Shear wave elastography combined with ALBI score enhances prediction of hepatic decompensation in chronic liver disease Authors: Jaejun Lee, Hyun Yang, Si Hyun Bae, Sunhong Yoo, Soon Woo Nam, Jiwoon Yoon, Jung Hyun Kwon, Soon Kyu Lee Published: 2026 ABSTRACT AbstractBackground & AimsHepatic decompensation is a key determinant of prognosis in chronic liver disease (CLD). Shear wave elastography (SWE) is widely used to assess fibrosis, but its prognostic value for decompensation remains uncertain. We evaluated SWE alone and combined with the albumin–bilirubin (ALBI) score for risk prediction.MethodsIn this retrospective two-center cohort study, we analyzed 2,759 patients with CLD who underwent SWE between 2019 and 2025. The endpoint was time to first decompensation event. We evaluated the predictive performance of SWE, alone and with ALBI, in training (n = 1,839) and test (n = 920) sets.ResultsDuring a median follow-up of 22.5 months, a total of 181 decompensation events and 53 deaths occurred. SWE alone demonstrated good performance for predicting decompensation (C-index 0.765) and overall survival. A 10 kPa threshold effectively stratified risk, with high-risk patients having higher hazards of decompensation and death (p <0.001). SWE and ALBI showed no statistical interaction, supporting their potential for complementary synergistic effects when combined. Integrating SWE with ALBI grade created a simple three-tier classification that further improved stratification, with high-risk patients (SWE >10 kPa and ALBI grade 2/3) having a 16.6-fold risk of decompensation and a 21.4-fold risk of death compared with low-risk patients (p <0.001). A Fine-Gray model in the training set generated a weighted linear formula (SWE + 18 × ALBI), termed the SALBI score, achieving the highest accuracy (C-index: training 0.804, test 0.820), outperforming SWE, ALBI, and MELD. The SALBI score stratified high-risk vs. low-risk groups, with 11.5-fold and 14.3-fold risks in train and test sets, respectively (p <0.001).ConclusionsSWE is a robust prognostic marker for hepatic decompensation. A linear score combining SWE with ALBI further improves prediction and provides a practical tool for stratification.Impact and ImplicationsHepatic decompensation is a life-threatening event that defines prognosis in CLD. SWE provides a non-invasive assessment of liver stiffness, and its role in predicting decompensation has not been fully validated. In this large cohort of 2,759 patients, SWE was confirmed as a robust prognostic marker for decompensation. We further developed the SALBI score, a simple composite index combining SWE and ALBI, which significantly outperformed existing models. These findings highlight SWE and the SALBI score as practical tools for individualized risk stratification in CLD and support their potential incorporation into future clinical practice and guideline development.Keywords: Shear wave elastography, Hepatic decompensation, Albumin, Bilirubin, Cirrhosis Introduction Chronic liver disease (CLD) results from persistent liver injury caused by chronic hepatitis B or C, metabolic dysfunction-associated steatotic liver disease, or alcoholic liver disease; it is characterized by progressive fibrotic remodeling of the liver parenchyma that may ultimately lead to advanced fibrosis and cirrhosis.1 Globally, cirrhosis and advanced stages of CLD affect an estimated 112 million individuals and account for approximately 2.4% of all deaths, imposing a substantial health and socioeconomic burden.2 Global studies examining trends from 1990 to 2019 have further reported an increasing burden of cirrhosis and other chronic liver diseases, underscoring their growing global impact.2,3 Progressive deterioration of hepatic function ultimately leads to serious liver-related outcomes, such as hepatocellular carcinoma (HCC) and hepatic decompensation, which necessitate active surveillance and timely clinical management.4,5 Hepatic decompensation, defined as a composite of complications resulting from hepatic functional deterioration, includes hepatic encephalopathy, variceal bleeding, and ascites.6 The occurrence of decompensation is a major prognostic determinant of overall survival (OS) in patients with cirrhosis or CLD; therefore, its management and prediction are of paramount importance.7,8 In this context, several studies have recently explored potential biomarkers for predicting decompensation in patients with CLD or cirrhosis.9 These include laboratory-based indices, such as the model for end-stage liver disease (MELD) score, albumin–bilirubin (ALBI) score, and fibrosis-4 index, as well as clinical parameters, such as the Child–Pugh score. Hepatic venous pressure gradient is also considered as a reliable marker for predicting decompensation; however, its clinical utility is restrained because of the procedure’s invasiveness.10,11 Notably, the Baveno VII consensus has proposed liver stiffness measurement (LSM) by vibration-controlled transient elastography (VCTE) as a potential predictive marker for portal hypertension.10 In this context, several studies have examined the prognostic and predictive role of LSM by VCTE for decompensation and other liver-related events in patients with CLD.12 Recently, Thorhauge et al.13 further demonstrated that LSM effectively predicted decompensation and all-cause mortality in patients with compensated CLD, underscoring its promising role in prognosis for cirrhosis. Meanwhile, shear wave elastography (SWE) is a promising ultrasound-based tool for non-invasive assessment of liver stiffness that provides real-time imaging and allows targeted measurement of parenchymal regions by selecting an appropriate region of interest.