Emerging Tools for Hepatocellular Carcinoma Surveillance

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related death worldwide, although curative options can afford long-term survival in patients with early-stage tumors. Given the association between early HCC detection and improved survival, professional society guidelines recommend semi-annual HCC surveillance in patients with chronic hepatitis B infection and those with cirrhosis from any etiology.^1–3 Abdominal ultrasound, with or without alpha fetoprotein (AFP), has remained the most recommended surveillance strategy for over two decades, although these tests in combination miss over one-third of HCC at an early stage.⁴ Further, ultrasound is highly operator-dependent, with variable performance based on the experience of individuals performing the exam and interpreting radiologists, as well as prone to poor visualization in patients with obesity or non-viral etiologies of cirrhosis.⁵ With an increasing proportion of HCC cases due to alcohol and nonalcoholic steatohepatitis (NASH), ultrasound-based surveillance effectiveness may decline further over time, highlighting a need for novel surveillance strategies. Herein, we discuss emerging imaging and blood-based tools for HCC surveillance.

Imaging-based surveillance tools

There has been increasing interest in alternative imaging modalities, including computed tomography (CT) and magnetic resonance imaging (MRI) (Table 1), although neither are recommended by current society guidelines.^2,3 A prospective cohort study comparing low-dose two-phase CT scan (arterial phase and 3-minute delayed phase) and ultrasound surveillance in 139 patients with cirrhosis followed for 1.5 years reported CT imaging having higher sensitivity for very early-stage HCC (82% vs. 18%, p<0.001) but no significant difference in early-stage detection (86% vs. 57%, p=0.13).⁶ Concerns about radiation exposure and contrast injury, particularly if repeated at semi-annual intervals, also diminish enthusiasm for CT-based surveillance. A prospective cohort study from South Korea, in which 407 patients with cirrhosis (majority HBV-infected) were followed for 1.5 years, reported MRI (using gadoxetate disodium contrast) also having higher sensitivity for early-stage HCC (86% vs. 26%, p<0.001) and higher specificity (97% vs. 94%, p=0.004) compared to ultrasound.⁷ Although MRI is regarded as safe, these data require validation in Western non-HBV populations, and other potential concerns about radiologic capacity, positive predictive value, patient acceptance, and cost-effectiveness need to be evaluated.

Table 1.

Emerging imaging-based surveillance strategies for early-stage HCC detection

Imaging Modality	Study Design	Number of Patients with HCC (% early-stage)	Performance characteristics from select studies
			Sensitivity	Specificity
Two-phase CT	Single arm cohort study6	24 (100%)	83.3% CT vs. 29.2% US	95.6% CT vs. 87.7% US
Multi-phase MRI**	Single arm cohort study6	43 (98%)	86.0% MRI vs. 27.9% US	97.0% MRI vs. 94.4% US
Abbreviated MRI**	Meta-analysis of case-control studies8*	917 (NR)	82% AMRI vs. 53% US	98% AMRI vs. 88% US
	Post-hoc analysis of single arm cohort study19	43 (98%)	86.0% AMRI vs. 27.9% US	95.6% AMRI vs. 96.3% US

^*Performance characteristics from the subgroup of studies evaluating both AMRI and ultrasound.

^**Studies performed with gadoxetate disodium contrast

AMRI – Abbreviated MRI; CT – computed tomography; HCC – hepatocellular carcinoma; MRI – magnetic resonance imaging

These latter concerns about MRI-based surveillance may be mitigated by abbreviated MRI (AMRI) protocols, using a subset of sequences from a full diagnostic protocol, which can shorten the exam from ~45 minutes to ~15 minutes. Meta-analyses of AMRI performance have reported pooled sensitivity and specificity estimates of 0.86 and 0.94–0.96, respectively, although AMRI performance may be overestimated given inclusion of patients without cirrhosis in several studies.⁸ Further, several studies simulated AMRI by selecting sequences from a diagnostic MRI exam, so AMRI was not independent from the reference standard. Results from an ongoing randomized clinical trial (NCT03731923) comparing annual AMRI to ultrasound for early-stage HCC detection are anticipated in the next couple years.

