Abstract
Background & Aims
Current guidelines recommend hepatocellular carcinoma (HCC) surveillance for patients with cirrhosis, though guidance on the use of non-invasive tests is lacking. Patients with compensated advanced chronic liver disease, defined non-invasively by liver stiffness measurement (LSM) ≥10 kPa, are at increased risk of hepatic decompensation and HCC. Although a fibrosis-4 (FIB-4) score ≥1.75 aligns with LSM ≥10 kPa in tertiary care, its utility for guiding HCC surveillance remains unclear. This study assessed the accuracy of FIB-4 vs. LSM in predicting HCC risk.
Methods
We retrospectively analyzed 6,143 patients with suspected chronic liver disease undergoing FIB-4 and LSM between 2007 and 2020 in Vienna. External validation was performed in participants from the UK Biobank with pre-existing liver disease. Patients were stratified by FIB-4 cut-offs (1.75 and 2.67) and corresponding LSM thresholds (10 kPa and 15 kPa).
Results
FIB-4 predicted de novo HCC with AUROCs of 0.79 (95% CI 0.65–0.94) at 1 year and 0.85 (95% CI 0.76–0.94) at 2 years, comparable to LSM. Two-year HCC incidence was 0.10% vs. 1.05% (0.05 vs. 0.53/100 person-years) for FIB-4 <1.75 vs. ≥1.75 (subdistribution hazard ratio 15.5, 95% CI 8.2–29.2), and 0.13% vs. 1.75% (0.07 vs. 0.89/100 person-years) for ≤2.67 vs. >2.67. Among UK Biobank participants with pre-existing liver disease, the 5 year cumulative incidence of HCC was 3.7% vs. 0% for FIB-4 ≥1.75 vs. <1.75 and 7.1% vs. 0.5% for >2.67 vs. ≤2.67.
Conclusions
FIB-4 (≥1.75 and >2.67) shows similar ability to predict HCC development as LSM (≥10 kPa and ≥15 kPa) in tertiary care and population-based cohorts. FIB-4 may serve as a practical tool to guide HCC surveillance.
Impact and implications
This study addresses the need for easily accessible non-invasive tools to stratify hepatocellular carcinoma risk in patients with (suspected) chronic liver disease and shows that the universally available fibrosis-4 (FIB-4) score performs comparably to liver stiffness measurement (LSM) across clinically relevant thresholds. FIB-4 ≥1.75 or LSM ≥10 kPa identify patients at risk, although thresholds of >2.67 or ≥15 kPa may be required for cost-effectiveness beyond a specialist setting. In clinical practice, physicians may use FIB-4 to guide hepatocellular carcinoma surveillance where LSM is unavailable. Researchers and policymakers can leverage these findings to develop risk-adapted surveillance strategies.
Keywords: Hepatocellular carcinoma, Non-Invasive Test, FIB-4, Compensated Advanced Chronic Liver Disease
Graphical abstract
Highlights
-
•
FIB-4 and LSM are valuable tools for predicting HCC risk in patients with suspected CLD.
-
•
FIB-4 (≥1.75/>2.67) shows similar ability to predict HCC development as LSM (≥10 kPa/≥15 kPa).
-
•
Findings were confirmed in those with pre-existing liver disease from a population-based cohort.
-
•
FIB-4 integration into clinical decision-making may enhance the implementation of HCC surveillance and increase early HCC detection.
Introduction
Hepatocellular carcinoma (HCC) is the most common primary liver malignancy and the third leading cause of cancer-related mortality worldwide, occurring predominantly in individuals with advanced fibrosis/cirrhosis (i.e. advanced chronic liver disease [ACLD]).1,2 Thus, regular surveillance for early detection of HCC is crucial, as the prognosis is significantly better when HCC is diagnosed at an early stage when curative treatments such as surgical resection, ablation, or liver transplantation are still available.3
Non-invasive tests (NITs), most importantly, liver stiffness measurement (LSM) by vibration-controlled transient elastography (VCTE), are widely used to diagnose compensated advanced chronic liver disease (cACLD) in clinical practice and provide important information on the associated HCC risk.3,4 Since the availability of VCTE is mostly restricted to tertiary care, blood-based NITs such as the fibrosis-4 (FIB-4) score have become triage tools for clinicians managing patients with suspected CLD.5,6 A FIB-4 cut-off of ≥1.75 has previously been identified to correspond to an LSM ≥10 kPa (suggestive of cACLD) in tertiary care,6 while FIB-4 values >2.67 are roughly equivalent to LSM values ≥15 kPa, which confirms cACLD.[6], [7], [8]
Current guidelines recommend regular HCC surveillance for patients with cirrhosis, unless they are at a high risk of death from non-HCC causes, but do not specify cut-offs for NITs.3 For patients with advanced fibrosis, HCC surveillance is not universally recommended due to their lower risk and concerns about the cost-effectiveness of such measures.3 Given that histology is no longer routinely used to stage fibrosis or diagnose cirrhosis, clear recommendations on NIT cut-offs to identify at-risk patients and guide decisions on HCC surveillance are urgently needed.3
Given the link between advanced fibrosis/cirrhosis and HCC risk, FIB-4 may serve as a non-invasive marker to guide HCC surveillance, ensuring timely identification of patients in whom screening is warranted, and thus, potentially improving its implementation. Our study aimed to evaluate the predictive utility of the FIB-4 vs. LSM for HCC in patients with suspected CLD.