[14], [15], [16] It can also be readily performed in primary care settings where ultrasound is available; moreover, its reliability in fibrosis assessment has been validated in several studies, with comparative analyses showing non-inferior performance to VCTE.[17], [18], [19] These observations indicate that SWE also holds potential for predicting decompensation in patients with CLD, thereby supporting its further exploration as a clinically relevant biomarker. Despite the promising performance of SWE in LSM and prior evidence supporting LSM as a predictive marker for portal hypertension, evidence on SWE-based prediction of decompensation remains limited. In this study, using a large hospital-based cohort, we aimed to assess the performance of SWE for predicting decompensation in patients with CLD. Furthermore, we sought to determine whether combining SWE with other established biomarkers such as ALBI or MELD score could enhance the risk prediction for decompensation. Patients and methods Patients This retrospective study included patients with CLD who underwent SWE between March 2019 and February 2025 at two university-affiliated tertiary hospitals in Korea. Eligible patients met the following criteria: (1) presence of CLD, defined as chronic liver disease of any etiology, including chronic viral hepatitis, metabolic dysfunction-associated steatotic liver disease, or alcohol-related liver disease; (2) age ≥18 years; and (3) available records of reliable SWE measurements during the study period. Patients were excluded if they had (1) a history of decompensation, including variceal bleeding, hepatic encephalopathy, or ascites, before SWE; (2) acute hepatitis or superimposed acute liver injury at the time of SWE assessment; or (3) no follow-up visits after SWE assessment. The study protocol was approved by the Institutional Review Board of The Catholic University of Korea (approval number: OC25RASI0147) and conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived owing to the retrospective design. Assessment of liver stiffness using SWE Two-dimensional SWE (2D-SWE; GE Healthcare, Chicago, IL, USA) was performed immediately after B-mode ultrasonography by experienced operators. Patients fasted for ≥8 h and were instructed to suspend respiration briefly at end-expiration during acquisitions.20 Using a right intercostal approach to the right hepatic lobe (segments VII/VIII), the SWE sampling box was positioned over a homogeneous parenchymal area located 1.5–2.0 cm below the liver capsule, avoiding large vessels, bile ducts, rib shadows, and focal lesions.21 A circular region of interest was placed within the reliability measurement index (RMI) color map, and acquisitions with RMI <0.8 or unstable color patterns were discarded and repeated.22 Ten valid elastography measurements were obtained, and the median liver stiffness (kPa) was used as the representative value. Measurement reliability was defined as an IQR-to-median ratio <30%.23 Definition of hepatic decompensation Hepatic decompensation was defined as the new onset of ascites, hepatic encephalopathy of any grade (per the West Haven criteria), or variceal bleeding, in accordance with recent international guidelines.6 To ensure compensation at baseline, the entire available medical record before the SWE examination was reviewed for prior decompensation events. Study endpoints The primary endpoint was the first occurrence of decompensation after SWE assessment. Subsequent episodes were not counted. Incident HCC before decompensation and death before decompensation were collected and treated as competing events. Patients without decompensation, HCC, or death were censored at the date of last follow-up. The secondary endpoint was OS, defined as the time from SWE to death from any cause. Patients alive at the end of follow-up or lost to follow-up were censored at the date of last contact. An additional endpoint is the evaluation of an SWE-based prediction model, assessed both as a standalone parameter and in combination with the ALBI score, with subsequent validation of its performance. Serum albumin and bilirubin levels obtained within 2 weeks of the SWE examination, using the measurement closest in time to the SWE date, were included in the analysis. Subdividing test and train cohort To derive and validate the SWE-based score in the study population, patients were randomly allocated to the training and test cohorts in a 2:1 ratio. To preserve covariate balance between the training and test cohorts, patients were assigned using a stratification-based random allocation scheme based on SWE, albumin, total bilirubin (TB), age, and etiology. Statistical analyses Statistical analyses were performed using appropriate methods based on the type of variable. Continuous variables with normal distributions are presented as means with SD. Differences in means among more than two groups were compared using ANOVA. Categorical variables were compared using the Chi-square test or Fisher’s exact test, depending on sample size and assumptions. A restricted cubic spline was applied to assess the continuous relationship between SWE or ALBI score and the subdistribution hazard ratio (sHR) for