Metabolic disorders across hepatocellular carcinoma in Italy

Metabolic disorders are well‐known risk factors for HCC. Conversely, their impact on the natural history of HCC is not established. This study aimed at evaluating the impact of metabolic disorders on clinical features, treatment and survival of HCC patients regardless of its aetiology.


Introduction
The incidence of hepatocellular carcinoma (HCC) is increasing in Europe [1]. Chronic infection with hepatitis C virus (HCV), hepatitis B virus and excessive alcohol consumption are the major risk factors in industrialized countries [2]. However, in concert with the recent epidemic of obesity and metabolic syndrome in developed countries, the incidence and prevalence of nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH) [3,4] have also increased, and today they represent rapidly growing causes of end-stage liver disease and HCC [5][6][7]. Moreover, regardless of the etiology of liver disease, metabolic disorders, such as obesity and diabetes, contribute to hepatocancerogenesis [8][9][10]. Indeed, there are well established pathophysiological mechanisms linking obesity, diabetes and HCC [11,12]. In obese individuals, the relative risk of HCC is higher than that of other cancer [13], and the cumulative incidence of HCC in diabetic patients is three times higher than in non diabetic patients [7]. Both obesity and diabetes are components of the metabolic syndrome (MS) [14,15], characterized by the presence of central obesity, dyslipidemia, diabetes, arterial hypertension [16]. The causal association between MS and HCC has been largely documented [17][18][19][20], whereas the impact of different metabolic features (obesity, diabetes mellitus, hypertension, hypertriglyceridemia and hypercholesterolemia) on clinical presentation, natural history, management and prognosis of HCC has been less investigated. We therefore conducted a field-practice study,

Accepted Article
This article is protected by copyright. All rights reserved.

Materials and methods
We retrospectively analyzed data of the Italian Liver Cancer (ITA.LI.CA) database. This registry collects data generated by the field-practice of 21 Italian centers spread throughout the Country between January 1987 and December 2014. Patients' data are collected prospectively and updated every 2 years. Data entry is regularly checked for consistency by the group coordinator and when clarification or additional information are deemed necessary, relevant cases are resubmitted to the recruitment center before final inclusion in the database. The ITA.LI.CA database management conforms to the past and the current Italian legislation regarding privacy and the present study conforms to the ethical guidelines of the Declaration of Helsinki. Approval for the study was obtained from the Institutional Review Board of the participating centers. Details on the ITA.LI.CA database management have been already reported [21].
For the purpose of this study, we analyzed the data of 1950 cirrhotic HCC patients, consecutively examined and managed during 5 years, from January 2009 to December 2014, in order to analyze the updated managing of HCC and to avoid any interference related to the use of the new direct acting antivirals (DAAs) for HCV treatment. Of them, 839 meet the selection criterion, i.e. the availability of all data for a metabolic evaluation.

Accepted Article
This article is protected by copyright. All rights reserved. and no other causes of liver damage were found. Patients were classified as having NAFLD if all other known etiologies of liver disease could be ruled out and if consistent present or past histological or ultra-sonographic features of fatty liver and alcohol intake < 30g/ day were present.
The diagnosis of HCC was based on the European Association for the Study of Liver (EASL) [22] or American Association for the Study of Liver Diseases (AASLD) [23] guidelines for HCC management.
Treatment options were considered according to the Barcelona-Clinic Liver Cancer (BCLC) staging system and their update when possible [24][25]. After 2011, the patient management followed the recommendations released by the Italian Association for the Study of the Liver (AISF) [26]. In each center, treatment decisions were taken by a multidisciplinary team, and were influenced by several factors, including comorbidities, specific contraindications for each procedure, local transplant criteria and patient's opinion. In particular, liver transplantation was also evaluated in the setting of a downstaging protocol [27] or a program using expanded selection criteria [28]. Patients were defined suitable for surgical resection according to the following criteria: single nodule or few contiguous nodules in the presence of a preserved liver function (Child-Pugh class A, MELD score ≤ 10); no evidence of portal vein infiltration or thrombosis, no evidence of extrahepatic metastasis, no general contraindications to surgery. Patients were defined suitable for local ablation (radiofrequency ablation, laser thermal ablation, percutaneous ethanol injection, cryoablation) in case of nonresectable nodules ≤ 4 cm in size and with no more than 4 lesions. Trans-catheter arterial chemoembolization (TACE) was indicated in paucifocal HCC not treatable with local ablation or multifocal HCC, involving less than 40% of the liver volume, Child-Pugh A-B, no portal vein

Accepted Article
This article is protected by copyright. All rights reserved. infiltration or thrombosis, no extrahepatic metastasis, no severe associated diseases, no general contraindications to TACE. Systemic therapy with sorafenib was indicated in advanced HCC, Child-Pugh A, adequate hematologic, hepatic and renal function, no severe associated diseases, no contraindications to sorafenib. Best supportive care (BSC) or different treatment options were considered in patients not amenable to (or refusing) the above mentioned treatments.
As a diagnosis of MS was not possible to make according to the data reported in the ITA.LI.CA. database, the presence of the following 5 different metabolic features was analyzed: BMI ≥ 25, diabetes, hypertension, hypercholesterolemia and hypertriglyceridemia. According to the number of metabolic features, the patients were divided into 3 groups: • Group 1: 0-1 metabolic features; • Group 2: 2 metabolic features; • Group 3: 3-5 metabolic features.
The section about the statistical analysis methodology is reported as supplementary methodology.

