Surveillance and Screening Strategies for Hepatocellular Carcinoma: Current Perspectives and Future Opportunities

Key Takeaways:

• Adherence remains a key challenge in the effective implementation of recommended surveillance strategies.
• Noncirrhotic patients with nonalcoholic fatty liver disease/nonalcoholic steatohepatitis comprise a large percentage of the population, and implementing surveillance in this particular group is controversial and will be an area of focus in the coming years.
• Identification and validation of more sensitive surveillance modalities is an ongoing area of investigation, with particular interest in serologic biomarkers for early detection.

Hepatocellular carcinoma (HCC) is the most common primary malignancy of the liver. Approximately 90% of cases develop among patients with cirrhosis, most commonly related to chronic viral hepatitis, alcohol-related liver disease, and nonalcoholic steatohepatitis (NASH). The rates at which HCC evolves vary across the different etiologies. The incidence of HCC in the United States has progressively increased over the past 3 decades but has begun to decline since 2013, and we in the field think that primarily has been driven by 2 factors: (1) of most importance, the availability of an all-oral, highly effective hepatitis C antiviral therapy, which has attenuated the risk of cancer in the highest-risk patients, and (2) some degree of attrition of older patients with chronic viral hepatitis, who are no longer in the population and able to develop cancer due to competing mortality related to aging. Unfortunately, incidence declines in HCC have not been uniform across racial/ethnic and age groups, with nonsignificant improvements or increases observed in non-Hispanic Black patients and indigenous peoples, respectively. In the United States and worldwide, as viral hepatitis-related liver cancer rates decline, nonalcoholic fatty liver disease (NAFLD)—specifically the more progressive form, NASH related—and alcohol-related liver cancers account for an increasing proportion of incident HCC cases.  

Principles of Surveillance
The liver parenchyma has no internal somatic nerves to sense pain, so HCC does not cause any symptoms in the vast majority of cases until it is quite large and distending the liver capsule. As such, waiting for patients to have symptoms to perform diagnostic testing is destined to fail to identify early cancers for which we have optimal therapy. Early diagnosis of HCC through surveillance in populations at risk for HCC clearly increases the likelihood that a patient may access curative therapies such as resection, ablation with curative intent, and/or liver transplantation. At present, surveillance programs have certain deficiencies, including low adherence and ineffective follow-up, and the modalities we currently use—ultrasound and alpha-fetoprotein (AFP)—suffer from relatively modest sensitivity. Surveillance incurs substantial costs from both testing and lost productivity related to multiple clinical visits. Given these factors, there is significant interest in identifying better approaches.

Given the evolution of risk factors for liver cancer and the limitations of currently recommended surveillance tools, key questions in the field remain: (1) On which patient subsets should surveillance resources be concentrated due to high risk for cancer evolution? and (2) What type of surveillance strategy, whether via imaging or serologic testing, will have the best sensitivity for identifying early lesions that are amenable to curative treatment?

Which Populations Should Undergo Surveillance?
To address the first question, we recommend—as do all major gastroenterology and hepatology societies—that all patients with a diagnosis of cirrhosis, irrespective of the underlying etiology, undergo surveillance for HCC. These would be patients who have chronic viral hepatitis C who had cirrhosis prior to being cured or those who have advanced fibrosis who are yet to be cured, patients with alcohol‑related liver disease and cirrhosis, and patients with NASH cirrhosis. Also, patients with autoimmune hepatitis, hemochromatosis, primary biliary cholangitis, primary sclerosing cholangitis, and Fontan‑associated liver disease remain at somewhat elevated risk.

The American Association for the Study of Liver Diseases (AASLD) has established a threshold for risk among patients with cirrhosis of 1.5% per year for recommending surveillance; this threshold has been in place for approximately 25 years and has not been recently validated. At this point, we do not have a different threshold to use, so we tend to accept the 1.5% per year threshold as dogma even though it should be better studied.