decompensation. For analyses of decompensation, incident HCC before decompensation and death before decompensation were considered competing events. Cumulative incidence functions were used to estimate the incidence of decompensation, and Gray’s test was used to compare cumulative incidence between groups. Fine-Gray competing-risk regression models were used to estimate sHRs for decompensation. Multivariable Fine-Gray models were constructed using backward stepwise selection based on the Bayesian information criterion (BIC). For OS analyses, the Kaplan-Meier method was used to visualize the survival curves, and the log-rank test was used to compare survival between groups. Cox proportional hazards models were used to estimate hazard ratio (HR) for OS across predefined risk groups. Time-dependent area under the receiver operating characteristic curve (AUROC) was calculated to evaluate the predictive performance of SWE, SWE-derived score, and other non-invasive indices for cumulative decompensation. Harrell’s concordance index (C-index) was also assessed to examine the predictive performance of SWE, SWE-derived score, and other non-invasive indices over time. Comparisons of time-dependent AUROCs were performed using inverse probability of censoring weighting-based methods appropriate for censored time-to-event data. For comparisons between nested models (e.g. SWE-derived score vs. SWE alone or ALBI alone), likelihood ratio tests were conducted using the corresponding regression models to assess improvements in model fit. Optimal cut-off values for SWE-derived score were determined in the training cohort using maximally selected Fine-Gray Wald statistics, with candidate cut-offs required to include at least 15% of patients in each group. A sequential approach was used to first identify the high-risk cut-off and then distinguish the low- and intermediate-risk groups among the remaining patients. A p value <0.05 was considered statistically significant. All analyses were performed using R statistical software (version 4.0.3; R Foundation Inc., Vienna, Austria).24 Results Baseline characteristics of the study population After applying the exclusion criteria, 2,759 patients were included in the final analysis (Fig. 1). The mean age was 55.7 years, and males accounted for 58.7% of the cohort (Table 1). Viral etiology was identified in 34.2% of patients, whereas 65.8% had non-viral etiologies. Regarding liver function, 86.2% had ALBI grade 1, whereas only 1.5% had ALBI grade 3, indicating relatively preserved liver function across the study population. Flowchart of patient selection and dataset allocation. SWE, shear wave elastography. Table 1Baseline characteristics.Table 1Total (N = 2,759)Age (years)55.7 ± 12.6Male1,620 (58.7)BMI (kg/m2)25.1 ± 4.1DM1,010 (36.6)HTN735 (26.6)Etiology HBV786 (28.5) AVT treated462 (58.8) HCV157 (5.7) Eradicated111 (70.7) Alcohol456 (16.5) MASLD1,198 (43.4) Others162 (5.9)WBC (1,000/μl)6.2 ± 2.2PLT (1,000/μl)211.9 ± 77.3TB (mg/dl)1.1 ± 1.5AST (IU/L)35.0 (26.0–54.0)ALT (IU/L)32.0 (20.0–56.0)Albumin (g/dl)4.4 ± 0.5PT (INR)1.0 ± 0.1Cr (mg/dl)0.8 ± 0.5SWE (kPa)7.6 ± 3.8 ≤8/8–10/>101,900 (68.9)/331 (12.0)/528 (19.1)ALBI grade (1/2/3)2,378 (86.2)/340 (12.3)/41 (1.5)Values are presented as median (IQR), mean ± SD, or number (%). Percentages for AVT-treated HBV- and HCV-eradicated cases were calculated among patients with HBV and HCV, respectively. ALBI, albumin–bilirubin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; AVT, antiviral treatment; Cr, creatinine; DM, diabetes mellitus; HTN, hypertension; MASLD, metabolic dysfunction-associated steatotic liver disease; PLT, platelet; PT (INR), prothrombin time (international normalized ratio); SWE, shear wave elastography; TB, total bilirubin. Baseline characteristics. Values are presented as median (IQR), mean ± SD, or number (%). Percentages for AVT-treated HBV- and HCV-eradicated cases were calculated among patients with HBV and HCV, respectively. ALBI, albumin–bilirubin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; AVT, antiviral treatment; Cr, creatinine; DM, diabetes mellitus; HTN, hypertension; MASLD, metabolic dysfunction-associated steatotic liver disease; PLT, platelet; PT (INR), prothrombin time (international normalized ratio); SWE, shear wave elastography; TB, total bilirubin. To further clarify the distribution of liver stiffness in the study population, patients were categorized according to SWE values using the 8 and 10 kPa thresholds. Overall, 1,900 patients (68.9%) had SWE values ≤8 kPa, 331 patients (12.0%) had SWE values 8–10 kPa, and 528 patients (19.1%) had SWE values >10 kPa. Baseline characteristics stratified by SWE category are presented in Table S1. Performance of SWE for predicting hepatic decompensation During a median follow-up period of 22.5 months, 181 cases of decompensation and 53 deaths were documented. The cumulative incidence of decompensation at 6, 12, 24, and 36 months was 3.7% (95% CI 3.0–4.5), 4.9% (95% CI 4.0–5.8), 6.6% (95% CI 5.5–7.7), and 8.0% (95% CI 6.7–9.2), respectively. We first evaluated the predictive ability of SWE for decompensation. For 12-month cumulative decompensation incidence, the AUROC was 0.770 (95% CI 0.723–0.816), with a mean AUC of 0.763 over 36 months (Fig. 2A,B). Regarding OS, SWE showed an AUROC of 0.729 (95% CI 0.621–0.836) for 12-month OS, with a mean AUC of 0.699 over 36 months, indicating fair performance for predicting both