Skip to content

Hepatocellular carcinoma — surveillance and early detection

TL;DR — Six-month ultrasound with AFP is the prevailing surveillance strategy for treatment-eligible people at sufficient HCC risk, but its mortality evidence is less secure than its guideline status may imply (Singal 2023, PMID 37199193). The only large positive randomized trial enrolled 18,816 people with chronic HBV in Shanghai and reported a 37% reduction in HCC mortality despite 58% adherence (Zhang 2004, PMID 15042359). Observational meta-analysis associates surveillance with earlier detection, more curative treatment, and longer survival, but lead-time, length-time, and healthy-user biases cannot be eliminated (Singal 2022, PMID 35139400). Ultrasound alone detects only about 47% of early HCC in meta-analysis; adding AFP raises sensitivity to about 63% at a specificity cost (Tzartzeva 2018, PMID 29425931). Abbreviated MRI and blood panels improve test performance in selected studies, but a better test is not yet proof of better population outcomes.

Surveillance is a programme, not a test

A surveillance programme must repeatedly perform five linked functions:

  1. identify an eligible at-risk population;
  2. deliver testing at the intended interval;
  3. document test quality and visualization;
  4. complete diagnostic evaluation of abnormal findings;
  5. connect early diagnosis to effective treatment.

Failure at any step erodes mortality benefit. Test sensitivity measured in a case-control study cannot compensate for low uptake or diagnostic delay.

Evidence for benefit

Evidence source Population Main finding Main limitation
Zhang randomized trial 18,816 HBV-infected adults in Shanghai Six-month AFP + ultrasound; HCC mortality rate ratio 0.63 (95% CI 0.41–0.98); adherence 58% Cluster/randomization and ascertainment concerns; HBV-specific setting (Zhang 2004, PMID 15042359)
Kansagara systematic review Chronic liver disease surveillance evidence through 2014 Evidence for mortality benefit rated very low strength Sparse randomized evidence and bias in observational studies (Kansagara 2014, PMID 24934699)
Singal observational meta-analysis 59 studies; 145,396 patients with cirrhosis Early detection RR 1.86; curative treatment RR 1.83; survival HR 0.67 after lead-time adjustment High heterogeneity and residual confounding (Singal 2022, PMID 35139400)

The correct synthesis is neither “surveillance is proven like a modern screening RCT” nor “surveillance has no evidence.” One randomized HBV trial and consistent observational associations support benefit, while design limitations leave uncertainty about the magnitude and transferability to contemporary nonviral cirrhosis.

Why six months

The six-month interval is a compromise between tumor growth, test performance, feasibility, and cost. Shorter intervals increase examinations and false positives; longer intervals increase the chance that a tumor crosses curative-treatment boundaries. The interval is a programme convention supported by comparative and observational evidence rather than a precisely individualized tumor-doubling-time calculation (Singal 2023, PMID 37199193).

Ultrasound and AFP

In a meta-analysis of 32 studies, ultrasound detected any-stage HCC with pooled sensitivity of 84%, but early-stage sensitivity was 47%. Ultrasound plus AFP increased early-stage sensitivity to 63%, with lower specificity than ultrasound alone (Tzartzeva 2018, PMID 29425931).

Strategy Early-stage sensitivity Operational advantages Main costs
Ultrasound alone ~47% No radiation; widely available; low direct cost Operator dependence; obesity/nodularity; misses early tumors (Tzartzeva 2018, PMID 29425931)
Ultrasound + AFP ~63% Adds a longitudinal blood signal More false positives; AFP affected by active liver disease (Tzartzeva 2018, PMID 29425931)
Diagnostic multiphase CT Not a routine surveillance test Broad access; standard diagnostic pathway Radiation and contrast; repeated-test burden
Full MRI High lesion sensitivity No ionizing radiation; liver-specific contrast options Cost, capacity, time, contrast, claustrophobia
Abbreviated MRI Higher pooled sensitivity than ultrasound in selected cohorts Shorter protocol Validation spectrum, access, cost (Maung 2024, PMID 38413315)
GALAD Age, sex, AFP, AFP-L3, DCP Operator-independent blood algorithm Case-control inflation; assay access; threshold transportability (Yang 2019, PMID 30464023)

AFP should be interpreted longitudinally and in clinical context. A single threshold trades sensitivity for specificity; persistent rise can matter even below a fixed cutoff, while inflammation and regeneration can elevate AFP without HCC (Singal 2023, PMID 37199193).

