LI-RADS Ultrasound Surveillance Version 2024: Comparison With Version 2017 for Hepatocellular Carcinoma Detection and...

BACKGROUND. The LI-RADS Ultrasound Surveillance algorithm was updated in 2024, incorporating α-fetoprotein (AFP) and a visualization score of VIS-C into management recommendations after nonpositive results. OBJECTIVE. The purpose of this study was to compare the diagnostic performance of LI-RADS Ultrasound Surveillance version 2017 (v2017) and version 2024 (v2024) for the detection of hepatocellular carcinoma (HCC) in at-risk patients and to identify predictors of VIS-C on follow-up surveillance examinations. METHODS. This retrospective analysis included 407 patients (230 men and 177 women; median age, 56 years) with cirrhosis who underwent rounds of semiannual surveillance ultrasound as part of a prospective trial from November 2011 to August 2014. Two radiologists independently assigned ultrasound categories to round 1 examinations and visualization scores to round 1 and round 2 examinations; a third radiologist adjudicated disagreements. The AFP level was considered positive if elevated or increasing from preenrollment values, per v2024 criteria. The reference standard for HCC was positive biopsy or an LR-5 observation on MRI. Diagnostic performance was compared between v2017 and v2024. Logistic regression analyses were performed to identify predictors of a round 2 VIS-C result, with attention given to risk factors for VIS-C described in v2024. RESULTS. HCC was diagnosed in 28 patients (6.9%). LI-RADS Ultrasound Surveillance v2024, in comparison with v2017, showed greater sensitivity for reader 1 (64.3% vs 42.9%, p = .03) and reader 2 (64.3% vs 39.3%, p = .02) and lower specificity for reader 1 (82.1% vs 92.6%, p < .001) and reader 2 (82.3% vs 92.9%, p < .001). All seven patients with HCC detected by v2024 but not v2017 by means of consensus assessments had increasing AFP; two also had elevated AFP. Among 299 patients who underwent round 2 ultrasound after negative round 1 v2024 surveillance results, the only independent predictor of a round 2 VIS-C result was a round 1 VIS-C result (adjusted OR = 21.04 [95% CI, 10.84-40.83], p < .001). For 88 of these patients with round 1 VIS-C, no v2024 risk factor showed a significant univariable association with repeat VIS-C. CONCLUSION. Compared with v2017, LI-RADS Ultrasound Surveillance v2024 had higher sensitivity but lower specificity for HCC detection, which was primarily related to increasing, rather than elevated, AFP. The only independent predictor of VIS-C on subsequent ultrasound was the initial VIS-C result. CLINICAL IMPACT. The findings support the use of LI-RADS Ultrasound Surveillance v2024 to improve HCC detection in at-risk patients.