Management of hepatocellular carcinoma

CLT, Cadaveric liver transplantation; LDLT, live donor liver transplantation; PEI, Percutanoeus ethanol injection; RF, radiofrequency; TACE, Transarterial chemoembolization; PS, Performance Status. These recommendations provide a data-supported approach to the diagnosis, staging and treatment of patients diagnosed with hepatocellular carcinoma (HCC). They are based on the following: (a) formal review and analysis of the recently-published world literature on the topic (Medline search through early 2005); (b) American College of Physicians Manual for Assessing Health Practices and Designing Practice Guidelines.1 (c) guideline policies, including the AASLD Policy on the Development and Use of Practice Guidelines and the AGA Policy Statement on Guidelines2; (d) the experience of the authors in the specified topic. We have also reviewed the guidelines prepared at the time of the Monothematic Conference of the European Association for the Study of the Liver (EASL)3 and the practice of authors experienced in the field. Intended for use by physicians, these recommendations suggest preferred approaches to the diagnostic, therapeutic, and preventive aspects of care. They are intended to be flexible, in contrast to standards of care, which are inflexible policies to be followed in every case. Specific recommendations are based on relevant published information. In an attempt to characterize the quality of evidence supporting recommendations, the Practice Guidelines Committee of the AASLD requires a category to be assigned and reported with each recommendation (Table 1). These recommendations are fully endorsed by the American Association for the Study of Liver Diseases. Over the last 5 to 8 years evidence has been accumulating in different countries that the incidence of hepatocellular carcinoma (HCC) is rising.4-9 Traditionally, the care of patients with HCC has been undertaken by hepatobiliary surgeons, interventional radiologists, and oncologists. Hepatologists in North America are not trained to perform the procedures required to treat HCC, such as alcohol injection, radiofrequency ablation, or hepatic artery catheterization, although hepatologists in Japan and elsewhere may perform many of these procedures. As a result, the role of the hepatologist traditionally has been limited to making the diagnosis and providing care of the underlying liver disease. However, more recently, the role of the hepatologist has changed. First, in many centers the development of multidisciplinary clinics has emphasized the role of the hepatologist in assessing the patient's liver disease status, and carefully managing the liver disease before and during treatment. The hepatologist has also become more actively involved in deciding what form of therapy is most appropriate and whether the patient's liver function would allow that form of therapy to be given. In addition, arising out of caring for patients with end stage liver disease, hepatologists also institute surveillance for HCC and manage the investigation of abnormal results. Finally, hepatologists are involved in the decision whether or not to offer liver transplantation to patients with HCC. There have been many reviews of various aspects of the care of patients with HCC, but only one clinical practice guideline has been published in the Western literature. The European Association for Study of the Liver (EASL) sponsored a single topic conference on HCC in 2000. The proceedings of this conference were published in 2001.3 This document largely reflected practices in Europe, and possibly North America, whereas practices in Japan are somewhat different. Definitions of the terms used in this section are given in Table 2. Surveillance for HCC involves more than simply applying a screening test or tests. Surveillance should be offered in the setting of a program or a process in which screening tests and recall procedures have been standardized and in which quality control procedures are in place. The process of surveillance also involves deciding what level of risk of HCC is high enough to trigger surveillance, what screening tests to apply and how frequently (surveillance interval), and how abnormal results should be dealt with (diagnosis and/or recall). Surveillance for HCC has become widely applied despite, until recently, the absence of evidence of benefit. There is a single randomized controlled trial of surveillance versus no surveillance that has shown a survival benefit to a strategy of 6-monthly surveillance with alphafetoprotein (AFP) and ultrasound.10 This study, which was performed in China, recruited 18,816 patients who had markers of current or prior hepatitis B infection. Adherence to surveillance was suboptimal (less than 60%) but in the subjects in the surveillance arm the HCC related mortality was reduced by 37%. These results probably represent the minimum benefit that can be expected from surveillance, because of poor compliance. In