Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
Contents lists available at ScienceDirect
Best Practice & Research Clinical Gastroenterology
3
Surveillance for hepatocellular carcinoma Morris Sherman, MB BCh PhD, FRCP(C), Professor of Medicine * University of Toronto, University Health Network, Toronto General Hospital, 200 Elizabeth Street, Toronto, ON M5G 2C4, Canada
a b s t r a c t Keywords: Hepatocellular carcinoma Cancer screening Ultrasonography Alpha fetoprotein Desgamma carboxy prothrombin
When hepatocellular carcinoma presents with symptoms cure is seldom possible and death usually follows within months. However, it is possible to detect HCC early, at which stage it is curable. This requires a surveillance program. The components of such a program include: identification of the at risk population, provision of appropriate surveillance tests, and an appropriate method of determining whether the abnormalities found on screening are cancer or not. Surveillance for liver cancer meets all these criteria. Unfortunately high quality evidence showing benefit of liver cancer surveillance is lacking, but lesser quality evidence is plentiful, including several cost efficacy analyses that all show that surveillance does decrease mortality. Therefore all the continental liver disease societies and all national liver disease societies have recommended that surveillance should be undertaken. © 2014 Elsevier Ltd. All rights reserved.
Introduction Hepatocellular carcinoma (HCC) has a well-deserved reputation as a rapidly progressive cancer that is almost invariably fatal. This is because in the absence of procedures to ensure early diagnosis HCC presents with symptoms, usually those of liver failure. By this stage the cancer is usually large, and untreatable. Tumour progression is apparently rapid, with the interval from diagnosis to death being about threeesix months. This is the pattern of disease that existed everywhere before the development of ultrasound and CT scan, and is still the pattern of presentation where these techniques are either not available or not applied.
* Tel.: þ1 416 340 4756; fax: þ1 416 591 2107. E-mail address: [email protected].
http://dx.doi.org/10.1016/j.bpg.2014.08.008 1521-6918/© 2014 Elsevier Ltd. All rights reserved.
784
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
However, HCC has a prolonged subclinical growth period [1e4], during which, if discovered, interventions are often possible, and cure can be achieved. This raises the possibility that surveillance could detect sub-clinical lesions that might be amenable to curative treatment. In this chapter we will examine the recommended HCC surveillance techniques and the controversy surrounding them. We will also see how the recommendations about surveillance rest not only on clinical evidence, but also on modelling studies using cost-efficacy techniques. Principles of cancer surveillance The objective of a cancer surveillance program is to decrease mortality from the disease. It is important to make the distinction between mortality (number of deaths/unit time) and survival (duration of life following diagnosis). Mortality is the only absolute proof of efficacy of a surveillance program. Survival is a surrogate endpoint because it is subject to several sources of bias that do not influence mortality. These include lead-time bias and length bias. Lead-time bias is the apparent increase in survival that comes exclusively from diagnosis at an earlier stage of disease. The duration of survival from diagnosis to death is increased, even if no intervention is applied. Length bias arises from the fact that surveillance is more likely to detect slow growing cancers than rapidly growing cancers, which might go from undetectable to death within the surveillance interval. There are a number of surrogate endpoints in addition to survival that have been evaluated in attempts to prove efficacy of surveillance. Stage migration is the ability of surveillance to find earlier stage disease compared to a population that does not undergo surveillance. Stage migration is a necessary outcome of screening programs but cannot be used to prove efficacy because stage migration paradoxically improves survival in both populations (Fig. 1). Furthermore, simply finding early stage disease is not sufficient. The early stage disease has to be curable with high frequency, or the effect on mortality will not be detected. Proof of cancer surveillance efficacy There is at present very little evidence to prove that HCC surveillance decreases mortality. The best evidence for the efficacy of surveillance has to come from randomized controlled trials. However, the study design is crucial. There have been several large-scale randomized trials, in China, each of which had faults in the design, or the execution or in the interpretation [5e7]. Furthermore, these were conducted