Ultrasound failure modes

Failure mode Consequence Possible response
Obesity or thick subcutaneous tissue Beam attenuation and incomplete visualization Record visualization; consider MRI-based strategy
Coarse/nodular cirrhotic parenchyma Small tumors blend with background nodules Alternative imaging in persistently limited examinations
Deep or dome lesion Acoustic window limitation Cross-sectional imaging when concern persists
Operator and equipment variation Site-dependent sensitivity Protocols, training, quality audit
Missed appointment No biological test opportunity Recall systems, navigation, same-day coordination
Abnormal test without follow-up Delayed diagnosis despite “surveillance” Closed-loop tracking

The growing MASLD population concentrates several failure modes—obesity, nonviral etiology, and potential non-cirrhotic disease—inside the population for which the evidence base is thinnest (Singal 2023, PMID 37884736).

Harms

Direct ultrasound and venipuncture harms are small; downstream harms are not. In a safety-net cohort, false-positive or indeterminate surveillance results triggered CT, MRI, biopsy, angiography, or other procedures without HCC diagnosis (Atiq 2017, PMID 27775821). A prospective cohort similarly evaluated benefits and physical harms together rather than counting early detection alone (Singal 2021, PMID 32920214).

Modeled over five years, surveillance of 1,000 patients with cirrhosis was estimated to avert approximately 13 HCC deaths while causing about 150 false-positive experiences; assumptions about incidence, test performance, and treatment benefit drive both estimates (Taylor 2017, PMID 28605060).

Harms include:

  • anxiety and diagnostic uncertainty;
  • radiation and contrast exposure;
  • biopsy bleeding or tumor seeding;
  • incidental findings;
  • cost, travel, and time;
  • overdiagnosis of indolent tumors;
  • false reassurance after a limited-quality examination.

Physical harms were estimated in 15%–20% of patients in the inputs to one cost-effectiveness model; ultrasound plus AFP remained the dominant strategy in its compensated-cirrhosis base case (Parikh 2020, PMID 32530829).

Underuse and implementation

Surveillance underuse is a larger population problem than small differences between tests. A meta-analysis of nine intervention studies involving 4,550 patients found that organized programmes improved uptake, but the underlying cohorts were predominantly viral cirrhosis and intervention designs varied (Ramai 2023, PMID 34999648).

Intervention Mechanism Evidence caveat
Electronic reminders Prompts ordering Does not ensure attendance or follow-up
Mailed outreach Reaches patients outside visits Address, language, and trust barriers
Patient navigation Resolves scheduling and transport Resource intensive
Standing orders Reduces clinician omission Requires accurate cirrhosis registry
Same-day ultrasound with liver visit Reduces extra travel Capacity and scheduling constraints
Population health registry Enables recall and closed-loop tracking Misclassification and data governance

Equity requires measuring the whole cascade by race/ethnicity, socioeconomic status, language, rurality, and etiology—not simply the proportion with an ultrasound order.

Abbreviated MRI

Abbreviated MRI protocols retain a subset of sequences to reduce scanner time. A 2024 systematic review and meta-analysis compared non-contrast and contrast-enhanced abbreviated MRI and found high pooled diagnostic performance, but heterogeneity reflected study location, cirrhosis prevalence, HCC prevalence, and protocol (Maung 2024, PMID 38413315).

Key uncertainties are:

  • whether performance holds in prospective surveillance rather than enriched retrospective cohorts;
  • how often to alternate MRI and ultrasound;
  • which patients should be selected by ultrasound visualization or risk score;
  • contrast safety and repeated exposure;
  • cost and scanner capacity;
  • whether detection gains translate into curative treatment and mortality benefit.

Blood-based early detection

GALAD combines sex, age, AFP, AFP-L3, and des-gamma-carboxy prothrombin. In a US/German study, GALAD outperformed ultrasound for HCC detection and the combined GALADUS model was proposed (Yang 2019, PMID 30464023). A Chinese multicentre study included 602 HCC cases and 923 controls and evaluated both cross-sectional performance and longitudinal monitoring, but only 34.1% of cancers were BCLC 0–A, limiting direct translation to a pure surveillance cohort (Huang 2022, PMID 34679250).

Biomarker development must separate:

  1. case-control discrimination—known cancer versus control;
  2. prospective detection—preclinical cancer in an at-risk cohort;
  3. clinical utility—fewer late cancers or deaths with acceptable harms.

Multitarget methylation, protein, and composite panels remain candidates rather than replacements for guideline surveillance (Singal 2022, PMID 35945907).

