Is liver disease a threat to patients with metabolic disorders?

Abstract

The association of metabolic disorders with liver disease is receiving increasing attention in the gastroenterological community. Cohort studies have shown that advanced liver disease may stem from metabolic disorders, via fatty liver, non‐alcoholic steatohepatitis, cryptogenic cirrhosis, and eventually hepatocellular carcinoma. In both obesity and diabetes, deaths from cirrhosis are higher than expected, mainly in subjects with no or moderate alcohol consumption, but high rates of fatty liver disease have been associated with all features of the metabolic syndrome. Also the risk of hepatocellular carcinoma is higher than normal, being dependent on body mass index (BMI) in obesity, and independent of age, BMI, gender and race in diabetes. Finally, metabolic liver disease may interact with hepatitis C virus infection, increasing the risk of steatosis and liver disease progression, as well as reducing the chances of an effective antiviral treatment. There is evidence that treatments aimed at reducing insulin resistance are also effective in improving liver histology. Although cardiovascular disease remains the major cause of increased morbidity and excess mortality in metabolic disorders, the risk of progressive liver disease should no longer be underestimated, being a threat to millions of people at risk in the present epidemics of obesity and diabetes, and therapeutic strategies need to be tested.

• Diabetes
• epidemiology
• fatty liver
• hepatocellular carcinoma
• liver disease
• metabolic syndrome

Abbreviations
ALF	=	acute liver failure
ALT	=	alanine aminotransferase
AR	=	attributable risk
AST	=	aspartate aminotransferase
BMI	=	body mass index
CC	=	cryptogenic cirrhosis
CI	=	confidence interval
DM	=	diabetes mellitus
FFA	=	free fatty acids
GGT	=	gamma‐glutamyltransferase
HBV	=	hepatitis B virus
HCC	=	hepatocellular carcinoma
HCV	=	hepatitis C virus
HR	=	hazard ratio
ICD‐9	=	international classification of diseases (9th edition)
IRS‐1	=	insulin receptor substrate‐1
MS	=	metabolic syndrome
NAFLD	=	non‐alcoholic fatty liver disease
NASH	=	non‐alcoholic steatohepatitis
OR	=	odds ratio
PI‐3K	=	phosphatidyl inosytol‐3 kinase
PPAR	=	peroxisome‐proliferator activated receptor
RR	=	relative risk
SIR	=	standardized incidence ratio
SMR	=	standardized mortality rate
T2DM	=	type 2 diabetes mellitus
TNF	=	tumor necrosis factor

Introduction

The association of liver disease with metabolic disorders is receiving increased attention 1. Hepatic steatosis is a common feature in obesity and type 2 diabetes mellitus (T2DM), but its potential significance as a cause of advanced liver disease has long been underestimated. More recently, the identification of non‐alcoholic fatty liver disease (NAFLD) as a clinical entity 2 and cohort studies suggesting a potential link between pure fatty liver, non‐alcoholic steatohepatitis (NASH), cryptogenic cirrhosis 3, and eventually hepatocellular carcinoma (HCC) 4 have suggested that metabolic disorders per se may cause advanced liver failure.

Although cardiovascular risk remains the main cause accounting for increased morbidity and excess mortality in obesity and T2DM, many studies are in keeping with the hypothesis that liver failure might also be a threat to the millions of people at risk, whose number is further increasing due to the epidemics of obesity and T2DM in western countries 5. Metabolic disorders also interact with viral infections in the pathogenesis of liver failure, particularly in areas where hepatitis C (HCV) virus is endemic.

The aim of the present report is to review the existing evidence that advanced liver disease may develop in patients with metabolic disorders, and to review the pathogenic and therapeutic relevance of the association of metabolic liver disease with viral infection.

Search strategy and selection criteria

We searched MEDLINE to May 2005, using different combinations of the following key‐words: obesity, diabetes, survival, cirrhosis, liver disease, hepatocellular carcinoma, hepatitis B, hepatitis C, fatty liver, and related terms. We also searched the bibliographies of most recent articles for relevant references. Because of the large number of articles identified and limitations for quoting references, the final decision on what to include was based on personal judgment, with preferences for the most recent articles.

Key messages

• Fatty liver, commonly observed in metabolic disorders, may progress to advanced liver disease, via non‐alcoholic steatohepatitis, cryptogenic cirrhosis, and eventually hepatocellular carcinoma.

• The prevalence of liver disease, death rates for cirrhosis and the risk of hepatocellular carcinoma are higher than expected in subjects with features of the metabolic syndrome, particularly in obesity and diabetes, and metabolic liver disease may also interact with hepatitis C virus infection, favoring disease progression and reducing the chances of an effective antiviral treatment.

• Although cardiovascular disease remains the major cause of increased morbidity and excess mortality in metabolic disorders, the risk of progressive liver disease should no longer be underestimated, being a threat to millions of people at risk in the present epidemics of obesity and diabetes.