years ago, using technology that is no longer up-to-date, and using clinical criteria that would be considered inadequate today. The ideal randomized controlled study would compare a group of at-risk subjects undergoing surveillance to a group that does not undergo surveillance. In order to increase efficiency of recruitment and study conduct only the patients with the highest risk of HCC should be included. Those in the arm undergoing surveillance should receive ultrasound every six months. The quality of ultrasound should be such that lesions up to 2e2.5 cm can be detected, even in a cirrhotic liver. Patients who have nodules on the liver should be investigated according to standard algorithms, such as those proposed by the guidelines from the American Association for Study of Liver Disease or the European Association for study of the Liver [8e10]. Once the diagnosis has been confirmed patients should be staged and undergo treatment according to recommendations from the same organizations. The first randomized HCC surveillance study to be published compared surveillance with alphafetoprotein (AFP) to no surveillance [5]. Subjects with AFP >200 g/mL were subject to ultrasound. As expected, in the group undergoing surveillance there were more tumours found and more that were found at an early stage. However, there was no survival benefit. The major criticisms of this study included that patients with diagnosed HCC did not necessarily get the required treatment, and that it is not clear that the proportion of all cancers defined as early stage were early stage as defined today (BCLC stage 0 or A). Even by the definitions used in the study only 27% of HCC's were diagnosed at an ‘early’ stage. In a second study, also in China, AFP and ultrasound was compared to no surveillance [6]. This study showed a survival benefit, and 76% of cases were early stage. However, the duration of the study was rather short, and only 38 out of more than 8000 subjects in the surveillance arm developed HCC. Finally, another study, also in China used a cluster randomization process and randomized
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
785
Fig. 1. A. Removal of patients with early cancer (yellow circle) from a population improves the median survival in the remaining population. B. Addition of a population with early stage good prognosis cancer to a population with mixed stage and mixed prognosis cancer improves the median survival in the overall group.
institutions to surveillance with AFP and ultrasound vs. no surveillance [7]. They showed a 37% reduction in mortality. However, this study has been criticized in that although randomization was done by institution analysis was done on a per subject basis. Today it is likely impossible to undertake another randomized controlled trial of surveillance vs. no surveillance. Poutschi et al asked patients who were candidates for surveillance whether they would consent to be in a study there was a possibility that they would not receive surveillance [11]. More than 90% of those asked indicated they would not consent to such a study. Evidence supporting HCC surveillance In the absence of data from a well-designed randomized controlled trial does evidence exist that suggests that surveillance might be effective?
786
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
The most impressive prospective data comes from a screening program in Taiwan [12]. This was a study in a population selected by virtue of a risk score. The risk score was constructed from usual blood tests, and a history of diabetes or chronic viral hepatitis. Those with a risk score above a cut-off were invited to participate in a once-off screening protocol. The mortality in this group was compared to the mortality of a similar group who did not respond to the invitation to be screened, and to the general population not invited to participate in the program. There was a reduction in mortality in the screened group compared to the unscreened high-risk group as well as to the general population. These results are difficult to generalize, but they do strongly suggest that surveillance would be effective in reducing HCC mortality. Furthermore, this study was conducted in a region of high HCC incidence, and the results would likely not be as impressive in an area where the HCC prevalence was lower, such as in the West. Studies of cohorts of screened patients who develop HCC compared to unscreened patients who develop HCC are subject to lead-time bias. Some, and theoretically all of the survival benefit could be ascribed to earlier diagnosis. Attempts have been made to correct for this bias by calculating lead-time [13]. There are several ways to calculate lead-time, but all are dependent on the tumour growth rate. Unfortunately this is not constant. One study identified the magnitude of error that could arise when lead-time is inaccurate [14]. They looked at two groups with HCC, one found by surveillance, and one not found by surveillance. Whether the difference in survival was statistically significant or not dependent was critically on the calculated lead-time. The longer the lead-time (essentially, the longer the doubling time), the less likely that the survival difference was statistically significant. Finally, many investigators have shown that patients who undergo surveillance present with smaller tumours, and are more likely to be able to undergo some sort of curative procedure, such as resection, local ablation or liver transplantation [15e17]. However, such studies do not necessarily show an improvement in survival, and even if they did, are still subject to lead-time bias, and to the inappropriateness of survival as an endpoint. Another type of analysis that can be used to support surveillance is cost-efficacy analysis. There have been many cost efficacy analyses of HCC surveillance [18e26]. No two are really similar, attesting to the difficulty of creating a really satisfactory model. Each uses different inception cohorts and follow them for varying periods of time. The surveillance techniques and surveillance intervals differ. The follow-up after identification of HCC also differs, as do the costs. Some look only at a single strategy, and others compare different surveillance strategies. Virtually all analyses show that some form of surveillance is effective, in that the life of the cohort is extended, i.e., mortality is decreased. Virtually all show that some form of surveillance is cost effective, although the various studies differ as to which is the most cost effective strategy. All also show that the cost efficacy of surveillance is critically dependent on the incidence of HCC. Although no study recommends a specific cut-off of incidence to institute surveillance, averaging out several studies suggests that for patients with cirrhosis if the incidence of HCC exceeds 1.5e2%/year surveillance is effective and cost effective [22,24]. There is much less information available for non-cirrhotic patients with chronic hepatitis B or non-cirrhotic nonalcoholic fatty liver disease (NAFLD), so it is more difficult to make recommendations about the degree of risk that should initiate surveillance in these groups. One of the first cost-efficacy analyses looked at a theoretical cohort of Child A cirrhotics, of unspecified aetiology who underwent six monthly surveillance with ultrasound and AFP, compared to no surveillance [22]. They showed that once HCC incidence exceeded about 1.5%/year surveillance was effective and cost effective. Nouso et al [24] in their study found that surveillance became effective and cost effective when the incidence exceeded 2%/year. These two studies are the basis for the AASLD recommendation that in cirrhotic populations surveillance should start if the HCC incidence exceeds 1.5e2%/year. This threshold is reached in chronic hepatitis B or C, in primary biliary cirrhosis (PBC) [27] and in some countries, with alcoholic hepatitis. This threshold is not reached with stage III fibrosis in hepatitis C or with autoimmune hepatitis [28]. Another noteworthy cost efficacy analysis that used a disease model that most closely mimics current practice found the surveillance with ultrasound alone was the most cost-effective strategy [21]. This analysis also showed that surveillance with CT scan or MRI were most costly and less effective, with an incremental cost efficacy ratio of more than $100,000 to $300,000. The ideal cost efficacy analysis is still to be done. Different analyses should be performed for hepatitis B vs. hepatitis C, since the severity of underlying liver disease is very different, and for NAFLD,
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
787
because the co-morbidities associated with NAFLD will have a substantial effect on outcome. The analyses should look at surveillance strategies that have a high likelihood of finding HCC smaller than about 2e2.5 cm in diameter. Treatment strategies should follow the BCLC recommendations. Treatment of the underlying liver disease should also be included in the model, given that hepatitis B can now be permanently suppressed in almost all, and hepatitis C can be cured in almost all. Risk scores Attempts have been made to develop risk scores that give a numerical cut-off that corresponds to a specified five and ten-year risk of developing HCC. These scores have been developed for hepatitis B and for hepatitis C separately, as well as for cirrhosis (of any aetiology). There are three risk scores that have been developed for hepatitis B [29e31]. The first was the socalled GAG HCC score [29]. This was developed in a cohort of 820 subjects undergoing surveillance. The Chinese university score (CU-HCC) also includes cirrhosis as part of the calculation [30]. The third score, the