contrast, an earlier study, also conducted in China failed to show benefit, largely because patients who were diagnosed with HCC did not undergo appropriate treatment.11 Ideally, these results should be validated in other geographical areas and therefore, additional randomized controlled trials (RCT) assessing the benefits of surveillance are still considered necessary. Such trials would be difficult to undertake, but are essential to unequivocally determine the benefit of surveillance in reducing HCC mortality. The objective of HCC surveillance must be to decrease mortality from the disease. Fewer people should die from HCC, or if this is not possible, surveillance should at a minimum provide a meaningful improvement in survival duration. Other endpoints, such as stage migration (detecting earlier disease) and 5-year mortality rates are not appropriate surrogate endpoints. This has clearly been shown by analysis of the Surveillance, Epidemiology and End Results (SEER) Program of the National Cancer Institute (NCI), which demonstrated that these endpoints did not correlate with a reduction in disease-specific mortality.12 There are several sources of bias to be considered in assessing reports of surveillance studies, such as lead-time bias and length bias. Only a RCT can eliminate these biases completely. Several studies have shown that surveillance does detect earlier disease (stage migration).13-16 However, as discussed above, this does not correlate well with reduction in disease-specific mortality. Uncontrolled studies, all subject to lead-time bias, have suggested that survival is improved after surveillance.13, 16 Surveillance for HCC is widely practiced and can generally be recommended for certain at-risk groups. HCC detected after the onset of symptoms has a dismal prognosis (0%-10% 5-year survival).17 In contrast, small HCCs can be cured with an appreciable frequency.17-21 Five-year disease-free survival exceeding 50% has been reported for both resection and liver transplantation.17, 22-30 Patients surviving free of disease for this duration must be considered cured. For these patients it is highly likely that surveillance did indeed decrease mortality. Since major advances in our ability to treat HCC are less likely to come from treating late stage disease it is therefore important to find early stage disease. The decision to enter a patient into a surveillance program is determined by the level of risk for HCC. This, in turn, is related to the incidence of HCC, and it is incidence that most people use to assess risk. However, there are no experimental data to indicate what level of risk or what incidence of HCC should trigger surveillance. Instead, decision analysis has been used to provide some guidelines as to the incidence of HCC at which surveillance may become effective. An intervention is considered effective if it provides an increase in longevity of about 100 days, i.e., about 3 months.31 Although the levels were set years ago, and may not be appropriate today, interventions that can be achieved at a cost of less than about $50,000/year of life gained are considered cost-effective.32 There are now several published decision analysis/cost-efficacy models for HCC surveillance. The models differ in the nature of the theoretical population being analyzed, and in the intervention being applied. Nonetheless, these models have several results in common. They all find that surveillance is cost-effective, although in some cases only marginally so, and most find that the efficacy of surveillance is highly dependent on the incidence of HCC. For example, Sarasin et al.33 studied a theoretical cohort of patients with Child–Pugh A cirrhosis and found that if the incidence of HCC was 1.5%/year surveillance resulted in an increase in longevity of about 3 months. However, if the incidence of HCC was 6% the increase in survival was about 9 months. This study did not include transplantation as a treatment option. Arguedas et al.,34 using a similar analysis which did include liver transplantation in a population of hepatitis C patients with cirrhosis and normal liver function, found that surveillance with either CT scanning alone or CT scanning plus ultrasound became cost-effective when the incidence of HCC was more than 1.4%. However, this study has to be interpreted cautiously, because the performance characteristics of CT scanning were derived from diagnostic studies, not surveillance studies (see Surveillance Tests). Lin et al.35 found that surveillance with AFP and ultrasound was cost-effective regardless of HCC incidence. Thus, for patients with cirrhosis of varying etiologies, surveillance should be offered when the risk of HCC is 1.5%/year or greater. Table 3 describes the groups of patients in which these limits are exceeded. These groups of patients are also discussed in more detail below. The above cost-efficacy analyses, which were restricted to cirrhotic populations, cannot be applied to hepatitis