Liver stiffness at baseline as a marker of hepatocellular carcinoma in cirrhosis: A matched analysis

Abstract

BACKGROUND

Early hepatocellular carcinoma (HCC) detection in cirrhosis remains suboptimal despite semi-annual ultrasound and alpha-fetoprotein surveillance. Advanced fibrosis is a central driver of hepatocarcinogenesis, and liver stiffness measurement (LSM) is a non-invasive measure of liver stiffness that may be associated with HCC risk.

AIM

To evaluate whether LSM by transient elastography (TE) is associated with the presence of HCC among patients with established cirrhosis.

METHODS

A retrospective, matched case-control study at a tertiary liver center including adults with cirrhosis and a valid TE (≥ 10 valid acquisitions, interquartile range < 30%, success rate > 60%) was conducted. Cirrhotic patients with diagnosed HCC per standard guidelines were frequency matched for age, sex, etiology, and Child-Pugh class with controls (cirrhotic patients without HCC). Multivariable logistic regression tested the association between LSM and HCC. Discrimination and cut-off points were assessed using receiver operating characteristic (ROC) analysis and Youden’s index.

RESULTS

A total of 262 patients (133 with HCC; 129 controls) were enrolled. Median LSM was higher in the HCC cohort than in controls (31.7 kPa vs 22.6 kPa; P < 0.001). Multivariate regression analysis revealed that only LSM was significantly associated with HCC (adjusted odds ratio 1.09; 95%CI: 1.05-1.13; P = 0.0001). A cut-off of ≥ 47 kPa had excellent discriminatory power (area under the ROC curve: 0.88; 95%CI: 0.84-0.93).

CONCLUSION

A liver stiffness threshold of approximately 47 kPa may serve as a marker associated with risk of HCC, justifying intensified screening in high-risk cirrhosis; prospective validation, integration with multivariable risk models, and cost-effectiveness analyses remain essential.

Key Words: Hepatocellular carcinoma; Liver cirrhosis; Liver fibrosis; Liver stiffness; Elastography; Liver stiffness measurement; Hepatocellular carcinoma; Cirrhosis; Risk stratification; Portal hypertension

Core Tip: Hepatocellular carcinoma (HCC) surveillance in cirrhosis remains imperfect, with a substantial proportion of tumors detected at advanced stages despite guideline-recommended ultrasound-based screening. In this matched case–control study, baseline liver stiffness measurement (LSM) assessed by transient elastography independently stratified HCC risk among patients with established cirrhosis. Higher LSM values were strongly associated with HCC, with excellent discriminative performance and a clinically relevant threshold identifying patients at particularly high risk. These findings support the integration of LSM into HCC risk stratification frameworks and suggest that liver stiffness–guided, individualized surveillance strategies may improve early detection in cirrhosis.

INTRODUCTION

Hepatocellular carcinoma (HCC) is a major global health burden, with approximately 747000 new cases reported annually and a global incidence of approximately 9.3 per 100000 population[1,2]. In India, registry-based data indicate a lower but rising burden, with an incidence of approximately 2.1-3.0 per 100000 and prevalence of approximately 2.27 per 100000[3]. HCC most commonly arises in the setting of cirrhosis, which represents the final common pathway of chronic liver injury irrespective of etiology. The principal risk factor for HCC is cirrhosis, irrespective of etiology, with underlying causes such as chronic viral hepatitis, alcohol-related liver disease, and metabolic dysfunction-associated steatotic liver disease (MASLD) acting as upstream drivers[1,4]. Notably, most of these factors are potentially preventable, underscoring the importance of disease control. Effective surveillance in at-risk populations (particularly those with cirrhosis) can detect HCC at early, curative stages and is therefore widely recommended. Current guidelines advocate semi-annual or bi-annual (6-month) surveillance using abdominal ultrasound with or without serum alpha-fetoprotein (AFP). This ultrasound-based strategy is non-invasive and cost-effective, but its performance is suboptimal[5]. A recent meta-analysis found that ultrasound alone detects only about 45% of early-stage HCC, rising to 63% sensitivity with the addition of AFP[6,7]. In other words, the standard surveillance regimen misses over one-third of early HCC cases[8]. These limitations, exacerbated in certain populations (e.g., patients with obesity or nonviral liver disease, in whom ultrasound visualization is often poor), have prompted interest in adjunctive tools to improve HCC surveillance outcomes[9]. There is a clear need for better biomarkers or risk stratification methods to identify the cirrhotic patients at highest risk for HCC that could benefit from enhanced surveillance or early intervention.

