl and 6.01 seconds in Group B respectively. Mean serum ferritin level was 215.1 ng/ml in Group A while it was 111.92 ng/ml in Group B. Diabetic patients were hospitalised for a mean of 3.82 times (24.03 days) while non-diabetics were admitted 3.24 times (16.91 days). The mean number of upper gastrointestinal bleed and spontaneous bacterial peritonitis were 1.76 and 0.63 versus 1.30 and 0.59 for Group A and B respectively. Baseline demographic, clinical and laboratory variables of patients are summarized in [Table/Fig-1].

Mean age of patients were 55.58 years and 54.51 years in HD and AD groups respectively. In the study, 41 (85.4%) of those with AD and 46 (78%) of those with HD were males. The mean duration of diabetes in months was 70.71 and 7.80 for HD and AD respectively. Duration of cirrhosis in months was 13.31 and 25.02 respectively for HD and AD groups. The CTP and MELD scores were 9.96 and 15.50 in HD patients while it was 9.90 and 15.85 in AD patients. The total serum bilirubin, albumin and PT prolongation were 3.90 mg/dl, 2.85 g/dl and 5.58 seconds for HD patients while were 3.45 mg/dl, 2.75 g/dl and 6.27 seconds in AD patients. The mean serum ferritin levels were 228.94 ng/ml in HD arm while it was 203.35 ng/ml in AD arm. Baseline demographic, clinical and laboratory variables of patients are summarized in [Table/Fig-2]. Complications and mortality were compared among the study patients and are depicted in [Table/Fig-3]. Urinary Tract Infection (UTI) was more common amongst diabetics [29.3% versus 7.5%; p=<0.001] while cellulitis and Lower Respiratory Tract Infections (LRTI) were similar across the groups. Gallstones were significantly more common amongst diabetic cirrhotics (27.6% vs 13.2%, p=0.008).

A total of 95 (42.8%) patients expired during the study period; 53 (45.7%) in the diabetic arm and 42 (39.6%) in the non-diabetic arm. The mortality rates were not statistically significant between the two arms (p=0.361). The mean duration of cirrhosis in months was significantly more in patients who died (24.79 versus 16.16; p=0.001). Majority of patients with cirrhosis who died belonged to child class C (n=77, 81.1%). UTI, LRTI and SBP were significantly more common in those cirrhosis patients who had a fatal outcome (p=0.002, 0.005, <0.001 respectively). [Table/Fig-4] compares patients who died versus those who survived in the current study. Logistic regression identified a model (Cox and Snell R² value is a goodness of fit statistics used in logistic regression to evaluate the goodness of model; A value of 0.387 indicate 38.7% variability in outcome can be explained by the model parameters together) with following variables-Child (CTP) score, MELD score, UTI, LRTI and duration of cirrhosis as predictors of mortality as shown in [Table/Fig-5].

Cirrhosis is known to cause dysglycaemia by a variety of mechanisms while diabetes predisposes patients to serious liver diseases [19]. DM has also been demonstrated to have significant impact in patients with cirrhosis making the association more ominous than expected. DM is found to be associated with greater incidence of complications and mortality in cirrhosis [19].

In cirrhotic patients with and without diabetes, most of the baseline characteristics were comparable. Diabetic patients had lower albumin compared to non diabetics, which maybe secondary to albuminuria in diabetes, which is more pronounced in cirrhosis [20,21].

The serum ferritin levels were significantly higher in cirrhosis with diabetes. High ferritin in diabetes is well described in studies [22,23]. Lecube A et al., had shown that patients with chronic hepatitis C infection had higher ferritin levels and was secondary to concomitant diabetes rather than the infection proper [24]. The exact mechanism of high ferritin levels in diabetes is still unclear. It has been demonstrated that the ferritin levels will be elevated in chronic inflammatory disease, which is one of postulated cause of diabetes [25-27].