B carriers without cirrhosis. These patients, particularly in Asia and Africa, are also at risk for HCC. A cost-efficacy analysis of surveillance of hepatitis B carriers using ultrasound and AFP levels suggested that surveillance became cost-effective once the incidence of HCC exceeded 0.2%/year (Collier J and Sherman M, unpublished observations). The subgroups of hepatitis B carriers in which the incidence of HCC exceeds 0.2%/year are given in Table 3. These groups are discussed in more detail below. Beasley et al., in a prospective controlled study showed that the annual incidence of HCC in hepatitis B carriers was 0.5%.36-38 The annual incidence increased with age, so that at age 70 the incidence was 1%. The incidence in patients with known cirrhosis was 2.5%/year. The relative risk of HCC was about 100, i.e., hepatitis B carriers were 100 times more likely to develop HCC than the uninfected. Sakuma et al.39 found the incidence of HCC in male Japanese railway workers was 0.4%/year. Both these populations were male and Asian, with the hepatitis B infection likely acquired at birth or in early childhood. Uncontrolled prospective cohort studies in North America, where the epidemiology of hepatitis B is different, i.e., hepatitis is acquired later in life, have indicated that the incidence of HCC in HBV carriers varies widely.40-42 Villeneuve et al.40 found no tumors in a cohort infected with HBV and followed for 16 years. McMahon et al.41 reported an incidence of HCC of 0.26%/year in a study of HBV-infected individuals in Alaska. Sherman et al.42 described an incidence of 0.46%/year in their cohort. In Europe HCC in hepatitis B carriers occurs mainly in patients with established cirrhosis.43, 44 Non- Asian chronic carriers who are anti-HBe-positive with long-term inactive viral replication and who do not have cirrhosis seem to have little risk of developing HCC.45-48 Whether surveillance is worthwhile in this population is not clear. This is not true for Asian hepatitis B carriers without cirrhosis, who remain at risk for HCC regardless of replication status.45, 49-51 Similarly, the risk of HCC persists in long-term HBV carriers from Asia who lose HBsAg, and these patients should continue to undergo surveillance.52 In Caucasian hepatitis B carriers who lose surface antigen the risk of HCC seems to decline dramatically.53, 54 The annual incidence of HCC in male hepatitis B carriers from South East Asia only starts to exceed 0.2% at about age 4038 irrespective of presence of cirrhosis or disease activity. In contrast, in Caucasians the risk is related to inflammatory activity and the presence of cirrhosis. Therefore Asian men should undergo surveillance from age 40 onwards. HCC will occur in younger patients, but the efficacy of providing surveillance to all carriers younger than age 40 is likely to be low. The incidence of HCC in women is lower than in men, although age-specific incidence rates are hard to come by. Nonetheless, it seems appropriate to start surveillance at about age 50 in Asian women. All hepatitis B carriers with cirrhosis, regardless of age should be screened for HCC. In the presence of a history of a first degree relative with HCC surveillance should start at a younger age than 40,55 although what that age should be is hard to define. Africans with hepatitis B seem to get HCC at a younger age.56, 57 Expert opinion suggests that surveillance in these populations should also start at a younger age. Whether this is true in Blacks born elsewhere is uncertain. In Caucasian hepatitis B carriers with no cirrhosis and with inactive hepatitis, as determined by a long term normal ALT and low HBV DNA concentration44, 46, 47, 58 the incidence of HCC is probably too low to make surveillance worthwhile. However, there are additional risk factors that have to be taken into account, including older age, persistence of viral replication and co-infection with hepatitis C or HIV, or the presence of other liver diseases. Nevertheless, even in the absence of cirrhosis, adult Caucasian patients with active disease are likely at risk for HCC, and should be screened. The risk of HCC in patients with chronic hepatitis C is highest and has been best studied in patients who have established cirrhosis,59-62 in whom the incidence of HCC is between 2%-8% per year. It should be noted that these data come from clinic-based studies. There is a single prospective population-based study of the risk of HCC in patients with hepatitis C.63 In this study of 12,008 men being anti-HCV-positive conferred a 20-fold increased risk of HCC compared to anti-HCV-negative subjects. The presence or absence of cirrhosis was not evaluated. Hepatitis C infected individuals who do not have cirrhosis have a much lower risk of developing HCC.64 However, the transition from bridging fibrosis to cirrhosis cannot be determined clinically so that the clinician cannot easily determine when these patients start to develop a significant increase in risk