Patients and methods
Tertiary care cohort
A retrospective analysis was conducted on 8,752 consecutive patients with suspected or confirmed CLD who underwent both FIB-4 assessment and LSM at the Medical University of Vienna between 2007 and 2020. As previously described, patients were excluded if they had a history of hepatic decompensation, HCC, or orthotopic liver transplantation.6 Additional exclusion criteria included congestive hepatopathy, incomplete clinical or laboratory data, advanced liver dysfunction (Child-Turcotte-Pugh stage B or C, who are known to have a high HCC risk), or insufficient clinical follow-up, and are detailed in.6
Demographic, clinical, and laboratory data were recorded at the time of first LSM. Patients were followed for HCC development until December 2022, with a minimum possible follow-up period of 2 years. HCC diagnosis was based on the current clinical practice recommendations of the European Association for the Study of the Liver (EASL). Mortality data were supplemented through linkage with the national death registry.
Population-based cohort
To validate our findings in a broader population that was not selected by LSM assessment, we analyzed data from the UK Biobank, compromising 502,133 individuals from the general population.9,10 Specifically, we identified three groups of individuals who are at risk for/have CLD: individuals with high metabolic risk (defined as BMI ≥30 kg/m2 or type 2 diabetes mellitus), pre-existing liver disease at baseline (defined by ICD-10 codes K70-K77), or alcohol use disorder (defined by ICD-10 code F10). Individuals were only included if they had available laboratory data for FIB-4 calculation. Notably, information on LSM was not available in this population-based cohort. Individuals with prevalent HCC or signs of decompensated cirrhosis (defined by ICD-10 codes K72.7, K76.82, I85.0x, or R18.x) were excluded from the analysis.
Incident cases of HCC were identified using a combination of general diagnosis records (data field 41270) and cancer registry data (data field 40006), based on the ICD-10 code C22.0. Date of diagnosis was retrieved from data fields 41280 and 40005; in cases of conflicting dates, the earlier date was used to ensure accurate time-to-event analyses.
Parameters
LSM by VCTE (FibroScan; Echosens, Paris, France) was performed by experienced operators adhering to established quality criteria.11 Laboratory parameters were assessed at the ISO-certified Department of Laboratory Medicine of the Medical University of Vienna. For the laboratory parameters, standard laboratory methods used in clinical routine were applied. The FIB-4 was calculated as previously described.7
Ethics
The study was approved by the ethics committee of the Medical University of Vienna (No. 1029/2023) and performed in conformity with the current version of the Helsinki Declaration. The need for informed consent was waived by the institutional ethics committee for patients who were retrospectively included. UK Biobank ethical approval and participant consent are detailed in the original cohort documentation. Data handling and analysis was conducted in accordance with UK Biobank guidelines under application number 148742.
Statistical analyses
All statistical analyses were performed using R version 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are presented as mean ± SD or median (IQR), as applicable. For comparisons between groups, we used the independent two-sample t-test for normally distributed continuous variables and the Mann-Whitney U test for non-normally distributed variables. Categorical variables were expressed as counts and percentages, and comparisons between groups were performed using the Pearson’s Chi-squared test.
Time-dependent ROC curves were generated using the ‘timeROC’ package to assess the predictive performance of FIB-4 and LSM for HCC at 1, 2, and 3 years, with the AUROC reported at each time point. AUROCs were compared using the DeLong test. Additionally, decision curve analysis based on a 2-year HCC risk horizon was conducted to evaluate the clinical utility of different FIB-4 and LSM cut-offs by calculating net benefit across a range of threshold probabilities using the ‘dcurves’ package. Log hazard ratios for HCC risk within the first 2 years were modelled using restricted cubic splines to capture non-linear effects, utilizing the ‘rms’ and ‘survival’ packages. Incidence ratios per 100 person-years, sensitivity and specificity, as well as false negative rates and false positive rates were calculated within 2 years to evaluate prognostic performance. To assess the association between FIB-4 and the risk of HCC, competing risk regression models using the ‘cmprsk’ package were applied over the full follow-up period, accounting for liver transplantation and death as a competing risk. Cumulative incidence functions were generated to illustrate the occurrence of HCC over a time period of 3 years, stratified by FIB-4 and LSM categories, and compared applying Fine and Gray’s subdistribution hazard ratios (SHRs).
In the population-based cohort, competing risk regression analyses, cumulative incidence functions, decision curve analyses, time-dependent ROC, and incidence ratios per 100 person-years were repeated. For all approaches, follow-up was censored at 5 years, reflecting a long time period before needing to repeat FIB-4-based HCC risk stratification. A two-sided p value <0.05 was considered statistically significant.
Results
Patient characteristics and outcomes
A total of 6,143 patients with suspected CLD were included in the analysis, divided into two groups based on FIB-4: those with FIB-4 <1.75 (n = 4,404; i.e. the equivalent of <10 kPa) and those with FIB-4 ≥1.75 (n = 1,739; i.e. the equivalent of ≥10 kPa).6 The characteristics of the two groups are summarized in Table 1, and differences described in the supplementary information. Median follow-up was 4.6 years, during which 91 individuals (1.5%) were diagnosed with HCC, including 20 (22.0%) and 37 (40.7%) cases within 2 and 3 years, respectively. Clinical characteristics of HCC cases are outlined in the supplementary information. Additionally, 41 (0.7%) participants underwent orthotopic liver transplantation and 526 died (8.6%).
Table 1.
