Abstract
Over the last decade, metabolic-associated fatty liver disease (MAFLD) has become an important public health issue worldwide. In many countries, MAFLD has become the most common cause of chronic liver disease. On the contrary, hepatocellular carcinoma (HCC) mortality is rising. Liver tumors have become the third cause of cancer mortality worldwide. HCC is the most frequent liver tumor. While the burden of HCC related to viral hepatitis is declining, the prevalence of MAFLD-related HCC is rising rapidly. Classical screening criteria for HCC consider cirrhotic, advanced fibrosis, and viral hepatitis patients. Metabolic syndrome with liver involvement or MAFLD is associated with a higher risk of HCC development, even in the absence of cirrhosis. The question about the cost effectiveness of surveillance for HCC in MAFLD is yet not fully answered. There are no guidelines that address the question of when to start or how to define the population who can benefit of surveillance for HCC in MAFLD patients. This review aims to revise the evidence of HCC development in MAFLD. It hopes to be a step closer to defining screening criteria for HCC in MAFLD.
Introduction
Hepatocellular carcinoma (HCC) is one of the most frequent solid tumors worldwide. According to the global cancer statistics 2020, lung cancer remains the leading cause of cancer death globally with an estimated 1.8 million deaths (18%), followed by colorectal (9.4%) and liver (8.3%). 1 There were 900,000 new cases of liver cancer reported in 2020 and liver cancer accounted to 830,000 deaths annually. 1 These numbers are on the rise. The incident cases of liver cancer increased 114.0% from 1990 to 2016. 2 This could be explained by the high prevalence and lack of antiviral treatment of chronic hepatitis B and chronic hepatitis C. Now, with the implementation of universal vaccination programs against hepatitis B virus and the availability of nucleotide analogues to control the infection as well as the use of potent direct antiviral agents to cure hepatitis C, the number of cases of HCC related to chronic viral hepatitis is decreasing. The proportion of cases due to alcohol remains about the same and represents the cause of about a quarter of HCC. 3 The most important cause for HCC is now metabolic-associated fatty liver disease (MAFLD). It has been recognized that MAFLD-related HCC is increasing as a cause of HCC presented at tumor boards 4 and the most rapidly growing indication of liver transplantation in patients with HCC.5,6 From 2002 to 2012, transplantation for HCC due to nonalcoholic steatohepatitis (NASH) was multiplied by a factor four. 5 In adult liver transplant candidates, the proportion of NASH-related HCC cases increased from 2 to 16% between 2002 and 2016, and the prevalence of HCC in liver transplant candidate with NASH increased more than 10-fold. 6 In Medicare patients diagnosed with HCC, nonalcoholic fatty liver disease (NAFLD) was the most common etiology of HCC accounting for 32% of hospitalized cases and 20% of outpatient cases. 7 In a Swiss study, Myers et al. 8 reported that the standardized incidence ratio for MAFLD-associated HCC increased from the years 2000–2004 to the years 2010–2014 to 1.92 for men and to 12.7 for women highlighting a significant sex difference.
Surveillance for HCC
The prognostic of HCC remains poor. In terms of mortality, this is the third cause of oncological cause of death after lung and colorectal cancer. 1 The prognosis of HCC is highly dependent on the stage of the disease at the time of diagnosis. 9 If the tumor is found early with a small size, curative treatments can be applied. If the diagnosis is done at later stages, curative treatments are not any more possible and the patients cannot be cured. It is therefore of utmost importance to diagnose HCC at an early stage. This is the rational for implementing surveillance programs. In a surveillance program, patients above a defined risk threshold are subjected to a screening test applied at specified intervals. Considering HCC, the population at risk is essentially defined by the presence of cirrhosis. Cost-effectiveness studies set the threshold for surveillance of HCC in cirrhotic patients at an incidence of 1.5%/year or greater. 10 In a systematic review and meta-analysis of cohort studies evaluating the benefits of HCC surveillance in patients with cirrhosis, surveillance was associated with improved early stage detection, curative treatment receipt, and overall survival after adjusting for lead-time bias. 11 For noncirrhotic patients, it has been estimated that there is a benefit to be included in a surveillance program if the risk to develop HCC is more than 0.2%/year. 9 The screening test for HCC is a sonography of the liver, and the interval between the tests has been set to 6 months. The concomitant determination of alpha-fetoprotein (AFP) in HCC surveillance programs is controversial, it is supported by the Asian Pacific Association for the Study of the Liver (APASL), not by the European Association for the Study of the Liver (EASL); the European Society for Medical Oncology (ESMO) and the American Association for the Study of Liver Disease (AASLD) leave it as optional. In the matter of supporting screening in MAFLD patients without cirrhosis, there still no clear guidelines or scores to better define patients with higher risk. Table 1 shows current recommendations of EASL, AASLD, APASL, the American Gastroenterological Association (AGA), and the Latin American Association for the Study of the Liver (ALEH).
