Skip to content

Hepatocellular carcinoma — cirrhosis and risk stratification

TL;DR — Cirrhosis is the dominant HCC risk substrate, but risk varies materially by etiology, age, sex, fibrosis severity, portal hypertension, viral control, diabetes, alcohol exposure, and prior HCC (Singal 2023, PMID 37199193). Risk stratification has two jobs: identify people whose incidence justifies surveillance and enrich higher-risk groups for better or more intensive tests. Etiology-specific scores such as PAGE-B can separate treated HBV populations into markedly different risk groups, but systematic review shows calibration and transportability vary across settings (Wu 2021, PMID 33667678). After HCV cure, cirrhosis retains substantially more risk than non-cirrhotic disease (Lv 2024, PMID 38965190). No score currently replaces fibrosis assessment, clinician judgment, or guideline-defined surveillance eligibility.

The cirrhotic field

HCC often develops through repeated injury, inflammation, cell death, regeneration, fibrosis, and clonal selection. Cirrhosis is therefore both a marker of accumulated carcinogenic exposure and an active tissue environment for cancer. It also creates competing risks—decompensation, infection, bleeding, and non-HCC death—that change whether earlier tumor detection can improve survival (Singal 2023, PMID 37199193).

Risk is not binary at the histologic boundary of cirrhosis. Advanced fibrosis and cirrhosis form a continuum, non-invasive tests misclassify some patients, and HBV or MASLD-related HCC can arise before cirrhosis is recognized (Huang 2021, PMID 33349658).

Variables that repeatedly predict HCC

Domain Higher-risk feature Why it matters Limitation
Fibrosis Cirrhosis; higher liver stiffness; portal hypertension Integrates duration and severity of injury Stiffness is affected by inflammation and congestion
Demography Older age; male sex Consistent across etiologies Does not define mechanism
Viral HBV viral activity; HCV without sustained response Ongoing inflammatory/oncogenic exposure Treatment changes risk over time
Liver function Lower albumin, higher bilirubin, thrombocytopenia Captures reserve and portal hypertension Also predicts competing death
Metabolic Diabetes, obesity, MASLD/MASH Adds inflammatory and metabolic exposure Common, so specificity is low
Alcohol Ongoing heavy exposure; alcohol-associated cirrhosis Continues injury and decompensation risk Exposure measurement is unreliable
Tumor history Prior HCC Field cancerization and residual microscopic disease Recurrence surveillance differs from primary surveillance
Imaging Indeterminate or non-characterized nodule May already represent early neoplasia Creates verification and lead-time bias

The purpose of a score determines which variables belong in it. A model optimized to predict HCC incidence may be poor at estimating net surveillance benefit if it ignores competing liver and cardiovascular death.

Etiology-specific incidence

Population Published estimate Key qualification
MASH/NASH cirrhosis 0.5%–2.6% per year Range across cohorts and definitions (Huang 2021, PMID 33349658)
Non-cirrhotic MASLD/NAFLD ~0.1–1.3 per 1,000 patient-years Population average; risk is concentrated (Huang 2021, PMID 33349658)
Alcohol-associated cirrhosis 0.9%–5.6% per year Wide between-study heterogeneity (Huang 2023, PMID 36258033)
HCV sustained response with cirrhosis 2.47 per 100 person-years Meta-analysis after antiviral treatment (Lv 2024, PMID 38965190)
HCV sustained response without cirrhosis 0.85 per 100 person-years Mixed fibrosis stages; lower than cirrhosis (Lv 2024, PMID 38965190)

These figures should not be compared as if they came from one cohort. Calendar era, fibrosis ascertainment, surveillance intensity, and competing death differ. Their value is to show why “cirrhosis” is a high-risk category and why non-cirrhotic MASLD remains a difficult population-level problem.

Who enters surveillance

AASLD guidance recommends surveillance for adults with cirrhosis who would be candidates for HCC-directed treatment and for selected non-cirrhotic chronic HBV populations whose risk is sufficiently high (Singal 2023, PMID 37199193). Surveillance is not automatically beneficial when severe decompensation or comorbidity makes transplant or HCC treatment impossible; the relevant endpoint is benefit from detection, not detection alone.

