AIM: To determine characteristics and prognostic predictors of patients with hepatocellular carcinoma (HCC) in association with non-alcoholic fatty liver disease (NAFLD).

METHODS: We reviewed the records of all patients with NAFLD associated HCC between 2000 and 2012. Data collected included demographics; histology; presence or absence of cirrhosis, size and number of HCC, alpha-fetoprotein, body mass index (BMI), and the presence of diabetes, hypertension, or dyslipidaemia.

RESULTS: Fifty-four patients with NAFLD associated HCC were identified. Mean age was 64 years with 87% male. Fifteen percent (8/54) were not cirrhotic. 11%, 24% and 50% had a BMI of < 25 kg/m², 25-29 kg/m² and ≥ 30 kg/m² respectively. Fifty-nine percent were diabetic, 44% hypertensive and 26% hyperlipidaemic. Thirty-four percent of the patients had ≤ 1 of these risk factors. Non-cirrhotics had a significantly larger mean tumour diameter at diagnosis than cirrhotics (P = 0.041). Multivariate analysis did not identify any other patient characteristics that predicted the size or number of HCC.

CONCLUSION: HCC can develop in NAFLD without cirrhosis. At diagnosis such tumours are larger than those in cirrhotics, conferring a poorer prognosis.

Non-alcoholic fatty liver disease (NAFLD) affects up to 30% of the population in industrialised countries and is becoming more prevalent in the developing world as living standards rise and dietary habits change[1,2]. Although most individuals with this condition do not develop serious liver disease, it is associated with an increased annual incidence of hepatocellular carcinoma (HCC) of 76-201 per 100000[3,4] compared to a background incidence of sporadic HCC of 4.9-16 per 100000[5,6]. Recent studies identified NAFLD or cryptogenic cirrhosis (which is thought to often represent end-stage NAFLD[7]) as the underlying cause in 13% to 38.2% of patients presenting with HCC[8-10].

Diabetes and obesity have been suggested as risk factors for HCC in large cohort and case-control studies[11-13] both with and without pre-existing NAFLD. A meta-analysis by Larsson et al[14] of 10 cohort studies found a relative risk of HCC of 1.89 in obese patients. Another meta-analysis of 17 case control studies and 32 cohort studies by Wang et al[15] found a relative risk of HCC of 2.31 in diabetics.

Whether other features of the metabolic syndrome such as hypertension and dyslipidaemia may also contribute to HCC risk is less well studied, though these conditions have been shown to independently correlate with NAFLD fibrotic severity which itself may increase HCC risk[16-18].

2012 American Association for the Study of Liver Diseases (AASLD) guidelines state that the risk of HCC in NAFLD is “likely limited to those with advanced fibrosis and cirrhosis”[19] and as such there are no established HCC screening guidelines for NAFLD patients in general[20]. However, recent international reports that HCC occurs in non cirrhotic patients with NAFLD[21] suggest that work is needed to identify factors other than severe fibrosis or cirrhosis that could be used to identify patients in whom HCC screening may be justified.

The aims of this study in a cohort of patients with NAFLD associated HCC were: (1) to describe the risk factor profile of these patients focussing on features of the metabolic syndrome; (2) to determine any demographic or risk factor profile differences between cirrhotics and non-cirrhotics; and (3) to determine if any risk factors correlated with poorer prognosis, in terms of HCC size and number.

This retrospective study was conducted at the Victorian Liver Transplant Unit which provides all liver transplant services to the states of Victoria and Tasmania covering a population of approximately 5 million people. As the sole liver transplant referral centre for this population, it has become the major tertiary referral centre for patients with HCC. The records of patients with “hepatocellular carcinoma”, “HCC”, “liver cell cancer” or “hepatoma” and “NAFLD”, “fatty liver”, “NASH” or “cryptogenic cirrhosis” between 2000 and 2012 based on International Classification of Diseases 10 coding were audited, using hospital medical records and pathology department databases.