One promising approach is to leverage non-invasive fibrosis assessment as a surrogate marker of HCC risk. The degree of hepatic fibrosis has recently been explored as a predictor of HCC development, and cirrhosis (advanced fibrosis) confers the greatest risk of HCC among chronic liver disease patients[10,11]. Liver stiffness measurement (LSM) by vibration-controlled transient elastography (TE) has emerged as a convenient, quantitative measure of fibrosis and portal hypertension[12]. LSM is widely used in clinical practice to noninvasively stage fibrosis and has proven useful in the prognostication of liver-related outcomes. Because higher liver stiffness reflects greater fibrotic burden, it is biologically plausible that elevated LSM in cirrhotic patients could correlate with an increased likelihood of HCC. Moreover, TE can be performed at the bedside or in the clinic, providing an immediate fibrosis assessment that could be integrated into HCC risk stratification models.

Despite growing evidence, LSM has not yet been incorporated into standard HCC surveillance guidelines, and no definitive liver stiffness cut-off for “high HCC risk” in cirrhosis has been established. Prior studies have been heterogeneous in patient population, etiologies, and LSM thresholds evaluated, and direct comparisons of LSM between those who develop HCC and those who do not are limited. In this context, further research is needed to clarify how LSM might be used to determine whether baseline LSM differs between cirrhotic patients with and without HCC and whether it is associated with the presence of HCC. The present study was designed to address this knowledge gap by evaluating LSMs in a case-control cohort of cirrhotic patients with and without HCC. We aimed to determine whether baseline LSM differs between patients with cirrhosis who develop HCC and those who remain HCC-free, and to identify an optimal LSM cut-off value that signals an increased risk of HCC in patients with cirrhosis. This information could lay the groundwork for integrating LSM into future HCC risk prediction models or surveillance algorithms, ultimately improving the early detection of HCC in high-risk patients.

MATERIALS AND METHODS

Study design and setting

This was a retrospective, matched case-control study at an academic tertiary liver care institution in northern India. The primary objective was to evaluate whether LSM is associated with the presence of HCC in patients with cirrhosis from January, 2024 to October, 2025. The study protocol was approved by the Institutional Ethics Committee and adhered to the STROBE guidelines for observational studies.

Participants

Patients were eligible for inclusion if they were at least 18 years of age with a confirmed diagnosis of cirrhosis, irrespective of etiology. Cirrhosis was diagnosed based on at least two concordant findings among clinical stigmata (such as splenomegaly, ascites, or esophageal varices), characteristic imaging features (e.g., nodular liver surface or caudate lobe hypertrophy on ultrasound, computed tomography (CT), or magnetic resonance imaging (MRI), or laboratory derangements suggestive of advanced liver disease, including thrombocytopenia, hypoalbuminemia, or prolonged prothrombin time. Histological confirmation of cirrhosis, where available, further substantiated the diagnosis. All participants were required to have undergone LSM via TE (FibroScan^® 502 Touch; Echosens, Paris, France) prior to study enrollment. A valid LSM was defined as one with at least ten successful measurements, a success rate > 60%, and an interquartile range (IQR) < 30% of the median stiffness value, ensuring data reliability. Only patients whose LSM was performed during a compensated or clinically stable phase, defined as absence of active decompensation (ascites, variceal bleeding, or hepatic encephalopathy) at the time of measurement were included, although some patients had a history of prior decompensation. This approach was adopted to minimize non-fibrotic influences on liver stiffness, including portal hypertension and acute clinical events

For the case group, additional eligibility criteria included an HCC diagnosis established by dynamic imaging (contrast-enhanced CT or MRI) demonstrating arterial phase hyperenhancement followed by venous or delayed phase washout, consistent with American Association for the Study of Liver Disease or European Association for the Study of the Liver guidelines. To ensure that LSM values represented true baseline hepatic stiffness rather than stiffness altered by tumor burden or vascular invasion, only patients whose LSM was performed at least 12 months prior to radiological diagnosis of HCC were included. This threshold was selected based on known HCC tumor growth kinetics, with a median tumor volume doubling time of approximately 4-5 months, such that a 12-month interval corresponds to multiple tumor doubling cycles and allows sufficient time for radiologically detectable tumor development[13]. This also allowed us to preserve temporal sequencing between exposure and outcome. For the control group, eligible participants were cirrhotic patients without a history or imaging evidence of HCC at the time of data collection. Controls were frequency-matched to cases based on age (± 5 years), sex, etiology of liver disease, and Child-Turcott Pugh (CTP) class. Controls were required to have undergone abdominal imaging (ultrasound, CT, or MRI) within 6 months of data collection to confirm the absence of HCC.