Relationship among tumor burden, liver function and metabolic features
Baseline demographics, clinical and laboratory characteristics of patients are summarized in Table 1.

Accepted Article
This article is protected by copyright. All rights reserved.
HCC developed in a setting of well-compensated liver cirrhosis (Child-Pugh class A) in 66% of cases and it was detected during a surveillance program in 430 patients (51%).According to BCLC staging system,44% of patients were included in early or very-early stages, 18% in the intermediate stage,30% in the advanced stage, and 8% in the terminal stage. Regarding HCC morphology, 403 patients had a single nodule (50%), 337 a multinodular (41%), 51 an infiltrating (6%) and 25 patients a massive (>10cm)(3%) tumor. The mean size of the main nodule was 3 cm (range 2-5). Metastasis and neoplastic portal vein thrombosis were present in 6% and 18% of cases, respectively. Table 2 reports the characteristics of the 3 metabolic groups of patients.
Overall, the groups of patients significantly differed for several clinical features, according to the Kruskal-Wallis analysis (Table2). As compared with Group 1 (reference group), patients with ≥ 3 metabolic disorders (outcome B) had higher prevalence of metabolic etiology (41% vs 9%, p<0.001), better liver function (Child-Pugh A, 72% vs 60%, p=0.007, and MELD < 10, 70% vs 59%, p=0.003), higher percentage of metastasis (9% vs 5%, p=0.024), lower percentage of HCC diagnosis on surveillance (46% vs 55%, p=0.021), larger tumors (3 cm vs 2.7 cm, p=0.038), lower percentage of portal vein thrombosis (14% vs 22%, p=0.010), higher platelet count (137000/ml vs 109000/ml, p<0.001), and lower AFP levels (8 ng/ml vs17 ng/ml, p=0.004).Differently, as compared with Group 1, patients with 2 metabolic disorders (outcome A) significantly differ for the following features: age, Child-Pugh class, MELD score, AFP levels, size of the largest nodule, and presence of portal vein thrombosis and metastasis. Finally, the comparison of Group 2 with Group 3 (outcome C) did not show significant differences except for the presence of metastasis (Table2). Despite these differences, the distribution of BCLC stages did not differ among the metabolic groups, due to the high heterogeneity of patients included in each stage of this classification.
In order to scrutinize the role played by metabolic disorders on the access to treatment options, a subanalysis by BCLC stage was performed ( Table 3).The only significant difference was found in BCLC stage B, where patients with ≥ 3 metabolic features had a less frequent access to loco-regional therapies and TACE as compared to Group 1 (47% vs 62%, p=0.012).This imbalance was associated

Accepted Article
This article is protected by copyright. All rights reserved. with a tendency toward a more frequent access to surgery (23% vs 15%), although the difference did not reach the statistical significance.
The 75 th centile OS according to BCLC stage did not significantly differ among the 3 groups, both before and after lead-time adjustment (Suppl. Figure 1). Furthermore, using a probabilistic sensitivity analysis approach, we evaluated the impact of five main metabolic features on survival: diabetes, obesity, arterial hypertension, hypercholesterolemia and hypertriglyceridemia. This analysis showed that diabetes marginally reduced the OS (p=0.046), while obesity, hypercholesterolemia, arterial hypertension and hypertriglyceridemia did not have a significant impact on prognosis (p=0.269, p=0.802, p=0.602, p=0.643 respectively).

Accepted Article
This article is protected by copyright. All rights reserved.

Univariate and multivariate analysis
In the whole population, at univariate analysis MELD score, BCLC stage, HCC morphology, nodule size, portal vein thrombosis, metastasis and diagnosis outside surveillance, but not male sex, age, etiology and presence of metabolic RFs were associated with the lead-time adjusted OS (Table 4,

Accepted Article
This article is protected by copyright. All rights reserved. three models and the BCLC class B-C and portal vein thrombosis and metastasis in each pertinent model (Table 4, section 4).