Patients were considered to have HCC if they had radiological and/or histological diagnoses according to AASLD guidelines[22]. Patients were included if they had characteristic radiological or histological features of NAFLD as recommended in Asia-Pacific Guidelines[23], or if the diagnosis of NAFLD or cryptogenic cirrhosis had previously been made in a patient with relevant risk factors and if concomitant causes of liver diseases including cleared or chronic hepatitis B (defined as having detectable hepatitis B core antibody, with or without positive surface antigen), chronic hepatitis C, Wilson’s disease, haemochromatosis, autoimmune hepatitis, alpha1-antitrypsin deficiency, cystic fibrosis, primary biliary cirrhosis, primary sclerosing cholangitis and other chronic biliary tract diseases and other hepatic malignancies had been excluded by relevant blood tests and/or liver histology. Patients were excluded if they had an alcohol intake of over 140 g per week for men and 70 g per week for women. These diagnostic criteria broadly concur with those in international (Chinese, Italian, European and American) guidelines as summarised by Nascimbeni et al[24].

Information was collected from the time of HCC diagnosis including patient sex, age, radiological or histological evidence of cirrhosis, Child-Pugh scores, histological NAFLD grade and stage where available according to Brunt criteria[25], size and number of HCCs, alpha-fetoprotein (AFP); metabolic profile (presence of diabetes, hypertension, dyslipidaemia and obesity) and medications. Obesity was measured via body mass index (BMI) instead of waist circumference which was not recorded in a majority of patients. No patients were of Asian background (in whom altered BMI cut-offs for overweight or obesity would have otherwise applied).

Diabetes was defined as having had a previous diagnosis of the disorder and/or being on relevant medication. Hypertension was defined as having a systolic blood pressure of over 130 mmHg and/or being on relevant medication. Dyslipidaemia was defined as being on relevant medication, having serum triglyceride levels > 2.0 mmol/L and/or total serum cholesterol levels > 5.5 mmol/L.

Univariate analysis was performed using Kruskal-Wallis testing, Fisher’s exact test and two-sample t testing to analyse for differences between cirrhotic and non-cirrhotic cohorts. Multivariate analysis was performed with Pearson correlation analysing for predictors of tumour size and number, and thus prognosis. Ethical approval was obtained by the Austin Hospital Human Research Ethics Committee.

This is the first cohort study to examine NAFLD-associated HCC in an Australian context, where the majority were obese and diabetic. Furthermore, it is the first, to our knowledge, to include patients with cryptogenic cirrhosis as a subset of NAFLD given evidence strongly linking the two conditions. Indeed, in our cohort, patients with NAFLD cirrhosis had similar demographic and risk factor profiles to those with cryptogenic cirrhosis. This concurs well with other studies which have found that patients with cryptogenic cirrhosis are more likely to be obese and up to four times as likely to have diabetes than other forms of cirrhosis[27]. Moreover, there is a high rate of NAFLD development after transplantation for cryptogenic cirrhosis[28].

Our study also adds to the increasing body of evidence that HCC can occur in patients with non-cirrhotic NAFLD, even in those who are non-obese. This concurs well with recent studies including two large Japanese studies of of 292[29] and 87 patients[30] with NAFLD and HCC that reported non-cirrhotic patients comprised 38% and 49% respectively of these patients. Duan et al[31], pooling 169 NAFLD patients from 25 smaller cohorts in the literature, found 40% were non-cirrhotic and a French series of 31 HCC patients with at least 2 features of the metabolic syndrome (25 of whom had histological evidence of hepatic steatosis) found 66% were non-cirrhotic[32].

There are multiple postulated mechanisms for HCC occurring in non-cirrhotic NAFLD. Hepatic steatosis and concomitant insulin resistance causes oxidative (via increased reactive oxygen species[33] and carcinogenic metabolites of lipid peroxidation such as trans-4-hydroxy-nonenal[34]), inflammatory (upregulation of tumour necrosis factor-α, interleukin-6 and nuclear factor kappa-light-chain-enhancer of activated B cells[35]), apoptotic and hormonal changes. Increased activity of c-Jun amino-terminal kinase 1 and consequent phosphorylation of insulin receptor substrate-1[36] increase hepatic inflammation and apoptosis through downstream modulation of such pathways as mitogen-active protein kinase and phosphatidylinositol-3 kinase[37]. Adiponectin levels are decreased in NAFLD with subsequent loss of its anti-angiogenic and anti-inflammatory effects[38]. There are also decreased levels of binding proteins of insulin-like growth factor-1, with its subsequent increased bioavailability promoting cellular proliferation[39]. More novel pathways being investigated include the phosphorylation of adenosine monophosphate-activated protein kinase and activation of target of rapamycin complex 1 which then inhibits hepatic autophagy and its “quality control” effects in the liver to remove carcinogenic material[40].