Recommendation for HCC screening across different societies.
AASLD, American Association for the Study of Liver Disease; AGA, American Gastroenterological Association; ALEH, Latin American Association for the Study of the Liver; APASL, Asian Pacific Association for the Study of the Liver; HCC, hepatocellular carcinoma; MAFLD, metabolic-associated fatty liver disease.
Prevalence of NAFLD and MAFLD
It was estimated that 25% of the world population has a nonalcoholic fatty liver disease. 12 Actually, this estimate bears all the limitation of the definition of NAFLD. Recently, a better definition for this entity has been proposed and widely accepted. 13 It is based on a positive definition with specific criteria of MAFLD, 14 a name proposed in the literature already in 2011. 15 The global prevalence of MAFLD among overweight/obese adults is estimated to be 50.7%. 16 The prevalence in lean individuals reached 5.4% and in nonobese individuals 29.8% giving an overall prevalence in the population of 38.8%. 17 Moreover, modeling studies found that the burden of this disease will continue to rise.18,19 Some categories of patients are more affected than others with a fatty liver. It is estimated that obese individuals have in 80% of the cases a fatty liver and that patients with a type 2 diabetes mellitus (T2DM) have in 65% of the cases a fatty liver. 20 The high prevalence of these two conditions worldwide explains the high global prevalence of fatty liver. The World Health Organization estimates that 39% of adults are overweight and 13% are obese (https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight). The global prevalence of T2DM increased from 211 million in 1990 to 476 million in 2017. 21 Fortunately, the majority of patients with a fatty liver does not have hepatic inflammation and sign of hepatocellular damage (ballooning) which define NASH, the progressive form of the disease. Patient with inflammation and ballooning are those at risk to develop fibrosis, cirrhosis, and liver complications. The prevalence of NASH in the general population is estimated to be 1.5–6.5%. 22 Like for fatty liver, this prevalence of NASH is higher in obese individuals and those with T2DM. NASH affects 10% of obese patients 23 and 37%% of diabetic patients. 24 The two subgroups of patients are also those at risk of advanced fibrosis. In a prospective cohort of 533 patients in a primary care setting, it has been found that 8.9% of the patients with obesity had an elevated transient elastography. 25 About 17% of patients with a type 2 diabetes have an advanced fibrosis.24,26,27
HCC in MAFLD
Patients with MAFLD are at risk to develop an HCC. This risk is dependent on the stage of fibrosis. It is the lowest in patients without fibrosis and increases as the degree of fibrosis increases. For patients with cirrhosis, the risk to develop HCC is above the threshold justifying surveillance. Kanwal et al. 28 conducted a retrospective cohort study from the Veterans Health Administration and found that the incidence of HCC in NAFLD patients is 0.21/1000 patient-years, and that in NAFLD patients with a cirrhosis, the incidence was 50 times higher at 10.6/1000 patient-year, whereas for patients without fibrosis, it was only 0.08/1000 patient-years. Ascha et al. 29 in a monocentric retrospective analysis of adult patients with cirrhosis seen between 2003 and 2007 reported that the yearly cumulative incidence of HCC was 2.6% in patients with NASH-related cirrhosis, compared with 4.0% in patients with hepatitis C virus (HCV)–related cirrhosis. In a meta-analysis of 86 studies including a sample size of 8.5 million patients from 22 countries, Younossi et al. 12 reported that the annual incidence of HCC in NAFLD patients was 0.44 per 1000 person-years, whereas the incidence rate for HCC in NASH patients was 10-fold higher at 5.29 cases per 1000 person-years. 12 Others reported an HCC risk in NAFLD patients between 0.1 and 0.8 per 1000 patient-years.30,31 Recently, one large study from the United States has shown that prevalence of HCC in noncirrhotic NAFLD is 0.46/1000 persons and 37.4/1000 among cirrhotic NAFLD patients. Age >65 years, presence of T2DM, smoking, male sex, and elevated alanine transaminase (ALT) were the risk factors identified for HCC development in noncirrhotic NAFLD patients. The presence of all these five factors increased the prevalence to 45/1000 persons. 32 Estes et al. 19 in a modeling study estimated the annual transition rates to HCC at different stage of fibrosis in NAFLD patients. For patients without fibrosis (F0), it is estimated at 0.0004%, for patients with F1 fibrosis, it is 0.011%, for patients with F2 fibrosis, it is 0.022%, for patients with F3 fibrosis, it is 0.044%, and for patients with a cirrhosis (F4), it is 0.34%. 19 The presence of NAFLD and components of metabolic syndrome, particularly obesity and T2DM, have been independently associated with HCC development and together contribute to the risk of HCC in the noncirrhotic liver.33–35
