Surveillance for Hepatocellular Carcinoma: Current Best Practice and Future Direction

Hepatocellular Carcinoma Surveillance Benefits

The overarching goal of HCC surveillance is to reduce HCC-related mortality and improve quality-adjusted life-years gained. However, several conditions are necessary for HCC surveillance to result in HCC-related mortality reduction. First, HCC risk must be high enough in the at-risk groups to justify surveillance. Second, surveillance tests must accurately detect HCC at an earlier stage than it would otherwise present because of symptoms, signs, or incidental imaging. Third, effective treatments, which improve outcomes in screen-detected HCC compared to treatments for non–screen-detected HCC, must be applied.

Two large randomized controlled trials, conducted in China, have demonstrated benefits of HCC surveillance in patients with HBV infection. The first study, including 17,820 HBV-infected persons, reported that patients randomized to surveillance were more likely to have HCC detected at an early stage and undergo curative therapy.⁷ The 1- and 2-year survival rates for HCC patients in the surveillance group were 88.1% and 77.5%, respectively, compared to 0% at 1 year for HCC patients in the no-surveillance group (P < .01). The second trial among 18,816 HBV carriers similarly reported higher early tumor detection and curative treatment receipt among patients randomized to HCC surveillance.⁸ HCC-related mortality of patients undergoing surveillance was significantly lower than that of the no-surveillance arm (83.2 vs 131.5 per 100,000; P < .01), with a hazard ratio of 0.63 (95% confidence interval [CI], 0.41–0.98). Taken together, these data provide level I evidence that HCC surveillance improves early tumor detection and reduces cancer-related mortality in HBV-infected individuals.

However, one cannot extrapolate surveillance data from these randomized controlled trials of patients with chronic HBV infection to patients undergoing HCC surveillance in Western countries; most of which have cirrhosis from HCV, alcohol, and/or NAFLD. There are several reasons underlying the limited generalizability of findings from available randomized controlled trials. Patients with cirrhosis are older, have more comorbidities, and a higher competing risk of liver-related mortality than patients with chronic HBV. Abdominal ultrasound may have lower sensitivity for HCC detection in a nodular cirrhotic liver and in patients with obesity; the latter is prevalent in populations at risk for HCC in the West. Last, cirrhosis patients also have fewer curative treatment options for HCC than patients with chronic HBV. For example, surgical resection can be aggressively applied to many patients with non-cirrhotic HBV infection, but the presence of portal hypertension often precludes this in patients with cirrhosis. Similarly, although several locoregional treatments are available for cirrhosis patients with liver-localized HCC, liver transplantation remains the only treatment that can cure HCC and underlying liver disease.

There have not been any randomized controlled trials of HCC surveillance in patients with cirrhosis. However, several cohort studies have demonstrated an association between HCC surveillance and increased early detection, curative treatment receipt, and improved overall survival.⁶ A systematic review of 47 studies (including 15,158 patients with cirrhosis) demonstrated that surveillance was associated with increased early tumor detection (odds ratio [OR], 2.08; 95% CI, 1.80–2.37) and curative treatment receipt (OR, 2.24; 95% CI, 1.99–2.52).⁶ Surveillance was also significantly associated with improved overall survival (OR, 1.90; 95% CI, 1.67–2.17), with a pooled 3-year survival rate of 50.8% among those who underwent surveillance vs 27.9% for those who presented symptomatically or were diagnosed incidentally. Among the subset of studies that adjusted for lead-time bias, the association between HCC surveillance and improved survival was sustained (3-year survival rate of 39.7% vs 29.1%). Another systematic review of 2 trials and 18 observational studies similarly concluded HCC surveillance was associated with early tumor detection; however, the authors were unable to conclude whether there was a survival benefit associated with surveillance.¹³ The available cohort studies have several inherent limitations that reduce the level of confidence in their results, including unmeasured confounders, length-time bias, lead-time bias, and ascertainment bias for mortality. Yet, the overall consistency in results from the available cohort studies suggests that HCC surveillance is likely beneficial in cirrhosis patients.