Association of liver disease with diabetes, obesity and the metabolic syndrome

Fatty liver is the marker of metabolic liver disease, with or without raised aminotransferase levels, namely alanine aminotransferase (ALT). Although an association between the severity of liver disease and raised aminotransferases has not been unequivocally demonstrated 6, 7, raised enzymes are the most common reason for patients' concern and extensive diagnostic workup. Several epidemiological data have convincingly associated the whole spectrum of NAFLD with the metabolic syndrome (MS) 6, 8–10, and with its individual components as suggested by Adult Treatment Panel III proposal 11. There is now evidence that fatty liver and insulin resistance may be early predictors of metabolic disorders, even in non‐obese, non‐diabetic subjects, particularly in the normal‐weight population 12.

Diabetes

The relationship of liver disease to T2DM is difficult to evaluate. End‐stage liver disease can cause glucose abnormalities and overt diabetes (the so‐called ‘hepatogenous diabetes’) 13, being a source of bias in prevalence studies. Old studies, based on liver biopsy, were also biased by the selection of cases with clinically suspected liver disease 14. Limiting the analysis to fatty liver, the estimated prevalence of diabetes ranges from 21% to 78% in recent studies, the difference being largely dependent upon diagnostic criteria 15–17. Ultrasonography is not sensitive enough to detect all cases with fatty liver infiltration, the detection limit being around 25%–30% of hepatocytes with fatty droplets 18, 19. No systematic studies are available on the prevalence of diabetes in subjects with more severe histological alterations (necroinflammation, fibrosis), potentially progressive to advanced liver disease, which would require the extensive use of liver biopsy.

Surprisingly, very few cohort studies have been published on the prevalence of clinically‐detectable liver disease or raised liver enzymes in subjects with T2DM. The overall incidence of elevated aminotransferase levels in T2DM treated with oral antidiabetic agents is reported in the order of 50/10,000 person‐years, but the criteria of ascertainment were questionable, and mild liver enzyme elevation was not considered 20. In the records of the Department of Veteran Affairs, the incidence of chronic non‐alcoholic liver disease was doubled among patients with diabetes 21. In turn, elevated ALT 22, 23 or GGT 24 were prospectively associated with the development of type 2 diabetes in the general population, and the predictive role of ALT was maintained after adjustment for insulin resistance 25.

Obesity

More data are available in obesity. Ultrasound studies show evidence of hepatic steatosis in 76% of cases not drinking alcohol in toxic amounts 26, and that liver histology is compatible with NAFLD of variable severity in over 80% of morbidly obese subjects 27, 28, where bariatric surgery provides an easy access to liver biopsy. Among 551 patients prospectively examined, steatosis was found in 86%, fibrosis in 74%, mild inflammation or NASH in 24%, and unexpected cirrhosis in 2% 27. In overweight subjects with abnormal aminotransferase levels but without overt liver disease, steatosis ⩾10% was observed in 79% of cases, fibrosis outside the portal tracts in 30%, necroinflammation in 81% 29. The prevalence of fatty degeneration is progressively higher in relation to glucose control (from normoglycemia, to impaired glucose tolerance, to T2DM) 30, 31.

A role of obesity in high aminotransferase cases has been repeatedly demonstrated. High levels of alanine and aspartate aminotransferases (ALT and AST), as well as raised gamma‐glutamyltranspeptidase (GGT) activity, are frequently observed, in association with body mass index (BMI) and raised insulin levels 32, 33, also in apparently healthy obese subjects with fatty liver at ultrasonography 34, the prevalence being higher in males and in relation to raised BMI. In a cohort study, Vozarova et al. 23 showed that raised liver enzymes were associated with body fat and insulin resistance, measured by the clamp technique. Prati et al. 35 confirmed that ALT levels are significantly associated with BMI in healthy individuals; their data suggested revising the upper limits of normal ALT to improve the sensitivity for identifying liver disease. In the general population, Stranges et al. 36 demonstrated an independent role of central adiposity in predicting increased levels of aminotransferases and GGT, a possible expression of an unrecognized hepatic disease.

Hyperlipidemia

In hyperlipidemic patients admitted to an urban hospital‐based clinic for the evaluation and management of hyperlipidemia, fatty liver was detected at ultrasounds in 50%, the prevalence being higher in subjects with severe hypertriglyceridemia and mixed hyperlipidemia 37. Hyperglycemia was an independent predictor of fatty liver, confirming the role of multiple metabolic disorders in the pathogenesis of steatosis.

Hypertension

In hypertension, the prevalence of ultrasonographically‐detected steatosis is more than doubled when compared to subjects with normal blood pressure 38. An association of hypertension with insulin resistance has long been documented 39. The higher prevalence of fatty liver in non‐obese hypertensive patients with normal liver enzymes was related to increased insulin resistance and body weight 38.

