Cost-effectiveness of HES V2.0 and GALAD for hepatocellular carcinoma surveillance in patients with cirrhosis

Abstract

Background:

Ultrasound-based surveillance for HCC is limited by suboptimal sensitivity and low adherence. Blood-based biomarkers are potential alternatives for routine HCC surveillance. We aimed to define optimal sensitivity, specificity, and cost parameters under which blood-based biomarkers could serve as cost-effective alternatives in patients with cirrhosis.

Methods:

We developed a microsimulation model of the natural history of HCC in individuals with compensated cirrhosis. We compared the cost-effectiveness of biannual surveillance using 2 blood-based biomarkers—HES V2.0 and GALAD—versus ultrasound-based surveillance across a range of sensitivity, specificity, test costs, and adherence levels. Outcomes included quality-adjusted life years (QALYs), costs, and incremental cost-effectiveness ratios (ICERs), with a willingness-to-pay threshold of $100,000/QALY.

Results:

At a $200 test cost, the optimal HES V2.0 performance was 83.8% sensitivity and 62.4% specificity, yielding more QALYs than ultrasound-based surveillance (6790 vs. 6780 per 1000 patients) with an ICER of $27,686/QALY. The optimal GALAD performance was 65.9% sensitivity and 76.4% specificity, yielding a modest QALY increase (6782 vs. 6780 per 1000 patients) with an ICER of $21,374/QALY. HES V2.0 detected more very early-stage HCCs than ultrasound-based surveillance (50.9 vs. 42.4 per 1000 patients) but increased downstream diagnostic imaging, whereas GALAD yielded smaller QALY gains with a moderate increase in diagnostic testing. At specificity ≥80%, neither biomarker was cost-effective, but cost-effectiveness was strongly influenced by adherence and test cost.

Conclusions:

Blood-based biomarkers such as HES V2.0 and GALAD can be cost-effective alternatives to ultrasound-based HCC surveillance under defined performance and cost conditions. These findings provide quantitative benchmarks to guide clinical evaluation and inform the development of future blood-based strategies for early HCC detection.

INTRODUCTION

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer, with a rising global incidence and poor prognosis when diagnosed at advanced stages. ¹ Early detection through effective surveillance improves patient outcomes and survival rates. ² Current clinical guidelines recommend biannual surveillance of individuals with cirrhosis using ultrasound in combination with alpha-fetoprotein (AFP). ³ However, ultrasound-based surveillance has well-documented limitations, including variable sensitivity, reduced accuracy in high-risk populations (eg, patients with obesity, steatosis), and low adherence rates.^4–8 These challenges highlight the need for alternative surveillance strategies for early detection of HCC that improve accuracy, increase adherence, and remain cost-effective.

Blood-based biomarkers have emerged as a potential alternative, offering the advantages of standardized measurement and greater accessibility.^6,9–13 Among these, GALAD (gender × age × log AFP × des-gamma-carboxy prothrombin) and HCC Early Detection Screening (HES) V2.0 are biomarker-based algorithms that integrate clinical and laboratory parameters. GALAD combines age, sex, AFP, AFP-L3, and des-gamma-carboxy prothrombin into a predictive score, ¹² while HES V2.0 incorporates AFP, AFP-L3, and des-gamma-carboxy prothrombin and their changes over time, age, ALT (alanine aminotransferase), platelets, and underlying liver disease etiology. ¹⁴ HES V2.0 is trained and validated for detecting HCC, and studies suggest that it improves sensitivity for early-stage HCC detection compared with existing biomarker panels.^15,16 However, the “optimal” sensitivity and specificity range of HES V2.0 has not been established.

Prior expert discussions have advocated for a minimum specificity of ≥80% for HCC surveillance biomarkers. ¹⁷ However, this threshold is derived from clinical expert opinion rather than from a formal consensus guideline issued by professional societies such as AASLD (American Association for the Study of Liver Diseases) or a formal cost–benefit analysis. While higher sensitivity could lead to increased early detection of HCC, it comes at the cost of lower specificity, which leads to unnecessary follow-up diagnostic tests and their harms. Therefore, the broader clinical adoption of this biomarker hinges on demonstrating its cost-effectiveness at different performance and cost threshold cutoffs compared with the current ultrasound-based standard.