In contrast, in a recent case–control study in the national US Department of Veterans Affairs (VA) health system, HCC surveillance was not associated with decreased HCC-related mortality.¹⁵ The authors found no significant difference in HCC surveillance receipt between cases who died from HCC and controls without HCC, further adding to the existing controversy surrounding the benefits of HCC surveillance.¹⁶ There are some potential aspects of the study that may account for the discrepant findings from prior cohort studies, highlighting the need for external validation of these results. Underuse of HCC surveillance and HCC therapies, which can contribute to lower surveillance effectiveness, were prevalent in this study. For example, >50% of HCCs were found at an early stage, but <15% underwent any curative therapies. Prior studies suggest higher rates of HCC surveillance and curative treatment in academic tertiary care referral centers, so it is unclear whether these results would apply to all health systems.¹⁷ Further, the authors measured exposure to ultrasound-based surveillance, but were unable to assess the quality of those examinations. Ultrasound is known to be operator-dependent, and prior studies have suggested up to 1 in 5 ultrasound examinations may be inadequate quality for evaluation of liver lesions.¹⁸ Accordingly, the American College of Radiology has recently adopted a measure of examination quality and visualization in ultrasound reporting (Liver Imaging Reporting and Data System [LI-RADS]) ranging from LI-RADS A (minimal issues with visualization) to LI-RADS C (significant issues with visualization). Regardless of the reasons underlying the discrepant results, these new data are important and highlight a need for continued evaluation of HCC surveillance effectiveness.

This ongoing debate would best be resolved with a randomized controlled trial. However, randomized data for HCC surveillance vs no surveillance are not likely to be forthcoming, mostly because patients and their clinicians strongly prefer HCC surveillance over no surveillance. In a pilot study of 205 patients, 204 (99.5%) patients declined randomization as part of a randomized controlled trial of HCC surveillance; 181 (88%) elected for a nonrandomized surveillance program, and almost all of the remaining patients chose usual care, which included ad hoc surveillance.¹⁹ Given this dilemma, expert society guidelines continue to rely on observational data to develop recommendations for HCC surveillance (see section on at-risk populations).

Hepatocellular Carcinoma Surveillance Harms

Multiple types of harms should be considered when evaluating a cancer surveillance program, including the potential for physical, financial, and psychological harms.²⁰ Physical harms can result from surveillance or follow-up testing and extend beyond medical complications to include discomfort. Financial harms can include anticipated or actual costs of surveillance and diagnostic evaluation, plus indirect costs such as missed work. Psychological harms can occur at any step of the surveillance process and include anticipation or fear of abnormal results; reactions of depression, anxiety, or cancer-specific worry after positive results; and psychological effects of being labeled with a diagnosis. Although HCC surveillance (using ultrasound and α-fetoprotein [AFP]) has minimal discomfort and no direct physical harms, there are potential “downstream” harms associated with diagnostic evaluation protocols.

To date, few studies have quantified HCC surveillance harms. In a single-center cohort study among 680 cirrhosis patients undergoing HCC surveillance over a 3-year period, surveillance-related physical harms were reported in 187 (27.5%) patients, with 66 (9.7%) having multiple computed tomography (CT)/magnetic resonance imaging (MRI) examinations and 3 (0.4%) undergoing invasive testing, such as biopsy.²¹ In another single-center study among 999 patients undergoing HCC surveillance over a median follow-up of 2.2 years, 256 patients had abnormal imaging—69 were diagnosed with HCC and 187 had an indeterminate nodule, that is, lesion ≥1 cm in diameter that could not be categorized as definitely benign or definite HCC on cross-sectional imaging).²² Among those with indeterminate nodules, 18 (9.6%) did not undergo diagnostic evaluation, 132 (70.6%) returned to ultrasound surveillance after negative CT/MRI imaging, and 37 (19.8%) continued CT/MRI-based imaging. Among those who underwent diagnostic evaluation with CT/MRI, 32 (17.1%) experienced severe harms, defined as ≤4 CT/MRI examinations or invasive testing, such as biopsy.