Decision Evidence needed Common error
Diagnose advanced fibrosis/cirrhosis Clinical history, platelets, elastography, imaging, prior biopsy Treating a single normal test as exclusion
Establish etiology Viral tests, alcohol history, metabolic evaluation, disease-specific work-up Assigning only one cause when exposures coexist
Estimate HCC incidence Etiology, age, sex, fibrosis, treatment response, risk score Importing a model outside its derivation population
Estimate net benefit Liver reserve, comorbidity, transplant/treatment candidacy, preferences Equating cancer risk with surveillance benefit
Select modality Ultrasound visualization, body habitus, prior findings, access Repeating inadequate ultrasound indefinitely

Incidence thresholds and cost-effectiveness

Traditional surveillance thresholds were often summarized as approximately 1.5% annual HCC incidence in cirrhosis, but contemporary modeling suggests lower thresholds may be cost-effective as treatments improve and costs change. In virologically cured HCV, a microsimulation found surveillance cost-effective above an HCC incidence of approximately 0.7% per year using a $100,000/QALY threshold (Chhatwal 2024, PMID 37302445).

This is not a universal biological cutoff. It depends on:

  • surveillance test sensitivity and specificity;
  • adherence and downstream diagnostic completion;
  • eligibility for curative treatment;
  • competing mortality;
  • test, treatment, and health-state costs;
  • willingness-to-pay threshold.

A separate Markov model incorporating surveillance harms found ultrasound plus AFP dominated ultrasound alone and no surveillance in the base case for compensated cirrhosis; physical harms were modeled because 15%–20% of patients may undergo downstream testing after false-positive or indeterminate results (Parikh 2020, PMID 32530829).

HBV prediction models

HBV models include combinations of age, sex, platelets, viral markers, cirrhosis, and treatment status. PAGE-B is intentionally simple—platelets, age, and sex—and was developed for patients receiving antiviral therapy.

A systematic review examined 14 HBV HCC-prediction models and externally validated them in a treated Chinese cohort. Discrimination and calibration varied; models built in untreated Asian cohorts did not necessarily transport to treated populations, and risk categorization was sensitive to the selected time horizon (Wu 2021, PMID 33667678).

Model family Typical inputs Intended population Main caution
PAGE-B / modified PAGE-B Platelets, age, sex; sometimes albumin Treated chronic HBV Ethnic and geographic transportability
REACH-B Sex, age, ALT, HBeAg, HBV DNA Initially non-cirrhotic untreated HBV Antiviral-era calibration
CU-HCC / GAG-HCC Viral, biochemical, cirrhosis variables Asian chronic HBV cohorts Different cirrhosis definitions
aMAP Age, male sex, albumin-bilirubin, platelets Multi-etiology chronic liver disease Thresholds and clinical action not fully standardized

Scores can identify a low-risk stratum, but stopping surveillance solely on a score demands prospective evidence that missed cancers and mortality remain acceptable.

Risk after HCV cure

Sustained virologic response sharply lowers HCC occurrence compared with treatment nonresponse, but fibrosis accumulated before cure remains consequential. In 29,395 treated patients, occurrence was threefold higher in cirrhosis than non-cirrhosis after sustained response (2.47 versus 0.85 per 100 person-years) (Lv 2024, PMID 38965190).

Imaging context also matters. A prospective analysis of patients with advanced liver disease after HCV cure showed that baseline non-characterized nodules identify a group needing careful diagnostic resolution rather than routine risk-score classification (Sanduzzi-Zamparelli 2022, PMID 34856322).

Risk is dynamic after cure. Age increases, fibrosis may regress, diabetes or alcohol exposure may progress, and liver stiffness can fall partly because inflammation resolves. A one-time post-cure score should not be assumed to remain calibrated indefinitely.

MASLD/MASH risk enrichment

The central MASLD problem is scale: non-cirrhotic incidence is low, but the exposed population is extremely large (Huang 2021, PMID 33349658). Fibrosis stage is the most useful enrichment axis, supplemented by age, sex, diabetes, and possibly genetic or longitudinal biomarker information.