Recall after an abnormal result

Surveillance result Typical next step Rationale
No lesion; adequate visualization Repeat in six months Continue programme
Sub-centimetre lesion Short-interval repeat ultrasound CT/MRI characterization is less reliable at very small size
Lesion ≥1 cm or suspicious finding Multiphase contrast CT or MRI Evaluate major HCC imaging features
Rising or markedly abnormal AFP without lesion Repeat/alternative imaging and etiologic assessment AFP is not lesion-localizing
Persistently limited visualization MRI or other alternative strategy A technically failed test should not be treated as negative

Diagnostic imaging belongs to the diagnosis and imaging pathway; it should not be repeated indefinitely as “surveillance” without resolving a suspicious result.

Biases in surveillance research

  • Lead-time bias: earlier diagnosis increases measured survival even if death is unchanged.
  • Length-time bias: slower tumors are more likely to be screen-detected.
  • Healthy-user bias: adherent patients may receive better care generally.
  • Confounding by treatment eligibility: healthier patients are more likely to undergo surveillance and curative therapy.
  • Immortal-time bias: classification can require survival long enough to receive surveillance.
  • Verification bias: only abnormal tests receive definitive pathology.

The 2022 meta-analysis adjusted for lead time where possible but remained highly heterogeneous, so its HR 0.67 should not be interpreted as a randomized causal effect (Singal 2022, PMID 35139400).

Prospective evidence beyond conventional ultrasound

Study Design and population Quantified result Interpretation
PREMIUM Recruiting randomized mortality trial; 4,700 US veterans with cirrhosis planned Ultrasound+AFP versus dynamic-contrast abbreviated MRI+AFP every six months; HCC mortality primary endpoint; completion estimated 2031 (NCT05486572; Ioannou 2026, PMID 41624484) Directly addresses outcome evidence but has no results as of 2026-08-30
Kim 2024 Prospective multicentre high-risk cohort; paired surveillance Annual non-contrast abbreviated MRI had higher sensitivity and diagnostic yield than biannual ultrasound, with the tradeoff of MRI capacity and false referral (Kim 2024, PMID 38636849) Comparative accuracy is not yet a mortality trial
Huang 2022 54 people with NAFLD cirrhosis; contemporaneous ultrasound and hepatobiliary-phase abbreviated MRI Severe visualization limitation: 35% with ultrasound versus 19% with abbreviated MRI; obesity predicted ultrasound limitation, OR 5.1 (95% CI 1.1–23.1) (Huang 2022, PMID 35229334) Supports visualization-triggered modality switching, not universal MRI
Vithayathil 2024 30 patients with prior VIS-B/C ultrasound 93.3% of abbreviated MRI scans were satisfactory; 20% showed abnormalities not seen on ultrasound, including one HCC (Vithayathil 2024, PMID 39123437) Small, enriched feasibility cohort
Singal 2022 Prospective-specimen, blinded biomarker evaluation; 397 cirrhosis patients, 42 HCC At 90% specificity, longitudinal GALAD sensitivity 66.7% versus AFP 40.5%; early-HCC sensitivity 69.2% (Singal 2022, PMID 34618932) Phase-2/early phase-3 performance needs programmatic outcome validation
HES V2.0 2025 Phase-3 biomarker cohort; 2,331 patients, 125 HCC, 71% early At a 10% false-positive rate, HES V2.0 exceeded GALAD true-positive rate by 7.2% overall and 14.6% at 24 months for early HCC (El-Serag 2025, PMID 38899967) Comparative biomarker performance does not establish replacement of imaging
Liou 2025 Prospective Asian surveillance cohort; 207 patients, 20 HCC For HCC within one year, GALAD AUC 0.84; cutoff −1.95 gave 75% sensitivity and 92.5% specificity, versus 12.5% sensitivity at cutoff −0.63 (Liou 2025, PMID 40346978) Threshold transportability is a major determinant of apparent performance

Interval evidence and its limits

A meta-analysis of 13 studies found six-month ultrasound increased early-stage detection versus 12-month surveillance (RR 1.17, 95% CI 1.08–1.26) and five-year survival (RR 1.39, 95% CI 1.07–1.82); three- or four-month schedules did not show clear improvement over six months (Yang 2023, PMID 36921104). Earlier non-randomized data in 559 HCV-infected people with haemophilia found no statistically significant difference in single-nodule detection between six- and twelve-month strategies, illustrating imprecision when event counts are small (Santagostino 2003, PMID 12649165).