Baseline characteristics of study participants in the tertiary care cohort stratified by FIB-4 cut-offs (<1.75, 1.75–2.67, and >2.67) and for the overall cohort.
| Overall cohort (N = 6,143) |
FIB-4 <1.75 (n = 4,404) |
FIB-4 1.75-2.67 (n = 841) |
FIB-4 >2.67 (n = 898) |
|
|---|---|---|---|---|
| Age (years) | 48.2 (14.2) | 44.6 (13.5) | 56.8 (11.6) | 58.0 (11.9) |
| Sex, male | 3,437 (55.9%) | 2,354 (53.5%) | 524 (62.3%) | 559 (62.2%) |
| Etiology | ||||
| Viral | 3,162 (51.5%) | 2,103 (47.8%) | 502 (59.7%) | 557 (62.0%) |
| MASLD | 1,790 (29.1%) | 1,513 (34.4%) | 164 (19.5%) | 113 (12.6%) |
| AIH/PBC/PSC | 497 (8.1%) | 370 (8.4%) | 64 (7.6%) | 63 (7.0%) |
| ALD | 352 (5.7%) | 161 (3.7%) | 73 (8.7%) | 118 (13.1%) |
| Other | 342 (5.6%) | 257 (5.8%) | 38 (4.5%) | 47 (5.2%) |
| BMI (kg/m2) | 27.4 (7.0) | 27.8 (7.6) | 26.6 (5.5) | 26.1 (5.0) |
| Diabetes | 854 (13.9%) | 510 (11.6%) | 148 (17.6%) | 196 (21.8%) |
| LSM (kPa) | 6.8 [5.1-10.7] | 6.1 [4.8-8.4] | 8.4 [6.1-13.5] | 17.1 [9.5-29.1] |
| <10 kPa | 4,419 (71.9%) | 3,679 (83.5%) | 501 (59.6%) | 239 (26.6%) |
| ≥10 kPa | 1,724 (28.1%) | 725 (16.5%) | 340 (40.4%) | 659 (73.4%) |
| FIB-4 | 1.18 [0.77-1.88] | 0.93 [0.66-1.25] | 2.08 [1.88-2.33] | 4.06 [3.19-5.82] |
| Platelets (G/L) | 221 [174-269] | 243 [206-286] | 183 [156-212] | 120 [85-158] |
| Albumin (G/L) | 44.0 (4.0) | 44.7 (3.6) | 43.3 (3.8) | 40.9 (4.4) |
| AST (U/L) | 35.0 [25.0-53.0] | 30.0 [23.0-43.0] | 46.0 [32.0-69.0] | 68.0 [43.0-108] |
| ALT (U/L) | 42.0 [27.0-72.0] | 40.0 [25.0-64.0] | 50.0 [28.0-88.0] | 60.0 [33.0-112] |
| GGT (U/L) | 51.0 [25.0-116] | 42.0 [22.0-90.0] | 66.0 [31.0-154] | 110 [55.0-226] |
Data are presented as median (IQR) for non-normally distributed continuous variables, mean (SD) for normally distributed continuous variables, and n (%) for categorical variables.
AIH, autoimmune hepatitis; ALB, albumin; ALD, alcohol-related liver disease; ALT, alanine aminotransferase; AST, aspartate aminotransferase; FIB-4, fibrosis-4; GGT, gamma-glutamyltransferase; LSM, liver stiffness measurement; MASLD, metabolic dysfunction-associated steatotic liver disease; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis.
FIB-4 and LSM as predictors of HCC
FIB-4 and LSM showed similar accuracy for the prediction of HCC within the first 3 years of follow-up. At 1 year, FIB-4 achieved an AUROC of 0.80 (95% CI 0.65–0.94). At 2 years, the AUROC increased to 0.85 (95% CI 0.76–0.94), and at 3 years, it reached 0.87 (95% CI 0.81–0.92), improving in predictive performance over time (Fig. 1). LSM was similarly able to predict HCC with an AUROC of 0.78 (95% CI 0.62–0.93) at 1 year, 0.86 (95% CI 0.76–0.96) at 2 years, and 0.85 (95% CI 0.80–0.91) at 3 years, without a significant difference between FIB-4 and LSM according to the DeLong test (all p >0.05).
Fig. 1.
Time-dependent ROC curves for FIB-4 and LSM to predict HCC at 1, 2 and 3 years in the tertiary care cohort.
The error bars represent the 95% CIs for each predictor. FIB-4, fibrosis-4; HCC, hepatocellular carcinoma; LSM, liver stiffness measurement; ROC, receiver-operating characteristic.
When modeled as continuous variables in competing risk regression, both FIB-4 and LSM were significantly associated with incident HCC (Table S1). In line, when allowing flexible modelling of the 2 year HCC risk associated with FIB-4 and LSM as continuous parameters (Fig. 2, i.e. using restricted cubic splines), the risk increased for both FIB-4 and LSM to around >2.67 and ≥15 kPa in a nearly linear manner and levelled off thereafter.
Fig. 2.
Log hazard ratios for a 2 year HCC risk in the tertiary care cohort across FIB-4 and LSM.
(A) FIB-4 and (B) LSM modelled using restricted cubic splines to capture non-linear effects. The vertical dashed lines represent predefined cut-offs. FIB-4, fibrosis-4; HCC, hepatocellular carcinoma; LSM, liver stiffness measurement.