Our objective was to determine the optimal conditions—range of sensitivity, specificity, cost—under which blood-based biomarkers, including HES V2.0 and GALAD, can be a viable and cost-effective alternative to ultrasound-based HCC surveillance. We used a simulation model to evaluate the cost-effectiveness of HES V2.0 and GALAD compared with ultrasound with alpha-fetoprotein (US+AFP) for HCC surveillance in patients with cirrhosis. By establishing the sensitivity and specificity ranges required for cost-effectiveness, this study aims to provide practical benchmarks for integrating biomarker-based HCC surveillance into clinical practice. This approach can also help guide future research on other blood-based biomarkers and implementation strategies.

METHODS

Overview

Our study presents a microsimulation Markov model—a detailed individual-level state-transition model—designed to simulate the natural history of cirrhosis and HCC across different liver etiologies and cancer stages.

We incorporated etiology-specific parameters [MASLD (metabolic dysfunction–associated steatotic liver disease), cured HCV (hepatitis C virus), and ALD (alcohol-associated liver disease), with ALD further divided into active and non-active drinkers]. These included cirrhosis progression, decompensation, and HCC incidence rates,^18–21 liver-related and competing non–liver-related mortality,^22–25 HCC treatment options, post-treatment survival, and the accuracy and costs of screening and diagnostic methods.^26–29 In addition, cost and utility values were tailored to each etiology, along with the economic and quality-of-life implications of contemporary HCC treatments.^29–33 To accurately capture the multifaceted risks inherent to each etiology, we also accounted for competing mortality risks, particularly extrahepatic manifestations such as cardiovascular events in individuals with MASLD.^22–25,34

The model evaluated the cost-effectiveness of biannual surveillance using HES V2.0 and GALAD compared with ultrasound with alpha-fetoprotein (US+AFP). For both biomarkers, we simulated outcomes under a wide range of sensitivity and specificity values as reported previously by the receiver operating characteristic (ROC) curve. ¹⁴ Primary outcomes included the number of HCC cases detected by stage, quality-adjusted life years (QALYs), total costs, and the incremental cost-effectiveness ratios (ICERs) of biannual HCC surveillance with HES V2.0 and GALAD compared with US+AFP. Additional outcomes included the number of screening and diagnostic tests performed and counts of test results by classification (true/false positives and negatives). The model follows a monthly cycle to accurately capture disease progression and surveillance outcomes, with all analyses conducted in Java and R.

Baseline population and natural history

We established a baseline population of 50-year-old individuals with compensated cirrhosis (F4) distributed as 38.45% with cured HCV, 30.08% with MASLD, and 31.47% with ALD, of which 4.00% are active drinkers and 27.47% are non-active drinkers. ¹⁴ This distribution reflects the prevalence of cirrhosis etiologies in contemporary clinical populations and serves as the starting point for the microsimulation model, capturing the heterogeneity within the population.

For each simulated individual, we modeled the clinical trajectory of their liver disease until death or age 100, considering the potential progression to HCC. Monthly, liver disease could either remain stable or progress, with the potential of decompensation and HCC development in various stages. The natural history of HCC was modeled using the Barcelona Clinic Liver Cancer staging system, categorizing HCC into “very early”, “early”, and “intermediate/advanced” stages. The schematic representation of the simulated natural history of liver disease and HCC is illustrated in Figure 1.

FIGURE 1.

Schematic representation of the natural history and treatment pathways of HCC in patients with cirrhosis. Abbreviations: ALD, alcohol-associated liver disease; DC, decompensated cirrhosis; F4, compensated cirrhosis; HCC, hepatocellular carcinoma.

Transition rates between liver disease health states were obtained from etiology-specific studies. Similarly, HCC incidence rates for decompensated cirrhosis were directly sourced from relevant literature, while incidence rates for compensated cirrhosis were derived from weighted averages of etiology-specific studies. Progression rates of HCC, however, were assumed to be independent of etiology.^35–37

Mortality risks included both liver-related and non–liver-related causes, adjusted by liver disease stage, age, and comorbidities. These mortality values were etiology-specific and considered as excess mortality on top of baseline age-adjusted and sex-adjusted mortality rates from unrelated causes. Similarly, post-treatment survival rates were dependent on etiology, liver disease stage, and HCC stage. Further details regarding model parameters are provided in Supplemental Table S1, http://links.lww.com/HC9/C413.