REACH-B score as derived from a prospective large-scale study that included a large number of subjects with hepatitis B (>3500). Initially the models produced three nomograms [31] to predict the development of HCC. Subsequently this was validated in other Asian populations and developed into a single score [32]. This score, the REACH-B score, has the advantage that it does not require knowledge about the presence of absence of cirrhosis, something that makes the score more practical for family practitioners, who often have difficulty recognizing the presence of cirrhosis [33]. Table 1 shows the different components of the three risk scores for hepatitis B. Table 2 shows the application of the REACH-B score. There is a risk score for hepatitis C as well, developed out of the HALT-C study [34]. This model includes age, alkaline phosphatase, Black race, albumin, varices and platelets. The calculation is complex. This score has not been validated, but can still be used, although given the uncertainty that comes with lack of validation it is probably wise to err on the side of caution and offer surveillance even to those who fall slightly outside of the appropriate risk score. The benefit of surveillance in other liver diseases is less clear. Although there is a well-recognized risk of HCC in alcoholic liver disease and in NAFLD there are no risk scores available to determine when the risk becomes high enough to warrant surveillance. Previous recommendations were that patients in these categories with cirrhosis should undergo surveillance [8,9]. However, since those recommendations were made one study in alcoholic liver disease suggests that the incidence of HCC might not be high enough to warrant surveillance [35]. We also know that the incidence of HCC in cirrhosis related to autoimmune hepatitis is quite low [36] and it may not be effective or cost effective to screen this population. In contrast, patients with cirrhosis secondary to primary biliary cirrhosis seem to have a risk of HCC that may be equivalent to that seen in cirrhotic hepatitis C [27]. If so, these patients should also undergo surveillance.
Table 1 Comparison between the different hepatitis B risk scores. Score
Components
GAG-HCC
Age Albumin Bilirubin HBV DNA Cirrhosis Age Gender HBV DNA Cirrhosis Age Gender ALT HBV DNA HBeAg/anti-HBe
CU-HCC
REACH-B
Cut-off
Negative predictive value at 10 years
5
97%
101
99%
8
98%
788
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
Table 2 REACH-B risk score. A. The allocation of points for each risk variable. B. The HCC risk associated with each score. A. Variable
Risk score
Gender
Male Female 30e34 35e39 40e44 45e49 50e54 55e59 >60
Contents lists available at ScienceDirect
Best Practice & Research Clinical Gastroenterology
3
Surveillance for hepatocellular carcinoma Morris Sherman, MB BCh PhD, FRCP(C), Professor of Medicine * University of Toronto, University Health Network, Toronto General Hospital, 200 Elizabeth Street, Toronto, ON M5G 2C4, Canada
a b s t r a c t Keywords: Hepatocellular carcinoma Cancer screening Ultrasonography Alpha fetoprotein Desgamma carboxy prothrombin
When hepatocellular carcinoma presents with symptoms cure is seldom possible and death usually follows within months. However, it is possible to detect HCC early, at which stage it is curable. This requires a surveillance program. The components of such a program include: identification of the at risk population, provision of appropriate surveillance tests, and an appropriate method of determining whether the abnormalities found on screening are cancer or not. Surveillance for liver cancer meets all these criteria. Unfortunately high quality evidence showing benefit of liver cancer surveillance is lacking, but lesser quality evidence is plentiful, including several cost efficacy analyses that all show that surveillance does decrease mortality. Therefore all the continental liver disease societies and all national liver disease societies have recommended that surveillance should be undertaken. © 2014 Elsevier Ltd. All rights reserved.
Introduction Hepatocellular carcinoma (HCC) has a well-deserved reputation as a rapidly progressive cancer that is almost invariably fatal. This is because in the absence of procedures to ensure early diagnosis HCC presents with symptoms, usually those of liver failure. By this stage the cancer is usually large, and untreatable. Tumour progression is apparently rapid, with the interval from diagnosis to death being about threeesix months. This is the pattern of disease that existed everywhere before the development of ultrasound and CT scan, and is still the pattern of presentation where these techniques are either not available or not applied.
* Tel.: þ1 416 340 4756; fax: þ1 416 591 2107. E-mail address: [email protected].
http://dx.doi.org/10.1016/j.bpg.2014.08.008 1521-6918/© 2014 Elsevier Ltd. All rights reserved.