There are also some populations who are noncirrhotic for whom we recommend surveillance; this is particularly for specific populations of individuals with chronic hepatitis B. For those individuals, we use a threshold of risk of 0.2% per year for recommending surveillance. The AASLD guidelines recommend that Asian males older than 40 years of age, Asian females older than 50 years, hepatitis B carriers with a family history of liver cancer, and hepatitis B carriers from sub‑Saharan Africa be started on surveillance at earlier ages. In particular, those from Africa are a very high–risk subgroup, due to either the African variant of genotype A infection that is most typical of chronic hepatitis B in Africa or, more likely, the high levels of aflatoxin exposure that potentiate the cancer risk in that population. In our practice, it is not uncommon to encounter relatively young patients from sub‑Saharan Africa, particularly West Africa, who already have liver cancer in their teens or 20s, so generally I start surveillance immediately for any adult patient from an endemic area in Africa who comes into my practice. Data suggest that patients with biopsy-confirmed NAFLD/NASH with noncirrhotic fibrosis also may have cancer incidence rates exceeding the AASLD threshold but are not included in current AASLD guidance. Parenthetically, the data for the 0.2% per year threshold have not been recently validated and should be viewed with some skepticism.

Among patients with cirrhosis, HCC incidence varies significantly by disease etiology, the presence of >1 etiology, and treatment status. The highest risk for HCC is among patients with chronic viral hepatitis and active infection with cirrhosis, and in those individuals the rate of HCC evolution is as high as 2.5% to 7% per year. Once we cure hepatitis C, the risk is reduced by 75% to 80%, but that then raises the question of whether these patients still have a risk over the 1.5% threshold and should continue doing surveillance. We do have good data out of the Veterans Affairs system that includes long‑term follow-up of patients with chronic hepatitis C who received treatment with all‑oral direct‑acting antiviral therapies and were followed for their HCC risk serially over time. In addition, now we have data out to approximately 7 years showing that among those who had diagnosed cirrhosis prior to achieving cure, and particularly those who had a Fibrosis-4 (FIB‑4) score >3.25, the incidence rate of HCC stays between 1.8% and 3.8% per year for up to 7 years following cure. The incidence rate goes down, but it never drops below the 1.5% threshold. For patients with cirrhosis but FIB‑4 <3.25, or patients with chronic liver disease or chronic viral hepatitis who were not diagnosed with cirrhosis but FIB‑4 was >3.25, the rate of HCC is still relatively high at 0.9% to 1.3% per year over a 7‑year follow-up. Although that is below that 1.5% per year threshold, this is probably the most common population in which I’m seeing new liver cancer cases evolve—and often, because these patients have not been undergoing surveillance, they present at a fairly advanced stage. It is exactly this population that makes me question the validity of the 1.5% threshold. So, for all patients with chronic hepatitis C who were cured, who had a diagnosis of cirrhosis or had FIB‑4 >3.25, or perhaps who had other evidence of advanced fibrosis such as a liver stiffness measurement >9-10 kilopascals prior to treatment (the last based on anecdotal experience more than hard data), we generally will recommend that they remain in a surveillance program due to ongoing risk that does not decline.

Whether this is cost-effective is a separate question. Limited work focusing on enrichment of surveillance to those at highest risk based on biomarkers including FIB-4 do suggest that tailoring surveillance might be a more cost-effective approach. I think many healthcare professionals take a very case-by-case approach to patients with advanced fibrosis and address the risks and benefits individually with those patients via shared decision-making to determine if they want to enter a surveillance program if they don’t meet the more stringent criteria. I think the consensus would be that patients with cirrhosis, despite cure, should remain in surveillance programs.

In chronic hepatitis B, suppression of viral replication by antiviral therapy also reduces the risk of liver cancer but requires 5 years for the impact to be significant. Patients with alcohol-related cirrhosis overall exhibit lower risk of liver cancer development than chronic viral hepatitis, but abstinence from alcohol after onset of cirrhosis has a less clear impact on cancer incidence.

Surveillance among patients with NAFLD/NASH who may have some degree of fibrosis but are without cirrhosis remains controversial. This affects a huge population—because nearly 25% of the world’s population has NAFLD, universal screening is wholly infeasible and could never be cost-effective. Those who do not have fibrosis have small rates of HCC, but the risk is not zero. These rates have been estimated to be approximately 0.03 to 0.08% per year—far too low to recommend HCC surveillance in a general fatty liver population. But if one were to biopsy, approximately 20% of patients with fatty liver would have some degree of fibrosis, and 1% to 3% would have cirrhosis at baseline. Among those higher‑risk populations with cirrhosis, the risk of HCC is significantly greater. We now appreciate that individuals with NASH cirrhosis most likely have incidence rates in the 2% to 4% per year range, which are similar to those seen with alcoholic liver disease and cured hepatitis C. So, there really is no argument that the patients with cirrhosis should undergo surveillance, but in patients who have noncirrhotic fibrosis, the risk is probably somewhere between 0.2% and 0.7% per year. Technically, this would be above the AASLD threshold for surveillance in a noncirrhotic population. 