Candidate enrichment tool Strength Unresolved issue
FIB-4 / platelet-based scores Cheap and scalable Designed for fibrosis, not HCC prediction
Elastography Continuous fibrosis/portal-pressure proxy Platform, inflammation, and cutoff variation
Imaging-defined cirrhosis Clinically interpretable Low sensitivity for compensated disease
Multi-variable HCC scores Can combine competing predictors External calibration and clinical thresholds
Germline variants Mechanistic enrichment potential Small incremental performance and equity concerns
Longitudinal biomarker trajectories May detect transition to cancer risk Requires prospective sampling and assay standardization

Blanket surveillance of all MASLD would generate a large false-positive burden. A successful strategy must show not only discrimination, but calibration, feasible ascertainment, improved early detection, and acceptable cost and harm.

Competing risk and surveillance benefit

HCC risk rises with cirrhosis severity, but the chance of benefiting from early detection can fall as decompensation and non-cancer mortality rise. This creates a non-monotonic relationship:

  1. low fibrosis → low HCC incidence;
  2. compensated cirrhosis → high enough incidence and sufficient treatment reserve;
  3. severe decompensation without transplant option → high incidence but limited actionable benefit.

Risk models that omit transplant candidacy and competing death answer “who develops HCC?” rather than “who benefits from surveillance?” AASLD therefore conditions surveillance on treatment eligibility and discourages it when life-limiting comorbidity precludes meaningful HCC therapy (Singal 2023, PMID 37199193).

Dynamic reassessment

Risk should be reconsidered when any of the following occurs:

  • new viral suppression or HCV cure;
  • fibrosis regression or progression;
  • new decompensation;
  • new diabetes, obesity, or alcohol exposure;
  • inadequate ultrasound visualization;
  • an indeterminate nodule;
  • change in transplant eligibility;
  • prior HCC treatment.

The interval and modality may change even when the underlying indication for surveillance remains.

What current scores cannot do

  • They do not diagnose HCC.
  • They do not reliably select among ultrasound, MRI, and biomarker panels.
  • They do not prove that score-guided surveillance reduces mortality.
  • They rarely model competing death and treatment eligibility jointly.
  • They can reproduce inequities if access, race, or laboratory availability proxy structural disadvantage.
  • They may be miscalibrated after etiologic treatment or in a different geographic population.

What risk models actually separate

Risk scores are most useful for separating broad risk strata, not declaring an individual “safe.” The aMAP model was derived and externally validated across chronic hepatitis populations using age, sex, albumin-bilirubin, and platelets; its appeal is etiologic breadth, while its limitation is that calibration and actionable thresholds still depend on setting (Fan 2020, PMID 32707225). HBV-specific models incorporate viral and treatment-era variables, and external validation shows that discrimination can shift when applied to treated Chinese cohorts rather than derivation populations (Lee 2016, PMID 27729738; Lou 2021, PMID 34218498).

Population Candidate stratifier Quantified result or key finding Why it does not yet replace eligibility rules
Mixed chronic hepatitis aMAP Low-risk group had a negative predictive value of 99.5% in development/validation Rare cancers still occur; calibration and acceptable miss-rate are policy choices (Fan 2020, PMID 32707225)
Biopsy-proven NAFLD FIB-4, NAFLD fibrosis score, APRI and histology In 1,389 patients, non-invasive markers separated long-term HCC risk Biopsy-selected cohorts and long prediction horizons limit population transportability (Toyoda 2023, PMID 39131553)
Non-cirrhotic NAFLD Non-invasive fibrosis tests Individual-patient and aggregate meta-analysis found risk gradients but insufficiently validated surveillance thresholds Low absolute incidence means specificity must be extremely high (Siriwong 2025, PMID 39919008)
Cured HCV with cirrhosis aMAP and other post-SVR models UK validation found variable calibration and discrimination across models Treatment-era, competing death, and changing fibrosis alter risk over time (Innes 2021, PMID 34805817)
High-risk HCV after SVR aMAP, THRI, PAGE-B-related approaches Comparative validation showed materially different performance among acknowledged scores A ranking within one cohort does not establish a universal clinical cut-point (Qiu 2023, PMID 36890735)
Veterans with cured HCV cirrhosis Post-SVR liver stiffness Liver stiffness improved HCC risk prediction beyond static baseline fibrosis measures Elastography platform, inflammation, and repeated-measure timing matter (John 2024, PMID 38061410)
Compensated cirrhosis Ultrasound/AFP and MRI-triggered strategies UK modeling found cost-effectiveness sensitive to test performance, cost, adherence, and incidence Economic results are health-system specific (Garay 2024, PMID 39127246)