784
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
However, HCC has a prolonged subclinical growth period [1e4], during which, if discovered, interventions are often possible, and cure can be achieved. This raises the possibility that surveillance could detect sub-clinical lesions that might be amenable to curative treatment. In this chapter we will examine the recommended HCC surveillance techniques and the controversy surrounding them. We will also see how the recommendations about surveillance rest not only on clinical evidence, but also on modelling studies using cost-efficacy techniques. Principles of cancer surveillance The objective of a cancer surveillance program is to decrease mortality from the disease. It is important to make the distinction between mortality (number of deaths/unit time) and survival (duration of life following diagnosis). Mortality is the only absolute proof of efficacy of a surveillance program. Survival is a surrogate endpoint because it is subject to several sources of bias that do not influence mortality. These include lead-time bias and length bias. Lead-time bias is the apparent increase in survival that comes exclusively from diagnosis at an earlier stage of disease. The duration of survival from diagnosis to death is increased, even if no intervention is applied. Length bias arises from the fact that surveillance is more likely to detect slow growing cancers than rapidly growing cancers, which might go from undetectable to death within the surveillance interval. There are a number of surrogate endpoints in addition to survival that have been evaluated in attempts to prove efficacy of surveillance. Stage migration is the ability of surveillance to find earlier stage disease compared to a population that does not undergo surveillance. Stage migration is a necessary outcome of screening programs but cannot be used to prove efficacy because stage migration paradoxically improves survival in both populations (Fig. 1). Furthermore, simply finding early stage disease is not sufficient. The early stage disease has to be curable with high frequency, or the effect on mortality will not be detected. Proof of cancer surveillance efficacy There is at present very little evidence to prove that HCC surveillance decreases mortality. The best evidence for the efficacy of surveillance has to come from randomized controlled trials. However, the study design is crucial. There have been several large-scale randomized trials, in China, each of which had faults in the design, or the execution or in the interpretation [5e7]. Furthermore, these were conducted years ago, using technology that is no longer up-to-date, and using clinical criteria that would be considered inadequate today. The ideal randomized controlled study would compare a group of at-risk subjects undergoing surveillance to a group that does not undergo surveillance. In order to increase efficiency of recruitment and study conduct only the patients with the highest risk of HCC should be included. Those in the arm undergoing surveillance should receive ultrasound every six months. The quality of ultrasound should be such that lesions up to 2e2.5 cm can be detected, even in a cirrhotic liver. Patients who have nodules on the liver should be investigated according to standard algorithms, such as those proposed by the guidelines from the American Association for Study of Liver Disease or the European Association for study of the Liver [8e10]. Once the diagnosis has been confirmed patients should be staged and undergo treatment according to recommendations from the same organizations. The first randomized HCC surveillance study to be published compared surveillance with alphafetoprotein (AFP) to no surveillance [5]. Subjects with AFP >200 g/mL were subject to ultrasound. As expected, in the group undergoing surveillance there were more tumours found and more that were found at an early stage. However, there was no survival benefit. The major criticisms of this study included that patients with diagnosed HCC did not necessarily get the required treatment, and that it is not clear that the proportion of all cancers defined as early stage were early stage as defined today (BCLC stage 0 or A). Even by the definitions used in the study only 27% of HCC's were diagnosed at an ‘early’ stage. In a second study, also in China, AFP and ultrasound was compared to no surveillance [6]. This study showed a survival benefit, and 76% of cases were early stage. However, the duration of the study was rather short, and only 38 out of more than 8000 subjects in the surveillance arm developed HCC. Finally, another study, also in China used a cluster randomization process and randomized
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
785
Fig. 1. A. Removal of patients with early cancer (yellow circle) from a population improves the median survival in the remaining population. B. Addition of a population with early stage good prognosis cancer to a population with mixed stage and mixed prognosis cancer improves the median survival in the overall group.
institutions to surveillance with AFP and ultrasound vs. no surveillance [7]. They showed a 37% reduction in mortality. However, this study has been criticized in that although randomization was done by institution analysis was done on a per subject basis. Today it is likely impossible to undertake another randomized controlled trial of surveillance vs. no surveillance. Poutschi et al asked patients who were candidates for surveillance whether they would consent to be in a study there was a possibility that they would not receive surveillance [11]. More than 90% of those asked indicated they would not consent to such a study. Evidence supporting HCC surveillance In the absence of data from a well-designed randomized controlled trial does evidence exist that suggests that surveillance might be effective?