Blood-based surveillance tools

Although there are several emerging blood-based biomarkers for HCC surveillance (Table 2), heterogeneity within and between HCC nodules limit the potential accuracy for any single biomarker, and panels with multiple biomarkers will likely be required to achieve adequate performance. A phase II study from the Early Detection Research Network (EDRN) demonstrated that AFP, AFP-L3%, and des-gamma-carboxy-prothrombin (DCP) each have insufficient accuracy in isolation. A case-control study with 308 patients with HCC (40% early-stage) and 740 patients with chronic liver disease found AFP and DCP had the highest area under the receiver operating characteristic (AUROC) among all biomarkers.⁹ Adding age, gender, and any of four potential biomarkers increased performance, achieving a sensitivity for early-stage HCC of 70%. Of these possible combinations, the most extensively evaluated is GALAD (incorporating Gender, Age, AFP-L3%, AFP, and DCP), which achieved sensitivities of 60.6% – 80.2% for early-stage HCC in a multinational case-control study including three patient samples with 6834 patients (2430 HCC and 4404 chronic liver disease).¹⁰ A subsequent case-control study in patients with NASH (125 HCC cases and 231 controls) showed GALAD achieved sensitivity and specificity for early-stage HCC detection of 72% and 95%, respectively.¹¹ GALAD was recently found to have a sensitivity of 54% for early-stage HCC, with specificity fixed at 90%, in a single center cohort study¹², and results from phase III validation in larger cohort studies are anticipated in 2022. These data will be critical to evaluate the performance of GALAD versus ultrasound and AFP and help determine if validation in phase IV and V biomarker studies is warranted (Table 3).

Table 2.

Emerging blood-based biomarker surveillance strategies for early-stage HCC detection

Biomarker	Study Design*	Number of Patients with HCC (% early-stage)	Performance characteristics from select studies
			Sensitivity for early-stage HCC	Specificity
AFP	EDRN Phase II biomarker study20	417 (50%)	53%	90%
AFP-L3	EDRN Phase II biomarker study20	417 (50%)	28%	97%
DCP	EDRN Phase II biomarker study20	417 (50%)	61%	70%
GALAD	Multinational Phase II biomarker study10	2183 (48%)	61% – 80%	89 – 96%
	Case-control study in NASH11	125 (23%)	72%	95%
	Pilot cohort study12		54%	90%
Oncoguard Liver	Phase II biomarker study13	156 (50%)	82%	87%
Helio	Phase II biomarker study14	122 (31%)	76%	91%
mSEPT9	Case-control study	60 (100%)	77%	64%
Multi-cancer platform	Subgroup of phase II biomarker study15	42 (31%)	81%	NR**

^*Phase II biomarker studies were adequately powered based on pre-specified outcomes, whereas case-control studies provide exploratory data that should be verified in larger phase II biomarker studies

^**Control patients in the study did not have cirrhosis so specificity in a cirrhosis population is likely overestimated

Table 3.

Phases of biomarker validation

Biomarker phase	Description
Phase I	Pre-clinical case-control study to identify potential markers
Phase II	Adequately powered case-control study to define sensitivity and specificity
Phase III	Cohort study in which samples are prospective collected and stored, with biomarkers beings retrospectively evaluated to validate sensitivity and specificity
Phase IV	Prospective cohort study in which biomarkers are acted upon to define true positive and false positive rate as well as tumor stage distribution
Phase V	Randomized clinical trial evaluating HCC-related mortality and screening-related harms

In parallel, there has been increasing interest in applying liquid biopsy techniques for early HCC detection. Liquid biopsy entails the analysis of tumor components -mainly fragments of circulating tumor DNA (ctDNA), extracellular vesicles and circulating tumor cells- released to the bloodstream and accessible for molecular characterization. It has been extensively applied in oncology for different clinical applications including cancer screening, prognostication, and prediction of treatment response. In HCC, a panel of ctDNA methylated CpG sites (HOXA1, TSPLY5, and B3GALT6) along with sex and AFP showed promising accuracy in a phase II biomarker validation study (156 patients with HCC [50% early-stage] and 245 controls [92% cirrhosis]). Sensitivity and specificity for early-stage HCC detection were 82% and 87%, respectively.¹³ Another ctDNA-based methylation marker panel (including age, sex, AFP, AFP-L3%, DCP, and 28 methylated DNA markers) demonstrated a similar sensitivity and specificity of 85% and 91%, respectively, in a phase II case-control study with 122 HCC cases (31% early-stage) and 125 controls (37% cirrhosis, 58% HBV infection).¹⁴ Both panels are completing phase III evaluation in patients with cirrhosis undergoing surveillance.