Metabolic syndrome

In a recent cross‐sectional analysis of 3,405 South Korean adults 40, an association between ALT and the metabolic syndrome was found in both sexes, independently of age, BMI, smoking, alcohol drinking, and sedentary life style. The association was maintained even in the range below the current upper limits of ALT. Similarly, in 799 obese Italian patients entering a weight‐reducing program, raised ALT levels were demonstrated in 21.0% of cases, with a lower prevalence of raised AST (8.6%) and GGT (13.7%). The median value of ALT increased slightly, but significantly, with obesity class (P for trend: = 0.001). Hyperglycemia (⩾110 mg/dl) and hypertriglyceridemia (⩾150 mg/dl) were the features of the metabolic syndrome most commonly associated with raised liver enzymes 41. In logistic regression analysis, after correction for age, gender, BMI and features of the metabolic syndrome, insulin resistance (homeostasis model assessment) maintained a highly predictive value for raised ALT, AST and GGT, suggesting that raised liver enzyme levels, indicative of subclinical liver disease, may be part of the insulin‐resistance (metabolic) syndrome.

As previously outlined, the presence of multiple metabolic disorders is strictly associated with a higher prevalence of liver disease. Both in subjects cared for in liver units 6 and in patients observed in metabolic/diabetes units 42, the association of obesity with diabetes and/or altered lipid metabolism leads to a multiplicative effect on the final prevalence of MS, and significantly increases the risk of more severe stages of liver disease 43–45.

Association between diabetes and viral liver disease

Diabetes in HCV

The prevalence of abnormal glucose regulation and diabetes is significantly elevated in HCV infected patients, after correction for the degree of liver cell failure. At population level (Third National Health and Nutrition Examination Survey), in persons older than 20 years of age, the presence of diabetes was over three times more prevalent in subjects with HCV infection than in those without HCV 46. In American‐Indian women, increasing age, obesity, and positive HCV status were each independently related to the diagnosis of diabetes 47. Older age, obesity, severe liver fibrosis and family history of diabetes help identify HCV patients at risk for development of T2DM 48. A threefold increase in the prevalence of glucose abnormalities was observed in HCV positive patients with chronic hepatitis in comparison with HCV negative subjects (32% versus 12%; P = 0.0003) 49. In contrast, among patients with cirrhosis the difference was not statistically significant, although the prevalence of both diabetes and impaired fasting glucose (110–125 mg/dL) were more prevalent in HCV positive patients (40% versus 36% in HCV negative). This suggests that the connection between HCV infection and diabetes starts at early stages of hepatic disease.

Diabetogenic effect of HCV

The mechanism accounting for the pro‐diabetogenic effect of HCV is under investigation. In liver biopsies of HCV‐infected subjects, defects in upstream insulin signaling pathways have been demonstrated at the level of insulin receptor substrate‐1 (IRS‐1) tyrosine phosphorylation, IRS‐1/p85 phosphatidylinositol 3‐kinase (PI3‐kinase) and IRS‐1‐associated PI3‐kinase enzymatic activity. They were not present in HCV negative biopsies, and might contribute to insulin resistance, which leads to progression to type 2 diabetes mellitus in patients with HCV infection 50. In a mouse model transgenic for the HCV core gene, plasma glucose levels during a glucose tolerance test were moderately higher than in control mice, in association with insulin resistance. The levels of tumor necrosis factor‐α were also elevated, as commonly observed in human infection. The administration of an anti‐tumor necrosis factor‐α antibody restored insulin sensitivity, providing a direct experimental evidence for the contribution of HCV to the development T2DM 51.

Virus‐induced steatosis

The association of steatosis with HCV infection is larger than expected by chance. In patients with untreated chronic hepatitis C, steatosis is significantly associated with BMI (P < 0.0001) 52, namely with visceral adiposity 53, but is particularly common in genotype 3a infection 54. Steatosis also correlates with viral load 55; it disappears in subjects with sustained response to interferon and reoccurs after liver transplantation in subjects with HCV‐genotype 3, whereas it does not change in patients with HCV genotype 1, irrespective of response 56, giving support for a causal association between genotype 3 infection and fat accumulation. The pathogenesis of HCV‐related steatosis remains poorly understood. Serfaty et al. 57 found that cholesterol levels are lower in HCV‐infected patients compared to a reference male population, as well as to patients with chronic hepatitis B or with non‐alcoholic fatty liver. Hypobetalipoproteinemia is corrected by HCV eradication, suggesting a direct involvement of HCV in both hypobetalipoproteinemia and steatosis. Apoprotein AII and HCV core colocalize in human HCV‐infected liver biopsies, and the hepatic overexpression of HCV core protein interferes with the hepatic assembly and secretion of triglyceride‐rich very‐low‐density lipoproteins in a transgenic murine model 58.

Metabolic disorders and HCV‐related disease severity

Altered glucose regulation is significantly associated with insulin resistance and with the staging of liver fibrosis 48. In a large series of patients with chronic HCV‐related hepatitis, age at infection, duration of infection, serum glucose and daily alcohol intake but not BMI were independently associated with significant fibrosis 59. Patients with high serum glucose had features of the metabolic syndrome, including a higher prevalence of steatosis, as well as faster fibrosis progression rates, suggesting that hyperglycemia is an independent co‐factor of fibrosis, with a higher pro‐fibrogenic impact than overweight 59. Finally, a BMI ⩾30 kg/m² is a negative predictor of sustained response to interferon treatment 60. The reason(s) might be lower maximal interferon concentrations in obese patients, as well as a weaker biologic response to exogenous interferon‐α 60, 61.

