Liver imaging reporting and data system category M: A systematic review and meta‐analysis
Dong Hwan Kim, Sang Hyun Choi, Seong Ho Park, Kyung Won Kim, Jae Ho Byun, So Yeon Kim, Seung Soo Lee, Yong Moon Shin, Hyung Jin Won, Pyo‐Nyun Kim
Abstract
Abstract Background and Aims The Liver Imaging Reporting and Data System (LI‐RADS) category M (LR‐M) was introduced to preserve the high specificity of LI‐RADS algorithm for diagnosing hepatocellular carcinoma (HCC). We aimed to systematically determine the probability of the LR‐M for HCC and non‐HCC malignancy, and to determine the sources of heterogeneity between reported results. Methods Original studies reporting the probability of LR‐M for HCC and non‐HCC malignancy on magnetic resonance imaging (MRI) were identified in MEDLINE and EMBASE. The meta‐analytic pooled percentages of HCC and non‐HCC in LR‐M were calculated. Meta‐regression analysis was performed to explore study heterogeneity. The meta‐analytic frequency of each LR‐M imaging feature was determined. Results We found 10 studies reporting the diagnostic performance of LR‐M (1819 lesions in 1631 patients), and six reporting the frequency of LR‐M imaging features. The pooled percentages of HCC and non‐HCC malignancy for LR‐M were 28.2% (95% confidence interval [CI], 23.8%‐33.1%; I 2 = 83%) and 69.6% (95% CI, 64.6%‐74.1%; I 2 = 83%) respectively. The study type and MRI scanner field strength were significantly associated with study heterogeneity ( P ≤ .04). Of the seven imaging features, rim arterial phase hyperenhancement showed the highest frequency in both non‐HCC (48.9%; 95% CI, 43.0%‐54.8%) and HCC groups (9.8%; 95% CI, 6.9%‐13.6%). Conclusions The LR‐M category most commonly included non‐HCC malignancy but also included 28.2% of HCC. Substantial study heterogeneity was noted, and it was significantly associated with study type and MRI scanner field strength. In addition, the frequency of LR‐M imaging features was variable.