HCC surveillance strategies

Two surveillance strategies were simulated: (1) standard-of-care surveillance using biannual ultrasound with AFP testing, (2a) biannual surveillance using HES V2.0, and (2b) biannual surveillance using GALAD. HCC detection could occur either through surveillance or incidentally/symptomatically under each strategy. Real-world adherence to biannual surveillance was modeled using published data: 25.2% of individuals did not undergo any surveillance, 52.0% underwent inconsistent surveillance (defined as having one or more ultrasounds during a 3-year period but less than annually), and 22.8% underwent consistent biannual surveillance. ³⁸ Related surveillance parameters are summarized in Table 1. In addition, we tested scenarios assuming improved adherence levels to biomarker-based surveillance.

TABLE 1.

Base values, ranges, and distributions for surveillance-related model parameters

Input parameters	Base case	Range	Distribution	Sources
Detection of HCC
Annual probability of symptomatic/incidental detection
Very early HCC	0.016	0.014–0.018	Beta (393.70, 24,212.36)	39
Early HCC	0.121	0.109–0.133	Beta (351.47, 2553.25)	39
Sensitivity
Ultrasound with AFP	0.63	0.48–0.75	Beta (31.60, 18.56)	26
MRI	0.88	0.80–0.93	Beta (87.10, 11.88)	27
CT	0.79	0.70–0.86	Beta (81.12, 21.56)	27
Specificity
Ultrasound with AFP	0.84	0.77–0.89	Beta (124.60, 23.73)	26
MRI	0.93	0.84–0.97	Beta (56.39, 4.24)	27
CT	0.94	0.87–0.98	Beta (69.16, 4.41)	27
Screening cost
Ultrasound	$ 175	$ 131–$ 219	Gamma (64.00, 2.73)	29
AFP	$ 55	$ 41–$ 69	Gamma (64.00, 0.86)	28
MRI	$ 646	$ 485–$ 808	Gamma (64.00, 10.09)	29
CT	$ 308	$ 231–$ 385	Gamma (64.00, 4.81)	27
Biopsy	$ 918	$ 689–$ 1148	Gamma (64.00, 14.34)	29
Biopsy complication	$ 7876	$ 5907–$ 9845	Gamma (64.00, 123.06)	40

Abbreviations: AFP, alpha-fetoprotein; CT, computed tomography; HCC, hepatocellular carcinoma; MRI, magnetic resonance imaging.

To evaluate the comparative cost-effectiveness of HES V2.0 and GALAD, we systematically simulated a wide range of test performance scenarios by varying sensitivity and specificity values as reported by the published ROC curve. ¹⁴ Each sensitivity–specificity combination was paired with a test cost and adherence level to assess the cost-effectiveness of HCC surveillance.

Individuals with abnormal surveillance results were simulated to undergo diagnostic evaluation with contrast-enhanced abdominal magnetic resonance imaging (MRI) or computed tomography (CT), similar to clinical practice. Diagnostic outcomes included negative results (no HCC or missed HCC), positive results (confirmed HCC), or inconclusive findings requiring additional testing. In cases of inconclusive results, up to one additional diagnostic test and a liver mass or nodule biopsy, if necessary, were performed (Supplemental Figure S1, http://links.lww.com/HC9/C413). Once diagnosed, treatment was initiated based on tumor size, underlying liver disease stage, and etiology.

We assumed that for individuals with intermediate/advanced HCC, diagnosis occurred incidentally/symptomatically, with confirmation requiring only one MRI. Transition to decompensated liver disease led to the cessation of HCC surveillance unless individuals were on the liver transplant waitlist. Those on the liver transplant waitlist underwent surveillance, alternating between ultrasound and MRI/CT, consistent with practices observed in most US transplant programs.

HCC treatment

Upon diagnosis, patients underwent either curative treatments (eg, ablation, resection, or liver transplantation) or non-curative treatments (eg, transarterial chemoembolization, Y90 radioembolization, systemic therapy, or palliative/best-supportive care). The choice of treatment was guided by published observational studies and expert opinions, considering HCC stage, liver disease severity, and baseline etiology at diagnosis (Supplemental Table S2, http://links.lww.com/HC9/C413).