Patients were excluded if they had incomplete, technically invalid, or unreliable LSM data, defined as fewer than ten valid readings, a success rate < 60%, or an IQR > 30% of the median stiffness value. Individuals with HCC diagnosed prior to or within 12 months of the LSM date were excluded to eliminate reverse causality and tumor-associated stiffness as confounders. Patients with poor-quality or unavailable imaging were also excluded if the presence or absence of HCC could not be verified. To avoid bias from treatment-modified liver stiffness, individuals who had previously undergone hepatic resection, liver transplantation, or locoregional therapy, such as trans-arterial chemoembolization or radiofrequency ablation, before the LSM date were excluded. Patients with concurrent malignancies other than HCC were also excluded to preserve disease-specificity of the outcome.

Further exclusions included individuals with acute liver failure, acute-on-chronic liver failure, or those requiring intensive care unit admission at the time of LSM, as such states could transiently affect liver stiffness unrelated to fibrosis burden. Patients with significant hepatic venous outflow obstruction (e.g., Budd-Chiari syndrome), cardiac-induced liver congestion, or obstructive biliary pathology were excluded due to potential confounding influences on elastographic values. In addition, patients in whom the use of an inappropriate probe (e.g., M probe used instead of XL probe in obese individuals) or morbid obesity [body mass index (BMI) > 35 kg/m²] led to LSM acquisition failure were omitted from analysis. Finally, individuals in the control group were excluded if they had < 6 months of follow-up without imaging confirmation of HCC-free status, as this could lead to misclassification of undiagnosed HCC cases.

Variables and data sources

The primary exposure variable was liver stiffness measured in kPa using TE (FibroScan^® 502 Touch). LSM values were obtained from electronic medical records. A valid LSM was defined as the median of ≥ 10 successful measurements with an IQR < 30% and success rate > 60%. All examinations were performed by trained operators, with patients fasting for at least 3 hours and positioned supine with right arm abducted. TE quality parameters, including number of valid measurements, success rate, and IQR, were recorded and used to define valid LSM values. However, probe type (M vs XL) and specific reasons for measurement exclusion were not consistently documented in the retrospective dataset, limiting the ability to verify probe-related exclusions.

Data were retrospectively extracted from electronic medical records using a standardized collection framework. Clinical, laboratory, imaging and elastography variables were obtained from institutional databases, with predefined definitions to ensure consistency. Other variables collected included demographics (age, sex), cirrhosis etiology (viral, alcoholic, nonalcoholic fatty liver disease, autoimmune, cryptogenic), BMI, comorbidities (including diabetes mellitus), and liver disease severity scores [Child-Pugh (CP), model for end-stage liver disease (MELD)]. Laboratory parameters closest to the LSM date were extracted, including serum albumin, bilirubin, international normalized ratio (INR), alanine aminotransferase (ALT), aspartate aminotransferase, platelet count, and AFP. CP classification used for matching reflected baseline clinical staging at the time of inclusion; however, clinical severity parameters may have evolved during follow-up.

Bias and matching

To minimize confounding and selection bias, cases and controls were matched for major confounders, including age, sex, cirrhosis etiology, and CP classification. To reduce information bias, all imaging and elastography data were independently verified by two authors blinded to case/control status. Missing data were minimal (< 5% for all variables) and were handled using listwise deletion.

Study size

Sample size was based on available data over the study period. All eligible HCC cases (n = 133) with prior valid LSM were included. Controls (n = 129) were matched in a near 1:1 ratio, yielding adequate power for logistic regression and receiver operating characteristic (ROC) analysis. A formal power calculation was not performed a prior due to the retrospective nature of the study.

Statistical analysis

Continuous variables are presented as mean ± SD or median (IQR), as appropriate, and categorical variables as n (%). Comparisons between groups were performed using the Student’s t-test or Mann-Whitney U test for continuous variables and the χ² or Fisher’s exact test for categorical variables.

Cases and controls were frequency matched on age, sex, etiology, and CTP class. Accordingly, unconditional logistic regression was used to evaluate factors associated with HCC. Univariable logistic regression was initially performed, and variables with P < 0.10 were considered for inclusion in the multivariable model.

To avoid collinearity, the MELD score was not included in the same multivariable model as its component laboratory variables (including INR and bilirubin). The final multivariable model therefore retained individual laboratory variables rather than MELD to allow clearer interpretation of independent associations. Matching variables were accounted for in the modeling strategy to preserve the validity of the frequency-matched design.