of HCC. For this reason the EASL conference3 suggested that surveillance may be offered to patients with hepatitis C and cirrhosis or with bridging fibrosis or transition to cirrhosis. The cost-efficacy of this recommendation has not been evaluated. Based on current knowledge, all patients with hepatitis C and cirrhosis should undergo surveillance. Whether patients with bridging fibrosis should also undergo surveillance remains controversial. There have been several attempts to develop non-invasive markers to predict the stage of fibrosis65-67 and if properly validated, these could be used to determine when to initiate surveillance. Similarly, several other markers may predict a significant risk of HCC. One such marker may be the platelet count. It has been suggested that the incidence of HCC in hepatitis C cirrhosis only increases substantially once the platelet count is less than 100×109/L,62, 68, 69 regardless of liver function. This needs to be validated. Others have attempted to develop predictive indices based on panels of commonly performed serological tests such as alpha 2-macroglobulin, apolipoprotein A1, haptoglobin, bilirubin and gamma-glutamyl-transpeptidase and the AST/ALT ratio.67, 70 However, these indices have still to be validated before entering general use and cannot be recommended at present. Patients who are co-infected with HIV and either hepatitis B or hepatitis C may have more rapidly progressive liver disease71 and when they reach cirrhosis they are also at increased risk of HCC.72 The MORTAVIC study indicated that HCC was responsible for 25% of all liver deaths in the post-HAART era.73, 74 The criteria for entering co-infected patients into programs for HCC screening are the same as for mono-infected patients, i.e., criteria based on the stage and grade of liver disease as described above. The incidence of HCC in cirrhosis caused by diseases other than viral hepatitis is, with some exceptions, not accurately known. Most of the studies of the incidence of HCC in alcoholic cirrhosis date from before the identification of the hepatitis C virus. Given that hepatitis C is relatively frequent in alcoholics75-77 most of the reported HCC incidence rates in earlier studies must be over-estimates. That alcoholic cirrhosis is a risk factor for HCC is clear. In one study alcoholic liver disease accounted for 32% of all HCCs.78 In an Austrian cohort with HCC alcoholic liver disease was the risk factor in 35% of subjects.79 In the United States the approximate hospitalization rate for HCC related to alcoholic cirrhosis is 8-9/100,000/year compared to about 7/100,000/year for hepatitis C.80 This study did not determine the incidence of HCC in alcoholic liver disease, but it does confirm that alcoholic cirrhosis is a significant risk factor for HCC, probably sufficient to warrant surveillance for HCC. With the recognition of steatohepatitis as a cause of cirrhosis, has come the suspicion that this too is a risk factor for HCC. No study to date has followed a sufficiently large group of such patients for long enough to describe an incidence rate for HCC. In one cohort study of patients with HCC81 diabetes was found in 20% as the only risk factor for HCC. Whether or not these patients were cirrhotic was not noted. Non-alcoholic fatty liver disease (NAFLD) has been described in cohorts of patients with HCC.82, 83 Since the incidence of HCC in cirrhosis due to NAFLD is unknown it is not possible to assess whether surveillance might be effective or cost-efficient. No recommendations can be made whether this group should be screened for HCC or not. This does not preclude the possibility that surveillance is beneficial in this group, and future data may change this recommendation. Patients with genetic hemochromatosis (GH) who have established cirrhosis have an increased risk of HCC.84-86 The relative risk of HCC is about 20. The standardized incidence ratio for HCC in cirrhotic GH is 92.9 (95% confidence interval [CI] 25-237.9). The incidence of HCC in cirrhosis due to GH is sufficiently high (about 3%-4%/year) that these patients should be included in surveillance programs. The incidence of HCC in stage 4 primary biliary cirrhosis is about the same as in cirrhosis due to hepatitis C.87 For cirrhosis due to alpha 1-antitrypsin (AAT) deficiency,88, 89 or autoimmune hepatitis there are insufficient data from cohort studies to accurately assess HCC incidence. There is as yet no convincing evidence that interferon treatment of chronic hepatitis B reduces the incidence of HCC. Studies in Europe suggested that interferon therapy for chronic hepatitis B improved survival and reduced the incidence of HCC.61, 90, 91 A study from Taiwan also indicated that successful interferon therapy, i.e., the development of anti-HBe, was associated with a reduced incidence of HCC.92 However, in these studies the event rate was low, and the sample sizes were relatively small. In contrast, a non-randomized, but matched controlled study