Our finding that HCCs in non-cirrhotic patients are significantly larger at diagnosis is concerning as tumour size correlates with worse prognosis according to the Barcelona Clinic Liver Cancer staging system[6] and likely reflects lead time bias due to a lack of screening. Also, importantly, none of these patients with HCC had cholangiocarcinoma. A similar result was found by Paradis et al[32] in a French audit of 31 patients in whom 20 (65%) were non-cirrhotic with F0-F2 fibrosis scores, with larger tumours than in cirrhotics (median diameter 10.1 cm vs 6.6 cm, P = 0.05). While this suggests that HCC screening programs in non-cirrhotic NAFLD patients may prevent significant morbidity, the high prevalence of NAFLD in the general population makes such screening difficult to justify in terms of cost-effectiveness.

Furthermore, the incidence of HCC development in non-cirrhotic NAFLD appears to be low. There is limited literature addressing this issue with many longitudinal cohort studies having either focussed on NAFLD patients who are cirrhotic at baseline, or are confounded by not analysing if non-cirrhotic NAFLD patients who develop HCC had also become cirrhotic in the interim. In 399 non-cirrhotic NAFLD patients followed up over a mean of 7.6 years who did not develop cirrhosis on serial biopsies, Adams et al[41] found no HCCs. Similarly, no HCCs were found in a study of 64 non-cirrhotic patients followed up over a mean of 13.7 years[42] and a cohort of 109 patients over 16.7 years[43]. In the largest cohort of NAFLD patients prospectively studied, Arase et al[3] found in 1600 ultrasound-diagnosed NAFLD patients followed up over a median of 8.2 years, a HCC incidence of 0.6% (10 patients) was found, which is equivalent to 0.08% per year. Out of these 10 patients, 7 had histology taken at the time of HCC diagnosis with 6 being non-cirrhotic (Stage 0-3). Unfortunately, this study did not specify the proportion of cirrhotics in the total cohort. These low rates correlate with the data from our large centre study which identified only 15 non-cirrhotic NAFLD patients in 13 years. Moreover, there is no way to establish a clear causal link between steatosis and carcinogenesis in all these patients. Some may have been patients who developed sporadic HCC in whom the presence of hepatic steatosis may have been coincidental. Others may have had HCC derived from a pre-existing adenoma, since there is increasing literature postulating that the metabolic syndrome may drive malignant transformation of adenomas[44].

If HCC screening of non-cirrhotic NAFLD patients were to be viable, this would require narrowing the screened population by targeting those with specific risk factor profiles associated with HCC. Unfortunately we could not identify such a profile that defined our non-cirrhotic group. Comparisons of previous large cohort studies suggest that patients with NAFLD-associated HCC are more likely to be diabetic than the general NAFLD population (up to 70%[29] vs 46%[45], respectively). However, the proportion of non-cirrhotics who develop HCC in these studies was not clear and they were performed in Japanese cohorts making their findings potentially less generalisable to a Western population. Moreover, the high prevalence of diabetes in the NAFLD population means this cannot be used alone as an indication for cost-effective HCC screening. We have also shown that the cumulative number of HCC risk factors is not useful in determining HCC risk, as many of our non-cirrhotic cohort had less than 2 risk factors. Similarly we were unable to identify patient characteristics or risk factors correlating with poorer tumour prognosis that theoretically could guide targeted screening. Finally we have found AFP poorly sensitive for the presence of HCC, reflecting previous literature and 2011 AASLD guidelines that recommend against the use of AFP to guide screening[22].

In conclusion, this is the first Australian study describing the risk factor profile of patients with NAFLD-associated HCC. Even though we have a small study size and a significant bias toward males, we have shown that HCC occurs in non-cirrhotic NAFLD patients. However this is of uncertain clinical significance in the context of the burgeoning NAFLD epidemic as rates of this phenomenon still appear low. In such patients, tumours are larger than in cirrhotics which may be due to delayed diagnosis due to lack of screening. In regards to screening, we suggest, firstly, that novel biomarkers with more accuracy than AFP should be identified to better complement radiological screening. Secondly, of more importance, HCC screening in non-cirrhotic NAFLD cannot be deemed cost-effective without further studies identifying specific risk factor profiles that significantly restrict the potential target population.

We gratefully acknowledge Panteha Rezvan for her expert statistical assistance and Angela Li for her assistance with patient records.