decompensation and OS (Fig. S1). Predictive performance of shear wave elastography (SWE) for hepatic decompensation in patients with chronic liver disease. (A) Area under the receiver operating characteristic curve (AUROC) for SWE in predicting hepatic decompensation at 12 months. (B) Time-dependent AUROC for SWE in predicting hepatic decompensation over a 36-month follow-up period. (C) Restricted cubic spline analysis showing the association between SWE and subdistribution hazard ratio (sHR) for hepatic decompensation. (D) Cumulative incidence of hepatic decompensation stratified by an SWE threshold of 10 kPa (high-risk vs. low-risk groups). (E) Overall survival (OS) curves according to SWE risk stratification at 10 kPa. Tables below panels D and E show cumulative hepatic decompensation rates and OS rates at 6, 12, 24, and 36 months. (F) Forest plot showing the sHR for hepatic decompensation according to rule-of-five-based SWE strata. (G) Forest plot showing the hazard ratio (HR) for OS according to rule-of-five-based SWE strata. Patients with SWE ≤5 kPa were used as the reference group in panels F and G. The p values were calculated using Gray’s test for cumulative hepatic decompensation and the log-rank test for OS. AUC, area under the curve; CI, confidence interval; kPa, kilopascal. Subgroup analyses by etiology revealed that, in patients with viral liver disease, the 12-month AUROC for decompensation was 0.713 (95% CI 0.612–0.814), with a mean AUC of 0.729 (Fig. S2A and B). Similarly, in those with non-viral etiologies, the 12-month AUROC was 0.799 (95% CI 0.742–0.856), with a mean AUC of 0.781 (Fig. S2C and D). Among patients with viral liver disease, further stratification according to viral control status, defined as HBV receiving antiviral therapy or HCV with documented viral eradication, showed similar SWE performance between the controlled and uncontrolled groups, with 12-month AUROC values of 0.715 (95% CI 0.578–0.853) and 0.710 (95% CI 0.562–0.858), and mean time-dependent AUC values of 0.727 and 0.736, respectively (Fig. S2A and B). Clinical outcomes stratified by SWE Next, we analyzed the association between SWE and sHR for decompensation to determine the optimal cut-off value of SWE for predicting decompensation (Fig. 2C). SWE showed a significant relationship with decompensation risk, demonstrating both linear (p <0.001) and non-linear (p <0.001) components. In the non-linear model, the risk of decompensation began to increase from approximately 8 kPa, whereas the lower bound of the 95% CI first exceeded 1.0 at around 10 kPa. At this point, the estimated sHR was 1.41 (95% CI 1.01–1.95). Using 10 kPa as the cut-off, patients were stratified into high-risk (n = 528) and low-risk (n = 2,231) groups. Compared with the low-risk group, the high-risk group exhibited significantly worse outcomes for both decompensation (sHR 4.74, p <0.001) and OS (HR 5.02, p <0.001) (Fig. 2D,E). To further evaluate the independent prognostic value of SWE, we constructed Fine-Gray competing-risk regression models. Backward elimination based on the BIC for competing risks (BICcr) was used to select variables for the final models. Model 1 included SWE as a continuous variable, whereas Model 2 included SWE as a categorical variable using the 10 kPa cut-off. In Model 1, continuous SWE remained independently associated with hepatic decompensation (sHR 1.079 per kPa, 95% CI 1.051–1.108, p <0.001). In Model 2, SWE >10 kPa was also retained as an independent predictor of decompensation (sHR 2.099, 95% CI 1.461–3.015, p <0.001) (Table S2). In an additional three-variable model comparison, the combination of TB, albumin, and SWE showed the lowest BICcr, supporting the rationale for combining SWE with ALBI (Table S3). To further evaluate the risk gradient according to rule-of-five-based SWE strata, patients were categorized into ≤5, 5–10, 10–15, and ≥15 kPa groups. Compared with patients with SWE ≤5 kPa, the risk of decompensation increased stepwise across higher SWE strata, with sHRs of 3.18 (95% CI 1.47–6.88), 11.01 (95% CI 5.04–24.03), and 18.70 (95% CI 8.23–42.52), respectively (Fig. 2F). A similar stepwise increase was observed for OS, with HRs of 3.31 (95% CI 0.78–14.04), 11.07 (95% CI 2.56–47.96), and 23.98 (95% CI 5.31–108.42), respectively (Fig. 2G). In subgroup analyses by etiology, patients with viral liver disease were effectively stratified by a 10 kPa threshold, with the high-risk group showing higher risks of decompensation (sHR 3.95, p <0.001) and OS (HR 4.92, p = 0.003) compared with the low-risk group (Fig. S3A and B). Similarly, in patients with non-viral etiologies, the high-risk group had significantly worse outcomes for both decompensation (sHR 5.33, p <0.001) and OS (HR 5.28, p <0.001) (Fig. S3C and D). Comparison and interaction of SWE with non-invasive indices in predicting decompensation Considering the significance of SWE in predicting decompensation, non-invasive indices including ALBI score and MELD score were evaluated alongside SWE for their predictive performance for decompensation (Table S4). SWE demonstrated a C-index of 0.765, which was comparable to that of ALBI score (0.790, p = 0.244). Conversely, MELD score showed inferior performance, with a C-index of 0.672, significantly lower than SWE (p <0.001). To explore whether combining SWE with other non-invasive indices could enhance predictive accuracy, interactions between SWE and these indices were assessed (Fig. S4). ALBI score showed no significant interaction with SWE (p = 0.250), suggesting potential synergy when the two are combined. Conversely, MELD score demonstrated a trend toward interaction with SWE (p = 0.068), indicating that MELD score may provide less additive predictive value when incorporated with SWE. Risk stratification incorporating SWE with ALBI grade Based on the interaction analysis, ALBI score combined with SWE was selected to enhance predictive performance for decompensation. ALBI score showed a clear linear association with decompensation risk showing p for linearity of <0.001 (Fig. S5). In the linear plot, an ALBI score of -2.72, which is close to the -2.60 cut-off separating ALBI grades 1 and 2, corresponded to an sHR of approximately 1.0. Consistently, patients with ALBI grade 2–3 had a significantly higher risk of decompensation than those with grade 1 (Fig. S5B and C). Patients were then stratified into three risk groups—low, intermediate, and high—based on the SWE threshold of 10 kPa and ALBI grade. Those with ALBI grade 1 and SWE ≤10 kPa were classified as low risk (n = 2,044), those with ALBI grade 2–3 and SWE >10 kPa as high risk (n = 194), and the remaining patients as intermediate risk (n = 521). Compared with the low-risk group, the high-risk group showed a significantly increased risk of decompensation (sHR 16.60, p <0.001), and the intermediate-risk group also demonstrated a higher risk (sHR 3.46, p <0.001) (Fig. 3A). Similarly, OS was significantly worse in the high-risk (HR 21.43, p <0.001) and intermediate-risk (HR 4.46, p <0.001) groups compared with the low-risk group, underscoring the effectiveness of this combined stratification in predicting clinical outcomes (Fig. 3B). Clinical outcomes according to risk stratification using SWE in combination with ALBI grade. Patients with SWE ≤10 kPa and ALBI grade 1 were classified as the low-risk group, whereas those with SWE >10 kPa and ALBI grade 2 or 3 were classified as the high-risk group. Patients who did not meet either criterion were classified as the intermediate-risk group. (A) Cumulative incidence of hepatic decompensation according to risk group. (B) Overall survival according to risk group. Tables below each panel show cumulative hepatic decompensation rates and overall survival rates at 6, 12, 24, and 36 months. ALBI, albumin–bilirubin; CI, confidence interval; HR, hazard ratio; sHR, subdistribution hazard ratio; SWE, shear wave elastography. Establishing a novel score combining SWE with ALBI score To develop a novel score integrating SWE and ALBI, we divided the study population into training (n = 1,839) and test (n = 920) cohorts, with no significant differences in baseline characteristics between the two groups (Table S5). In the training cohort, Fine-Gray competing-risk regression analysis for decompensation identified β-coefficients of 0.0835 for SWE and 1.4806 for ALBI (Table S6). Dividing and rounding these coefficients yielded the formula: SWE + ALBI score × 18. The score was named SALBI, derived from the initial letter of SWE combined with ALBI. To better visualize component contributions, the SALBI score was decomposed into SWE, albumin, and TB components for each patient (Fig. 4). Stacked bar charts illustrated patient-level contributions of each factor (Fig. 4A). The standardized mean absolute contributions of albumin, SWE, and TB in the training cohort were 0.712, 0.714, and 0.377, respectively (Fig. 4B,C). In the test cohort, the contributions were consistent, at 0.720, 0.721, and 0.308, respectively (Fig. 4D–F). Contribution of SWE, albumin, and total bilirubin to the SALBI score in the training and test sets. (A–C) Training set: (A) Individual score contributions for each patient, separated by score component (albumin, total bilirubin, SWE). (B) Distribution of raw values for each score component shown as box plots. (C) Standardized mean absolute contribution of each component to the model. (D–F) Test set: (D) Individual score contributions, (E) distribution of raw values, and (F) standardized mean absolute contributions, as described for the training set. SALBI, shear wave elastography plus albumin–bilirubin; SWE, shear wave elastography. Performance of SALBI score on predicting hepatic decompensation and overall survival The predictive performance of SALBI score for decompensation was evaluated in both the training and test cohorts (Fig. 5). For 12-month decompensation, the AUROC of SALBI was 0.818 in the training cohort, the highest among all non-invasive indices. In the test cohort, the 12-month AUROC was similarly high at 0.824 (Fig. 5A,B). Over a 36-month period, the mean AUC was 0.797 in the training cohort and 0.801 in the test cohort, indicating consistently strong predictive accuracy (Fig. 5C,D). The overall predictive performance across time points is summarized in Table 2. To clarify the incremental value of SALBI, we compared the continuous SALBI score with rule-based SWE + ALBI stratification based on the SWE cut-off of 10 kPa and ALBI grade. Although both the continuous SALBI score and the rule-based SWE + ALBI stratification demonstrated good predictive performance, the continuous SALBI score consistently showed superior overall discrimination in both the training and test sets. Specifically, the SALBI score achieved higher C-index values than the rule-based SWE + ALBI stratification in the training set (0.804 vs. 0.757, p <0.001) and the test set (0.820 vs. 0.773, p = 0.006). In the training cohort, the C-index of SALBI score was 0.804, significantly outperforming SWE (p = 0.011), ALBI (p = 0.018), and MELD (p <0.001). In the test cohort, the SALBI score also demonstrated a superior C-index of 0.820 relative to other indices. Calibration plots confirmed excellent agreement between predicted and observed risks (Fig. S6). Comparative predictive performance of the SALBI score and other non-invasive indices for hepatic decompensation at 12 months in the training and test sets. Time-dependent receiver operating characteristic (ROC) curves for decompensation at 12 months in (A) the training set and (B) the test set. Time-dependent area under the ROC curve (AUROC) over a 36-month follow-up period in (C) the training set and (D) the test set. ALBI, albumin–bilirubin; AUC, area under the curve; CI, confidence interval; HD, hepatic decompensation; kPa, kilopascal; MELD, model for end-stage liver disease; SALBI, shear wave elastography plus albumin–bilirubin; SWE, shear wave elastography. Table 2Time-dependent predictive performance of SALBI and other non-invasive tests on hepatic decompensation in the training and test sets.Table 2Time-dependent AUROC (95% CI)C-index (95% CI)p values6-monthp values12-monthp values24-monthp values36-monthp valuesTraining setSALBI0.819 (0.752–0.889)0.818 (0.758–0.875)0.802 (0.744–0.860)0.786 (0.726–0.845)0.804 (0.756–0.852)SWE + ALBI0.766 (0.698–0.834)0.0050.772 (0.712–0.831)0.0080.763 (0.707–0.820)0.0240.745 (0.687–0.804)0.0120.757 (0.709–0.805)<0.001SWE0.755 (0.689–0.822)0.0300.754 (0.694–0.815)0.0090.767 (0.715–0.819)0.1110.743 (0.683–0.803)0.0560.750 (0.702–0.799)0.011ALBI0.799 (0.729–0.868)0.0810.785 (0.720–0.851)0.0790.769 (0.707–0.831)0.0260.757 (0.696–0.818)0.0880.778 (0.726–0.829)0.018MELD0.680 (0.595–0.765)0.0030.646 (0.570–0.722)<0.0010.663 (0.596–0.731)<0.0010.639 (0.569–0.710)<0.0010.652 (0.589–0.716)<0.001Test setSALBI0.813 (0.735–0.892)0.824 (0.754–0.894)0.817 (0.752–0.883)0.824 (0.764–0.884)0.820 (0.765–0.876)SWE + ALBI0.776 (0.689–0.862)0.0920.770 (0.690–0.850)0.0170.772 (0.701–0.843)0.0170.779 (0.709–0.848)0.0210.773 (0.710–0.836)0.006SWE0.782 (0.706–0.858)0.2600.786 (0.717–0.855)0.1310.779 (0.713–0.844)0.0980.773 (0.707–0.838)0.0210.779 (0.722–0.836)0.045ALBI0.793 (0.710–0.877)0.1950.805 (0.730–0.881)0.1840.790 (0.718–0.863)0.1070.806 (0.741–0.870)0.2320.800 (0.741–0.861)0.112MELD0.714 (0.622–0.807)0.0770.704 (0.615–0.794)0.0170.705 (0.621–0.789)0.0220.705 (0.622–0.788)0.0020.707 (0.635–0.779)0.004Values are presented as AUROC or C-index with 95% CIs. The p values were calculated using the continuous SALBI score as the reference. SWE + ALBI indicates risk stratification based on the SWE cut-off of 10 kPa and ALBI grade. ALBI, albumin–bilirubin; AUROC, area under the receiver operating characteristic curve; MELD, model for end-stage liver disease; SALBI, shear wave elastography plus albumin-bilirubin; SWE, shear wave elastography. Time-dependent predictive performance of SALBI and other non-invasive tests on hepatic decompensation in the training and test sets. Values are presented as AUROC or C-index with 95% CIs. The p values were calculated using the continuous SALBI score as the reference. SWE + ALBI indicates risk stratification based on the SWE cut-off of 10 kPa and ALBI grade. ALBI, albumin–bilirubin; AUROC, area under the receiver operating characteristic curve; MELD, model for end-stage liver disease; SALBI, shear wave elastography plus albumin-bilirubin; SWE, shear wave elastography. To assess the robustness of the SALBI score in patients with higher liver stiffness values, we performed sensitivity analyses in patients with SWE >8 kPa and >10 kPa. In patients with SWE >8 kPa and >10 kPa, the 12-month AUROC of SALBI was 0.810 (95% CI 0.756–0.865) and 0.810 (95% CI 0.747–0.872), respectively; the mean time-dependent AUC over 36 months was 0.808 in both groups (Fig. S7). Risk stratification and outcomes using the SALBI score Two SALBI cut-offs were determined using maximally selected rank statistics: -37.77 (high cut-off) and -45.00 (low cut-off) (Fig. S8). Based on these thresholds, patients were stratified into low-risk (SALBI <-45.00), intermediate-risk (-45.00-≤ SALBI <-37.77), and high-risk (SALBI ≥-37.77) groups. In the training cohort, the high-risk group demonstrated a significantly increased risk of cumulative decompensation incidence (sHR 11.45, p <0.001), and the intermediate-risk group also showed a significantly elevated risk (sHR 2.29, p = 0.003) compared with the low-risk group (Fig. 6A). Regarding OS, the HRs for the high- and intermediate-risk groups were 25.39 and 4.46, respectively, relative to the low-risk group (Fig. 6B). Similar findings were observed in the test cohort, where the high-risk group and the intermediate-risk group showed sHR of 14.29 (p <0.001) and 2.82 (p = 0.011), respectively, for decompensation compared with the low-risk group (Fig. 6C). For OS, the high-risk group demonstrated an HR of 13.13 compared with the low-risk group, confirming the consistent risk stratification capability of the SALBI score for both decompensation and OS (Fig. 6D). Clinical outcomes according to risk stratification using the SALBI score in the training and test sets. Patients were categorized into three risk groups based on the SALBI score cut-off values: high-risk (SALBI score ≥-37.77), intermediate-risk (-45.00≤ SALBI score <-37.77), and low-risk (SALBI