Cumulative incidence of HCC according to FIB-4 and LSM strata
Next, patients were stratified according to established FIB-4 and LSM cut-offs. As shown in Fig. 3A and Table 2, patients with FIB-4 ≥1.75 had a ∼15-fold higher risk of developing HCC compared to those with a FIB-4 <1.75 (SHR 15.5, 95% CI 8.2–29.2), with a 2 year cumulative incidence of 0.1% vs. 1.1% and annual incidence rates of 0.05 and 0.53 per 100 person-years (Table S2). Similarly, patients with LSM ≥10 kPa had a ∼16-fold higher risk of developing HCC compared to those with LSM <10 kPa (SHR 16.0, 95% CI 8.5–30.0), with corresponding 2 year cumulative incidences of 0.1% vs. 1.1% and annual incidence rates of 0.04 and 0.56 per 100 person-years. Similar patterns were observed for higher thresholds (FIB-4 >2.67 and LSM ≥15 kPa), and further stratification into three categories for each NIT (FIB-4 <1.75, 1.75–2.67, >2.67 and LSM <10 kPa, 10–14.9 kPa, ≥15 kPa) revealed a stepwise increase in HCC risk that was comparable between both NITs (Table 2, Fig. 4, and supplementary information). Notably, within each pair of FIB-4 and LSM-based cut-offs, sensitivity and specificity were comparable, with FIB-4 ≥1.75 and LSM ≥10 kPa showing a sensitivity of 80-85%, while FIB-4 >2.67 and LSM ≥15 kPa showed a specificity of 87%. A subgroup analysis in patients with MASLD (metabolic dysfunction-associated steatotic liver disease) is provided in the supplementary information.
Fig. 3.
Cumulative incidence of HCC over a time period of 3 years in the tertiary care cohort compared between binary FIB-4 and LSM cut-offs.
(A) FIB-4 ≥1.75 and <1.75, and (B) LSM ≥10 kPa and <10 kPa. SHRs were derived from Fine & Gray competing risk regression analysis over the full follow-up period, with death and liver transplantation as competing events. FIB-4, fibrosis-4; HCC, hepatocellular carcinoma; LSM, liver stiffness measurement; SHR, subdistribution hazard ratio.
Table 2.
Risk associated with HCC across FIB-4 and LSM strata in the tertiary care cohort.
| SHR | 95% CI | p value | |
|---|---|---|---|
| FIB-4 | |||
| ≥1.75 | 15.5 | 8.2–29.2 | <0.001 |
| >2.67 | 12.2 | 7.7–19.4 | <0.001 |
| <1.75 (reference) | |||
| 1.75–2.67 | 6.2 | 2.9–13.4 | <0.001 |
| >2.67 | 23.8 | 12.5–45.2 | <0.001 |
| LSM | |||
| ≥10 kPa | 16.0 | 8.5–30 | <0.001 |
| ≥15 kPa | 11.4 | 7.3–18.1 | <0.001 |
| <10 kPa (reference) | |||
| 10–14.9 kPa | 7.1 | 3.3–15.2 | <0.001 |
| ≥15 kPa | 23.4 | 12.3–44.4 | <0.001 |
SHR derived from Fine & Gray competing risk regression analysis over the full follow-up period, with death and liver transplantation as competing events. FIB-4, fibrosis-4; LSM, liver stiffness measurement; SHR, subdistribution hazard ratio.
Fig. 4.
Cumulative incidence of HCC over a time period of 3 years in the tertiary care cohort compared between ternary FIB-4 and LSM cut-offs.
(A) FIB-4 <1.75, 1.75-2.67 and >2.67, and (B) LSM <10 kPa, 10-14.9 kPa and ≥15 kPa. FIB-4, fibrosis-4; HCC, hepatocellular carcinoma; LSM, liver stiffness measurement; SHR, subdistribution hazard ratio.
Net benefit to detect HCC based on FIB-4 and LSM cut-offs
As shown in Fig. 5, identifying the at-risk group for surveillance based on LSM (≥10 kPa or ≥15 kPa) and FIB-4 (≥1.75 or >2.67) showed a net benefit within threshold probabilities of ∼0.3% to 2% (i.e. corresponding to justified surveillance if 0.3-2.0 HCC cases are detected per 100 patients), and exceeded the “surveil none” strategy (preferred strategy below 0.3%) and the “surveil all” strategy (preferred above 2%). Supporting their comparable applicability in a tertiary care setting, a tailored at-risk group for surveillance based on FIB-4 ≥1.75 or LSM ≥10 kPa showed a comparable net benefit (5 and 7 additional HCCs per 1,000 individuals, respectively, at a 1% threshold probability), as did stratifying based on FIB-4 >2.67 or LSM ≥15 kPa (15 and 17 HCCs per 1,000, respectively). As such, the latter showed the highest net benefit in terms of increasing true-positives without increasing false-positives.
Fig. 5.
Decision curve analysis for HCC risk stratification in the tertiary care cohort for a 2 year HCC risk based on FIB-4 and LSM cut-offs.
This figure illustrates the net benefit of various strategies for risk stratification of HCC within 2 years across a range of threshold probabilities (0–3%). The light gray line represents the “surveil all” strategy, which assumes surveillance for all individuals, while the black line (horizontal at 0) corresponds to “surveil none.” FIB-4, fibrosis-4; HCC, hepatocellular carcinoma; LSM, liver stiffness measurement.
Validation of FIB-4 cut-offs in risk groups from a population-based cohort
To assess whether these findings are generalizable to lower levels of care, we evaluated the performance of FIB-4 in three at-risk groups from the UK Biobank. Baseline characteristics are shown in the supplementary information.