Liver disease and survival in metabolic patients

Whatever the origin of liver disease, the key question is whether liver disease may harm patients with metabolic disorders. These patients remain at very high risk of cardiovascular diseases, but evidence is emerging suggesting that liver disease needs to be tackled as well. The study of the natural history of NAFLD recently suggested that the presence of diabetes and high BMI are significantly associated with higher rates of liver fibrosis progression, also in the absence of raised aminotransferases 62. Metabolic liver disease is slowly progressive 63, and liver‐related morbidity and mortality are expected to occur only in aged cohorts.

Liver‐related mortality and cirrhosis

The prognostic studies on liver disease‐related mortality in metabolic disorders are reported in Table I, 64–70. In a registry‐based cohort study, Jepsen et al. 68 studied overall and cause‐specific standardized mortality ratios in Danish subjects discharged from hospital with a diagnosis of fatty liver. During follow‐up, mortality was increased 2.6‐fold (95% CI 2.4–2.9) in patients with non‐alcoholic or unspecified fatty liver, independently of diabetes and after censoring patients upon diagnosis of liver cirrhosis. However, survival estimates in patients with pure fatty liver, without inflammation, were not different from those observed in the general Danish population 70, suggesting the need for additional factors favoring disease progression. In T2DM patients observed in the Verona cohort study, the standardized mortality ratio for cirrhosis during a 5‐year follow‐up was remarkably high (2.5; 95% CI, 2.0–3.2) 65, the ratio being even higher than that reported for cardiovascular disease. Less clear data had previously been reported in a Japanese cohort with a low prevalence of diabetes 64. Contrary to what is commonly believed, obesity is frequently associated with cirrhosis, even in subjects without significant alcohol intake. In a multicenter Italian study of 1402 patients with cirrhosis 29% of females and 18% of males appeared to be overnourished 71. In obesity‐related cryptogenic cirrhosis, severe liver disease was as frequent as in HCV‐related cirrhosis, and survival was even lower than in age‐ and sex‐matched cirrhosis of viral origin 72. These data were not confirmed in a U S series of NASH patients, where prognosis was either similar or less severe than in HCV‐cirrhosis 73. In a long‐term follow‐up of the NHANES I cohort, cirrhosis‐related deaths or hospitalizations were more common in obese persons (adjusted hazard ratio, 1.7; 95% CI, 1.0–3.0), after exclusion of subjects with evidence of cirrhosis at entry or during the first 5 years of follow‐up 69. The association was particularly strong among obese persons who did not consume alcohol (4.1; 1.4–11.4). In the presence of cirrhosis, diabetes is a risk factor for mortality 13, the larger mortality rate being related to an increased risk of hepatocellular failure, heralded by low platelet counts and high bilirubin 73, not to diabetes or diabetes‐related complications. In subjects followed for NAFLD, overall mortality and mortality related to liver disease (relative risk, 22.8; P = 0.003) were both more common in subjects with diabetes 74. Obesity 67 and T2DM 66 also increase the risk of acute liver failure in a few cases. On the basis of the clinical and histological findings, Caldwell and Hespenheide 67 suggested that these patients had undiagnosed NASH with silent progression to cirrhosis followed by subacute liver failure.

Table I. Liver disease‐related mortality associated with metabolic disorders.

First author (ref)	Area	Population	Result	Comment
Sasaki 64	Japan (Osaka prefecture)	Population‐based study on 32,222 death certificates of patients with diabetes; observation period, 30 years	Death rate for liver cirrhosis increases from 2.1% in the period 1960–64 to 4.6% in 1985–89.	No data on other risk factors for liver disease in the population.
De Marco 65	Italy (Verona Diabetes Study)	7,158 patients with T2DM; follow‐up, 5 years	SMR for cirrhosis as high as 2.5 (95% CI, 2.0–3.2), and higher than that observed for cardiovascular disease.	No correction for prevalence of hepatitis virus infection and cirrhosis prior to diabetes
El‐Serag 66	U S A. (Dept. of Veterans Admin. Affairs)	173,643 patients with and 650,620 without DM; follow‐up, 10–15 years.	Acute liver failure identified by ICD‐9 570. The cumulative risk of ALF significantly higher among patients with diabetes (HR, 1.44; 95% CI, 1.26–1.63). This risk is maintained after excluding patients with liver disease or hepatitis during follow‐up.	After excluding the first year of follow‐up, diabetes increases the risk of ALF, independently of recognized underlying chronic liver disease or viral hepatitis.
Caldwell 67	U S A (Liver disease registry)	2,380 patients, 167 with NASH, 215 with CC; follow‐up, 8 years.	Five middle‐aged women, with no history of liver disease, died of subacute liver failure	Obesity‐related liver disease may infrequently present as severe, subacute illness.
Jepsen 68	Denmark (Danish National Registry)	7,372 subjects with a diagnosis of fatty liver; follow‐up, up to 16 years	All causes SMR for non‐alcoholic or unspecified liver disease, 2.6 (95% CI, 2.4–2.9). In the same group, hepatobiliary SMR are as high as 19.7 (15.3–25.0). No differences in relation to T2DM.	Mortality remains increased after censoring patients upon diagnosis of liver cirrhosis. Cases with unspecified fatty liver might have biased the results
Ioannou 69	U S A (NHANES I cohort)	11,465 persons aged 25–74 years without evidence of cirrhosis at entry; mean follow‐up, 12.9 years	Cirrhosis‐related deaths or hospitalizations more common in obese persons (adjusted HR: 1.69; 95% CI, 1.0–3.0). The association is confined to persons with no or moderate alcohol consumption (adjusted HR: 4.1, 95% CI 1.4–11.4).	Obesity is a risk factor for cirrhosis‐related death or hospitalization among persons who consume little or no alcohol. The use of a combined primary outcome is subject to criticism
Dam‐Larsen 70	Denmark	215 patients with biopsy‐proven pure fatty liver (109 non‐alcoholic); median follow‐up, 16.7 years	Survival estimates different between alcoholic and non‐alcoholic fatty liver. Survival estimates in the non‐alcoholic fatty liver group are not different from the Danish population.	The study supports the concept that pure fatty liver may have a benign clinical course.