The feasibility, effectiveness, and cost-effectiveness of systematically screening this population—in whom ultrasound would be expected to perform poorly due to body habitus and whose major causes of death will be cardiovascular disease and nonhepatic cancers—remains dubious. That is a daunting task in terms of the absolute number of patients who would need to be accessing surveillance resources, and it is unlikely our healthcare system in the United States could handle an influx of these patients without compromising access to surveillance for patients with cirrhosis. My personal bias at present is that we should not be doing surveillance in noncirrhotic NASH. I do not think the evidence base is currently strong enough to justify that this would be effective or cost-effective. There is significant risk that expanding surveillance to lower‑risk populations could compromise access to higher‑risk populations, so I think there are probably more harms than benefits in this population, although it still needs to be studied. The problem is that, despite low incidence rates, due to the sheer size of the at-risk population, 38% of cancers that arise out of patients with NAFLD/NASH emerge in those who do not have cirrhosis. This will be a challenge over the next 5‑10 years as the field really grapples with how to handle this population. To make screening feasible, tools are needed to identify precirrhotic patients with NAFLD/NASH at high enough annual risk to warrant surveillance, perhaps based on genetic risk factors.  

Current Surveillance Modalities
The current surveillance modalities that we use based on AASLD recommendations are ultrasound and AFP every 6 months. The National Comprehensive Cancer Network (NCCN) clarifies and supports this recommendation by including that all surveillance should be done as part of a formal surveillance program and, as such, should have standardized recall processes and quality control procedures. The main quality control procedure that is currently used is the Liver Imaging Reporting and Data System (LI‑RADS) from the American College of Radiology, which uses a visualization score for ultrasounds to ensure that there is good evaluation of the entire liver. It is graded A, B, and C: A for minimal limitations, B for moderate limitations, and C for severe limitations. Factors associated with poor-quality studies include body habitus, liver echotexture, and operator experience. In general, most studies are good quality, but up to approximately 20% of ultrasounds are judged to be of moderate or severely reduced quality based on the B or C scores, and a significant amount of work has been done to try to understand what happens during follow-up of those: Are the following studies good quality? Are the following studies still poor quality? In addition, there is a fairly high rate of patients who have a poor‑quality study whose next study is also poor quality, and there are now data showing that poor‑quality studies increase the rate of false‑negative and false‑positive results. In general, my experience with the LI‑RADS visualization is that it is somewhat subjective, and some of the reasons for poor visualization scores may not be clinically significant based on exams. To date, there have not been any data looking at the interobserver reliability of the visualization score, so although it is important to look at, there are no data at present to guide whether every patient with a B or C ultrasound should have altered surveillance intervals or switch to CT- or MRI-based surveillance. It creates the potential for a very high cost for what is most likely a very small, incremental benefit, and at this point I don’t recommend changing surveillance modality or interval based on visualization scores, although there are experts who would recommend doing so.

Challenges With Implementing Current Surveillance Strategies
Part of my reason for not making that recommendation is that our efforts really need to be focused on adherence. The biggest barrier to surveillance programs is that patients have challenges adhering to the surveillance. I feel that one of the key factors for patient adherence to a surveillance program is emphasis by a healthcare professional that the surveillance is important and potentially beneficial. It has been repeatedly shown that patients who see gastrointestinal and liver specialists are approximately 4 times more likely to stay in a surveillance program than those who are followed in primary care. The most likely reason for that is education about the importance of surveillance and what the benefit to the patient might be. Another major issue in terms of instituting large surveillance programs is that it is not an accepted surveillance program by the US Preventive Services Task Force, largely driven by the absence of randomized clinical trials demonstrating that surveillance improves survival. However, completing such randomized clinical trials is neither ethical nor feasible. Subsequently, although surveillance is recommended by all of the major gastrointestinal and liver organizations, as well as the NCCN, it does not have broad uptake. There also are many system problems that make it challenging for patients to get in for surveillance. Patients frequently must travel long distances to get an ultrasound. These are usually scheduled separately from their clinic appointment, so it involves 2 trips, resulting in the need to take leave from work or other responsibilities twice. In addition, out-of-pocket expenses may be a barrier. Surveillance incurs significant costs, including those of radiologic tests, serologic tests, patient time and lost productivity, and costs related to infrastructure needed to maintain patient adherence. Most studies suggest that adherence to surveillance in cirrhotic populations is approximately 25% at best. During the COVID-19 pandemic, we saw a marked reduction in access to surveillance, both because systems were shut down and because patients were unwilling to come into healthcare facilities due to concerns about infection risk. Most modeling studies suggest that, at a population level, unless 40% of the at‑risk population is undergoing surveillance, it is unlikely that a survival benefit would be found from surveillance. Our current adherence rates in the 20% range are likely relatively ineffective due to low penetrance. Anything we can do to improve adherence can have a big impact.