Annual community screening also has supportive—but non-randomized—evidence. Among 14,426 HBsAg-positive adults in China, adjusted mortality HRs after lead- and length-time corrections were 0.74 (95% CI 0.60–0.91) for prevalence-round and 0.52 (0.40–0.68) for incident-round screen-detected cancers (Zeng 2023, PMID 37667043). This does not resolve whether annual surveillance is sufficient for cirrhosis or whether results transport outside HBV-endemic community programmes.

Values, harms, and test substitution

In a discrete-choice study, patients assigned 51.3% (95% CI 49.0–53.4) of decision importance to early-detection sensitivity, versus 15.2% to out-of-pocket cost, 16.7% to logistics, 9.3% to convenience, and 7.6% to physical harms; simulated preferences favored abbreviated MRI (29.0%) over ultrasound alone (3.4%) (Woolen 2022, PMID 33618022). These preferences are informative but do not remove system-level scarcity or equity constraints.

Psychological harms are measurable and generally modest. In a multicentre surveillance-outreach trial, true-positive results temporarily increased anxiety; false-positive and indeterminate results produced intermittent mild depressive effects, while decisional regret remained low across groups (Narasimman 2024, PMID 37401857). The controversy is therefore not “harms versus no harms,” but whether added detection from a more sensitive test justifies downstream imaging, biopsies, anxiety, cost, and capacity use.

Explicit controversies

  • Ultrasound-first versus risk-adapted MRI. Paired prospective studies support MRI when ultrasound visualization is poor. A 2026-08-30 search found no completed MRI mortality trial; PREMIUM is recruiting 4,700 participants and is powered for HCC mortality, converting the former absence into an active evidence gap (NCT05486572; Ioannou 2026, PMID 41624484; Kim 2024, PMID 38636849; Huang 2022, PMID 35229334).
  • Biomarker replacement versus augmentation. Longitudinal panels outperform AFP in selected cohorts, yet thresholds, assay availability, etiology, and verification design materially change sensitivity (Singal 2022, PMID 34618932; El-Serag 2025, PMID 38899967; Liou 2025, PMID 40346978).
  • Six months versus individualized intervals. Six months has pooled comparative support, while optimal intervals may vary with tumor growth and baseline risk; evidence for safe extension in low-risk strata is not yet definitive (Yang 2023, PMID 36921104).
  • Survival association versus causal effect. Lead-time, length-time, healthy-adherer, and treatment-access biases remain, even after statistical correction; the Shanghai randomized HBV trial remains unusually important because most later evidence is observational (Zhang 2004, PMID 15042359; Zeng 2023, PMID 37667043).

Open questions

  • Will PREMIUM confirm a mortality benefit and quantify harms in contemporary cirrhosis, and will its veteran population transport to other health systems (NCT05486572; Ioannou 2026, PMID 41624484)?
  • Can visualization-triggered abbreviated MRI outperform uniform ultrasound without worsening inequity or cost (Maung 2024, PMID 38413315)?
  • Which biomarker threshold and longitudinal rule should trigger imaging in prospective cohorts (Huang 2022, PMID 34679250)?
  • Can closed-loop outreach reduce late-stage HCC more than switching surveillance modality (Ramai 2023, PMID 34999648)?
  • How should overdiagnosis be estimated when HCC growth and competing cirrhosis mortality are heterogeneous (Taylor 2017, PMID 28605060)?