Eligibility for liver transplant was assessed using the Milan criteria and re-evaluated monthly, with probabilities of receiving a transplant determined by median wait times from Organ Procurement and Transplantation Network data. ⁴¹

Costs and quality of life

All costs associated with surveillance, HCC treatment, and liver disease management were sourced from published literature and adjusted to 2024 US dollars using the Consumer Price Index. The model accounted for direct medical costs, including the cost of diagnostic tests, treatment modalities, and follow-up care. Quality-of-life utility weights were assigned to each health state and treatment intervention, accounting for factors such as age, liver disease stage, and year of treatment (initial year vs. subsequent years). Utility weights were tailored to each etiology to reflect disease-specific burden.

Model outcomes

To enhance precision in the simulation and account for variability inherent in the modeling process, virtual cohorts of 5 million individuals were created for each simulation run, with each run replicated 10 times. Within this framework, we projected the expected number of HCC cases detected by stage, total QALYs, associated healthcare costs, and the number of screening and diagnostic tests performed, along with counts of test results by classification (true/false positives and negatives) under both surveillance strategies. All future QALYs and costs were discounted at an annual rate of 3% to reflect present value considerations.

Subsequently, ICERs were calculated to assess the cost-effectiveness of HES V2.0-based and GALAD-based surveillance relative to US+AFP. A willingness-to-pay threshold of $100,000 per QALY was used, consistent with established health economic guidelines. ⁴² The model identified conditions under which each biomarker strategy was cost-effective, considering biomarker performance characteristics and test costs. In addition, we considered a scenario where adherence to biannual HCC surveillance would improve by using a blood-based biomarker, in contrast to ultrasound-based surveillance.

Subgroup and sensitivity analyses

We conducted subgroup analyses stratified by liver disease etiology, evaluating the cost-effectiveness of biomarker-based surveillance strategies separately for MASLD, cured HCV, and ALD (active and non-active drinkers).

In addition, we performed a probabilistic sensitivity analysis to evaluate the impact of joint uncertainty across all model parameters on the primary outcomes. We simultaneously sampled 5000 parameter sets from their respective statistical distributions and simulated each set using a population size of 2 million individuals.

RESULTS

Cost-effective test characteristics of HES V2.0 and GALAD

We examined different test costs and adherence levels to identify the conditions under which HES V2.0 and GALAD were cost-effective compared with US+AFP. Figure 2 presents these findings using a two-dimensional grid, where each point corresponds to a combination of true positive rate (TPR, ie, sensitivity) and false positive rate (FPR, ie, 1—specificity). The grid is divided into cost-effective (ie, ICER ≤ $100,000 per QALY) and not cost-effective regions. The ROC curves for both biomarkers are overlaid, with segments falling in the cost-effective region indicating test performance levels where biomarkers become viable alternatives to ultrasound-based surveillance.

FIGURE 2.

Cost-effectiveness grids overlaid with ROC curves. (A) $200 test cost and baseline adherence, (B) $200 test cost and 20% improved adherence, (C) $300 test cost and baseline adherence, (D) $300 test cost and 20% improved adherence. The gray area indicates combinations of sensitivity and specificity for which the biomarkers are cost-effective compared with US+AFP (ICER ≤ $100,000/QALY), whereas the brown area indicates combinations that are not cost-effective (US+AFP performs better or is recommended). Solid and dashed lines represent the ROC curves for HES V2.0 and GALAD, respectively. Abbreviations: GALAD, gender × age × log AFP × des-gamma-carboxy prothrombin; FPR, false positive rate; HES V2.0, HCC Early Detection Screening version 2.0; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year; ROC, receiver operating characteristic; TPR, true positive rate; US+AFP, ultrasound with alpha-fetoprotein.

At a test cost of $200 and baseline adherence levels to surveillance, both biomarkers were cost-effective across broad ranges of performance. HES V2.0 was cost-effective for specificity values between 14.3% and 77.3% (corresponding sensitivity: 63.3%–98.9%), with the most favorable ICER of $27,686 per QALY at 83.8% sensitivity and 62.4% specificity. GALAD showed a similar range, with specificity between 11.8% and 77.8% (sensitivity: 62.5%–98.9%), and the most favorable ICER of $21,374 per QALY at 65.9% sensitivity and 76.4% specificity (Figure 2A). With a 20% improvement in adherence, the cost-effective regions expanded for both biomarkers (Figure 2B) with the most favorable ICERs of $31,271 per QALY for HES V2.0 (73.5% sensitivity, 69.5% specificity) and $24,352 per QALY for GALAD (58.0% sensitivity, 83.7% specificity).

