Shear wave elastography demonstrates similar risk stratification performance to vibration-controlled transient elastography in FIB-4–based two-step algorithms for MASLD
Jaejun Lee, Hyun Yang, Si Hyun Bae, Sun Hong Yoo, Soon Woo Nam, Jiwoon Yoon, Jung Hyun Kwon, Soon Kyu LeeBackground & Aims:
The role of shear wave elastography (SWE) in two-step MASLD risk stratification remains unclear. We compared SWE with vibration-controlled transient elastography (VCTE) using American Gastroenterological Association (AGA) and European Association for the Study of the Liver (EASL) algorithms.
Approach & Results:
Patients with MASLD who underwent both VCTE and SWE during the same session between 2019 and 2025 were included. Risk stratification was performed using AGA- and EASL-based two-step approaches. The primary endpoint was liver-related events (LRE). Among 2,817 patients, the AGA-SWE algorithm classified 2,196, 140, and 481 patients as low-, intermediate-, and high-risk, respectively. Using the AGA-SWE algorithm, both intermediate- and high-risk groups showed increased risks of LREs compared with the low-risk group (subdistribution hazard ratio [sHR] 7.60 and 9.86, respectively). The EASL-SWE algorithm also stratified LRE risk relative to the low-risk group, with higher sHRs observed for the intermediate-low, intermediate-high, and high-risk groups (sHR 2.99, 10.83, and 18.87, respectively). The integrated time-dependent area under the curve through 60 months was 0.770 (95% confidence interval [CI]: 0.709, 0.827) and 0.775 (95% CI: 0.712, 0.835) for AGA-SWE and AGA-VCTE, respectively, and 0.804 (95% CI: 0.753, 0.849) and 0.805 (95% CI: 0.750, 0.852) for EASL-SWE and EASL-VCTE, respectively, with no significant differences between SWE and VCTE in either algorithm. No significant net reclassification improvement was observed between SWE- and VCTE-based AGA algorithms. SWE and VCTE demonstrated a strong correlation (Spearman’s ρ = 0.713, 95% CI: 0.691–0.735,
Conclusions:
SWE-based two-step risk stratification showed similar predictive performance to VCTE-based models for predicting LREs in MASLD, although these findings do not establish formal equivalence.