Liver cancer screening in high-risk populations

Abstract

Patients with liver disease have a risk of developing hepatocellular carcinoma (HCC), but the risk varies between different diseases and is also different according to several other variables, such as age, type of underlying liver disease and severity of disease. Several risk scores have been developed to more adequately quantitate HCC risk in individual patients. Each risk score is applicable to a specific population (chronic hepatitis B or C, patients on the transplant list, cirrhosis in general, and so on). Most publications on risk scores do not provide clear guidance as to what level of measured risk is sufficient to trigger surveillance.

Practice points.

• Patients with chronic liver disease are at risk for the development of hepatocellular carcinoma.

• Patients at risk should undergo surveillance for hepatocellular carcinoma.

• Guidelines from professional societies are based large scale epidemiological studies and cost efficacy modeling.

• Populations identified by the guidelines are too broad. More sophisticated assessments of risk are possible.

• Because different risk scores have been developed in different populations care must be taken to apply the appropriate risk score to patients.

Most hepatologists accept the need to screen patients who are at-risk for hepatocellular carcinoma (HCC) using ultrasound with or without biomarkers. This is because HCC has a prolonged subclinical development and does not call attention to itself until very late in the course of the disease, by which time effective treatment is seldom possible, and because there are effective treatments for early stage disease. However, in the absence of convincing data from randomized controlled trials many still doubt that HCC surveillance is effective or necessary [1]. They argue that there is considerable potential for harm and the costs cannot be justified in the absence of evidence.

The lower the incidence of HCC in a population under surveillance the greater the number of patients who will not benefit. Given the lack of a randomized controlled trial it behooves those of us who advocate surveillance to identify with as much accuracy as possible those who are at highest risk of developing HCC and who stand to have the greatest benefit from early detection and at the same time not subjecting those who are assessed as unlikely to develop HCC to surveillance.

In this review we will look at how the HCC risk was assessed in the past and how this assessment of risk was incorporated into guidelines. We will also review the current status of risk assessment and how it might translate in future into better identification of the at-risk population.

Historical assessment of HCC risk

The relationship between cirrhosis and the development of HCC was recognized as early as the 1950s. This was in an era before the identification of hepatitis B and hepatitis C and the recognition of nonalcoholic fatty liver disease. Subsequently, the relationship between hepatitis B and HCC was identified by Wolf Szmuness in the 1970s [2]. These two findings have colored our assessment of HCC risk ever since, and only recently has this been challenged.

The link between cirrhosis and HCC was well recognized but it was also recognized that a proportion of patients with hepatitis B developed cancer in the absence of cirrhosis, and some patients without hepatitis B also developed cancer [3]. Kamal Ishak reported that a large proportion of a series of cases in North America did not have cirrhosis [3]. Similarly, in chronic hepatitis C most HCCs develop on a background of cirrhosis, but HCC has been reported in patients with stage 3 fibrosis and with even lesser degrees of fibrosis [4]. Finally, the growing population with diabetes and nonalcoholic fatty liver disease also may develop HCC in the absence of fibrosis [5], although the frequency with which this happens is probably quite low.

Knowing degree of risk is not enough

The decision to offer a patient surveillance depends in part on the degree of risk, but the level of risk that is necessary to trigger surveillance, and how this should be determined remains unclear. Even patients with no liver disease, no diabetes and none of the usual risk factors still have a finite, albeit very small risk of developing HCC. The decision to offer surveillance then depends at what point along the spectrum of risk from negligible to more than 5%/year the benefit exceeds the harms and whether this can be achieved at a reasonable cost. Markov modeling can help to assess this.

One of the earliest models looked at a population of cirrhotics without considering etiology [6]. This analysis used alpha-fetoprotein (AFP) and ultrasound as the modalities for surveillance. The analysis showed that surveillance became effective and cost–effective if the incidence of HCC exceeded 1.5%/year. Another unpublished analysis looking at hepatitis B as the starting population showed that surveillance was effective and cost–effective once the HCC incidence exceeded 0.2%/year. The difference was likely due to the fact that patients with hepatitis B were less likely to have cirrhosis and were less likely to have advanced liver disease.

