Diagnostic Performance of CT/MRI LI-RADS Version 2018 Major Feature Combinations: Individual Participant Data Meta-Analysis.

Abstract

Background The CT/MRI Liver Imaging Reporting and Data System (LI-RADS) diagnostic algorithm classifies liver observations in patients with high-risk hepatocellular carcinoma (HCC) using imaging features. However, data regarding the diagnostic performance of specific LI-RADS major feature combinations is limited. Purpose To conduct a systematic review and individual participant data (IPD) meta-analysis to establish the positive predictive values (PPVs) of LI-RADS major feature combinations using CT/MRI LI-RADS version 2018 in patients at risk for HCC. Materials and Methods Medline, Embase, Cochrane Central, and Scopus were searched for studies published from January 2014 to February 2023. Studies reporting HCC percentages for LI-RADS categories in patients at high risk for HCC were included. A one-stage random-effects IPD meta-analysis was used to calculate the PPV for HCC diagnosis and 95% CIs of major feature combinations. Wald test was used to compare combinations. Risk of bias (RoB) was assessed using Quality Assessment of Diagnostic Accuracy Studies 2, known as QUADAS-2 (protocol: https://osf.io/ah5kn). Results Forty-six studies including 6765 patients (mean age, 59 years ± 10.69 [SD]; 75% male patients [5081 of 6765]; age range, 18-93 years) with 7500 liver observations were analyzed. High RoB in at least one domain was found in 80% of studies (37 of 46). The pooled PPV estimate for major feature combinations was 58.28% in LR-3 (95% CI: 44.00, 71.29), 80.82% in LR-4 (95% CI: 71.04, 87.86), and 95.81% in LR-5 (95% CI: 91.06, 98.09). The majority of LI-RADS major feature combinations had PPVs that did not differ from others within the same category, supporting the current categorization (P value ranges: LR-3, .17-.73; LR-4, .10 to >.99; LR-5, .08 to >.99). Notably, five major feature combinations differed from the pooled PPV of the LR category. LR-3 was lower without nonrim arterial phase hyperenhancement (APHE) measuring smaller than 20 mm without additional major features (14.81%; 95% CI: 6.35, 30.85; P < .001), and higher with APHE measuring 10-19 mm without additional major features (68.33%; 95% CI: 53.94, 79.90; P = .01). LR-4 was lower without APHE measuring 20 mm or larger with enhancing capsule (50.81%; 95% CI: 28.92, 72.39; P = .009). LR-5 was lower with APHE measuring 10-19 mm with threshold growth (74.40%; 95% CI: 51.06, 89.00; P < .001), and with APHE measuring 20 mm or larger with threshold growth (82.35%; 95% CI: 57.29, 94.20; P = .02). Conclusion This meta-analysis showed that most major feature combinations in the same CT/MRI LI-RADS category had similar PPVs for HCC in patients at high risk for HCC, with the exception of five combinations within LR-3 through LR-5. © RSNA, 2025 Supplemental material is available for this article. See also the editorial by Johnson in this issue.