Odds ratios (ORs) with 95%CIs were reported. Discriminatory performance of LSM was assessed using ROC analysis, and the area under the ROC curve (AUROC) was calculated. A two-sided P value < 0.05 was considered statistically significant. All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp., Armonk, NY, United States). A two-tailed P < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics

A total of 262 cirrhotic patients were included in the final analysis, comprising 133 patients with HCC (cases) and 129 patients without HCC (controls). The two groups were comparable in age (mean age: 51 ± 11 years in HCC group vs 52 ± 9 years in controls) and sex distribution (male: Female ratio 96:37 vs 92:37, respectively). The mean duration since cirrhosis diagnosis was 5.2 ± 3.6 years in the HCC group compared to 5.8 ± 3.2 years in the control group. The median duration from baseline LSM to HCC diagnosis in cases was 6.7 years (IQR: 5.4-8.3), while in controls, the median duration from LSM to last imaging confirming absence of HCC was 6.4 years (IQR: 5.3-9.2), indicating comparable at-risk follow-up between groups. The underlying etiologies of cirrhosis were similar between groups, with alcohol-related liver disease, MASLD, and metabolic dysfunction and alcohol-associated liver disease comprising the majority of cases. The proportion of patients with hepatitis B virus (HBV) and hepatitis C virus (HCV)-related cirrhosis was non-significantly higher in the HCC group. Liver disease severity, as assessed by CP classification, was also balanced between groups (CP-A: 57.9% in HCC vs 53.5% in controls; CP-B: 23.3% vs 27.9%; CP-C: 18.8% vs 18.6%) depicted in Table 1.

Table 1 Baseline characteristic of patients, n (%)/mean ± SD/median (interquartile range).

Characteristics	CLD with HCC, n = 133	CLD without HCC, n = 129	P value
Age (years)	51 ± 11	52 ± 9	0.19
Sex (male:female)	96:37	92:37	0.99
Duration of cirrhosis (years)	5.2 ± 3.6	5.8 ± 3.2	0.15
Etiology
Alcohol	42 (31.6)	39 (30.2)	0.81
MASLD	41 (30.8)	37 (28.7)	0.72
MetALD	19 (14.3)	19 (14.7)	0.93
HBV	7 (5.3)	9 (7.0)	0.56
HCV	18 (13.5)	23 (17.8)	0.34
Others	6 (4.5)	2 (1.6)	0.16
Severity
CP-A	77 (57.9)	69 (53.5)	0.77
CP-B	31 (23.3)	36 (27.9)
CP-C	25 (18.8)	24 (18.6)
History of decompensation during disease course
Ascites	27 (20.3)	29 (22.5)	0.78
Jaundice	29 (21.8)	33 (25.6)	0.57
Variceal bleed	33 (24.8)	29 (22.5)	0.77
Esophageal	21 (15.8)	19 (14.7)	0.80
Gastric	12 (9.0)	8 (6.2)	0.39
Ectopic	0	2 (1.6)	0.242
Hepatic encephalopathy	19 (14.3)	29 (22.6)	0.12
Presentation at diagnosis of cirrhosis
Jaundice	24 (18.0)	27 (20.9)	0.56
Bleed	33 (24.8)	24 (18.6)	0.24
Hepatic encephalopathy	4 (3.0)	7 (5.4)	0.502
Incidental	56 (42.1)	49 (38.0)	0.51
Lab parameters
Hemoglobin (g/dL)	11.4 ± 2.2	11.3 ± 1.4	0.66
Platelet count (103/µL)	116.9 ± 56.1	112.2 ± 52.5	0.23
Bilirubin (mg/dL)1	1.27 (0.9-2.6)	2.4 (1.0-3.8)	0.06
Albumin (g/dL)	3.8 ± 0.6	3.7 ± 0.6	0.18
Creatinine (mg/dL)	0.8 ± 0.4	0.9 ± 0.3	0.12
INR	1.9 ± 0.3	1.8 ± 0.4	0.23
Alanine transaminase (IU/L)	69.6 ± 34.9	53.7 ± 19.3	0.06
Aspartate transaminase (IU/L)	60.1 ± 29.1	46.9 ± 29.1	0.01

¹Bilirubin compared using Mann-Whitney U test. Categorical variables compared using χ² test.

²Fisher’s exact test used where expected cell counts < 5.

Decompensation variables reflect prior or cumulative events and do not represent status at the time of LSM. CLD: Chronic liver disease; CP: Child-Pugh; HBV: Hepatitis B virus; HCV: Hepatitis C virus; INR: International normalized ratio; MASLD: Metabolic associated steatotic liver diseases; MetALD: Metabolic dysfunction and alcohol-associated liver disease; HCC: Hepatocellular carcinoma.

Laboratory parameters demonstrated slightly higher median serum bilirubin levels in patients without HCC [2.4 (IQR: 1.0-3.8) mg/dL vs 1.27 (IQR: 0.9-2.6) mg/dL; P < 0.01], and a modestly elevated INR (1.8 ± 0.4 vs 1.9 ± 0.3). Platelet counts were similar in both cohorts (116.9 ± 56.1 × 10³/μL vs 112.2 ± 52.5 × 10³/μL; P = 0.03). Serum albumin and creatinine levels were comparable between groups.