ALF = acute liver failure; BMI = body mass index; CC = cryptogenic cirrhosis; CI = confidence interval; DM = diabetes mellitus; HR = hazard ratio; ICD‐9 = International classification of diseases (9^th edition); NASH = non‐alcoholic steatohepatitis; SMR = standardized mortality rate; T2DM = type 2 diabetes mellitus.

Hepatocellular carcinoma (HCC)

The prognostic studies on progression to HCC and HCC‐related mortality in metabolic disorders are reported in Table II, 64, 75–89. In a recent mortality study of the American Cancer Society, the risk of death from HCC during 16 years of follow‐up was increased both in males (P for trend <0.001) and in females (P < 0.04) with increasing BMI class 83. These data expand previous reports linking obesity to HCC 75, 90.

Table II. Progression to hepatocellular carcinoma and HCC‐related mortality in metabolic disorders.

Author (ref)	Area	Population	Result	Comment
Sasaki 64	Japan (Osaka prefecture)	Population‐based study on 32,222 death certificates of patients with diabetes; observation period, 30 years	Death rate for HCC increases from 0.5% in the period 1960–64 to 3.2% in 1985–89.	No data on other risk factors for HCC in the population. The risk of ascertainment bias is elevated.
Moller 75	Denmark	37,957 obese persons, compared to the total Danish population; follow‐up, 4.8 years.	Relative risk for HCC, 1.9 (95% CI, 1.5–2.5). Results are unchanged by restriction to a subcohort of 8,207 persons where obesity is the primary discharge diagnosis, and are also similar irrespective of first year of follow‐up.	Overall, the incidence of cancer is increased by 16% in the cohort. Selection bias is not likely to have influenced the results.
Adami 76	Sweden (Swedish in‐patient and Cancer Registry)	153,852 patients with diabetes compared to nationwide estimates; follow‐up, 6–24 years	SIR for HCC in diabetes, 4.1 (95% CI, 3.8–4.5). The risk is higher in men (4.7, 4.2–52) than in women (3.4, 2.9–3.9).	After exclusion of subjects with alcoholism, cirrhosis and hepatitis, the risk continues to be three‐fold increased.
La Vecchia 77	Italy	Point‐prevalence case‐control study (428 with incident HCC and 1,502 controls); observation period, 13 years	In diabetes, OR for HCC, 2.1 (95% CI, 1.4–3.2) after adjustment for confounders, including history of hepatitis and liver cirrhosis, body mass index and history of liver cancer in first‐degree relatives.	A history of diabetes mellitus can explain about 8% (95% CI 5–11) of cases of liver cancer in this population.
Wideroff 78	Denmark	109,581 hospitalized subjects with diabetes; follow‐up, 4–16 years.	SIR for HCC, 4.0 (95%, 3.5–4.6) in males and 2.1 (95% CI = 1.6–2.7) in females. SIR remains elevated after exclusion of patients with reported risk factors (e.g., cirrhosis and hepatitis) or patients whose cancers were diagnosed at autopsy.	Patients hospitalized with a diagnosis of diabetes are at higher risk of developing HCC. The elevated risk may be confounded by obesity.
Braga 79	Italy(Northern Area)	Case‐control of 320 HCC and 1,408 controls; observation period, 1984–93)	Population‐ARs of HCC, 18% for cirrhosis, 16% for hepatitis, 8% for diabetes and 7% for heavy alcohol consumption. Compared with females, males had higher AR for cirrhosis (21% versus 11%), diabetes (10% versus 2%) and heavy alcohol intake (9% versus 1%).	Even if available information on dietary factors was limited, and the AR were based on arbitrary assumptions, diabetes control might significantly reduce the burden of HCC in Northern Italy.
Lagiou 80	Greece	Point‐prevalence case‐control study (333 incident HCC and 360 controls); 4‐year observation period.	History of diabetes associated with HCC (adjusted OR, 1.86; 95% CI, 0.99–3.51). The association is not confounded by any major risk factor (hepatitis B and C, alcohol consumption)	The authors speculate that diabetes may increase HCC risk either via hyperinsulinemia, or via NASH and non‐alcoholic cirrhosis