786
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
The most impressive prospective data comes from a screening program in Taiwan [12]. This was a study in a population selected by virtue of a risk score. The risk score was constructed from usual blood tests, and a history of diabetes or chronic viral hepatitis. Those with a risk score above a cut-off were invited to participate in a once-off screening protocol. The mortality in this group was compared to the mortality of a similar group who did not respond to the invitation to be screened, and to the general population not invited to participate in the program. There was a reduction in mortality in the screened group compared to the unscreened high-risk group as well as to the general population. These results are difficult to generalize, but they do strongly suggest that surveillance would be effective in reducing HCC mortality. Furthermore, this study was conducted in a region of high HCC incidence, and the results would likely not be as impressive in an area where the HCC prevalence was lower, such as in the West. Studies of cohorts of screened patients who develop HCC compared to unscreened patients who develop HCC are subject to lead-time bias. Some, and theoretically all of the survival benefit could be ascribed to earlier diagnosis. Attempts have been made to correct for this bias by calculating lead-time [13]. There are several ways to calculate lead-time, but all are dependent on the tumour growth rate. Unfortunately this is not constant. One study identified the magnitude of error that could arise when lead-time is inaccurate [14]. They looked at two groups with HCC, one found by surveillance, and one not found by surveillance. Whether the difference in survival was statistically significant or not dependent was critically on the calculated lead-time. The longer the lead-time (essentially, the longer the doubling time), the less likely that the survival difference was statistically significant. Finally, many investigators have shown that patients who undergo surveillance present with smaller tumours, and are more likely to be able to undergo some sort of curative procedure, such as resection, local ablation or liver transplantation [15e17]. However, such studies do not necessarily show an improvement in survival, and even if they did, are still subject to lead-time bias, and to the inappropriateness of survival as an endpoint. Another type of analysis that can be used to support surveillance is cost-efficacy analysis. There have been many cost efficacy analyses of HCC surveillance [18e26]. No two are really similar, attesting to the difficulty of creating a really satisfactory model. Each uses different inception cohorts and follow them for varying periods of time. The surveillance techniques and surveillance intervals differ. The follow-up after identification of HCC also differs, as do the costs. Some look only at a single strategy, and others compare different surveillance strategies. Virtually all analyses show that some form of surveillance is effective, in that the life of the cohort is extended, i.e., mortality is decreased. Virtually all show that some form of surveillance is cost effective, although the various studies differ as to which is the most cost effective strategy. All also show that the cost efficacy of surveillance is critically dependent on the incidence of HCC. Although no study recommends a specific cut-off of incidence to institute surveillance, averaging out several studies suggests that for patients with cirrhosis if the incidence of HCC exceeds 1.5e2%/year surveillance is effective and cost effective [22,24]. There is much less information available for non-cirrhotic patients with chronic hepatitis B or non-cirrhotic nonalcoholic fatty liver disease (NAFLD), so it is more difficult to make recommendations about the degree of risk that should initiate surveillance in these groups. One of the first cost-efficacy analyses looked at a theoretical cohort of Child A cirrhotics, of unspecified aetiology who underwent six monthly surveillance with ultrasound and AFP, compared to no surveillance [22]. They showed that once HCC incidence exceeded about 1.5%/year surveillance was effective and cost effective. Nouso et al [24] in their study found that surveillance became effective and cost effective when the incidence exceeded 2%/year. These two studies are the basis for the AASLD recommendation that in cirrhotic populations surveillance should start if the HCC incidence exceeds 1.5e2%/year. This threshold is reached in chronic hepatitis B or C, in primary