Dynamic risk after etiologic cure

HCV cure changes the hazard rather than resetting it. In a large treated cohort, HCC risk declined as time accrued after eradication, but the decline was heterogeneous and did not justify stopping surveillance solely because several event-free years had passed (Vutien 2024, PMID 37955206). Post-treatment fibrosis measurements can outperform pretreatment staging because they partly capture regression, persistent portal hypertension, and ongoing injury; this makes a one-time “cirrhosis yes/no” label biologically crude (John 2024, PMID 38061410).

Cirrhosis itself remains the dominant cross-etiology enrichment state. A meta-analytic assessment found substantially higher HCC development with cirrhosis across HBV, HCV, alcohol-related, and other liver diseases, but the magnitude varied by etiology and follow-up (Tarao 2019, PMID 30791221). In metabolic disease, diabetes, age, male sex, and advanced fibrosis repeatedly increase risk, yet the enormous non-cirrhotic denominator makes a small absolute-risk subgroup difficult to identify efficiently (Xiao 2024, PMID 39047929; Younossi 2016, PMID 26707365).

Competing positions and unresolved thresholds

  • Universal cirrhosis surveillance versus risk-based surveillance. Universal eligibility is simple and avoids model misses; individualized eligibility could reduce low-yield testing but requires externally calibrated thresholds and explicit tolerance for false negatives (Fan 2020, PMID 32707225; Chhatwal 2024, PMID 37302445).
  • Post-SVR discontinuation. Declining time-varying risk supports reassessment, while persistent fibrosis, older age, diabetes, and portal hypertension argue against an unqualified stopping rule (Vutien 2024, PMID 37955206; John 2024, PMID 38061410).
  • F3 fibrosis. Some models identify subgroups above conventional cost-effectiveness thresholds, but imprecise fibrosis classification and low average incidence make blanket surveillance contentious (Siriwong 2025, PMID 39919008; Garay 2024, PMID 39127246).
  • HBV without cirrhosis. Viral suppression lowers incidence but does not eliminate integration-related carcinogenesis; HBV-specific scores can refine rather than replace age-, ancestry-, family-history-, and region-based policies (Lee 2016, PMID 27729738; Campbell 2021, PMID 33305479).

Open questions

  • Can risk-guided surveillance intensity improve mortality or cost-effectiveness compared with uniform six-month surveillance in cirrhosis (Chhatwal 2024, PMID 37302445)?
  • At what sustained post-HCV-cure risk can surveillance safely stop, and how should fibrosis regression be measured (Lv 2024, PMID 38965190)?
  • Can an HCC-specific MASLD model enrich non-cirrhotic risk enough for prospective surveillance trials (Huang 2021, PMID 33349658)?
  • Should competing decompensation and cardiovascular death be incorporated into all HCC surveillance scores?
  • How should indeterminate nodules be represented in risk models without converting occult cancer into a “predictor” (Sanduzzi-Zamparelli 2022, PMID 34856322)?