These two models informed the first version of the AASLD guidelines for the management of HCC [7]. Data from a large population study of HCC in hepatitis B conducted in Taiwan many years ago [8] (which looked only at males) found that the HCC risk exceeded 0.2% in the cohort over age 40. There was no equivalent data for women, but it is known that the incidence of HCC in hepatitis B carrier women is lower than for men. Thus it was more or less arbitrarily decided that the risk became high enough in women over age 50. For all other forms of cirrhosis the 1.5% incidence was suggested as the cutoff for introducing surveillance. Subsequently other analyses suggested that the risk of HCC made surveillance worthwhile if the incidence exceeded 2% [9]. Thus the 2010 AASLD recommendations suggested that for cirrhotic patients surveillance should begin if the HCC incidence exceeded 1.5–2% [10].

Better assessment of HCC risk

It has long been recognized that within the broadly defined risk groups of hepatitis B and cirrhosis most patients will not develop HCC and may not need surveillance. The challenge has been to identify these individuals. This has led to the development of a number of risk scores derived from Cox regression analysis of risk factors associated with the development of HCC. The different models have looked at different populations and identified different risk factors. They also vary in their ease of application.

Risk scores in hepatitis B

One of the earliest models was the GAG-HCC score, derived by multivariate analysis of risk factors in untreated patients with chronic hepatitis B who developed HCC [11]. The score consists of age, gender, HBV DNA, presence of cirrhosis and presence of the core-promoter mutation. The report described a continuous S-shaped curve of risk, and prescribed a cutoff above which risk was considered high. The study did not determine a cutoff risk score at which surveillance should be instituted. Presumably all those at high risk should undergo surveillance, but whether the risk in the low risk group is so low that surveillance is not necessary is not clear.

The CU-HCC score was also derived from a similar cohort of patients with chronic hepatitis B using similar methodology as before [12]. The prediction score consisted of age, albumin, bilirubin, HBV DNA and the presence or absence of cirrhosis. More recently the authors removed cirrhosis as a categorical variable and substituted liver stiffness measured by FibroScan^® separated into three strata [13]. This improved the predictive ability of the score.

The REACH-B score was derived from the REVEAL study population [14,15]. This was a large-scale community study in Taiwan among whom there were more than 4000 subjects with hepatitis B who were followed for more than 10 years. Analysis of the potential risk factors in those who developed HCC and those who did not initially lead to the development of several nomograms [15] but this was ultimately simplified into a score based on age, gender, HBV DNA concentration, ALT and HBeAg/anti-HBe status [16]. The score was externally validated in Hong Kong and South Korea [16]. Like the GAG-HCC and CU-HCC score this score is applicable only to hepatitis B. The REACH-B score does not include cirrhosis as a variable, which has advantages and disadvantages. The advantage is that it makes the score easier to use since the noninvasive diagnosis of cirrhosis may be inaccurate. Older age then becomes a surrogate for cirrhosis. The disadvantage is that it may not perform as well in younger patients with cirrhosis. More recently quantitative HBsAg levels have been added to the score with improvement in predictability [17] but making it more complex to use.

All the models discussed above were developed in untreated cohorts of patients with chronic hepatitis B. There is now ample evidence that treatment of hepatitis B (and hepatitis C) reduces HCC incidence. Whether the risk scores remain applicable after treatment has just started to be addressed. Wong et al. evaluated three risk scores for patients with hepatitis B in patients treated with entecavir [18]. These were the REACH-B score, the GAG-HCC score and the CU-HCC score. They found that each of the scores at baseline prior to treatment predicted the development of HCC with varying degrees of accuracy (based on ROC curve analysis). As expected, risk and risk scores decreased over time while on treatment. On treatment risk scores predicted relatively well who would and who would not develop HCC, although the three risk scores did not perform equally well. However, the scores were less reliable predictors in a more mixed European or North American population [19,20]. CU-HCC still showed high accuracy (AUROC 0.89) but the other two did not perform as well.