Within the first 5 years, 56 (0.05%) cases of HCC occurred in people with metabolic risk (n = 115,342), 6 (0.23%) in those with alcohol use disorder (AUD; n = 2,643), and 21 (1.17%) in individuals with pre-existing liver disease (n = 1,799). FIB-4 showed good discrimination with AUROCs ranging from 0.75-0.85 for HCCs in the first 5 years in the metabolic risk group, 0.74-0.82 in patients with AUD, and 0.86-0.91 in patients with pre-existing liver disease (Fig. S5).
FIB-4 cut-offs of ≥1.75 and >2.67 identified individuals at higher risk of HCC across all subgroups. A cut-off of ≥1.75 showed particularly good discrimination in patients with pre-existing liver disease (perfect separation) and in those with AUD, with a sensitivity >80%, whereas discrimination was lowest in individuals with metabolic risk (Table S4). Importantly, in the merged cohort, as well as in the metabolic risk and AUD subgroups, individuals with FIB-4 ≥1.75 or >2.67 still had a comparatively low absolute risk compared with the tertiary care cohort, likely reflecting the low background event rate. In contrast, in individuals with pre-existing liver disease recruited from the community, 5-year cumulative incidences were 0% vs. 3.7% and 0.5% vs. 7.1% for FIB-4 </≥1.75 and ≤/>2.67, which corresponds to annual incidence rates of 0% vs. 0.71% and 0.1% vs. 1.31%, thereby closely resembling observations in the tertiary care cohort.
When evaluating the net benefit of FIB-4 cut-offs to guide surveillance, a FIB-4 >2.67 yielded a net benefit of 0.000019 at a 1% threshold probability, corresponding to 1.9 additional HCC cases per 100,000 individuals allocated to surveillance. FIB-4 ≥1.75 yielded a negative net benefit at a threshold probability of 1%. Net benefit was highest in individuals with pre-existing liver disease (net benefit between 0.007 and 0.008, corresponding to 7 and 8 additional HCCs per 1,000 individuals), followed by AUD and metabolic risk with negative net benefits at a 1% threshold probability (Fig. S6).
Discussion
In this study, we evaluated the performance of FIB-4 and LSM for predicting HCC in patients with suspected CLD presenting at a tertiary care center, as well as the performance of FIB-4 in risk groups with steatotic liver disease in the general population. We confirmed the predictive performance of FIB-4 for HCC beyond complications related to portal hypertension.6 Importantly, FIB-4 showed a similar discriminative ability as LSM, and may therefore be equally applied to guide HCC surveillance.
Although LSM is the most well-established surrogate parameter for fibrosis/cirrhosis,12 it has limited availability and high associated costs, relevant considerations in more rural areas or regions of lower economic wealth. Here, FIB-4 has the advantage of being widely and easily available at low cost, making it an attractive alternative to LSM.
While FIB-4 and LSM have previously been associated with HCC development,[13], [14], [15] recommendations regarding certain thresholds for indicating HCC surveillance are missing. Specifically, EASL clinical practice guidelines do not recommend any cut-offs for the NIT-based diagnosis of ‘cirrhosis’, leaving the decision to the treating physician.3 Here, we show that FIB-4 ≥1.75, corresponding to LSM ≥10 kPa, identified a population at increased risk, with annual incidences above usual cost-effectiveness thresholds (e.g. 0.4% per year). These findings are in line with previous work.13,16,17 Specifically, a cohort focusing on patients with chronic hepatitis C showed that LSM reliably predicted HCC over 3 years, with a stepwise increase in hazard ratios for higher LSM strata.13 Furthermore, patients with MASLD from the Veterans Analysis of Liver Disease cohort with an LSM ≥10 kPa and diabetes showed annual HCC incidences above cost-effectiveness thresholds for surveillance.14 In a large multicenter cohort across various liver disease etiologies, LSM demonstrated high discriminative ability for incident HCC (C-index >0.80), with a stepwise increase in HCC incidence with LSM cut-offs ≥10 kPa, supporting its use to guide monitoring strategies.16 Another recent multicenter study recommended HCC surveillance for patients with FIB-4 ≥3.25 or LSM ≥20 kPa, or when an elevated FIB-4 in the primary setting is confirmed by LSM ≥15 kPa.17 Overall, these studies consistently support LSM-based thresholds for identifying patients at elevated HCC risk, with ≥10 kPa and ≥15 kPa being commonly suggested as sensitive cut-offs to guide HCC surveillance. FIB-4 >2.67, corresponding to LSM ≥15 kPa, stratified HCC risk in the present tertiary care setting with similar accuracy and cost-effectiveness, and could therefore guide HCC surveillance.
While LSM might be more accurate in diagnosing advanced fibrosis/cirrhosis, FIB-4 has the advantage that it incorporates age, with increasing FIB-4 values at increasing age. While this might lead to a higher number of false-positive results in terms of fibrosis screening,18,19 this seems less relevant when predicting the occurrence of hepatic decompensation or death, as both are closely related to age.6,20 This also applies to the development of HCC, which is clearly age-related,21,22 and was evident by an increased risk of HCC already in the group with FIB-4 1.75-2.67, whereas the risk in those with LSM 10–14.9 kPa was not statistically different from <10 kPa.
Importantly, all tools to inform on the need for HCC surveillance must carefully balance sensitivity and specificity at different thresholds. While lower cut-offs such as FIB-4 ≥1.75 increase sensitivity and reduce the number of HCC cases missed, they also substantially increase financial and patient burden due to screening. Conversely, higher cut-offs, such as FIB-4 >2.67, improve specificity and reduce false positives, but may compromise sensitivity, potentially missing HCC cases in patients not subjected to surveillance.
