Hepatocellular carcinoma (HCC) is a leading cause of cancer-related death in Hong Kong, with chronic hepatitis B (CHB) infection and liver cirrhosis (LC) as dominant risk factors. Early-stage detection enables potentially curative treatments such as resection, ablation, or transplantation, but HCC is often asymptomatic until advanced stages. To improve early detection, guidelines recommend biannual surveillance of high-risk patients.
The current recommended standard in Hong Kong is biannual ultrasound (US) plus alpha-fetoprotein (AFP). However, ultrasound access and performance can be limited by capacity, equipment quality, operator experience, and patient factors such as hepatic steatosis. AFP can increase sensitivity but reduces specificity, producing false positives that prompt confirmatory testing. For patients with constrained access to ultrasound, biomarker-based surveillance using PIVKA-II (protein induced by vitamin K absence or antagonist-II) combined with AFP has been proposed as an alternative to improve diagnostic performance and feasibility.
The objective was to evaluate the cost-utility of biannual surveillance using PIVKA-II + AFP compared with the recommended US + AFP in Hong Kong, and to compare both with US alone. The study adapted a previously described health economic Markov microsimulation model to Hong Kong clinical settings and data. Outcomes included diagnostic performance, direct healthcare costs, quality-adjusted life years (QALYs), and cost-effectiveness metrics based on a lifetime horizon from a health system payer perspective.
A state-transition microsimulation with 6-month cycles simulated individual patient health journeys. The base-case population was a mixed cohort: 85% non-cirrhotic CHB and 15% LC, consistent with local guideline populations. Starting ages varied by etiology (mean 45–56 years) and routine surveillance was assumed to stop at age 70 if no HCC was detected.
Each simulated patient was evaluated four times to compare surveillance strategies: biannual PIVKA-II + AFP, US + AFP (standard of care), and US alone. For each cycle, patients could remain in their state or transition to decompensated cirrhosis, early-stage HCC (BCLC 0/A), late-stage HCC (BCLC B/C/D), or death. Positive surveillance tests led to confirmatory CT; false positives therefore incurred additional diagnostic costs. For confirmed HCC, first-line treatments modeled included orthotopic liver transplantation (OLT), resection, radiofrequency ablation (RFA), transarterial chemoembolization (TACE), systemic therapy (sorafenib), or best supportive care, with treatment probabilities depending on HCC stage at detection.
Diagnostic accuracy for US + AFP and US alone were sourced from a published meta-analysis. Accuracy for PIVKA-II + AFP differed by cirrhosis status: non-cirrhotic performance was taken from a retrospective study, and cirrhotic performance from another published study. Alternative accuracy estimates from a different source were used in scenario analyses. Model inputs prioritized Hong Kong data, supplemented by Asia-Pacific sources where local data were unavailable. Surveillance compliance in the base case was 52% and was varied in scenario analyses.
Costs included surveillance, confirmatory diagnostics, treatments, and prognosis extracted from local publications and inflated to 2024 HK$ values. A lifetime horizon was applied with a 3% annual discount rate. In the absence of an official Hong Kong cost-effectiveness threshold, the authors used HK$422,242 per QALY gained (GDP per capita in 2024) as a reference. Utility values for health states were taken from a Chinese patient utility study; the model assumed health-state utilities were constant over time and not treatment-specific. Where stage-specific treatment costs were not available, average first-line treatment costs and their distribution by stage were used.
Across 50,000 microsimulations, an estimated 1,249 HCC cases developed per 10,000 surveilled individuals. In the mixed base-case population, PIVKA-II + AFP detected 676 early-stage HCC cases (54.1%) versus 619 cases (49.5%) with US + AFP—an absolute increase of 57 early detections per 10,000 individuals (4.6 percentage points). The biomarker combination also reduced false positive diagnoses by 7.2% relative to US + AFP, potentially lowering the need for confirmatory CT scans.
On average, surveillance with PIVKA-II + AFP was associated with a mean lifetime gain of 0.019 QALYs per person and mean cost savings of HK$4,144 per person versus US + AFP in the base case. The results were particularly favorable among non-cirrhotic CHB patients. Full tabulated diagnostic, cost, and health outcomes were reported in the source's tables.
The analysis included one-way sensitivity analyses to identify key model drivers and probabilistic sensitivity analysis (PSA) to assess overall uncertainty. Scenario analyses tested alternative diagnostic accuracy sources for PIVKA-II + AFP, different surveillance compliance rates, and different target populations (CHB-only or LC-only). Detailed sensitivity and scenario results were provided in the study appendices referenced in the source.
The adapted microsimulation suggests that biannual surveillance using PIVKA-II + AFP could be a feasible alternative for CHB and LC patients in Hong Kong who have limited access to ultrasound. The combination may modestly increase early-stage HCC detection, reduce false positives, improve access to early treatments, and yield both QALY gains and cost savings in the base-case mixed population. The authors highlight particularly favorable outcomes for non-cirrhotic CHB patients.
The study used local cost, treatment distribution, and utility inputs and validated structural assumptions with Hong Kong clinical experts. Detailed uncertainty analyses were performed; results and sensitivities are presented in the source appendices.
Limitations: The source article reports data sources and sensitivity analyses but full numeric details, all parameter values, and extensive appendix tables are contained within the original manuscript and supporting information files referenced by the authors.