Risk scores in hepatitis C

There have been fewer modeling studies in patients with chronic hepatitis C. The first was derived from the HALT-C study [21]. Using the same methodology as before they developed a score using age, black race, alkaline phosphatase, esophageal varices, ever smoked and platelets. This score clearly separated out those at high risk from those at minimal risk. Although the authors did not specifically address this, the incidence of HCC in the low risk group was so low as to make surveillance unlikely to be beneficial if the AASLD criteria were applied [22].

El Serag et al. [23] looked at patients with hepatitis C cirrhosis and took advantage of the fact that AFP levels in these patients are less informative about HCC being present when the ALT is elevated. They developed a model that included age, platelets AFP and ALT. The study population was a cohort of patients with cirrhosis and HCC, and the model was based on blood tests done 6 months before the HCC was diagnosed. The implication of this study is that surveillance without ultrasound is possible. However, this report did not comment on the stage of HCC at detection, nor on the outcome of treatment. These are crucial issues, because there is little value to early detection of asymptomatic HCC that is of a size that is not amenable to curative treatment, nor of finding HCC with an aggressive phenotype. Tumors that are AFP-positive are more likely to carry such a phenotype.

Risk scores in cirrhosis

Others have attempted to assess those at highest risk within the cirrhotic population, again recognizing that not all cirrhotics will develop HCC. The ADRES-HCC model was derived from patients on the transplant waiting list [24]. This scoring system is therefore not applicable to any other situation, since patients on the transplant waiting list have the most advanced disease and are therefore at the highest risk for the development of HCC. This score includes age, race, presence of diabetes, fibrosis stage (F3 or F4), and platelets. The value of this score is not clear, since all patients on the transplant waiting list would undergo regular surveillance, and it is not clear that a low risk score would result in surveillance not being offered. It is also not clear why there were patients with F3 fibrosis on the transplant waiting list unless HCC was the indication for treatment, in which case this would not be a predictive score, but a diagnostic score.

Effect of etiology of cirrhosis on risk of HCC

Not all patients with cirrhosis will develop HCC, yet the recommendations suggest that all should undergo surveillance. This is because until recently it has not been possible to identify those etiologies associated with a higher or lower likelihood of developing HCC. A Danish study was the first that evaluated the risk of HCC in a population of alcoholic cirrhotics to determine whether the risk was high enough to warrant surveillance [25]. This study showed that the risk of HCC was below the AASLD 1.5% cutoff for non-hepatitis B cirrhosis. The authors however did note that there seemed to be a North–South gradient of HCC risk in alcoholics lowest in Northern Europe and highest in Southern Europe. Thus it may be that elsewhere than Denmark the risk of HCC in alcoholics is high enough to warrant surveillance.

The effect of etiology of cirrhosis was evaluated by Sharma et al. [Unpublished Data]. The incidence was highest in patients with hepatitis B and lowest in those with autoimmune disease. These factors were included in their risk score. This is one of the few risk scores that takes etiology into account. Once again no information was provided as to the level of risk that would warrant surveillance.

In Taiwan and other places in Asia HCC is one of the most common causes of cancer death. There is therefore interest in screening whole populations. One approach was to initially define risk using blood tests, and those found to be at sufficiently high risk would undergo surveillance with ultrasound [26]. A different approach was taken by Hung et al. [27]. They assessed the HCC risk in the general population, showing the importance of elevated transaminases as risk factors. Adding hepatitis B or hepatitis C to the model improved predictability. They developed five different models, all but one of which performed well in ROC curve analysis. The variables included in the models were health history (diabetes, hypertension, myocardial infarction), ALT, AFP, HBV history and HCV history. This approach may be feasible in countries where the HCC incidence is high, but in Western countries this is not likely to be a cost-effective approach.