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. Wu S, et al. Hepatocellular carcinoma prediction models in chronic hepatitis B: a systematic review of 14 models and external validation. Clin Gastroenterol Hepatol. 2021;19:2499-2513. PMID 33667678
  3. Lv GJ, et al. Risk of hepatocellular carcinoma occurrence after antiviral therapy for patients with chronic hepatitis C infection: a systematic review and meta-analysis. Hepatol Int. 2024;18:1459-1471. PMID 38965190
  4. Chhatwal J, et al. Hepatocellular carcinoma incidence threshold for surveillance in virologically cured hepatitis C individuals. Clin Gastroenterol Hepatol. 2024;22:91-101.e6. PMID 37302445
  5. 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
  6. Sanduzzi-Zamparelli M, et al. Liver cancer risk after HCV cure in patients with advanced liver disease without non-characterized nodules. J Hepatol. 2022;76:874-882. PMID 34856322
  7. Huang DQ, El-Serag HB, Loomba R. Global epidemiology of NAFLD-related HCC: trends, predictions, risk factors and prevention. Nat Rev Gastroenterol Hepatol. 2021;18:223-238. PMID 33349658
  8. Huang DQ, et al. Global epidemiology of alcohol-associated cirrhosis and HCC: trends, projections and risk factors. Nat Rev Gastroenterol Hepatol. 2023;20:37-49. PMID 36258033
  9. Tarao K, et al. Real impact of liver cirrhosis on the development of hepatocellular carcinoma in various liver diseases: meta-analytic assessment. Cancer Med. 2019;8:1054-1065. PMID 30791221
  10. Lee HW, et al. Prediction models of hepatocellular carcinoma development in chronic hepatitis B patients. World J Gastroenterol. 2016;22:8314-8321. PMID 27729738
  11. Lou T, et al. External validation of hepatocellular carcinoma risk scores in patients with chronic hepatitis B virus infection in China. J Viral Hepat. 2021;28:1373-1380. PMID 34218498
  12. Fan R, et al. aMAP risk score predicts hepatocellular carcinoma development in patients with chronic hepatitis. J Hepatol. 2020;73:1368-1378. PMID 32707225
  13. Toyoda H, et al. Validation of noninvasive markers for HCC risk stratification in 1,389 patients with biopsy-proven NAFLD. Gastro Hep Adv. 2023;2:1093-1102. PMID 39131553
  14. Siriwong N, et al. Role of noninvasive tests on the prediction of hepatocellular carcinoma in nonalcoholic fatty liver disease patients without cirrhosis: a systematic review and meta-analysis. Eur J Gastroenterol Hepatol. 2025;37:358-369. PMID 39919008
  15. Innes H, et al. Performance of models to predict hepatocellular carcinoma risk among UK patients with cirrhosis and cured HCV infection. JHEP Rep. 2021;3:100384. PMID 34805817
  16. Qiu L, et al. Comparison of acknowledged hepatocellular carcinoma risk scores in high-risk hepatitis C patients with sustained virological response. J Viral Hepat. 2023;30:559-566. PMID 36890735
  17. Vutien P, et al. Hepatocellular carcinoma risk decreases as time accrues following hepatitis C virus eradication. Aliment Pharmacol Ther. 2024;59:361-371. PMID 37955206
  18. John BV, et al. Liver stiffness measurement and risk prediction of hepatocellular carcinoma after HCV eradication in veterans with cirrhosis. Clin Gastroenterol Hepatol. 2024;22:778-788.e7. PMID 38061410
  19. Garay OU, et al. Cost-effectiveness of hepatocellular carcinoma surveillance strategies in patients with compensated liver cirrhosis in the United Kingdom. Value Health. 2024;27:1698-1709. PMID 39127246
  20. Campbell C, et al. Risk factors for the development of hepatocellular carcinoma in chronic HBV infection: a systematic review and meta-analysis. J Viral Hepat. 2021;28:493-507. PMID 33305479
  21. Younossi ZM, et al. Global epidemiology of nonalcoholic fatty liver disease: meta-analytic assessment of prevalence, incidence, and outcomes. Hepatology. 2016;64:73-84. PMID 26707365
  22. Xiao S, et al. Modifiable risk factors for hepatocellular carcinoma in patients with metabolic dysfunction-associated steatotic liver disease: a meta-analysis. Am J Med. 2024;137:1072-1081.e32. PMID 39047929
  23. Wang J, et al. Risk factors for hepatocellular carcinoma: an umbrella review of systematic review and meta-analysis. Ann Med. 2025;57:2455539. PMID 39834076
  24. Cao M, et al. Hepatocellular carcinoma attributable to hepatitis B, hepatitis C and other risk factors at global, regional and national levels: an updated systematic review and meta-analysis. Gut. 2026. PMID 42135055
  25. Stockdale AJ, et al. The global prevalence of hepatitis D virus infection: systematic review and meta-analysis. J Hepatol. 2020;73:523-532. PMID 32335166