HCC characteristics and stage at diagnosis

Among patients with HCC, 37 (28.7%) were diagnosed incidentally during surveillance imaging, while the remaining 92 (71.3%) presented with symptoms such as jaundice (n = 42), ascites (n = 14), hepatic encephalopathy (n = 8), or gastrointestinal bleeding (n = 4). Barcelona-Clinic Liver Cancer (BCLC) staging at diagnosis revealed that only 19 (14.7%) patients had very early-stage (BCLC-0) disease, 21 (16.3%) had early-stage (BCLC-A), 29 (22.4%) had intermediate-stage (BCLC-B), 23 (17.8%) had advanced-stage (BCLC-C), and 37 (28.7%) had end-stage disease (BCLC-D). Thus, only one-third of patients were diagnosed at potentially curative stages, underscoring limitations in current surveillance strategies (Table 2).

Table 2 Baseline characteristics of hepatocellular carcinoma patients, n (%)/median (interquartile range).

Characteristics	Patients with HCC, n = 133
Duration	2.6 (1.2-3.0)
Presentation at time of diagnosis of HCC
Screening/incidental	37 (28.7)
Jaundice	42 (31.6)
Hepatic encephalopathy	8 (6.1)
Bleed	4 (3.1)
Ascites	14 (10.5)
BCLC grading (stage)
0	19 (14.7)
A	21 (16.3)
B	29 (17.8)
C	23 (17.8)
D	37 (28.7)
Treatment received
Resection	6 (4.5)
Liver transplant	7 (5.4)
Microwave ablation	18 (13.5)
SBRT	11 (8.3)
TACE	19 (14.3)
Tyrosine kinase inhibitors	34 (25.6)
Immunotherapy	5 (3.8)
Palliative therapy	29 (21.8)

BCLC: Barcelona-Clinic Liver Cancer; HCC: Hepatocellular carcinoma; SBRT: Stereotactic body radiation therapy; TACE: Transarterial chemoembolization.

Liver stiffness and discriminatory performance

Median liver stiffness values were significantly higher among patients with HCC compared to those without HCC [31.7 (IQR: 25.9-36.2) kPa vs 22.6 (IQR: 18.4-28.1) kPa, P < 0.001]. On univariate logistic regression, elevated LSM, higher INR, elevated serum bilirubin, and lower platelet count were associated with increased odds of HCC. Specifically, each 1 kPa increase in LSM was associated with a 12% increase in the odds of HCC (unadjusted OR: 1.12; 95%CI: 1.08-1.17; P = 0.0001). Other significant univariate predictors included INR (OR: 1.75; 95%CI: 1.10-2.79; P = 0.02) and serum bilirubin (OR: 1.35; 95%CI: 1.10-1.67; P = 0.004).

In multivariate analysis, LSM remained an independent and robust predictor of HCC (adjusted OR: 1.09 per 1 kPa increase; 95%CI: 1.05-1.13; P = 0.0001). INR (adjusted OR: 1.6; 95%CI: 0.95-2.57; P = 0.06) and serum bilirubin (adjusted OR: 1.28; 95%CI: 0.93-1.60; P = 0.08) did not retain statistical significance after adjustment. Platelet count and MELD score trended toward significance, while AFP levels were not independently associated with HCC in the multivariate model.

ROC analysis demonstrated excellent discriminatory ability of LSM for predicting HCC, with area under the curve of 0.88 (95%CI: 0.84-0.93). The optimal cut-off derived from Youden’s index was ≥ 47.0 kPa, yielding a sensitivity of 95%, specificity of 76%, predictive value (PPV) of 57.6%, negative predictive value (NPV) of 97.8%, and an overall accuracy of 81.1% (Figure 1).

Figure 1 Receiver operating characteristic curve demonstrating the discriminatory performance of liver stiffness measurement for identifying hepatocellular carcinoma. The area under the receiver operating characteristic curve is shown along with corresponding sensitivity and specificity at selected thresholds. Prevalence-dependent measures are not emphasized given the case–control design. Area under the curve = 0.88 (95%CI: 0.84-0.93). The optimal threshold by Youden’s index was 47 kPa. The diagonal line denotes the no-discrimination reference. ROC: Receiver operating characteristic; AUC: Area under the curve.

A progressive improvement in specificity and PPV was observed with increasing LSM thresholds. For example, a threshold of 47 kPa offered a sensitivity of 95.0% (95%CI: 83.5%-98.6%) and a specificity of 76.5% (95%CI: 68.1%-83.2%), while a threshold of 75 kPa achieved perfect specificity (100%) and PPV (100%) but very low sensitivity (7.5%). These data suggest that while higher stiffness values confer greater predictive certainty for HCC, lower cut-offs offer better screening utility. Detailed performance metrics for various LSM thresholds are presented in Table 3.