What Is Coming Next?

A pragmatic randomized controlled trial is under way that is randomizing patients to HCC surveillance outreach with patient navigation services vs “opportunistic” surveillance at 3 diverse health systems, including an academic tertiary care referral center, a large urban VA health system, and an integrated safety-net health system (ClinicalTrials.gov ID NCT02582918). Although the primary outcome of the trial is receipt of HCC surveillance, secondary outcomes include early detection and receipt of treatment. The trial is also measuring surveillance-related physical, psychological, and financial harms. While awaiting results from this trial, we recommend following expert society guidelines for HCC surveillance.

Current efforts evaluating HCC surveillance in large cohorts should consider both benefits and harms to better characterize its overall value of surveillance. Although false-positive ultrasound and AFP results are the most common causes for HCC surveillance harms, non-guideline concordant management of indeterminate ultrasound results can also account for a substantial proportion of surveillance harms. This includes diagnostic testing, such as liver biopsy being performed for sub-centimeter lesions and/or heterogeneous echotexture on ultrasound. Guidelines currently recommend close observation with short interval ultrasound in these cases. Interventions such as provider education may reduce unnecessary evaluation of indeterminate results and improve the value of HCC surveillance in patients with cirrhosis.

What Is Coming Next?

There is consensus regarding the benefit of HCC surveillance after sustained virologic response in HCV patients with cirrhosis. However, there is uncertainty whether HCC risk will sufficiently decrease over time, such that HCC surveillance can be discontinued. Recent data suggest that the risk may remain high enough in the intermediate term to require ongoing HCC surveillance.⁴³

Although HCC has been reported increasingly in some patients with non-cirrhotic NAFLD, HCC surveillance is not recommended in NAFLD patients in the absence of cirrhosis until future studies identify biomarkers that can reliably identify at-risk subgroups among patients with non-cirrhotic NAFLD.

HCC surveillance is currently recommended in all patients with cirrhosis; however, HCC risk varies widely between patients. There are currently only a few existing risk stratification tools for HCC. Most were derived from cohorts of mostly untreated (ie, active) HBV or HCV patients^44,45—virtually all are obsolete because of recent changes in the treatment paradigm for viral hepatitis and the growth of NAFLD and metabolic syndrome. A recent study developed and validated models predicting HCC risk after antiviral treatment in HCV patients and may be useful to HCC risk prediction in this specific subgroup.⁴⁶ Studies are needed that evaluate and incorporate new as well as established factors to create novel risk-stratification indices that are applicable to patients with cirrhosis from all etiologies. These will allow personalized estimation of the downstream risk of HCC in patients with cirrhosis across diverse etiologies. A novel tissue-based gene signature has been shown to accurately risk stratify patients in terms of HCC risk and may identify a subgroup of low-risk patients who do not require HCC surveillance⁴⁷; however, it requires a blood-based surrogate to facilitate adoption in clinical practice. Recent epidemiologic evidence supports the association between current smoking and HCC risk. Incorporating this information, along with other emerging risk factors, may enhance the predictive ability of risk-stratification indices.

Differences in society recommendations for at-risk individual populations require further study. For example, European Association for the Study of the Liver and American Association for the Study of Liver Diseases–Infectious Disease Society of America guidelines include HCV-infected persons with F3 fibrosis among at-risk patients, but this subgroup is not included in other guidelines. Similarly, there is uncertainty regarding cost-effectiveness of HCC surveillance in all patients with F3 fibrosis after sustained virologic response.

Last, updated cost-effectiveness models are needed to allow long-term comparative effectiveness analyses of HCC surveillance and to determine the circumstances when HCC surveillance is cost-effective in the newer cohorts of patients with virologically cured HCV, suppressed HBV, and NAFLD.