Our study has found the occurrence of complications were more in the diabetes group than the non-diabetes group. The incidence of urinary tract infections and gall stones were significantly more in diabetic cirrhotics than non-diabetics. Previous studies in cirrhosis patients have demonstrated the higher incidence of bacterial infections in those with DM, owing to its immunosuppressive effects [28-32]. In the current study, the incidence of LRTI and SBP was however not different between the diabetic and non-diabetic patients, cirrhosis per se is associated with a greater incidence of gall stones but mostly asymptomatic [33,34]. The postulated mechanisms for the increased risk of gall stones in cirrhosis include haemolysis secondary to hypersplenism, hyperestrogenism, alterations in the composition of bile due to aberrations in hepatic synthesis and transport of bile salts as well as unconjugated bilirubin [35]. Likewise, DM by itself is a risk factor for gall stone disease even in the general population [36]. Considering the increased risk of gall stones in DM and cirrhosis it is logically expected that the incidence of gall stones is more in cirrhosis with DM than those without diabetes. In the current study, only 10.9% patients had symptoms pertaining to gall stone disease, while the rest where silent gall stones detected during abdominal sonogram.

Another finding from this study is the increased incidence of portal hypertensive upper gastrointestinal bleeds in patients with cirrhosis and diabetes. A recent study had concluded that insulin resistance and hepatogenous diabetes were having a positive correlation with portal hypertension and variceal bleeding [37].

In our study, there was no statistically significant difference in the incidence of HCC or death even though both were slightly higher in the diabetic group. There is mounting body of evidence in medical literature to suggest an increased risk of HCC in type 2 DM. Patients with cirrhosis related to alcohol, hepatitis viruses B or C when associated with DM are found to have higher risk of HCC [38,39]. Some studies have demonstrated an increased mortality in cirrhosis patients with diabetes [40-43]. The increased mortality was however not due to the classical diabetes-related complications, but mostly due to liver disease related complications. The prognostic significance of DM was relevant only when patients who died of gastrointestinal bleeding, were excluded [43]. Also, most of the studies that had identified DM as a prognostic variable were long term studies with a mean follow up period more than five years suggesting that DM might be a long term prognostic variable [43-46]. The shorter follow up duration in the current study might have contributed to the non-significance of the potential adverse prognostic impact of DM on cirrhosis.

In cirrhotic patients with antecedent and hepatogenous diabetes, incidence of complications was almost similar except for upper GI bleeds, hospitalizations, HCC and deaths. HD is postulated to increase the morbidity and mortality associated with cirrhosis [3]. The postulated mechanisms are however similar to those operating in patients with cirrhosis and AD [3,43]. The current study did not reveal any difference in complications or prognosis of liver disease between the two groups.

Greater duration of cirrhosis, higher CTP and MELD scores were associated with mortality. More PT prolongation, increased total bilirubin, creatinine, low levels of platelets, sodium and albumin were associated with poorer outcome. UTI and LRTI as well as SBP were also associated with higher mortality. Number of hospital admissions was also more common in patients who expired. Diabetes, antecedent or hepatogenous was not found to be a predictor of mortality in cirrhosis. Logistic regression revealed CTP score, MELD score, UTI, and duration of cirrhosis to be independent predictors of mortality. The findings in the current study are consistent with published literature with respect to most of the analysed variables [43-47].

The major limitation of the study was shorter follow up. Cirrhosis and DM both being long term diseases the current short duration of study might have missed atleast some of the impacts of DM on cirrhosis liver. Again it is always possible that either of cirrhosis or diabetes might have developed initially and had remained undiagnosed until evaluated for the other, thereby confounding the diagnosis of HD versus AD. Yet another limitation is that an oral glucose tolerance test has not been performed for any of the patients included in the study. If it was performed it might have led to identification of a few patients with impaired glucose tolerance but without overt DM. Inadvertently this might have led to misclassifying atleast some patients as having no diabetes.

In conclusion, co-existent diabetes increases the incidence of complications in patients with cirrhosis. Although, AD and HD did not show significant differences in outcome, it is worth mentioning separately to characterize natural history of diabetes in cirrhotics. Antecedent and hepatogenous diabetes are less described entities in most of the studies and requires longer follow up studies to understand the progression of disease.

All the authors have contributed enough towards this publication to justify authorship criteria. TMR designed the study, recruited patients, collected the clinical and laboratory data and participated in manuscript preparation. AHRR participated in patient recruitment and data collection. GSZ prepared the manuscript. RPA edited the manuscript. The manuscript has been read and approved by all the authors. We warrant that this manuscript is original and has not been in part or in whole simultaneously submitted to or published in another journal.