Table 3 Various liver stiffness measurement cut-off values for predicting risk of hepatocellular carcinoma in patients with cirrhosis with sensitivity, specificity, positive predictive value and negative predictive value, 95%CI.

LSM cut-off in kPa	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Accuracy (%)
30	100.0 (91.2-100.0)	19.3 (13.2-27.3)	29.4 (22.4-37.6)	100.0 (85.7-100.0)	39.6 (32.4-47.4)
35	97.5 (87.1-99.6)	46.2 (37.5-55.2)	37.9 (29.1-47.5)	98.2 (90.6-99.7)	59.1 (51.4-66.5)
40	97.5 (87.1-99.6)	52.1 (43.2-60.9)	40.6 (31.3–50.6)	98.4 (91.5–99.7)	63.5 (55.8-70.6)
45	95.0 (83.5-98.6)	67.2 (58.4-75.0)	49.4 (38.5-60.3)	97.6 (91.5-99.3)	74.2 (66.9-80.4)
47	95.0 (83.5-98.6)	76.5 (68.1-83.2)	57.6 (45.6-68.8)	97.8 (92.5-99.4)	81.1 (74.3-86.5)
50	77.5 (62.5-87.7)	82.4 (74.5-88.2)	59.6 (46.1-71.8)	91.6 (84.8-95.5)	81.1 (74.3-86.5)
55	57.5 (42.2-71.5)	86.6 (79.3-91.6)	59.0 (43.4-72.9)	85.8 (78.5-91.0)	79.2 (72.3-84.8)
60	42.5 (28.5-57.8)	91.6 (85.2-95.4)	63.0 (44.2-78.5)	82.6 (75.2-88.1)	79.2 (72.3-84.8)
65	32.5 (20.1-48.0)	95.8 (90.5-98.2)	72.2 (49.1-87.5)	80.9 (73.6-86.5)	79.9 (73.0-85.4)
70	20.0 (10.5-34.8)	97.5 (92.8-99.1)	72.7 (43.4-90.3)	78.4 (71.1-84.2)	78.0 (70.9-83.7)
75	7.5 (2.6-19.9)	100.0 (96.9-100.0)	100.0 (43.9-100.0)	76.3 (69.0-82.3)	76.7 (69.6-82.6)

LSM: Liver stiffness measurement; NPV: Negative predictive value; PPV: Predictive value.

DISCUSSION

The current case-control study of cirrhotic patients found that higher baseline LSM by TE was strongly associated with the presence of HCC. Cirrhotic patients who developed HCC had significantly greater LSM values at baseline compared to matched cirrhotic controls without HCC. On multivariate logistic regression adjusting for other risk factors, LSM was independently associated with future risk of HCC with an adjusted OR of approximately 1.09 per 1 kPa increase (95%CI: 1.05-1.13, P < 0.001) (Table 4). It had excellent discriminatory performance with an area under the curve of about 0.88 (95%CI: 0.84-0.93). Furthermore, we identified an optimal LSM threshold of 47 kPa by Youden’s index, which yielded a sensitivity of 95% and specificity of 76%, with about 80% overall diagnostic accuracy. Interestingly, bilirubin levels were slightly higher in the non-HCC group. This may reflect greater underlying hepatic dysfunction or decompensation in a subset of controls, whereas HCC cases may have been detected under closer surveillance or at relatively preserved liver function. This observation should be interpreted cautiously given the heterogeneity of the cohort. However, traditional HCC predictors like serum AFP did not remain significant in our multivariable model, whereas LSM retained a robust association with HCC risk. These findings suggest that higher LSM reflects a phenotype associated with increased likelihood of HCC, postulating that HCC surveillance in cirrhotics can be further stratified on the basis of liver stiffness values.

Table 4 Univariate and multivariate logistic regression analysis for various variables for hepatocellular carcinoma risk.

Variable	Univariate logistic regression		Multivariate logistic regression
	Unadjusted OR (95%CI)	P value	Adjusted OR (95%CI)	P value
LSM per 1 kPa increase	1.12 (1.08-1.17)	0.0001	1.09 (1.05-1.13)	0.0001
INR	1.75 (1.10-2.79)	0.02	1.6 (0.95-2.57)	0.06
Platelet count per 10³/μL	0.98 (0.96-1.00)	0.06	0.99 (0.97-1.01)	0.09
Serum bilirubin (mg/dL)	1.35 (1.10-1.67)	0.004	1.28 (0.93-1.60)	0.08
AFP (ng/mL)	1.01 (0.99-1.02)	0.09	1.0 (0.98-1.02)	0.42

AFP: Alpha-fetoprotein; INR: International normalized ratio; LSM: Liver stiffness measurement; MELD: Model for end-stage liver disease; OR: Odds ratio.