Radiographic Surveillance Tests

Liver ultrasound has been long regarded as a standard surveillance test for HCC and is one of the most commonly used. Its advantages include being non-invasive and relatively inexpensive, and it poses no risk of contrast or radiation exposure. In one of the randomized trials evaluating surveillance in HBV patients, the sensitivity of ultrasound was 84% for any stage HCC and 63% for early-stage HCC.⁴⁸ However, the performance of ultrasound appears to be suboptimal in patients with cirrhosis. A meta-analysis of cohort studies performed in patients with cirrhosis reported a pooled sensitivity and specificity of 84% (95% CI, 76%–92%) and 91% (95% CI, 86%–94%), respectively, for detection of HCC at any stage. In studies that examined early-stage HCC detection, the pooled sensitivity of ultrasound was only 47% (95% CI, 33%–61%) for detection of early-stage HCC.⁴⁹

Detection of HCC in the background of a nodular cirrhotic liver is particularly challenging due to the presence of fibrous septa and regenerative nodules, which appear as a coarse pattern on ultrasound and may mask the presence of a small tumor. In a retrospective analysis of the HALT-C (Hepatitis C Antiviral Long-Term Treatment against Cirrhosis) trial data, the most common reason for not diagnosing HCC at an early stage was “absence of detection,” that is, ultrasound failing to detect HCC lesions.⁵⁰ The operator dependency of ultrasound and the importance of high-quality equipment for good performance has been emphasized in consensus guidelines.⁵¹ To maximize ultrasound efficacy, special training for those performing ultrasounds for HCC surveillance has been advocated.

The performance of ultrasound can also be impacted by several patient-level factors, with particularly suboptimal performance in obese patients (such as those with NAFLD) and those with more advanced cirrhosis. In a retrospective analysis of 941 cirrhotic patients undergoing HCC surveillance, >20% of ultrasound examinations were determined to be of inadequate quality for surveillance. Factors correlated with ultrasound inadequacy included male sex, obesity and morbid obesity, alcohol or nonalcoholic steatohepatitis etiology of cirrhosis, advanced liver disease (Child–Pugh class B cirrhosis), inpatient status, and elevated alanine transaminase.¹⁸ Likewise, another study found male sex, Child–Pugh class B cirrhosis, and elevated AFP levels were significantly correlated with surveillance failure in patients with cirrhosis, which was defined as a tumor detected beyond an early stage or missed on ultrasound and later identified by CT or MRI.⁵² These data highlight a need for better HCC surveillance tools to improve detection of early-stage tumors.

Cross-sectional imaging modalities, such as CT or MRI, would be anticipated to have high accuracy based on data for HCC diagnosis. However, there have been few data evaluating CT or MRI for surveillance purposes, with only 2 studies evaluating CT-based surveillance and 2 evaluating MRI-based surveillance. In a single-center randomized controlled trial comparing ultrasound- and CT-based surveillance in patients with cirrhosis, the sensitivity and specificity of CT for any stage detection were 87.5% (95% CI, 50.8%–99.9%) and 87.5% (95% CI, 77.7%– 93.5%), respectively; however, the sensitivity of CT for early HCC detection was only 62.5% (95% CI, 30.4%–86.5%) and did not differ significantly from that of ultrasound.⁵³ Further, CT-based surveillance is also likely limited by potential harms, including radiation exposure and contrast-induced nephrotoxicity. The 2 studies evaluating MRI had a pooled sensitivity and specificity for any HCC detection of 83.1% (95% CI, 72.0%–90.5%) and 89.1% (95% CI, 86.5%–91.3%), respectively. In the PRIUS (Gadoxetic Acid-MRI Versus Ultrasonography for the Surveillance of Hepatocellular Carcinoma in High-Risk Patients) study, a prospective cohort study of 407 cirrhotic patients comparing MRI- and ultrasound-based surveillance, MRI-based surveillance resulted in significantly higher early-stage HCC detection compared with ultrasound (83.7% vs 25.6%; P < .001).⁵⁴ In addition, MRI had significantly fewer false-positive findings (3.0% vs 5.6%; P = 0.004). However, MRI is too time-consuming and costly for widespread adoption among all cirrhosis patients, with prior cost-effectiveness analyses demonstrating MRI is not cost-effective for HCC surveillance in all patients with cirrhosis.⁵⁵ Abbreviated MRI had been proposed as a shorter version to reduce in-scanner time, significantly save costs, and make it a more cost-effective strategy.⁵⁶ In a study among 174 patients, abbreviated MRI had a sensitivity of 81% and specificity of 96%.⁵⁷ In a subsequent pilot study among 19 patients at University of California–San Diego, abbreviated MRI had a sensitivity of 90% for early HCC and specificity of 89% compared to only 50% and 67%, respectively, for ultrasound.