When the test cost increased to $300, the cost-effective regions narrowed. Under baseline adherence, HES V2.0 remained cost-effective only within a narrow range (specificity 53.4%–62.8%, sensitivity: 81.5%–86.1%), with the most favorable ICER of $80,793 per QALY at 83.8% sensitivity and 62.4% specificity, whereas GALAD was not cost-effective in this setting (Figure 2C). With a 20% improvement in adherence, the cost-effective region expanded again (Figure 2D), with optimal ICERs of $60,641 per QALY for HES V2.0 (83.8% sensitivity, 62.4% specificity) and $74,503 per QALY for GALAD (65.9% sensitivity, 76.4% specificity). Results for additional test performance combinations are provided in Supplemental Tables S3 and S4, http://links.lww.com/HC9/C413. Subgroup analyses stratified by liver disease etiology, including MASLD, cured HCV, and ALD (active and non-active drinkers), showed qualitatively similar cost-effectiveness patterns across populations (Supplemental Figures S2–S5, http://links.lww.com/HC9/C413).

Our analysis showed that both biomarkers required minimum sensitivity thresholds to achieve QALY gains. Under baseline adherence, QALY improvements were observed only when sensitivity exceeded 63.3% for HES V2.0 (corresponding specificity 77.3%) and 62.5% for GALAD (corresponding specificity 77.8%). With a 20% improved adherence, these sensitivity thresholds were 44.2% for HES V2.0 (specificity 88.6%) and 43.2% for GALAD (specificity 89.3).

Consistent with prior biomarker and surveillance studies, we evaluated test performance across clinically relevant thresholds defined by fixed specificity levels (eg, 80%, ⁴³ and 90%^11,15,44). Neither biomarker was cost-effective at a specificity threshold of ≥90%, nor at a specificity ≥80% under baseline adherence.

However, cost-effectiveness emerged only under scenarios with improved adherence while maintaining specificity ≥80%. Under 20% improved adherence and $200 test cost, both biomarkers became cost-effective, with ICERs of $37,269 per QALY for HES V2.0 (58% sensitivity, 80% specificity) and $24,352 per QALY for GALAD (58% sensitivity, 84% specificity). At a $300 test cost, both biomarkers were cost-effective under improved adherence, though with higher ICERs.

Clinical outcomes

We also analyzed the clinical outcomes comparing HES V2.0 (at 83.8% sensitivity, 62.4% specificity, $200 test cost) and GALAD (at 65.9% sensitivity, 76.4% specificity, $200 test cost) with standard US+AFP surveillance (Table 2). The HES V2.0 strategy yielded a gain of 10.2 QALYs per 1000 patients (6790.0 vs. 6779.8) with an increase in lifetime costs of $282 per patient ($78,071 vs. $77,789), resulting in an ICER of $27,686 per QALY. This benefit was driven by a shift toward earlier diagnosis, with more “very early” stage HCC detected (50.94 vs. 42.42 per 1000) and consequently fewer intermediate/advanced stage HCC detected (43.06 vs. 44.74 per 1000). However, this higher sensitivity and lower specificity led to an increase in downstream diagnostic testing, more than doubling the lifetime number of diagnostic MRIs (1648.8 vs. 753.5) and CTs (1100.9 vs. 490.2) per 1000 patients.

TABLE 2.

Clinical and economic outcomes comparing US+AFP, GALAD, and HES V2.0 surveillance at selected sensitivity and specificity values

Outcome	US+AFP (Sens.=63.0%, Spec.=84.0%)	GALAD (test cost $200) (Sens.=65.9%, Spec.=76.4%)	HES V2.0 (test cost $200) (Sens.=83.8%, Spec.=62.4%)
QALY	6.7798	6.7817	6.7900
Cost	$77,789	$77,831	$78,071
ICER	—	$21,374	$27,686
Number of detected HCCs (per 1000 patients)
Very early HCC	42.42	43.74	50.94
Early HCC	52.60	51.78	47.40
Intermediate/advanced HCC	44.74	44.48	43.06
Total HCC	139.76	140.01	141.40
Number of tests performed (per 1000 patients, lifetime)
Surveillance test	6751.2	6748.8	6740.1
Diagnostic MRI	753.5	1084.3	1648.8
Diagnostic CT	490.2	715.9	1100.9
Biopsy	4.9	6.5	9.3

Note: ICERs are calculated relative to US+AFP. Test performance values for HES V2.0 and GALAD correspond to representative points on their respective ROC curves under baseline adherence.