El‐Serag 81	U S A (Dept. of Veterans Affairs)	Point‐prevalence, case‐control study (823 HCC, 3,459 controls); observation period, 3 years	In diabetes adjusted OR for HCC, 1.57(95% CI, 1.08–2.28) only in the presence of other etiologic factors (virus, alcohol). The combined presence of HCV and alcoholic cirrhosis further increases the risk (OR 79.2, 60.3–103.4).	Diabetes increases the risk of HCC only in the presence of other risk factors. Alcohol exerts a primary role in this population
Nair 82	U S A (National transplant registry)	19,271 patients with cirrhosis (28% obese; 19% alcoholic, 13% cryptogenic cirrhosis)	OR for HCC in obese patients with alcoholic cirrhosis (3.2; 95% CI, 1.5‐6.6) and cryptogenic cirrhosis (11.1, 1.5‐87.4)	No effect of obesity on HCC in cirrhosis of different etiology
Calle 83	U S A (American Cancer Society Study)	Prospective study of 900,053 adults; follow‐up, 16 years	Death risk for HCC increased both in males (P for trend<0.001) and in females (P<0.04) with increasing BMI class. Compared to normal weight, a BMI in the range 30‐35 kg/m^2 gives a RR of 1.90 (95% CI, 1.46‐2.47) in males and 1.40 (0.97‐2.00) in females.	Increased body weight is associated with increased death rates for all cancers combined and for cancers at multiple specific sites. The risk for HCC is particularly relevant in severely obese persons
Verlato 84	Italy (Verona Diabetes Study)	7,148 patients with T2DM; follow‐up, 10 years	SMR for overall malignancies increased in females, but not in men. Excess mortality from hepatic cancer (SMR, 1.86, 95% CI, 1.44‐2.38) observed in both men and women.	No adjustment for other etiologic factors (alcohol, virus)
Sorensen 85	Denmark (Danish National Registry)	7,326 subjects with fatty liver; follow‐up, 1‐16 years	SIR for HCC elevated in patients with alcoholic (9.5, 95% CI, 5.7‐14.8) as well as non‐alcoholic fatty liver (4.4; 1.2‐11.4), compared with general Danish population.	Patients were censored upon diagnosis of cirrhosis; accordingly, the potential sequence of fatty liver to cirrhosis to HCC was missed.
Regimbeau 86	France	210 patients with liver resection for HCC; 18 (8.6%) with no underlying cause of liver disease	The prevalence of obesity (50% vs. 14%) and diabetes (56% vs. 11%) are higher in patients with cryptogenic liver disease than in patients chronic viral hepatitis (P<0.0001).	Obesity and diabetes may be important risk factors for cryptogenic chronic liver disease progressing to HCC
Coughlin 87	U S A (Cancer Prevention Study II)	467,922 men and 588,321 women; follow‐up, 16 years	Diabetes significantly associated with fatal HCC in men (RR, 2.19; 95% CI, 1.76‐2.72), but not in women (RR, 1.37; 0.94‐2.00)	The association is not explained by high body mass index, suggesting that diabetes is an independent predictor of mortality from cancer
Jee 88	Korea (National Health Insurance Co.)	1,298,385 Koreans; follow‐up, up to 10 years	Death rates from all cancers combined increased in subjects with fasting glucose > 140 mg/dl (HR, 1.29; 95% CI, 1.22‐1.37 in men and 1.23, 1.09‐1.39 in women). By cancer site, the association is confirmed for HCC (Males: HR, 1.59; 1.45‐1,74; Females: 1.28, 1.00‐1.66)	In Koreans, a population with a low BMI, elevated fasting serum glucose levels and a diagnosis of diabetes are independent risk factors for HCC.
Davila 89	U S A (Medicare database)	2,061 HCC and 6,183 controls; population‐based, case‐control study; age⩾65 years; observation period, 1994‐99	After adjustment for HBV, HCV, alcohol and hemochromatosis, OR for diabetes, 2.87 (95% CI, 2.49‐3.30). Similar data in analysis restricted to diabetes between 2‐3 years from HCC	Diabetes is independently associated with HCC, regardless of other major risk factors in the elderly. Several features of MS also increase HCC risk.

AR = attributable risk; BMI = body mass index; CI = confidence interval; HCC = hepatocellular carcinoma; HR = hazard ratio; MS = metabolic syndrome; NASH = non‐alcoholic steatohepatitis; OR = odds ratio; RR = relative risk; SIR = standardized incidence ratio; SMR = standardized mortality rate; T2DM = type 2 diabetes mellitus.