from Hong Kong that included a larger cohort followed for longer periods found that the incidence of HCC was not decreased in the treated group.93 A single RCT suggests that lamivudine treatment of chronic hepatitis B carriers with cirrhosis does reduce the incidence of HCC,94 but whether the risk reduction is sufficient that surveillance becomes unnecessary is not clear. If a patient is a candidate for surveillance before the institution of treatment, it seems prudent to continue to offer surveillance even after therapy-induced seroconversion or therapy-induced remission of inflammatory activity. There are a number of studies evaluating the effect of treatment of chronic hepatitis C on the incidence of HCC. A single RCT in Japan suggested that the incidence of HCC was reduced in both responders and non-responders to interferon.95 These results could not be confirmed in a second RCT from France.96 The results of these other studies were summarized in a meta-analysis, which concluded that the benefit is mainly seen in those who were successfully treated, i.e., had a sustained virological response, and even then, the effect was small.97 A number of studies in Japan compared the incidence of HCC in treated patients with that in historical controls.15, 64, 98-103 These have suggested that there is a reduced incidence of HCC in treated patients. However, no data demonstrate that treating or eradicating hepatitis C completely eliminates the risk for HCC. Thus it seems that patients with hepatitis C and cirrhosis who have achieved viral clearance on therapy should, at least for now, continue to undergo surveillance. Note that patients with treated or spontaneously inactivated chronic hepatitis B or C may show regression of fibrosis sufficient to suggest reversal of cirrhosis. The risk of HCC in these patients probably does not decrease proportionately with the improvement in fibrosis. There are many theories about the pathogenesis of HCC in these patients, but one common factor seems to be that repeated rounds of necrosis and regeneration are necessary. The steps required to initiate the carcinogenic pathway probably occur many years before the disease becomes inactive, and so the threat of HCC remains even if fibrosis decreases. Regressed fibrosis is not a reason to withhold surveillance. There are a number of factors associated with an increased risk of HCC that are seen in patients at risk for developing HCC. These include an elevated AFP concentration,104-106 presence of macroregenerative nodules,107 small and large cell dysplasia on biopsy,62, 108, 109 irregular regeneration (irregular margins to regenerative nodules)110 and increased labeling index for proliferating cell nuclear antigen or silver staining of the nucloeolar organizing region.111-115 Although such patients are at more immediate risk of developing HCC they will likely already be in surveillance programs because of other recognized risk factors such as cirrhosis or chronic hepatitis B. The increased risk, however, does not require a change in surveillance protocol. 1. Patients at high risk for developing HCC should be entered into surveillance programs (Level I). The at-risk groups are identified in Table3. There are several reasons for screening patients on the liver transplant waiting list. Patients should be screened for HCC to identify small tumors that might require therapy, and to identify patients who develop cancer that exceeds the guidelines for transplantation. In addition, in the United States, under the current UNOS criteria the development of HCC provides liver transplant priority. Thus, it would seem to be in a patient's interest to have a small HCC diagnosed while on the liver transplant waiting list. One cost-efficacy analysis has suggested that the increase in longevity over the whole cohort of patients awaiting transplant is negligible, because although there may be an increase in longevity in those who develop HCC, it is countered by the loss of longevity in other patients on the waiting list whose transplants are delayed so that the patient with HCC can have priority.116 In contrast, identification of HCC that exceeds guidelines, and resultant de-listing of such patients, is beneficial to other patients on the waiting list. Another analysis suggested that there were benefits to treating patients with HCC on the transplant waiting list with either resection or local ablation.117 The benefit depended in part on the length of the waiting list. The longer the wait, the greater the benefit of intervention. In the United States UNOS currently allows patients to be listed for liver transplantation with an elevated AFP without histological confirmation of HCC, even in the absence of a mass on imaging. It is important to note that AFP is being used here for diagnosis, not surveillance. Nonetheless, the performance characteristics of AFP, even as a diagnostic test are inadequate, particularly in the absence of a mass on imaging (see below). 