This study also looked at how to assess what level of risk made surveillance worthwhile. This study used a technique called decision curve analysis. Essentially, this is a modeling study using real life data, in which the potential benefits and harms from surveillance are quantitated, usually in terms of quality adjusted life years. A level of ‘net benefit’ (benefit minus harms) is chosen based on societal and/or cost considerations. The threshold level of risk for which surveillance provides a net benefit is then calculated. In this particular study they showed that there was a category of patients, particularly younger patients who would be excluded from surveillance by AASLD criteria, but would nonetheless benefit from surveillance.

Others have attempted to define HCC risk by liver stiffness measurements [28,29] and by incorporating these measurements into models combining other variables. Once more, these maneuvers can separate patients into high and low risk categories, but they do not indicate whether either group meets the threshold incidence of HCC that would make surveillance worthwhile.

The AASLD and EASL criteria to identify patients at risk for HCC and who should undergo surveillance are crude tools that were based on incidence in large population studies and on cost efficacy analysis. More sophisticated HCC risk models are now available, many of which are applicable only to specific subpopulations of patients. Only the REACH-B and ADRES-HCC models have been validated externally. Therefore, in applying these models it is important to ensure that the population to which the model is being applied matches the population described in the report. So, for example, the ADRES-HCC model should only be used in patients on the transplant waiting list, the REACH-B model should only be used in Asian patients with hepatitis B.

With the exception of a single study from Taiwan studies do not provide information on who might not need surveillance, nor what the cutoff of risk might be that should trigger surveillance. There remains work to be done, mostly modeling, to determine the thresholds for surveillance. This is the next phase of research in HCC surveillance. In particular, although there have been several cost efficacy analyses previously none have closely modeled our current staging and management strategies for HCC, and this remains a priority.

In future risk might be better assessed using a combination of clinical, laboratory and molecular markers. Studies of gene expression in the nontumorous liver of patients with HCC have identified several ‘signatures’ that are associated with the development of HCC [30]. As yet there are no studies that incorporate these factors into risk assessment, and these markers require liver tissue, which is a drawback in clinical practice. Attempts have been made to identify markers from circulating tissue (liquid biopsy), but at present this is still in early stages of development.

Conclusions & future perspectives

Although there are several risk scores the data so far are not convincing that these are ready for general use. Additional external validation is required for most scores. Additional validation is required to determine whether these scores can be used in patients who have been successfully treated for viral hepatitis. However, in future these scores or others like them will be widely used to quantify an individual's risk for HCC.

Additional work also needs to be done to determine, for each risk score, the cutoff below which surveillance is not necessary and above which surveillance should be implemented. This will require extensive modeling studies. In addition, to do this properly requires better disease models than has been published so far (Table 1).

Table 1. . Predictive scores for hepatocellular carcinoma risk.

Population	Variables	Validation	Ref.
Chronic hepatitis B	Age, gender, HBV DNA cirrhosis, core promoter mutation	No	[11]
Chronic hepatitis B	Age, albumin, bilirubin, HBV DNA, cirrhosis (yes or no)	Variable results in European and North American populations	[12,13]
Chronic hepatitis B	Age, ALT, HBeAg status, gender, HBV DNA	Yes (only in Asia)	[16]
Chronic hepatitis B	Age, albumin, HBV DNA, liver stiffness by Fibroscan	No	[13]
Chronic hepatitis B	Age, gender, ALT, HBV DNA, quantitative HBsAg, HBV genotype, HBe status	No	[17]
Chronic hepatitis C F3 and F4	Age, race, alkaline phosphatase, esophageal varices, smoking, platelets		[21]
Hepatitis C cirrhosis	ALT, AFP, age platelets	No	[23]
General population	Age, gender, alcohol consumption, BMI, diabetes (yes or no), coffee consumption, hepatitis B, hepatitis C	No	[31]
Liver transplant waitlist	Age, diabetes, race, etiology of liver disease, sex, severity (CTP score)	Yes	[24]
General population	Age, gender, ALT, liver disease, family history of HCC, cumulative smoking history	No	[27]
HCV post-SVR	Age, gender, platelet count, AFP, fibrosis stage HCV genotype		[32]