Serologic Surveillance Tests

The best-studied biomarker to date is AFP, with a level of 20 ng/mL being the most commonly used cutoff to trigger further evaluation in clinical practice. A systematic review of 5 studies evaluating AFP at a cutoff value of 20 ng/mL in cirrhotic patients showed sensitivities ranging from 41% to 65% and specificities ranging from 80% to 94% for HCC at any stage.⁵⁸ However, the sensitivity of AFP for early-stage tumors is lower, at only 32%–49%. AFP may be elevated in the setting of chronic liver disease, particularly in patients with significant elevations of transaminases, and in some patients with non-HCC malignancies, such as cholangiocarcinoma.⁵⁹ Although current serologic biomarkers are insufficiently accurate when used alone, the addition of biomarkers to ultrasound for surveillance has been proposed as a means of increasing the sensitivity of surveillance, particularly with regard to early tumor detection. A meta-analysis of studies comparing the accuracy of ultrasound with or without AFP found ultrasound with AFP had significantly higher sensitivity than ultrasound alone (relative risk, 1.23; 95% CI, 1.08–1.41).⁴⁹ The pooled sensitivities of ultrasound with and without AFP for early-stage HCC were 63% (95% CI, 48%–75%) and 45% (95% CI, 30%–62%), respectively (P = .002). The benefit of AFP as an adjunct test to ultrasound was consistent across subgroups, including prospective studies, studies conducted in the United States, and studies conducted after the year 2000. These data suggest that, among currently available tests, ultrasound in combination with AFP may be the most effective strategy for HCC surveillance in patients with cirrhosis.

There have been several methods proposed to further improve AFP accuracy for HCC detection. Gopal and colleagues⁶⁰ suggested tailoring AFP cutoffs by liver disease etiology may maximize accuracy, including cutoffs of 59 ng/ mL for HCV-positive patients and 11 ng/mL for HCV-negative patients. Another proposed method to improve AFP accuracy is incorporating other patient factors to develop AFP-adjusted algorithms. In another study among 11,721 patient with HCV-related cirrhosis, an AFP-adjusted algorithm incorporating platelet count, alanine transaminase level, and patient age was developed to improve predictive value of AFP for identifying patients likely to develop HCC within 6 months.⁶¹ Finally, longitudinal AFP measurements have been evaluated to discriminate benign changes in AFP from changes that reflect true development of HCC and improve accuracy compared to single-threshold measurements.⁶² In a secondary analysis of the HALT-C trial, a parametric empirical Bayes screening algorithm was similarly shown to be more effective at detecting HCC in patients with cirrhosis than the single threshold method (77.1% vs 60.4%; P < .005).⁶³

In combination with AFP, 2 biomarkers, lens culinaris agglutinin-reactive AFP (AFP-L3) and des-γ-carboxy prothrombin are widely used in Asia for early detection of HCC. Both have been approved by the US Food and Drug Administration for risk stratification for HCC. In a US cohort, the sensitivity and specificity of AFP-L3% for detecting HCC at a cutoff of 1.7% were 37% and 94%, respectively. The combination of AFP with AFP-L3% or des-γ-carboxy prothrombin improved AFP performance for early HCC diagnosis. However, the sensitivity for early HCC for each biomarker individually has remained low in studies to date.⁶⁴ A model including sex, age, AFP-L3, AFP, and des-γ-carboxy prothrombin (GALAD) was evaluated in a multinational study with 2430 HCC patients and 4404 patients with chronic liver disease, demonstrating sensitivity >70% for overall HCC detection and >60% for early HCC detection.⁶⁵