An association of T2DM with HCC was first established by comparison with the prevalence of diabetes observed in subjects with other tumors 91. It has been confirmed in several recent reports 21, 77, 80, 86, 92. In a case‐control study, El‐Serag et al. 81 found that diabetes increases the risk of HCC in the presence of hepatitis B, hepatitis C, or alcoholic cirrhosis, whereas in chronic HCV‐related hepatitis, steatosis per se was a risk for HCC 93. The temporal relationship between diabetes and HCC is important, due to the high prevalence of diabetes in cirrhosis. In a large cohort of subjects admitted to Veteran Administration hospitals 87, the incidence of HCC was over twice as high among patients with diabetes, and was higher in subjects with a longer follow‐up. Diabetes was a risk factor independent of age, BMI, gender and race, with a hazard ratio of 2.16. A high BMI further increases the risk of HCC. In the Verona cohort, where 70% of cases were overweight or obese, the hazard ratio of HCC was 1.86 (95% CI, 1.43–2.38) 84. The majority of these studies have been carried out on referred cohorts, and are subject to referral biases. Only recently have population‐based studies become available. In a Korean cohort with low average BMI (23.2 kg/m², only approximately 25% overweight or obese), the age‐adjusted incidence rate of HCC increased progressively with increasing fasting glucose and BMI, reaching the top in men with diabetes (hazard ratio, 1.66; 95% CI, 1.53–1.79) 88. In this population with high HBV carrier state, neither NAFLD nor HBV were demonstrated as confounders, suggesting a primary role of hyperglycemia. In a large U S Medicare study in subjects aged 65 years or over, the risk of HCC was increased 2 to 3 times in the presence of diabetes, after adjustment for HBV, HCV, alcohol and hemocromatosis, and did not change in analyses restricted to subjects with diabetes detected between two and three years prior to HCC diagnosis 89. Thus diabetes may be considered an additional risk factor for HCC, when other major risk factors are ruled out.

Finally, the association of NAFLD with HCC is well known, and is probably mediated by cryptogenic cirrhosis 4, 72, 85, 93–95. Decreased surveillance 96 might be implicated in the poor risk of subjects with NAFLD‐related HCC, compared with HCC occurring in cirrhosis of different etiology 72.

Proposed mechanism(s) for liver disease and disease progression

The mechanism(s) implicated in fatty liver have been outlined in several recent reviews 16, 97–99; they have insulin resistance as the key factor (Figure 1). The family clustering of diseases strongly supports a genetic inheritance of decreased insulin sensitivity, but the search for genes has so far been disappointing. Polygenic transmission is very likely in metabolic disorders, and determining the genetics of insulin resistance will not be easy 100, 101. There are hundreds of gene interactions involved in fat storage 102 and in insulin action 103, and abnormalities in at least 23 genes have been detected in NAFLD 103. Further studies should also identify the site of insulin resistance (liver versus adipose tissue and skeletal muscle). Lifestyle maintains the primary role in fatty liver, via excessive food intake and low physical exercise, promoting obesity and further decreasing insulin sensitivity. Finally, the contribution of HCV in selected populations should not be forgotten. Disease progression and carcinogenesis may stem from hyperglycemia, either generating oxidative stress 104–106 or facilitating septal fibrosis 59. Also hyperinsulinemia and increased insulin‐like growth factor‐1 107–109 have been associated with fibrogenesis and the generation of hepatic hyperplasia, through the up‐regulation of connective‐tissue growth factor in stellate cells demonstrated in human liver biopsies and in in vitro experiments 108. Finally, obesity per se may increase oxidative stress, causing a dysregulation of adipocytokines promoting the development of MS 110.

Figure 1 Interaction between genes, virus and lifestyle in the initiation and progression of metabolic liver disease. Insulin resistance is the core of the mechanism. Genes, through various and largely undefined polymorphisms, may cause or favor insulin resistance. Hepatitis‐C virus, either by its viral genome, or via increased TNF‐α production, may interfere with insulin signaling. Finally lifestyle, through obesity, adipokines and increased release of TNF‐α, produces or enhances insulin resistance. Increased FFA flux promotes steatosis (possibly aggravated by HCV‐dependent interference with the hepatic assembly and secretion of triglyceride‐rich very‐low‐density lipoproteins); hyperglycemia induces oxidative stress, hyperinsulinemia favors fibrosis and hepatic hyperplasia, both leading from fatty liver to non‐alcoholic steatohepatitis, fibrosis, cirrhosis and eventually to hepatocellular carcinoma. FFA = free fatty acids; HCV = hepatitis C virus; TNF = tumor necrosis factor; NASH = non‐alcoholic steatohepatitis; HCC = hepatocellular carcinoma; IRS‐1 = insulin receptor substrate‐1; PI‐3K = phosphatidyl inosytol‐3 kinase; PPAR = peroxisome‐proliferator activated receptor.