score <-45.00) groups. (A) Cumulative incidence of hepatic decompensation according to SALBI risk groups in the training set. (B) Overall survival according to SALBI risk groups in the training set. (C) Cumulative incidence of hepatic decompensation according to SALBI risk groups in the test set. (D) Overall survival according to SALBI risk groups in the test set. HR, hazard ratio; SALBI, shear wave elastography plus albumin–bilirubin; sHR, subdistribution hazard ratio. The 12-month cumulative incidence of decompensation in the training cohort was 18.7%, 3.6%, and 1.4% in the high-, intermediate-, and low-risk groups, respectively; in the test cohort, the corresponding incidences were 20.3%, 4.2%, and 1.7%, respectively (Table S7). The distribution of outcomes (event-free vs. hepatic decompensation) across SALBI risk groups in the entire, training, and test cohorts further highlights the enhanced predictive ability of the SALBI score (Fig. S9). Discussion In the present study, SWE demonstrated good predictive value for hepatic decompensation, with a mean AUC of 0.763, and effectively stratified patients with CLD into high- and low-risk groups using 10 kPa as the cut-off. Moreover, the predictive performance was further improved when SWE was combined with the ALBI score. The resulting SALBI score, a novel composite index integrating SWE and ALBI, showed excellent performance in predicting decompensation and successfully classified patients into high-, intermediate-, and low-risk groups. Collectively, these findings highlight the potential of SWE and SALBI score as practical, non-invasive tools for risk stratification in patients with CLD. In our study, we demonstrated that LSM by SWE, an easily accessible tool in clinical practice, effectively predicted decompensation in patients with CLD. Previously, several investigations have demonstrated the utility of non-invasive biomarkers, including ALBI and MELD, for predicting decompensation, mortality, and other liver-related events. More recently, LSM by VCTE has emerged as a reliable biomarker for stratifying high-risk patients for decompensation in compensated advanced CLD as reported in the Baveno VII criteria.[25], [26], [27] VCTE-based indices such as the Agile score have demonstrated strong predictive performance for liver-related outcomes, further underscoring the prognostic value of LSM.28,29 In line with these findings, we finally demonstrated that LSM by SWE, an easily accessible tool in clinical practice using ultrasonography, also effectively predicted decompensation in patients with CLD. The AUC for 12-month decompensation exceeded 0.7 across different etiologies, and SWE also showed good performance for predicting 12-month OS (AUC 0.729). In addition, SWE outperformed the MELD score, a traditional and widely used prognostic model, thereby highlighting its clinical utility in predicting decompensation. Notably, in addition to its standalone predictive ability, combining SWE with the ALBI score further enhanced the predictive ability of decompensation according to our results. The ALBI score is a composite score of serum albumin and TB levels, reflecting liver functional reserve. Originally developed to assess mortality risk in patients with HCC, the ALBI score has shown reliable performance in predicting liver-related outcomes,30 similar to our results. Interestingly, SWE, which measures liver stiffness, showed no significant interaction with ALBI, supporting their possible synergistic role in predicting decompensation in our results. Indeed, when patients with CLD were stratified according to an SWE cut-off of 10 kPa and ALBI grade, this stratification effectively classified the cohort into three distinct risk groups (low, intermediate, and high), showing a 16.6-fold higher hazard for decompensation in the high-risk group (SWE >10 kPa and ALBI grade 2/3) compared with the low-risk group (SWE ≤10 kPa and ALBI grade 1). This simple stratification using SWE and ALBI showed promising results for decompensation risk prediction and highlights its potential applicability in routine clinical practice. Most importantly, based on the synergistic effect of ALBI and SWE for predicting decompensation, we newly developed the reliable predictive model for decompensation: SALBI score (SWE [kPa] + 18 × ALBI). Among its components, SWE contributed the most, followed by serum albumin and TB. Both albumin and bilirubin have been consistently reported as independent risk factors for hepatic decompensation in patients with CLD, which supports the biological plausibility of our findings.[31], [32], [33] In addition, previous studies have shown that SWE is not only useful for assessing fibrosis but also provides prognostic value for outcomes such as OS.34 Extending these observations, our study is the first to demonstrate that SWE, when combined with ALBI, significantly improves risk prediction for decompensation. The SALBI score demonstrated excellent performance for predicting decompensation, with AUROC values of 0.818 and 0.824 for 12-month decompensation in the training and test sets, respectively. Overall predictive ability, assessed by Harrell’s C-index, was improved compared with SWE or ALBI alone in both datasets. In addition, stratification into three groups (low, intermediate, and high risk) again effectively distinguished the risk of decompensation. Finally, beyond predictive accuracy, clinical applicability is a key consideration