Accuracy of the Liver Imaging Reporting and Data System in Computed Tomography and Magnetic Resonance Image Analysis of Hepatocellular Carcinoma or Overall Malignancy—A Systematic Review

Background & Aims: The Liver Imaging Reporting and Data System (LI-RADS) categorizes observations from imaging analyses of high-risk patients based on the level of suspicion for hepatocellular carcinoma (HCC) and overall malignancy. The categories range from definitely benign (LR-1) to definitely HCC (LR-5), malignancy (LR-M), or tumor in vein (LR-TIV) based on findings from computed tomography or magnetic resonance imaging. However, the actual percentage of HCC and overall malignancy within each LI-RADS category is not known. We performed a systematic review to determine the percentage of observations in each LI-RADS category for computed tomography and magnetic resonance imaging that are HCCs or malignancies. Methods: We searched the MEDLINE, Embase, Cochrane CENTRAL, and Scopus databases from 2014 through 2018 for studies that reported the percentage of observations in each LI-RADS v2014 and v2017 category that were confirmed as HCCs or other malignancies based on pathology, follow-up imaging analyses, or response to treatment (reference standard). Data were assessed on a per-observation basis. Random-effects models were used to determine the pooled percentages of HCC and overall malignancy for each LI-RADS category. Differences between categories were compared by analysis of variance of logit-transformed percentage of HCC and overall malignancy. Risk of bias and concerns about applicability were assessed with the Quality Assessment of Diagnostic Accuracy Studies 2 tool. Results: Of 454 studies identified, 17 (all retrospective studies) were included in the final analysis, consisting of 2760 patients, 3556 observations, and 2482 HCCs. The pooled percentages of observations confirmed as HCC and overall malignancy, respectively, were 94% (95% confidence interval [CI] 92%–96%) and 97% (95% CI 95%–99%) for LR-5, 74% (95% CI 67%–80%) and 80% (95% CI 75%–85%) for LR-4, 38% (95% CI 31%–45%) and 40% (95% CI 31%–50%) for LR-3, 13% (95% CI 8%–22%) and 14% (95% CI 9%–21%) for LR-2, 79% (95% CI 63%–89%) and 92% (95% CI 77%–98%) for LR-TIV, and 36% (95% CI 26%–48%) and 93% (95% CI 87%–97%) for LR-M. No malignancies were found in the LR-1 group. The percentage of HCCs and overall malignancies confirmed differed significantly among LR groups 2–5 (P <.00001). Patient selection was the most frequent factor that affected bias risk, because of verification bias and case–control study design. Conclusions: In a systematic review, we found that increasing LI-RADS categories contained increasing percentages of HCCs and overall malignancy based on reference standard confirmation. Of observations categorized as LR-M, 93% were malignancies and 36% were confirmed as HCCs. The percentage of HCCs found in the LR-2 and LR-3 categories indicate the need for a more active management strategy than currently recommended. Prospective studies are needed to validate these findings. PROSPERO number CRD42018087441.