Few studies have evaluated the risk stratification of carcinogenesis in patients with cirrhosis. Masuzaki et al[14] first reported that baseline liver stiffness strongly stratified HCC incidence in a prospective cohort of hepatitis C patients. Patients with LSM > 25 kPa had a 45-fold higher hazard of developing HCC compared to those with LSM ≤ 10 kPa. Similarly, another study in chronic hepatitis B observed that HCC incidence progressively increased with higher LSM; HBV patients with LSM > 23 kPa had about a 6.6-fold greater hazard of HCC vs those with LSM < 8 kPa[15]. While these prospective studies support a relationship between increasing liver stiffness and incident HCC, our case–control analysis demonstrates that higher LSM is strongly associated with the presence of HCC across diverse etiologies of cirrhosis. In our study, each 1 kPa increase in LSM was associated with an approximately 10% higher odds of HCC, consistent with prior meta-analytic estimates (relative risk 1.11 per 1 kPa increase)[16]. Taken together, these findings support the concept that higher liver stiffness reflects more advanced underlying liver disease and is associated with HCC; however, our design does not permit inference regarding prediction of incident HCC, and these observations should be interpreted as hypothesis-generating pending prospective validation. However, the optimal LSM cut-off (≥ 47 kPa) identified in this study is higher than thresholds commonly reported in the literature. This likely reflects that, in established cirrhosis, very high liver stiffness values may indicate more advanced architectural distortion and greater portal hypertensive burden rather than fibrosis alone. As this threshold was derived using statistical optimization within this cohort, it should be interpreted cautiously and requires prospective validation before clinical application.

Moreover, the prognostic value of LSM may vary across liver diseases and patient factors. Lai et al[17] emphasized that non-invasive fibrosis tests such as TE cannot be applied using a single universal scale, noting that “a common cut-off value does not imply the same degree of fibrosis across all disease settings.” For instance, an LSM of 20 kPa in a patient with compensated viral hepatitis may not indicate the same fibrosis severity as 20 kPa in MASLD, underscoring why different studies adopt distinct thresholds.

Huang et al[18] further highlighted that interpreting LSM requires clinical context, particularly regarding hepatic inflammation and viral activity. In their analysis of patients with HBV-related HCC, LSM correlated well with liver function only when ALT levels were normal; this relationship disappeared when ALT exceeded 40 U/L. They also observed differing optimal LSM thresholds for early cirrhosis-approximately 9.3 kPa in low-viremia vs 7.4 kPa in active HBV infection-demonstrating that necro-inflammatory activity can transiently inflate stiffness values. Collectively, these studies suggest that LSM-based HCC risk assessment should be performed when ALT and viral replication are stable and interpreted in light of disease-specific factors.

Similarly, in a large MASLD study, 5-year HCC incidence sharply increased with higher LSM strata, ranging from only 0.3% at < 12 kPa to 13.3% > 38 kPa, demonstrating a graded, non-linear risk pattern. In chronic hepatitis B, Shili-Masmoudi et al[19] proposed an LSM-based HCC risk score that enhanced predictive accuracy (AUROC 0.83-0.89) compared with clinical models alone and identified a very-low-risk subgroup with near-zero 5-year incidence.

Our study adds new evidence to this literature by focusing on a direct comparison of HCC-positive vs HCC-negative patients with cirrhosis under matched conditions. Prior studies were often prospective cohorts in specific diseases (e.g., untreated HCV or HBV). Here, we included a real-world mix of cirrhosis etiologies (viral, alcohol, MASLD, etc.) and used careful matching (age, sex, CP class, etiology) to isolate the impact of LSM on HCC risk. Moreover, our study is the first to report data from the Indian sub-continent, which has different risks and natural history of cirrhosis compared to those worldwide. The findings confirm that the association of high LSM with HCC is not confined to viral hepatitis but extends to mixed cirrhosis populations, consistent with newer data in MASLD. In fact, recent large studies in metabolic fatty liver disease have highlighted the value of fibrosis markers for HCC stratification. A recent study proposed that HCC surveillance be initiated in MASLD patients if the fibrosis-4 index (FIB-4) score is ≥ 3.25 or LSM is ≥ 20 kPa, as these criteria identified a subgroup with sufficiently high HCC risk to justify screening[20]. Such risk-based strategies echo the implications of our study: Liver stiffness can be used to enrich for higher-risk patients who may benefit from enhanced surveillance intensity. By demonstrating a clear stiffness cut-off (about 30-40 kPa) above which HCC odds rise sharply, our results support the concept of incorporating elastography into HCC risk prediction models.