There is a dearth of other sensitive and specific HCC biomarkers that are well-validated in prospective studies. Although the hope is that better imaging and serologic biomarkers will be available in the future, these strategies still require extensive evaluation before routine adoption in clinical practice. In the interim, we will continue to depend on ultrasound-based surveillance, with or without AFP, and efforts should focus on maximizing ultrasound quality and surveillance utilization. In our centers, we routinely use ultrasound and AFP as our surveillance strategy in patients with cirrhosis. For those in whom ultrasound-based surveillance is not effective, for example, obese patients in whom ultrasound has a visualization score of C, alternative surveillance tests, such as MRI, can be considered; however, there are few data evaluating the effectiveness of this strategy.

What Is Coming Next?

Although MRI-based surveillance could improve early tumor detection, this is likely not cost-effective if implemented in all at-risk patients, and might best be reserved for the subset of patients who are prone to ultrasound failure or have sufficiently high risk of HCC in which this strategy may be cost-effective. Other strategies, such as abbreviated MRI, may reduce cost while maintaining sufficient sensitivity for early detection but still require further evaluation.

One efficient strategy may be to improve existing blood-based biomarkers for early HCC detection. For example, platforms using combinations of biomarkers with other clinical data as well as changes over time may achieve acceptable test performance (similar to AFP algorithm and GALAD).^66,67 This framework (biomarker + demographic and clinical data + surveillance history) is likely to be an essential platform for future application of biomarkers for early HCC detection, including novel biomarkers.

Several HCC-related biomarkers, including DNA, messenger RNAs, non-coding RNAs, proteins, and posttranslational protein modifications have been identified or tested in preclinical (phase 1) studies. However, few have been evaluated in case–control (phase 2) studies, so their utility as HCC surveillance markers is not yet known. Glypican 3, GP73, osteopontin, squamous cell carcinoma antigen, human hepatocyte growth factor, and insulin growth factor-1 are examples of biomarkers that are currently being evaluated. More biomarkers are likely to be identified using proteomics-based approaches. Plasma-based methylation markers found in the cell-free DNA (ie, liquid biopsy) are also promising markers for detection of HCC. In a case– control (phase 2 biomarker) study including 95 HCC cases, 51 cirrhosis controls, and 98 healthy controls, a panel of 6 methylation markers had a sensitivity of 95% overall, 75% (3 of 4) for stage 0, and 93% (39 of 42) stage A, 93% (13 of 14) stage B, 96% (27 of 28) stage C, and 100% (7 of 7) sensitivity for stage D HCC at a specificity of 86% for cirrhotic controls.⁶⁸ Further studies, based on guidelines for biomarker development, are necessary to better evaluate the potential role of these biomarkers during surveillance.

There are at least 2 large cohorts of cirrhosis patients in the United States, including National Cancer Institute–funded Early Detection Research Network and Cancer Prevention and Research Institute of Texas–funded Texas HCC Consortium that will allow phase 3 biomarker evaluation studies. Additionally, the National Cancer Institute recently funded the Consortium on Translational Research in Early Detection of Liver Cancer to accelerate testing and validation of promising biomarkers for HCC detection. Eventually, randomized studies that assess the clinical utility of the newer biomarkers and biomarker combinations for HCC surveillance are needed.

In the era of precision medicine, we may witness programs that tailor HCC surveillance, precisely matching surveillance intensity and/or type with patients’ underlying risk and their potential for a survival benefit. This approach can maximize benefits of surveillance by shifting resource-intensive efforts toward high-risk patients, while reducing harms of testing (including false-positive results) in patients at low risk for HCC.

What Is Coming Next?

With the introduction of newer surveillance tests with higher sensitivity for very early-stage HCC detection, studies will be needed to see if surveillance intervals can be prolonged.

What Is Coming Next?

Studies are needed to evaluate more intensive interventions to increase HCC surveillance utilization with the current imaging and AFP-based tests, ensure quality of ultrasound testing, and to optimize follow-up of surveillance results.