We evaluated the generalizability of our findings to the UK Biobank, a large, population-based cohort not pre-selected by level of care or LSM assessment. Here, FIB-4 showed a similar discrimination as in tertiary care and both FIB-4 ≥1.75 and >2.67 stratified HCC risk across individuals with metabolic risk, AUD, or pre-existing liver disease. These findings potentially support the utility of FIB-4 in triaging at-risk individuals in non-specialist settings, where LSM may not be available. However, decision curve analyses suggested only modest net benefit for FIB-4-based strategies. Importantly, those with pre-existing liver disease and a FIB-4 of ≥1.75 and >2.67 showed an HCC incidence above the cost-effectiveness threshold (e.g. 0.4% per year),23 validating our findings in a CLD cohort that stems from the general population rather than specialized care. Thus, in patients with CLD, a FIB-4 of >2.67 should prompt HCC screening regardless of the level of care, while the presence of risk factors and increased FIB-4 should first trigger specialist referral for a detailed assessment of CLD, which is in line with recommendations of national24 and international25 societies. Crucially, the inherently low incidence of HCC in general population cohorts raises questions about the cost-effectiveness of using FIB-4 as a stand-alone test to guide HCC surveillance. Nevertheless, the ordinal HCC risk along FIB-4 strata and similar discrimination as observed in tertiary care supports its association with HCC outside of tertiary care settings.
This study has several limitations. First, the tertiary care cohort comprised several etiologies from a real-life cohort in tertiary care, with predominantly viral etiologies. Second, the relatively low number of HCC events within the first 2 years (n = 20) and 3 years (n = 37) may have limited the confidence of our estimates in this timeframe, particularly in subgroups with lower risk. Nevertheless, predictions beyond this timeframe are less relevant for clinical practice, as updated NITs for updated risk prediction will be obtained. Third, while FIB-4 and LSM performed similarly in the tertiary care cohort, further studies are needed to validate this finding in external populations and cohorts with predominantly steatotic liver disease. Fourth, 20% and 30% of patients who developed HCC during the subsequent 2 years were not captured by the FIB-4 1.75 and 2.67 thresholds, respectively. Thus, although non-invasive markers of fibrosis/cirrhosis identify an appropriate target population, further improvements in risk prediction tools are needed, as also endorsed by EASL.26 Fifth, HCC surveillance in the tertiary care and population-based cohorts was not standardized and depended on the treating physician. Thus, we cannot rule out detection/surveillance bias. Sixth, unstandardized HCC surveillance or noncompliance with recommendations could have resulted in HCC being diagnosed only after symptoms appeared, making the data subject to interval censoring.
In conclusion, our study confirms that FIB-4 and LSM are valuable tools for predicting HCC risk in patients with suspected CLD. FIB-4 ≥1.75 equaling LSM ≥10 kPa and >2.67 equaling ≥15 kPa identify patients at progressively increased HCC risk. Given its simple calculation, FIB-4 is a broadly accessible alternative for HCC risk prediction, particularly in resource-limited settings where LSM may not be available. Its integration into clinical decision-making may enhance the implementation of HCC surveillance, and thus, increase early HCC detection. Validation in a large, population-based cohort points towards a certain utility in stratifying HCC risk outside specialist settings.
Abbreviations
(c)ACLD, (compensated) advanced chronic liver disease; AUD, alcohol use disorder; CLD, chronic liver disease; EASL, European Association for the Study of the Liver; FIB-4, fibrosis-4; HCC, hepatocellular carcinoma; LSM, liver stiffness measurement; NIT, non-invasive test; SHR, subdistribution hazard ratio; VCTE, vibration-controlled transient elastography.
Authors’ contributions
Study concept and design (G.S. and M.M.), acquisition of data (J.E., G.S., L.B., P.T., C.S., N.D., G.K., B.S.H., L.H., B.Si., B.Sc., M.P., M.T., T.R., and M.M.), analysis and interpretation of data (J.E., G.S., K.R., and M.M.), drafting of the manuscript (J.E., G.S, and M.M.) critical revision of the manuscript for important intellectual content (all authors).
Data availability
Data are available from the corresponding authors upon reasonable request and adhering to European data protection laws and local regulatory restrictions.
Declaration of AI and AI-assisted technologies in the writing process
During the preparation of this work the author(s) used ChatGPT4o to maximize readability, eloquence and grammatical correctness. After using this tool, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.
Financial support
J.E., M.J., L.B., P.T., C.S., N.D., G.K., B.S.H., L.H., B.Sim., B.Sch., M.T., T.R., G.S., and M.M. are supported by the Clinical Research Group MOTION – a project funded by the Clinical Research Groups Program of the Ludwig Boltzmann Gesellschaft (Grant Nr: LBG_KFG_22_32) with funds from the Fonds Zukunft Österreich.