It is important to note that current clinical practice guidelines recommend bi-annual HCC surveillance in all cirrhotics with AFP and ultrasonography of liver[7,9,21]. In recent times, Baveno-VII has endorsed LSM as a surrogate marker for portal hypertension risk (variceal screening) rather than HCC and explicitly note that LSM is not yet a recommended HCC prediction tool in practice. While current guidelines do not incorporate LSM for HCC risk stratification, our findings suggest that liver stiffness warrants further evaluation in prospective studies as a potential adjunctive risk marker. This supports recent calls for a more individualized, risk-based surveillance strategy (“precision HCC screening”) that could incorporate non-invasive fibrosis markers. For example, patients with extremely high LSM values might merit shorter surveillance intervals or the use of more sensitive modalities (e.g., MRI/CT), whereas those with lower stiffness, especially if compensated and with other low-risk features, might be managed with standard ultrasound intervals. It is encouraging that new risk models (e.g., the “two-step” FIB-4 plus LSM algorithm in MASLD) are being proposed to refine HCC surveillance criteria. Our study adds to the evidence base needed for societies to consider LSM thresholds in future guideline updates, complementing histologic fibrosis stage and clinical factors in gauging HCC risk.

This study’s strengths include a rigorously matched case-control design (age, sex, etiology, CP class), strict TE quality criteria (≥ 10 valid shots, IQR < 30%, > 60% success rate), and exclusion of conditions that spuriously elevate stiffness, all of which enhance internal validity. This might also be the reason that LSM cut-off in our study is higher than previous studies. LSM remained independently associated with HCC after multivariable adjustment for key covariates (bilirubin, INR, platelets, MELD score, AFP), underscoring its robust prognostic signal. The cohort was moderately large and etiologically diverse, improving applicability across common cirrhosis phenotypes.

Our study had few important limitations. First, the retrospective, single-center design, with potential selection and information bias and restricted generalizability. Temporality cannot be proven; although LSM preceded HCC diagnosis by ≥ 12 months, the analysis used a single baseline LSM without longitudinal trajectories, precluding assessment of dynamic risk. Second, the diagnostic performance metrics must be interpreted within the context of a retrospective matched case-control design. While sensitivity, specificity, and AUROC reflect intrinsic test characteristics, PPV, NPV, and overall accuracy are influenced by disease prevalence and may not be generalizable to real-world populations with lower HCC prevalence. Third, exclusion of individuals with BMI > 35 kg/m² and those with TE acquisition failure may limit generalizability, particularly in MASLD populations. However, high BMI is known to increase TE failure rates and reduce measurement reliability due to greater skin-to-liver capsule distance and technical limitations. Fourth, transient elevations in liver stiffness can occur due to inflammatory activity, cholestasis, hepatic congestion, or infiltrative disorders. While documented cases were excluded, the possibility of residual confounding from subclinical or unrecognized conditions cannot be entirely excluded. Lastly, although our regression analysis treated LSM as a continuous variable, emerging evidence suggests the association between LSM and HCC risk is non-linear, with an inflection around the advanced-fibrosis range. Model calibration using spline or category-based approaches may therefore better capture risk gradients and prevent overestimation at lower stiffness values. Despite these limitations, our study adds valuable information to the medical literature and addresses an important question and warrants placement in the medical literature.

Importantly, this threshold was derived from a cohort in the Indian subcontinent with heterogeneous etiologies and a substantial proportion of patients with more advanced liver disease (including CP-B/C cirrhosis). These population-specific factors may influence the distribution of LSM values and limit the generalizability of this cut-off. Its applicability to other populations, including those with predominantly compensated disease or different etiological profiles, remains uncertain and requires external validation.

CONCLUSION

In conclusion, TE–measured LSM is strongly associated with the future risk of development of HCC among patients with cirrhosis and demonstrates excellent discriminatory performance. These findings support incorporating LSM into risk-stratification algorithms to identify patients who warrant intensified surveillance or early intervention. Prospective validation is needed to define absolute risk across thresholds (e.g., about 20-30 kPa and above), refine optimal cut-offs, and test whether dynamic LSM trajectories further enhance prediction. Future work should also evaluate multivariable models that integrate LSM with clinical and laboratory markers (i.e. age, sex, liver function, platelets, AFP, e.g., age-male-albumin-bilirubin-platelets-like scores) to improve accuracy and calibration. Clinically, a risk-tailored approach could individualize interval and modality or shorter intervals for very high LSM, standard ultrasound for lower risk, pending evidence of cost-effectiveness and outcome benefit. If validated, embedding LSM into guidance could enable more precise allocation of surveillance resources, earlier-stage detection, and potentially improved survival for cirrhosis.