Abbreviations: CT, computed tomography; HCC, hepatocellular carcinoma; ICER, incremental cost-effectiveness ratio; MRI, magnetic resonance imaging; QALY, quality-adjusted life year; Sens., sensitivity; Spec., specificity; US+AFP, ultrasound with alpha-fetoprotein.

The GALAD strategy yielded a smaller QALY gain (1.3 per 1000 patients; 6781.7 vs. 6779.8) and a modest increase in lifetime costs of $42 per patient ($77,831 vs. $77,789), resulting in an ICER of $21,374 per QALY. Similar to HES V2.0, this was associated with increased early-stage detection (“very early” stage HCC detected: 43.74 vs. 42.42 per 1000; intermediate/advanced stage HCC detected: 44.48 vs. 44.74 per 1000), but with a smaller magnitude of benefit and a more moderate increase in diagnostic testing (diagnostic MRIs: 1084.3 vs. 753.5 per 1000; diagnostic CTs: 715.9 vs. 490.2 per 1000 patients). Results for additional test performance combinations are provided in Supplemental Tables S5 and S6, http://links.lww.com/HC9/C413.

Cost threshold contours

To illustrate how biomarker test cost influences cost-effectiveness, we plotted contours of the maximum allowable test cost at which HES V2.0 remains cost-effective for different combinations of sensitivity and specificity (Figure 3). At baseline adherence levels, HES V2.0 remains cost-effective up to a test cost of $336 when sensitivity is 83.8%, and specificity is 62.4% (Figure 3A). With 20% improved adherence, this value increases to $439 (Figure 3B). GALAD remains cost-effective up to a test cost of $254 when the sensitivity is 88.6%, and the specificity is 38.9% (Figure 3A). With 20% improved adherence, this value increases to $359 (Figure 3B).

FIGURE 3.

The maximum test cost at which biomarkers remain cost-effective across sensitivity and specificity values. (A) Baseline adherence and (B) 20% improved adherence. Shaded areas indicate the maximum biomarker cost at which a biomarker remains cost-effective compared with US+AFP (ICER ≤ $100,000/QALY), with darker green shades representing higher allowable costs. Gray indicates regions where the biomarker is not cost-effective. Solid and dashed lines represent the ROC curves for HES V2.0 and GALAD, respectively. Abbreviations: FPR, false positive rate; TPR, true positive rate.

Probabilistic sensitivity analysis

To evaluate the robustness of our findings under parameter uncertainty, we conducted a probabilistic sensitivity analysis by running 5000 simulations across varying input values. Supplemental Figure S6, http://links.lww.com/HC9/C413, displays probability contour plots, where each contour represents the likelihood that HES V2.0 is cost-effective for a given combination of sensitivity and specificity.

DISCUSSION

Our study demonstrates that HES V2.0 and GALAD can be cost-effective alternatives to ultrasound-based surveillance. The optimal performance of HES V2.0 was observed at a sensitivity of 83.8% and specificity of 62.4%, while GALAD achieved its most favorable performance at 65.9% sensitivity and 76.4% specificity. Both biomarkers improved early detection of HCC, shifting diagnoses toward the “very early” stage and consequently reducing intermediate/advanced stage. However, these gains were accompanied by increased downstream diagnostic testing. Importantly, our findings highlight that constraining biomarker performance to an arbitrary higher specificity threshold (eg, ≥80%) may reduce sensitivity and limit QALY gains and push biomarkers out of the cost-effectiveness range. The results illustrate how lowering the specificity—and increasing the sensitivity—is not only acceptable but more cost-effective when positive results function as risk flags that trigger more accurate confirmatory imaging. This study provides a practical framework for biomarker evaluation that moves beyond traditional ROC analysis by integrating specific sensitivity–specificity profiles with patient adherence and test cost to determine optimal cutoffs for HCC surveillance.