for risk stratification tools. Because VCTE requires a dedicated device for LSM, this novel score offers greater clinical utility in diverse practice settings where ultrasound is readily available. Several limitations should be acknowledged. First, the retrospective design may have introduced selection bias and residual confounding, highlighting the need for prospective studies to confirm these results. Second, potential inter- and intra-operator variability in SWE measurement should be considered. In our study, all examinations were performed by experienced operators, which likely minimized measurement variability. In addition, previous studies have demonstrated good to excellent inter-operator reliability for SWE, supporting the reproducibility of our findings in clinical practice.35 However, because a head-to-head comparison between SWE and VCTE was not performed in this study, future prospective studies incorporating both modalities are needed to clarify their comparative performance and clinical roles. Third, serum albumin and bilirubin values were collected within a 2-week window around the SWE examination, and liver function may fluctuate even over short intervals, potentially resulting in misclassification of ALBI grade or SALBI categories. Moreover, because SALBI was based on a single baseline SWE and laboratory assessment, its long-term predictive performance may be limited by dynamic changes in liver stiffness and liver function. Repeated assessment of SALBI should therefore be evaluated in future prospective studies. Finally, the SWE and SALBI cut-offs were derived from the same dataset in which their performance was evaluated, raising the possibility of overfitting; external validation in independent cohorts will be necessary to confirm the robustness of these thresholds. Nonetheless, this study also has important strengths. It incorporated the largest cohort to date with SWE measurements, thereby improving the robustness and generalizability of the findings. In addition, despite the absence of external validation, the random division of the cohort into training and test sets allowed for internal validation, further strengthening the reliability of our conclusions. In conclusion, this study demonstrated the prognostic value of SWE for predicting decompensation in patients with CLD and introduced the SALBI score, a novel SWE-based index that further improved the predictive accuracy. Given that SWE is widely available in routine clinical practice, its use may facilitate individualized risk stratification and support tailored management strategies in CLD. These findings highlight the potential of SWE-based approaches to refine prognostication and optimize clinical decision-making for patients at risk of decompensation. Abbreviations ALBI, albumin–bilirubin; AUROC, area under the receiver operating characteristic curve; BIC, Bayesian information criterion; BICcr, BIC for competing risks; CLD, chronic liver disease; HCC, hepatocellular carcinoma; HR, hazard ratio; LSM, liver stiffness measurement; MELD, model for end-stage liver disease; OS, overall survival; RMI, reliability measurement index; ROC, receiver operating characteristic; SALBI, shear wave elastography plus albumin–bilirubin; sHR, subdistribution hazard ratio; SWE, shear wave elastography; TB, total bilirubin; VCTE, vibration-controlled transient elastography. Authors’ contributions Study concept and design: JL, SKL. Software: JL. Validation and formal analysis: JL, SKL. Resources: SHY, SWN, JY, JHK, SKL. Data collection: JL, SWN, JHK, SKL. Writing – original draft: JL. Writing – review and editing: JL, JHK, SKL. Visualization: JL. Supervision: HY, SHB, SWN, JHK, SKL. Project administration and funding acquisition: SKL. Final approval of the version to be published: all authors. Data availability The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed at the corresponding author. Financial support This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (Ministry of Science and ICT, MSIT) (RS-2024-00451810 to SKL). This work was also supported by the Technology Innovation Program (or Industrial Strategic Technology Development Program) (20024163, Development of microbiome-based treatment technology to improve the treatment and prognosis of liver transplant patients) funded by the Ministry of Trade, Industry and Energy (MOTIE, Korea) (to SKL). This research was also supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2025-25459150). Conflicts of interest The authors declare no conflicts of interest. Please refer to the accompanying ICMJE disclosure forms for further details. Acknowledgements We gratefully acknowledge Sumin Lee, Hyejin Yun, Jaeseon Kim, A Hyun Won, Seung Min Kim, Ryun Ah Kim, Soyoung Yun, Adsom Lee, Songhwa Che, Arum Mun, Mihee Lee, and Hyewon Lim, members of the liver center teams at Incheon St. Mary's Hospital and Eunpyeong St. Mary's Hospital, for their valuable technical and logistical support in liver stiffness assessment. Footnotes 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.2026.101963. Supplementary data The following are the Supplementary data to this article: Multimedia component 1 Multimedia component 2 Multimedia component 3 References Associated Data Supplementary Materials Multimedia component 1 Multimedia component 2 Multimedia component 3 Data Availability Statement The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed at the corresponding author.