Conflicts of interest
M.P. received speaker honoraria from Bayer, BMS, Eisai, Ipsen, Lilly, MSD, and Roche; he is a consultant/advisory board member for AstraZeneca, Bayer, BMS, Eisai, Ipsen, Lilly, MSD, and Roche; he received grants from AstraZeneca, Bayer, BMS, Eisai, and Roche; he received travel support from Bayer, BMS, Ipsen, and Roche. M.T. received speaker fees from Agomab, BMS, Chemomab, Falk, Gilead, Intercept, Ipsen, Jannsen, Madrigal, MSD, and Roche; he advised for AbbVie, Albireo, BiomX, Boehringer Ingelheim, Cymabay, Falk, Genfit, Gilead, Hightide, Intercept, Ipsen, Janssen, MSD, Novartis, Phenex, Pliant, Rectify, Regulus, Siemens, and Shire. He further received travel support from AbbVie, Falk, Gilead, Intercept, and Jannsen and research grants from Albireo, Alnylam, Cymabay, Falk, Gilead, Intercept, MSD, Takeda and UltraGenyx. He is also a co-inventor of patents on the medical use of norUDCA filed by the Medical Universities of Graz and Vienna. T.R. received grant support from AbbVie, Boehringer Ingelheim, Gilead, Intercept/Advanz Pharma, MSD, Myr Pharmaceuticals, Philips Healthcare, Pliant, Siemens, and W. L. Gore & Associates; speaking/writing honoraria from Abbvie, Echosens, Gilead, GSK, Intercept/Advanz Pharma, Pfizer, Roche, MSD, Siemens, and W. L. Gore & Associates; consulting/advisory board fees from AbbVie, AstraZeneca, Bayer, Boehringer Ingelheim, Gilead, Intercept/Advanz Pharma, MSD, Resolution Therapeutics, Siemens; and travel support from AbbVie, Boehringer Ingelheim, Falk, Gilead, and Roche. P.S. reports receiving grants or contracts from Arrowhead Pharmaceuticals, CSL Behring, and Grifols Inc; consulting fees from Arrowhead Pharmaceuticals, BioMarin Pharmaceutical, Dicerna Pharmaceuticals, GSK, Intellia Pharmaceuticals, NovoNordisk, BridgeBio, Takeda Pharmaceuticals, and Ono Pharmaceuticals; honoraria from Alnylam Pharmaceuticals, CSL Behring, and Grifols Inc; participating in leadership or fiduciary roles in other boards for Alpha1-Deutschland and Alpha1 Global; and material transfer support for Vertex Pharmaceuticals and Dicerna Pharmaceuticals. G.S. received travel support from Amgen. M.M. received grant support from Echosens, served as a consultant and/or advisory board member and/or speaker for AbbVie, AstraZeneca, Echosens, Eli Lilly, Falk, Gilead, Ipsen, Takeda, and W. L. Gore & Associates and received travel support from AbbVie and Gilead. The remaining authors have nothing to disclose.
Please refer to the accompanying ICMJE disclosure forms for further details.
Footnotes
Given their role as Associate Editor, Mattias Mandorfer had no involvement in the peer-review of this article and had no access to information regarding its peer-review. Full responsibility for the editorial process for this article was delegated to the Guest Editor Annalisa Berzigotti and Editor in Chief Josep M. Llovet.
Author names in bold designate shared co-first authorship
Supplementary data to this article can be found online at https://doi.org/10.1016/j.jhepr.2026.101739.
Contributor Information
Georg Semmler, Email: georg.semmler@meduniwien.ac.at.
Mattias Mandorfer, Email: mattias.mandorfer@meduniwien.ac.at.
Supplementary data
The following are the Supplementary data to this article:
References
- 1.Forner A., Reig M., Bruix J. Hepatocellular carcinoma. Lancet. 2018 Mar 31;391(10127):1301–1314. doi: 10.1016/S0140-6736(18)30010-2. [DOI] [PubMed] [Google Scholar]
- 2.Bray F., Laversanne M., Sung H., et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229–263. doi: 10.3322/caac.21834. [DOI] [PubMed] [Google Scholar]
- 3.European Association for the Study of the Liver EASL Clinical Practice Guidelines on the management of hepatocellular carcinoma. J Hepatol. 2025 Feb 1;82(2):315–374. doi: 10.1016/j.jhep.2024.08.028. [DOI] [PubMed] [Google Scholar]
- 4.Pons M., Rodríguez-Tajes S., Esteban J.I., et al. Non-invasive prediction of liver-related events in patients with HCV-associated compensated advanced chronic liver disease after oral antivirals. J Hepatol. 2020 Mar;72(3):472–480. doi: 10.1016/j.jhep.2019.10.005. [DOI] [PubMed] [Google Scholar]
- 5.Stebbing J., Farouk L., Panos G., et al. A meta-analysis of transient elastography for the detection of hepatic fibrosis. J Clin Gastroenterol. 2010 Mar;44(3):214–219. doi: 10.1097/MCG.0b013e3181b4af1f. [DOI] [PubMed] [Google Scholar]
- 6.Semmler G., Hartl L., Mendoza Y.P., et al. Simple blood tests to diagnose compensated advanced chronic liver disease and stratify the risk of clinically significant portal hypertension. Hepatology. 2024 Oct 1;80(4):887–900. doi: 10.1097/HEP.0000000000000829. [DOI] [PubMed] [Google Scholar]
- 7.Vallet-Pichard A., Mallet V., Nalpas B., et al. FIB-4: an inexpensive and accurate marker of fibrosis in HCV infection. comparison with liver biopsy and fibrotest. Hepatology. 2007 Jul;46(1):32–36. doi: 10.1002/hep.21669. [DOI] [PubMed] [Google Scholar]
- 8.Mandorfer M., Aigner E., Cejna M., et al. Austrian consensus on the diagnosis and management of portal hypertension in advanced chronic liver disease (Billroth IV) Wien Klin Wochenschr. 2023 Sep;135(Suppl 3):493–523. doi: 10.1007/s00508-023-02229-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Schneider C.V., Gross S., Balasubramani S., et al. Insulin-like growth factor-1 reflects liver disease stage and improves prediction of liver-related mortality. Clin Gastroenterol Hepatol. 2025 May;14 doi: 10.1016/j.cgh.2025.02.030. S1542-3565(25)00407-0. [DOI] [PubMed] [Google Scholar]