Prior studies, including Singal et al, ⁴⁵ Kao et al, ⁴⁶ and international analyses by Pinjaroen et al ⁴⁷ and Hui et al, ⁴⁸ have evaluated the cost-effectiveness of HCC surveillance strategies at a fixed biomarker specificity or range. In contrast, our study systematically evaluates the full sensitivity–specificity trade-off provided by the ROC curves for both GALAD and HES V2.0, identifying the specific operating points at which each biomarker becomes cost-effective relative to ultrasound-based surveillance. This approach provides quantitative benchmarks—including sensitivity–specificity trade-offs, maximum allowable test costs, and the impact of adherence levels—that are not available from prior analyses. In addition, HES V2.0 has not previously been evaluated in any cost-effectiveness analysis.

Our findings extend beyond individual biomarkers and provide applicable performance and cost-effectiveness benchmarks for blood-based surveillance strategies. Applying this framework to both HES V2.0 and GALAD yielded consistent insights, reinforcing that our approach is broadly applicable across different biomarker platforms rather than specific to a single test. Prior to implementing any new biomarker for HCC surveillance, it is essential to assess its cost-effectiveness compared with the current standard-of-care. A fundamental trade-off exists between sensitivity and specificity: while higher sensitivity increases early detection of HCC, it can also result in a greater number of follow-up diagnostic procedures and increased downstream costs due to false positives. Our simulation model quantifies these trade-offs, offering insights into the conditions under which biomarker-based surveillance becomes a viable strategy. As additional biomarkers enter clinical evaluation, this framework can inform development decisions and regulatory pathways.

Our study also highlights the importance of improving adherence to HCC surveillance with blood-based biomarkers. We found that even modest improvements in adherence expanded the range of conditions under which both HES V2.0 and GALAD would be deemed cost-effective. Unlike imaging-based surveillance, blood-based biomarkers can be integrated into routine outpatient visits, reducing logistical barriers and potentially improving early HCC detection rates. Though we note that these adherence improvement scenarios do not assume or imply patient preference for blood-based surveillance over imaging. Biomarker development should prioritize not only analytical performance but also implementation features that enhance patient compliance. Future biomarker research should incorporate implementation science principles to maximize real-world effectiveness and cost-effectiveness. This adherence advantage extends beyond convenience to address healthcare equity concerns. ⁴⁹ Patients in rural or underserved areas who face geographic or economic barriers to regular imaging-based surveillance may benefit disproportionately from accessible, blood-based alternatives.

Our findings are important in the context of limitations of ultrasound-based surveillance, particularly its reduced sensitivity in patients with obesity and steatosis. Our estimates for ultrasound sensitivity reflect pooled real-world data, largely from the US. Blood-based biomarkers offer a promising alternative by providing more consistent sensitivity and specificity across diverse patient groups. Improvements in adherence with blood-based biomarkers were modeled as hypothetical scenarios, acknowledging that actual uptake will depend on healthcare access, patient education, insurance coverage, and provider engagement. Previous studies suggest that biomarkers may improve adherence while reducing false positives and indeterminate findings, but cost-effectiveness evaluations remain limited.^6,45

Despite these insights, our analysis has several limitations. First, we focused on individuals with compensated cirrhosis and did not model earlier fibrosis stages, as HCC risk is low and surveillance is not routinely recommended in these groups. Second, while our model incorporates first-line treatment modalities for HCC, we did not simulate treatment sequences or recurrent disease, which may underestimate long-term clinical outcomes and associated costs, though this effect would be nondifferential across biomarker strategies. Third, we assumed fixed test costs and constant test performance characteristics, whereas in practice, these may vary depending on market adoption, reimbursement policy, or laboratory infrastructure. Lastly, while the model accounted for major liver disease etiologies, we assumed population-level adherence patterns that did not vary by patient characteristics, and further stratification by patient demographics, comorbidities, and healthcare access may improve the precision of cost-effectiveness estimates. Future work should explore patient subgroups most likely to benefit from biomarker-based surveillance.

In conclusion, this study establishes performance and cost thresholds under which blood-based biomarkers such as HES V2.0 and GALAD represent viable alternatives to ultrasound-based HCC surveillance. We offer practical guidance for clinical decision-making, health policy development, and future biomarker research. By defining specific cost-effectiveness benchmarks, this study helps facilitate the integration of evidence-based biomarker surveillance into future prospective comparative studies ⁵⁰ and supports the development of more effective HCC detection strategies.

Supplementary Material

hc9-10-e01005-s001.docx ^{(776KB, docx)}