Given the very high prevalence of MAFLD, when one looks at the numbers from the perspective of patients with HCC, one notes that a significant proportion of these patients has have MAFLD or NASH as an underlying liver disease. One peculiarity of these cases is that a significant proportion develops HCC in a noncirrhotic liver. This was first reported by Valerie Paradis. In her landmark publication, the authors compared 3 groups of HCC patients, 31 patients with metabolic syndrome as the only risk factor for liver disease, 81 patients with overt cause of chronic liver disease, and 16 without causes of chronic liver disease. Sixty-five percent of the patient in the metabolic syndrome group had not advanced fibrosis (F0–F2) while only 26% in the chronic liver disease had not advanced fibrosis. 36 This observation has been since then confirmed by many publications, and it is estimated that about 30–50% of the patient with NAFLD-related HCC does not have a cirrhosis.37–39 This implies that noncirrhotic NAFLD patients have a higher risk of developing HCC in comparison to noncirrhotic patients with other chronic liver diseases. This increase in risk has been estimated to be 2.5-folds. 39
Risk factors for HCC
Therefore, we cannot rely only on the stage of fibrosis to decide to include patients in a surveillance program; other parameters to select patients for a surveillance when they do not have (yet) a cirrhosis have to be selected. This discussion already took place in other areas of hepatology, for example, in patients with chronic hepatitis B. The decision to include or not to include patients with a chronic hepatitis B in an HCC surveillance program can be based on the determination of the PAGE-B score. This score considers platelet count, age, and sex; it classifies patients in low, intermediate, and high HCC risk classes. Patients in the low HCC risk class, which represent about 25% of the hepatitis B virus (HBV)–infected patients, rarely develop HCC, and therefore do not overcome the 0.2%/year threshold for starting surveillance. 40 This classification has been endorsed by the EASL clinical practice guidelines on the management of HCC. 9 We need to have such a validated score for patients with MAFLD. It is noteworthy to MAFLD increases the risk to develop HCC in patients with chronic hepatitis B with an adjusted hazard ratio of 1.41 41 and that the presence of MAFLD in HBV-associated HCC worsens the prognosis. 42 In search of the factors being associated with HCC in the context of a metabolic liver disease, Tobari et al. 43 have shown that among patients with noncirrhotic NAFLD, male sex, alcohol consumption, and high fibrosis-4 (FIB-4) index were risk factors for HCC development. Elevated ALT along with increasing age has been identified as independent risk factors for HCC development among NAFLD patients. 44 Not surprisingly, these algorithms are taking into account parameters that have been associated with cancers, like obesity and T2DM. High body mass index (BMI) is associated with an increased risk of different types of cancers, including liver and colon cancer.45,46 Even if the relative risk of HCC in case of T2DM and obesity is not as high as the risk associated with HBV or HCV, the high prevalence of obesity and T2DM account for higher population attributable fractions with 37% for obesity and T2DM in comparison to 22% for HCV and 6% for HBV.28,47,48 Several propositions to decide of the inclusion or not in a surveillance program of MAFLD/NASH patients are worth to be mentioned. Johnson et al. 49 proposed the GALAD (gender, age, alpha-fetoprotein L3% [AFP-L3], AFP, des-gamma-carboxy prothrombin) score that is based on sex, age, and three serologic biomarkers AFP, AFP-L3, and des-gamma-carboxyprothrombin. In a retrospective cohort of patients with NAFLD, the GALAD score was able to identify those with HCC with an area under the receiver operating characteristic (AUROC) of 0.96, focusing the analysis to the patients without cirrhosis, the GALAD score kept its accuracy with an AUROC of 0.98. 50 Chalasani et al. 51 developed an algorithm based on methylation of circulating biomarkers (HOXA1, TSPYL5, and B3GALT6), AFP, and sex. This algorithm demonstrated a high sensitivity to detect HCC at an early stage (82%). In a study from Korea, an HCC risk score based on six clinical variables (age, sex, ALT, total cholesterol, DM, and smoking) was evaluated in a general population cohort excluding individuals with cirrhosis, alcohol abuse, and viral hepatitis. Its AUROC was 0.83 and 0.92 in a validation cohort. 52 Several gene polymorphisms affect the risk to develop HCC. Scores combining genetic information to clinical parameters have also been proposed. Using age, sex, obesity, T2DM, and fibrosis, Donati et al. 53 in a cohort of 765 NAFLD patients without cirrhosis obtained an AUROC of 0.93. Adding relevant polymorphisms of three implicated genes (PNPLA3 I148M, TM6SF2 E167K, and MBOAT7 rs641738 C > T), the AUROC to detect HCC increased to 0.96, which was not a significant improvement. Focusing the analysis to the patients without severe fibrosis, this trend of nonsignificance remained. In a large study with cohorts from Italy and the United Kingdom with a validation cohort from Germany, Bianco et al. 54 used the genetic information of five genes adding GCKR and HSD17B13 to the three mentioned above; its accuracy was moderate with an AUROC of 0.65. From these publications, it seems that HCC risk scores based on clinical parameters are more helpful than genetic scores. Moreover, the importance of genetic polymorphisms is ethnicity-dependent. If age, sex, obesity, and type 2 diabetes mellitus are obvious parameters to include in these scores, a parameter such a vibration-controlled transient elastography is also an obvious one that remains to be taken into account. Liver stiffness captures the stage of fibrosis and is also affected by inflammation and steatosis, features associated with HCC risk.