biliary cirrhosis (PBC) [27] and in some countries, with alcoholic hepatitis. This threshold is not reached with stage III fibrosis in hepatitis C or with autoimmune hepatitis [28]. Another noteworthy cost efficacy analysis that used a disease model that most closely mimics current practice found the surveillance with ultrasound alone was the most cost-effective strategy [21]. This analysis also showed that surveillance with CT scan or MRI were most costly and less effective, with an incremental cost efficacy ratio of more than $100,000 to $300,000. The ideal cost efficacy analysis is still to be done. Different analyses should be performed for hepatitis B vs. hepatitis C, since the severity of underlying liver disease is very different, and for NAFLD,
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
787
because the co-morbidities associated with NAFLD will have a substantial effect on outcome. The analyses should look at surveillance strategies that have a high likelihood of finding HCC smaller than about 2e2.5 cm in diameter. Treatment strategies should follow the BCLC recommendations. Treatment of the underlying liver disease should also be included in the model, given that hepatitis B can now be permanently suppressed in almost all, and hepatitis C can be cured in almost all. Risk scores Attempts have been made to develop risk scores that give a numerical cut-off that corresponds to a specified five and ten-year risk of developing HCC. These scores have been developed for hepatitis B and for hepatitis C separately, as well as for cirrhosis (of any aetiology). There are three risk scores that have been developed for hepatitis B [29e31]. The first was the socalled GAG HCC score [29]. This was developed in a cohort of 820 subjects undergoing surveillance. The Chinese university score (CU-HCC) also includes cirrhosis as part of the calculation [30]. The third score, the REACH-B score as derived from a prospective large-scale study that included a large number of subjects with hepatitis B (>3500). Initially the models produced three nomograms [31] to predict the development of HCC. Subsequently this was validated in other Asian populations and developed into a single score [32]. This score, the REACH-B score, has the advantage that it does not require knowledge about the presence of absence of cirrhosis, something that makes the score more practical for family practitioners, who often have difficulty recognizing the presence of cirrhosis [33]. Table 1 shows the different components of the three risk scores for hepatitis B. Table 2 shows the application of the REACH-B score. There is a risk score for hepatitis C as well, developed out of the HALT-C study [34]. This model includes age, alkaline phosphatase, Black race, albumin, varices and platelets. The calculation is complex. This score has not been validated, but can still be used, although given the uncertainty that comes with lack of validation it is probably wise to err on the side of caution and offer surveillance even to those who fall slightly outside of the appropriate risk score. The benefit of surveillance in other liver diseases is less clear. Although there is a well-recognized risk of HCC in alcoholic liver disease and in NAFLD there are no risk scores available to determine when the risk becomes high enough to warrant surveillance. Previous recommendations were that patients in these categories with cirrhosis should undergo surveillance [8,9]. However, since those recommendations were made one study in alcoholic liver disease suggests that the incidence of HCC might not be high enough to warrant surveillance [35]. We also know that the incidence of HCC in cirrhosis related to autoimmune hepatitis is quite low [36] and it may not be effective or cost effective to screen this population. In contrast, patients with cirrhosis secondary to primary biliary cirrhosis seem to have a risk of HCC that may be equivalent to that seen in cirrhotic hepatitis C [27]. If so, these patients should also undergo surveillance.
Table 1 Comparison between the different hepatitis B risk scores. Score
Components
GAG-HCC
Age Albumin Bilirubin HBV DNA Cirrhosis Age Gender HBV DNA Cirrhosis Age Gender ALT HBV DNA HBeAg/anti-HBe
CU-HCC
REACH-B
Cut-off
Negative predictive value at 10 years
5
97%
101
99%
8
98%
788
M. Sherman / Best Practice & Research Clinical Gastroenterology 28 (2014) 783e793
Table 2 REACH-B risk score. A. The allocation of points for each risk variable. B. The HCC risk associated with each score. A. Variable
Risk score
Gender
Male Female 30e34 35e39 40e44 45e49 50e54 55e59 >60