Discussion
Features of the metabolic syndrome, such as diabetes and obesity, have been associated with both increased risk of cancer development and cancer-related mortality regarding both gastrointestinal (colorectal, esophagus, stomach, and pancreas) and extra-intestinal (kidney and breast) malignancies [29][30]. The increased oncologic risk would be due to a number of molecular mechanisms (insulin resistance, chronic inflammation, reduced apoptosis, and imbalance of gut microbiota) that initiate and fuel carcinogenesis. In recent years, several studies have shown a strong correlation also between metabolic features and the development of HCC. In particular, the presence of diabetes would increase the incidence and aggressiveness of HCC [12,14]. However, studies describing the impact of metabolic disorders on clinical presentation, management and survival of patients with HCC are lacking.
Our study showed that, at the time of HCC diagnosis, cirrhotic patients with ≥ 3 disorders had a more preserved liver function than those with no or one metabolic risk factors, suggesting that, in metabolic patients, HCC develops at an earlier stage of underlying liver disease. This information is in line with the data reported in literature, showing that in NAFLD patients HCC frequently occurs before the appearance of liver cirrhosis [18,31]. Moreover, we found that these patients more frequently presents with an advanced stage of tumor, as documented by the larger nodule size and the higher frequency of metastasis, even if with a lower frequency of portal vein thrombosis. This dismal oncologic presentation can be fundamentally attributed to the lower use of surveillance as compared to viral patients. However, it might also reflect a "reinforced" carcinogenesis and faster neoplastic progression related to hyperinsulinemia, lipotoxicity, oxidative stress and imbalance in the relative proportion of pro-inflammatory/anti-inflammatory cytokines [31][32][33].

Accepted Article
This article is protected by copyright. All rights reserved.
We also observed no differences in the access to therapies by the presence of metabolic disorders with the exception of BCLC B stage, where patients with several metabolic disorders less frequently underwent ablative/trans-arterial treatments and a slightly greater access to surgical therapies in comparison to those without metabolic problems. This shift toward surgical procedures, and principally to hepatic resection, is probably related to the better liver function and the larger size of the tumor we found in these patients. Since there is no similar data in the literature, larger and dedicated studies are required to confirm that metabolic comorbidities prompt a treatment shift toward surgery in HCC patients.
The survival rates of our patients at 1, 3 and 5 years was 75% ± 1.5, 52% ± 1.9 and 49% ± 2.0, respectively. These data are comparable with those already reported in literature in Italy and worldwide [1,34,35]. The survival analysis showed no impact of metabolic features even after lead time adjustment. The absence of impact on survival was confirmed also when survival analysis was made according to BCLC stage. This finding could be explained by the fact that the adverse impact on survival of metabolic disorders was counterbalanced by the occurrence of HCC at an earlier stage of underlying liver disease with respect to patients without o with an incomplete metabolic syndrome.
In addition, the study allowed us to perform a cross-sectional evaluation of the impact of the five main metabolic features (diabetes, obesity, arterial hypertension, hypercholesterolemia and hypertriglyceridemia) on survival with a probabilistic sensitivity analysis approach. This analysis showed that only diabetes marginally reduces the overall survival of HCC patients. Our result supports and extends to Western HCC patients -managed in an updated way -the findings of quite old studies usually based on small surgical cohorts of Asiatic patients, reporting an adverse impact of diabetes on the outcome of patients with cirrhosis and HCC [36][37][38][39][40]. Nevertheless, despite these evidences and the well know pathogenic role of diabetes on cardiac, renal, and neurologic diseases, other studies report a better prognosis of patients with HCC and diabetes as compared to non-diabetic patients [8,41]. This surprising and difficult to understand discrepancy probably relies on the effect of selection biases and confounding factors on results. Therefore, future studies on this topic should take

Accepted Article
This article is protected by copyright. All rights reserved. into account, for instance, insulin and hypoglycemic drugs (in particular metformin that seems to have strong antineoplastic effects [42,43]) as factors potentially capable to influence results.
Finally, the multivariate analysis selected BCLC stage as the only factor related to OS in Group 1 and 2, whereas, in Group 3, the prognosis was dictated by HCC morphology, portal vein thrombosis and presence of metastasis. Taken together, these results indicate that metabolic disorders This study presents some limitations. First, this is a retrospective and multicenter study and therefore it may suffer of unintended biases (for example, treatment choice influenced by different local facilities or conviction of the HCC team leader). Second, the incomplete clinical and laboratory information, for about 60% of our initial population, limit the complete transferability of our results to the whole patient population. However, it should be pointed out that enrolled and excluded patients did not differ for main clinical characteristics, as resulted comparing the two groups by means of univariate statistical analysis. Third, a selection bias may derive from the fact that ITA.LI.CA centers include hospital and academic centers in which surveillance programs are well applied and, for this reason, only a 30% of our patients had an advanced HCC. In particular, a high adherence rate to current recommendations for surveillance of at high risk patients likely favored, in terms of cancer presentation and survival, cirrhotic patients (particular the viral ones) rather than non-cirrhotic metabolic patients (who do not represent a group to be surveyed). However, we tried to minimalize this unavoidable bias by the adjustment of survival for the lead-time and by assessing survival in each BCLC stage.
In conclusions our study, conducted on a large series of cirrhotic patients with HCC managed in the "real world" of clinical practice, documented that: a) HCC cirrhotic patients with metabolic disorders show a worse neoplastic picture, but this does not impact on survival probably because of a more preserved liver function; b) diabetes marginally reduces survival of HCC cirrhotic patients.

Accepted Article
This article is protected by copyright. All rights reserved. This article is protected by copyright. All rights reserved.