Limitations of surveillance programs
There are several important limitations in the way surveillance is currently conducted. First, patients at risk must be recognized and enrolled. In a retrospective study, enrollment in a surveillance program of patients at risk for HCC has been reported to be as low as 20%. The main reason for this failure is the lack of surveillance orders (38%) followed by an unrecognized liver disease (20%) and an unrecognized cirrhosis (19%). 55 Nearly 10 years later, the same group performed a larger similar study and obtained similar results. 56 Second, the use of sonography as a screening test has pitfalls. Sensitivity of ultrasound for early stage HCC detection was reported to be at 51.6%, and it increased to 74% with the addition of AFP; however, this was offset by decreased specificity (87.9% vs 83.9%, respectively). 57 The quality of liver ultrasound is dependent on the expertise of the performer and is affected by characteristics of the patients. Reviewing the quality of ultrasound exams of 941 patients with cirrhosis, Simmons et al. 58 found that in 20% of the cases, the quality was inadequate to exclude HCC. Predictors of inadequate exams were an NASH cirrhosis, a high BMI, and a Child-Pugh B or C. Inadequate visualization concerns up to one-third of patients with decompensated cirrhosis and a BMI above 35 kg/m2. 58 Third, there are barriers for patients to participate in a surveillance program. The geographical distance, the need to regularly take time off work, and the fear of the exam are all reasons for patients not to stick to regular abdominal ultrasounds. 56 Adherence to screening ultrasound has been reported to be as low as 24% in a meta-analysis. 59 Fourth, there are also harms to be in a surveillance program. The discovery of a new hepatic lesion leads to an imaging escalation with cross-sectional computerized tomography (CT) or magnetic resonance imaging (MRI). These are anxiogenic tests for the patients and their families.60,61 False positive findings have been reported to affects up to 25% of patients undergoing ultrasound screening over a median of 1.5 years. 62 Fifth, the health system may not have the resources to put in place a surveillance program. The uneven geographical distribution of medical coverage may impair the inclusion in surveillance programs of patients living is poorly medicalized regions. The quality of the program should have in place a recall system and a cascading algorithm for subsequent radiological exams in case of doubt or in presence of a new lesion. All these difficulties call for developing alternatives to the current surveillance strategy.
Better screening tests should be investigated. The paradigm of ultrasound-based HCC screening bears important limitations. It can only find tumors that are visible, and it requires to regular visits to a clinic with equipment and trained staff. Blood-based noninvasive, new generation tests hold great promise. Technological tools permit to measure circulating tumor cells, tumor DNA, specific metabolites which in combination with clinical information can detect early cancer in an agnostic manner. A blood test based on circulating tumor DNA methylation was reported to detect cancer years before conventional diagnostic in asymptomatic individuals. 63 Such approach has the potential to transform how we are searching for early HCC in MAFLD patients. Its careful development should consider not only test sensitivity but also adherence, as a better adherence improves the overall effectiveness of the test. 64
In conclusion, the answer to the title question ‘Should we undertake surveillance for HCC in patients with MAFLD?’ is yes in patients with a cirrhosis. For patients with a certain degree of fibrosis, but not (yet) with a cirrhosis, risk factors such as obesity, T2DM, age, and sex have to be considered in validated scores with cut-off values helping clinicians in their decisions to enroll or not MAFLD patients in a surveillance program.