References

  1. Singal AG, et al. AASLD Practice Guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma. Hepatology. 2023;78:1922-1965. PMID 37199193
  2. Zhang BH, Yang BH, Tang ZY. Randomized controlled trial of screening for hepatocellular carcinoma. J Cancer Res Clin Oncol. 2004;130:417-422. PMID 15042359
  3. Kansagara D, et al. Screening for hepatocellular carcinoma in chronic liver disease: a systematic review. Ann Intern Med. 2014;161:261-269. PMID 24934699
  4. Singal AG, et al. HCC surveillance improves early detection, curative treatment receipt, and survival in patients with cirrhosis: a meta-analysis. J Hepatol. 2022;77:128-139. PMID 35139400
  5. Tzartzeva K, et al. Surveillance imaging and alpha fetoprotein for early detection of hepatocellular carcinoma in patients with cirrhosis: a meta-analysis. Gastroenterology. 2018;154:1706-1718.e1. PMID 29425931
  6. Atiq O, et al. An assessment of benefits and harms of hepatocellular carcinoma surveillance in patients with cirrhosis. Hepatology. 2017;65:1196-1205. PMID 27775821
  7. Singal AG, et al. Benefits and harms of hepatocellular carcinoma surveillance in a prospective cohort of patients with cirrhosis. Clin Gastroenterol Hepatol. 2021;19:1925-1932.e1. PMID 32920214
  8. Taylor EJ, et al. Modeling the benefits and harms of surveillance for hepatocellular carcinoma. Hepatology. 2017;66:1546-1555. PMID 28605060
  9. Parikh ND, et al. Cost-effectiveness of hepatocellular carcinoma surveillance: an assessment of benefits and harms. Am J Gastroenterol. 2020;115:1642-1649. PMID 32530829
  10. Ramai D, et al. Utilization of hepatocellular carcinoma surveillance programs in patients with cirrhosis: a systematic review and meta-analysis. J Clin Gastroenterol. 2023;57:198-203. PMID 34999648
  11. Maung ST, et al. Abbreviated MRI for hepatocellular carcinoma surveillance: a systematic review and meta-analysis. Acad Radiol. 2024;31:3142-3156. PMID 38413315
  12. Yang JD, et al. GALAD score for hepatocellular carcinoma detection in comparison with liver ultrasound and proposal of GALADUS score. Cancer Epidemiol Biomarkers Prev. 2019;28:531-538. PMID 30464023
  13. Huang C, et al. Validation of the GALAD model for early diagnosis and monitoring of hepatocellular carcinoma in Chinese multicenter study. Liver Int. 2022;42:210-223. PMID 34679250
  14. Singal AG, et al. Comparison of a multitarget blood test to ultrasound and alpha-fetoprotein for hepatocellular carcinoma surveillance: results of a network meta-analysis. Hepatol Commun. 2022;6:2925-2936. PMID 35945907
  15. Singal AG, Kanwal F, Llovet JM. Global trends in hepatocellular carcinoma epidemiology. Nat Rev Clin Oncol. 2023;20:864-884. PMID 37884736
  16. Kim DH, et al. Comparison of non-contrast abbreviated MRI and ultrasound as surveillance modalities for HCC. J Hepatol. 2024;81:461-470. PMID 38636849
  17. Huang DQ, et al. Comparative efficacy of an optimal exam between ultrasound versus abbreviated MRI for HCC screening in NAFLD cirrhosis: a prospective study. Aliment Pharmacol Ther. 2022;55:820-827. PMID 35229334
  18. Singal AG, et al. GALAD demonstrates high sensitivity for HCC surveillance in a cohort of patients with cirrhosis. Hepatology. 2022;75:541-549. PMID 34618932
  19. El-Serag HB, et al. HES V2.0 outperforms GALAD for detection of HCC: a phase 3 biomarker study in the United States. Hepatology. 2025;81:465-475. PMID 38899967
  20. Liou WL, et al. Performance of the GALAD model in an Asian cohort undergoing hepatocellular carcinoma surveillance: a prospective cohort study. J Gastroenterol Hepatol. 2025;40:1818-1824. PMID 40346978
  21. Yang J, et al. Comparative effectiveness of different hepatocellular carcinoma screening intervals or modalities: a systematic review and meta-analysis. Chin Med J (Engl). 2023;136:1322-1330. PMID 36921104
  22. Santagostino E, et al. A 6-month versus a 12-month surveillance for hepatocellular carcinoma in 559 hemophiliacs infected with hepatitis C virus. Blood. 2003;102:78-82. PMID 12649165
  23. Zeng H, et al. Performance and effectiveness of hepatocellular carcinoma screening in individuals with HBsAg seropositivity in China: a multicenter prospective study. Nat Cancer. 2023;4:1382-1394. PMID 37667043
  24. Woolen SA, et al. Patient preferences for hepatocellular carcinoma surveillance parameters. Clin Gastroenterol Hepatol. 2022;20:204-215.e6. PMID 33618022
  25. Narasimman M, et al. Hepatocellular carcinoma surveillance may be associated with potential psychological harms in patients with cirrhosis. Hepatology. 2024;79:107-117. PMID 37401857
  26. Vithayathil M, et al. Prospective study of non-contrast abbreviated MRI for hepatocellular carcinoma surveillance in patients with suboptimal hepatic visualisation on ultrasound. Cancers (Basel). 2024;16. PMID 39123437
  27. Ioannou GN, et al. Practice changing RCT design and rationale: abbreviated MRI plus AFP vs ultrasound plus AFP for HCC surveillance in cirrhosis (PREMIUM study). JHEP Rep. 2026;8:101666. PMID 41624484