2. Patients on the transplant waiting list should be screened for HCC because in the USA the development of HCC gives increased priority for OLT, and because failure to screen for HCC means that patients may develop HCC and progress beyond listing criteria without the physician being aware (level III). Any assay that is used to determine the presence or absence of a disease must be validated using a series of analyses that determine how well the test performs in diagnosing the disease (since no test is 100% accurate). The simplest measures are the sensitivity (true-positive rate) and specificity (true-negative rate), which are inversely related. For any single test and the underlying disease, as sensitivity increases, specificity decreases. Furthermore, the diagnostic accuracy of any test is related to the frequency of the underlying disease in the population being studied. This is measured by the positive and negative predictive values, i.e., the rates at which positive or negative results are correct. An estimate of the efficiency of a test can also be obtained free of the influence of disease incidence by using the Youden Index. This is a measure of the combined sensitivity and specificity (sensitivity+specificity-1). Finally since the performance characteristics of a test vary across the range of the test results the optimal cut-off for diagnosis can be obtained from the Receiver Operating Characteristics (ROC) curve, a plot of sensitivity vs. 1-specificity over the entire range of the test results. An important additional consideration is that the natural history of sub-clinical liver cancer is not the same as for clinical cancer. In particular growth rates of sub-clinical cancer may be very different than tumor growth rates in clinically observed cancers. Second, sub-clinical cancer may not progress to clinically detectable cancer in all cases. Thus it cannot be assumed that all sub-clinical lesions found on surveillance will ultimately develop into cancer. Similarly, the performance characteristics of a test used to diagnose sub-clinical disease (i.e., as a screening test) are not the same as when the test is used for diagnosis. Therefore one cannot take the performance characteristics of a test used in diagnosis (e.g., CT scan) and extrapolate the sensitivity and specificity to the surveillance situation. Screening tests fall into two categories, serological and radiological. Of the serological tests the performance characteristics of AFP have been best studied.42, 118-121 Receiver operating curve analysis of AFP used as a diagnostic test suggests that a value of about 20 ng/mL provides the optimal balance between sensitivity and specificity.118 However, at this level the sensitivity is only 60%, i.e., AFP surveillance would miss 40% of HCC if a value of 20 ng/mL is used as the trigger for further investigation. This is inadequately sensitive for general use. If a higher cut-off is used a progressively smaller proportion of HCCs will be detected. If the AFP cut-off is raised to, e.g., 200 ng/mL the sensitivity drops to 22%. Conversely, reducing the cut-off means that more HCCs would be identified, but at the cost of a progressive increase in the false-positive rate. This analysis was performed in a case control study where the prevalence of HCC was artificially set at 50%. At this prevalence the positive predictive vaule of an AFP of 20 ng/mL was 84.6%. However, if the HCC prevalence rates were more like those seen in most liver clinics, i.e., about 5%, the positive predictive value of an AFP of 20 ng/mL is only 41.5%, and even at a cut-off of 400 ng/mL the PPV is only 60%.118 In cohorts undergoing surveillance the incidence of HCC may be even lower than 5%, depending on the criteria for entry into surveillance. For example, in non-cirrhotic hepatitis B carriers infected at birth the incidence of HCC is usually less than 1%. Therefore, AFP is an inadequate screening test.122 AFP still has a role in the diagnosis of HCC, since in cirrhotic patients with a mass in the liver an AFP greater than 200ng/mL has a very high positive predictive value for HCC.118 Furthermore, a persistently elevated AFP has been clearly shown to be a risk factor for HCC.104, 105 Thus, the AFP can be used to help define patients at risk, but appears to have limited utility as a screening test. Another serological test used to diagnose HCC is the des-gamma-carboxy prothrombin (DGCP), also known as Prothrombin Induced by Vitamin K Absence II (PIVKA II).121, 123-126 Most reports on the use of DGCP have evaluated the use of this test in a diagnostic mode, rather than for surveillance. Although there are reports of its use in a surveillance mode, these do not yet provide sufficient justification for routine use of this marker. There are also reports that DGCP is a marker for portal vein invasion by tumor.127 If confirmed, this would also suggest that DGCP is not a good screening test. A screening test should be able to identify early dis