Emerging Surveillance Modalities
How can we make surveillance better? One approach is to try to improve the sensitivity of the modalities that are used, and an area of active research is using abbreviated MRI as an alternative surveillance approach. Abbreviated MRI is an MRI protocol that still involves IV contrast, but only approximately one third of the imaging sequences are used, such that the study can be reduced to an approximately 15‑minute exam. There are some exciting early data that this approach might be much more sensitive for small and early‑stage cancers that may have been missed on ultrasound, and this approach is going to be tested in randomized clinical trials. The largest randomized trial, the PREMIUM study (NCT05486572), is projected to start in the spring of 2023 and will compare ultrasound and AFP to abbreviated MRI in a very large cohort of veterans with cirrhosis.

Although improving the imaging modality is potentially beneficial and may reduce the risk of false positives and false negatives, procedures such as abbreviated MRI or contrast‑enhanced ultrasound still require patients to come into healthcare facilities, still require an IV, and do not mitigate other barriers to surveillance adherence, such as time and travel limitations and cost. A serologic screening platform that obviated the need for surveillance imaging would most likely increase uptake and adherence to HCC surveillance; unfortunately, no optimal test yet has been validated. There is a lot of interest in the development of serologic biomarkers for early HCC detection that might supplant imaging biomarkers, and this is an active area of investigation. Currently, at least 2 major phase III biomarker validation studies are recruiting patients with cirrhosis using proprietary biomarker technologies. Lens-culinaris agglutinin-reactive AFP, known more commonly as AFP-L3, and des-carboxy prothrombin (DCP), although more specific than AFP for detection of HCC, both suffer from lower sensitivity and do not appear suitable as isolated surveillance tests. In an attempt to make these agents perform better and have higher sensitivity, they have been combined in different composite models. The most commonly studied is GALAD, a composite risk score combining gender, age, log-transformed AFP, and DCP, which appears to have a higher sensitivity than AFP when used as a single-point or serial test in phase II biomarker development. In a phase III study comparing GALAD, AFP-L3, DCP, and Hepatocellular Carcinoma Early Detection Screening (HES) scores all were superior to AFP and had similar discrimination for early-stage HCC, with a modestly higher sensitivity for GALAD. Although there is interest in integrating GALAD into the standard surveillance with ultrasound, I think there is more interest in newer biomarkers that are identifying perhaps more specific precancerous changes, such as changes in protein fucosylation, which is a glycomic modification of proteins—particularly a marker called fucosylated kininogen—and other markers such as methylated tumor DNA, which are the agents being used primarily in the proprietary products that are being tested at present. These are promising biomarkers being investigated in prospective trials and should be compared with GALAD, AFP-L3, DCP, and HES to inform the best serologic screen for future use. 

So, there is lots of interest here because these may be sensitive to much earlier precancerous changes that may identify patients at risk of tumor development even before imaging shows a nodule. If a serologic test can identify a patient with cancer 1 or 2 years before a tumor can be detected on imaging, it would be very helpful in stratifying patients who require even more intense imaging surveillance, and it may obviate the need for any imaging in individuals who are low risk. In addition, if we could move to a blood‑based marker that patients can get done at their local lab, that would markedly reduce the barriers to participation in surveillance programs that are the biggest limitation of our current screening modalities.

Your Thoughts?
What are your thoughts and questions regarding HCC screening and surveillance? Answer the polling question and join the conversation in the discussion box below.