Performance of CEUS LI-RADS v2017 major feature combinations: individual patient data meta-analysis.

BACKGROUND: The LI-RADS diagnostic algorithm uses imaging features in contrast-enhanced ultrasound (CEUS) to standardize the diagnosis of hepatocellular carcinoma (HCC) in at-risk patients. However, the diagnostic performance of specific major feature combinations has not been comprehensively evaluated. PURPOSE: To evaluate the diagnostic performance of CEUS LI-RADs version 2017 major feature combinations for (HCC) in at-risk individuals across LI-RADS categories 3-5. MATERIALS AND METHODS: A living systematic review and individual participant data (IPD) meta-analysis was conducted, including studies using CEUS LI-RADS v2016 or v2017 in at-risk patients, identified through database searches updated to February 2024. Eligible observations were categorized per LI-RADS guidelines, and PPVs for HCC were calculated for all major feature combinations in LI-RADS categories 3-5 using a random-effects one-step model. Risk of bias was assessed independently using a customized QUADAS-2 tool. RESULTS: Thirteen studies were included, comprising 1575 patients (mean age, 62.8 ± 11.2 years; 79.3% male) with 1594 liver observations (median size, 37.1 mm). Pooled PPVs for HCC increased with higher CEUS LI-RADS v2017 categories: LR-3, 40.4% (95% CI: 27.2-55.1); LR-4, 69.7% (95% CI: 49.7-84.3); and LR-5, 95.1% (95% CI: 90.2-97.6). Major feature combinations did not differ from others within the same category. Most studies were at moderate to high risk of bias, primarily due to retrospective design, but sensitivity analysis restricted to low-risk observations yielded similar findings. CONCLUSION: CEUS LI-RADS demonstrated progressively higher PPVs from LR-3 to LR-5, supporting its ability to stratify risk across HCC categories. Major feature combinations performed similarly within each category, indicating internal consistency of the system. Although only PPVs were assessed, the results align with trends seen in CT/MRI LI-RADS, with wider confidence intervals in CEUS reflecting smaller sample sizes.

Duke Scholars

Author Mustafa Shadi Rifaat Bashir Radiology, Abdominal Imaging