In parallel with HCC-specific biomarkers described above, multi-cancer detection platforms are being developed. One such platform was examined in a large, multi-center phase II case-control study among 2823 patients with cancer and 1254 healthy patients without cancer.¹⁵ The overall sensitivity and specificity of the panel were 51.5% and 99.5%, respectively, with sensitivity increasing from 16.8% for stage I disease to 40.4% for stage II, 77.0% for stage III, and 90.1% for stage IV disease. Sensitivity and positive predictive value of this multi-cancer detection panel for liver or bile duct cancer were both high at 93.0% and 81.6%, respectively, although the study only included 43 patients with liver cancer and controls did not have underlying cirrhosis. Therefore, further studies are needed to evaluate the performance of these panels for identifying early-stage liver cancer among at-risk patients with cirrhosis or chronic HBV infection.

The potential for biomarker strategies to improve surveillance effectiveness extends beyond performance characteristics, as ease of implementing of a biomarker-based strategy could also increase adherence (Figure 1). Several studies have demonstrated underuse of ultrasound-based surveillance, related to both patient- and provider-level barriers, such as transportation barriers, uncertainty where to get ultrasound exams completed, and difficulty with scheduling.¹⁶ Many of these barriers would be addressed by blood-based biomarkers, which could be performed the same day as a clinic visit.

Figure 1.

Surveillance Effectiveness is driven by test performance and utilization

Compared to ultrasound surveillance (strategy 1), biomarker-based surveillance may increase overall effectiveness by decreasing surveillance barriers and increasing utilization despite similar if not lower sensitivity for early-stage HCC (strategy 2). Conversely, imaging-based surveillance may have higher sensitivity for early-stage HCC but achieve lower test effectiveness given difficulty implementing in broad at-risk populations resulting in decreased utilization (strategy 3).

Summary and Future Direction

Although there are several emerging surveillance modalities, most have only been evaluated in phase II biomarker studies and still require validation in phase III and IV studies (Table 3). This step is important given phase II studies, in which cases with known HCC are enrolled, can overestimate biomarker performance compared to phase III and phase IV cohort studies, in which the at-risk population is enrolled and followed. Further, while phase II and phase III biomarker data establish clinical validation of test performance, prospective clinical utility (phase IV and V) studies, in which the biomarker is tested and acted upon, are necessary to demonstrate that surveillance strategies improve clinical outcomes including early detection and survival.¹⁷ This step is particularly important as some blood-based biomarkers, including AFP, have potential to not only act as surveillance tests (i.e., detecting pre-clinical HCC) but also have a role in risk stratification (i.e., identifying those at risk of HCC prior to its presence). Therefore, some blood-based biomarkers may turn positive well in advance of cancer diagnosis, resulting in potential physical, financial, and psychological harms.¹⁸

Maturation of large well-characterized prospective cohort studies with standardized blood collection protocols, such as the EDRN HCC Early Detection Strategy (HEDS) Study, Texas HCC Consortium (THCCC), and ANRS CirVir Cohort, will facilitate phase III evaluation of several biomarkers. However, alternative strategies will need to be considered for validation of liquid biopsy techniques and imaging-based surveillance modalities. Certain molecular analyses in liquid biopsy (e.g., extracellular vesicle purification) require specific processing protocols beyond what is used in most blood banking efforts, and many existing cohorts did not capture radiologic images.

Marked interest in emerging surveillance tools, particularly liquid biopsy, paired with industry pressure to be early-to-market has spurred commercialization of surveillance tools prior to robust evidence for clinical utility. This underscores the importance of professional societies to adequately frame the role of these new surveillance tools, considering emerging data to evaluate the appropriateness of their use in clinical settings. While awaiting prospective validation data for novel biomarkers and imaging techniques, ultrasound with or without AFP remains the current gold standard HCC surveillance strategy.