Treatment of liver disease in metabolic patients

The liver disease of subjects cared for in metabolic/diabetic units is by no means less severe than that observed in hepatology units 42. Although the burden of cardiovascular disease largely outweighs that of liver failure, progressive liver disease is expected to reduce cardiovascular risk, by reducing several diabetes‐related risk factors (arterial pressure, hyperlipidemia, platelets and clotting factors) 111. As a consequence, these subjects may ultimately be at risk because of their liver disease, and treatment may be worthwhile.

NAFLD treatment

Several review articles are available on NAFLD treatment 112. Therapeutic options range from vitamins, antioxidants and cytoprotective agents to lipid‐lowering drugs, phlebotomy and insulin‐sensitizers 112, 113. In all NAFLD subjects, weight‐reducing programs and life‐style interventions remain the first line of treatment (Table III) 114–120. They were reported as effective in reducing aminotransferase levels and improving liver histology in the short‐term 116, 120, but the results were not long‐lasting. Similar approaches are being considered as background treatment in HCV‐positive patients with obesity and/or T2DM before and during anti‐viral therapy.

Table III. Clinical studies on treatment of non‐alcoholic fatty liver disease with lifestyle/weight loss modifications and insulin‐sensitizers, published in extension in peer‐reviewed journals.

Treatment	No. of cases	Type of study	Duration of treatment	Result
First author (ref)
Weight loss or lifestyle modifications
Eriksson 114	3	Observational	6–12 months	Weight loss decreases ALT and improves histology.
Palmer 115	39	Observational	Unspecified	In overweight adults, weight loss > 10% corrects hepatic abnormalities
Ueno 116	15	Controlled against cases who refused treatment (n = 10)	3 months	Weight‐losing diet and intensive physical activity. Biochemistry and histology (only steatosis) significantly better in the experimental arm
Franzese 117	38	Observational	6 months	In obese children, ALT and ‘bright liver’ improve with weight loss
Knobler 118	49	Observational	24 months	Improved biochemistry with dietary intervention and drugs (if indicated to treat associated diseases)
Kugelmans 119	16	Observational, with/without vitamin E	3 months	Improved biochemistry with lifestyle changes (diet and physical exercise); no extra advantage of vitamin E
Hickman 120	14	Observational	15 months	Improved biochemistry, health‐related quality of life, and histology with lifestyle modifications
Insulin‐sensitizers
Glitazones
Caldwell 121	10	Observational	6 months	Biochemistry improves significantly: no significant differences in histology
Neuschwander –Tetri 122	25	Observational	12 months	Both biochemistry and histology (including fibrosis and necroinflammation) improve
Promrat 123	18	Observational	12 months	Both biochemistry and histology (including fibrosis and necroinflammation) improve
Metformin
Marchesini 124	14	Controlled against cases who refused treatment (n = 6)	4 months	Metformin significantly improves ALT, reduces liver volume and decreases insulin resistance
Uygun 125	17	RCT versus diet alone (n = 17)	6 months	Metformin significantly better on ALT. Liver histology similarly improves in both arms
Nair 126	15	Observational	12 months	No differences in the long term
Bugianesi 127	55	RCT versus prescriptive diet or vitamin E (n = 55)	12 months	All subjects received nutritional counseling. Metformin significantly favors the general improvement in ALT. Histology (including fibrosis and necroinflammation) improves in a subgroup of subjects of the metformin arm

ALT = Alanine aminotransferase; RCT = randomized controlled trial.

Insulin‐sensitizing agents have proved effective as well 121–127, independently of glucose levels, with a single exception 126. Data on short‐term treatment 121, 124, 128 have now been expanded to one year with both metformin 127 and glitazones 122, 123; aminotransferase levels return within normal values, steatosis at histology is markedly reduced, apparently more with the use of glitazones 129, and also necroinflammation and fibrosis improve, but aminotransferase tends to return to pre‐treatment levels after stopping treatment 121, 124. We need long‐term, placebo‐controlled trials to validate therapeutic interventions on lifestyle and pharmacological treatment against hard outcomes (disease progression to cirrhosis and ultimately death).

Conclusions

Epidemiological studies suggest that 20% to 30% of adults in U S and other Western countries have excess fat accumulation in the liver, that about 10% of these individuals meet current diagnostic criteria for NASH, and that one third of them progress to fibrosis and cryptogenic cirrhosis 113. These estimates may become even higher in the next few years, considering the rapid spread of obesity and T2DM in the general population 5, thus increasing the global burden of theoretically‐preventable metabolic diseases in Western countries 130.

Prevention and treatment strategies of metabolic liver disease are largely based on the same approaches used to tackle the metabolic syndrome, but we all know how difficult it is to implement lifestyle changes 131. Cognitive approaches to the treatment of patients with the metabolic syndrome need to include liver disease as an additional complication, preventable by a healthy lifestyle. More importantly, physicians need to consider the importance of surveillance for detection and early treatment of complications. The potential for disease prevention and health gain from tackling major known risk factors is substantial in metabolic disorders 132, and also the risk of progressive liver disease should no longer be underestimated in these subjects.

Conflict of interests

The authors declare that they have no conflict of interests in relation to this review.