- 10.Schneider C.V., Fromme M., Schneider K.M., et al. Mortality in patients with genetic and environmental risk of liver disease. Am J Gastroenterol. 2021 Aug 1;116(8):1741–1745. doi: 10.14309/ajg.0000000000001326. [DOI] [PubMed] [Google Scholar]
- 11.Schwabl P., Bota S., Salzl P., et al. New reliability criteria for transient elastography increase the number of accurate measurements for screening of cirrhosis and portal hypertension. Liver Int. 2015 Feb;35(2):381–390. doi: 10.1111/liv.12623. [DOI] [PubMed] [Google Scholar]
- 12.Mózes F.E., Lee J.A., Selvaraj E.A., et al. Diagnostic accuracy of non-invasive tests for advanced fibrosis in patients with NAFLD: an individual patient data meta-analysis. Gut. 2022;71:1006–1019. doi: 10.1136/gutjnl-2021-324243. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Masuzaki R., Tateishi R., Yoshida H., et al. Prospective risk assessment for hepatocellular carcinoma development in patients with chronic hepatitis C by transient elastography. Hepatology. 2009 Jun;49(6):1954–1961. doi: 10.1002/hep.22870. [DOI] [PubMed] [Google Scholar]
- 14.John B.V., Bastaich D.R., Deng Y., et al. Use of liver stiffness measurement for HCC risk stratification in metabolic dysfunction-associated steatotic liver disease. Hepatology. 2025 Aug 14 doi: 10.1097/HEP.0000000000001498. Online ahead of print. [DOI] [PubMed] [Google Scholar]
- 15.Siriwong N., Sriphoosanaphan S., Decharatanachart P., 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 of aggregate and individual patient data. Eur J Gastroenterol Hepatol. 2025 Mar 1;37(3):358–369. doi: 10.1097/MEG.0000000000002912. [DOI] [PubMed] [Google Scholar]
- 16.Hegmar H., Wester A., Aleman S., et al. Liver stiffness predicts progression to liver-related events in patients with chronic liver disease - a cohort study of 14 414 patients. Liver Int. 2024 Jul;44(7):1689–1699. doi: 10.1111/liv.15919. [DOI] [PubMed] [Google Scholar]
- 17.Lai J.C.T., Yang B., Lee H.W., et al. Non-invasive risk-based surveillance of hepatocellular carcinoma in patients with metabolic dysfunction-associated steatotic liver disease. Gut. Nov 10 2025;74(12):2050–2057. doi: 10.1136/gutjnl-2025-334981. [DOI] [PubMed] [Google Scholar]
- 18.Sung S., Al-Karaghouli M., Tam M., et al. Age-dependent differences in FIB-4 predictions of fibrosis in patients with MASLD referred from primary care. Hepatol Commun. 2024 Dec 11;9(1) doi: 10.1097/HC9.0000000000000609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Graupera I., Thiele M., Serra-Burriel M., et al. Low accuracy of FIB-4 and NAFLD fibrosis scores for screening for liver fibrosis in the population. Clin Gastroenterol Hepatol. 2022 Nov;20(11):2567–2576.e6. doi: 10.1016/j.cgh.2021.12.034. [DOI] [PubMed] [Google Scholar]
- 20.Anstee Q.M., Berentzen T.L., Nitze L.M., et al. Prognostic utility of Fibrosis-4 Index for risk of subsequent liver and cardiovascular events, and all-cause mortality in individuals with obesity and/or type 2 diabetes: a longitudinal cohort study. Lancet Reg Health Eur. 2024 Jan;36 doi: 10.1016/j.lanepe.2023.100780. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Semmler G., Meyer E.L., Kozbial K., et al. HCC risk stratification after cure of hepatitis C in patients with compensated advanced chronic liver disease. J Hepatol. 2022 Apr;76(4):812–821. doi: 10.1016/j.jhep.2021.11.025. [DOI] [PubMed] [Google Scholar]
- 22.Petrick J.L., Florio A.A., Znaor A., et al. International trends in hepatocellular carcinoma incidence, 1978-2012. Int J Cancer. 2020 Jul 15;147(2):317–330. doi: 10.1002/ijc.32723. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Parikh N.D., Singal A.G., Hutton D.W., et al. Cost-effectiveness of hepatocellular carcinoma surveillance: an assessment of benefits and harms. Am J Gastroenterol. 2020 Oct;115(10):1642–1649. doi: 10.14309/ajg.0000000000000715. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Mandorfer M., Semmler G., Aigner E., et al. Austrian multisociety consensus on metabolic dysfunction-associated steatotic liver disease: Austrian Society of Gastroenterology and Hepatology (ÖGGH), Austrian Society of Diabetology (ÖDG), Austrian Society of Obesity (ÖAG) Wien Klin Wochenschr. 2025 Oct 17;137(Suppl 10):307–319. doi: 10.1007/s00508-025-02617-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.European Association for the Study of the Liver EASL–EASD–EASO Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD) J Hepatol. 2024 Sep 1;81(3):492–542. doi: 10.1016/j.jhep.2024.04.031. [DOI] [PubMed] [Google Scholar]
- 26.EASL Policy Statement: risk-based surveillance for hepatocellular carcinoma among patients with cirrhosis [Internet]. EASL - Home Hepatol. [cited 2025 Apr 24]. Available from: https://easl.eu/publication/easl-policy-statement-risk-based/.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data are available from the corresponding authors upon reasonable request and adhering to European data protection laws and local regulatory restrictions.