Abstract
Non-contrast MRI is used for identifying patients with hepatocellular carcinoma (HCC), especially among high-risk patients with cirrhosis or chronic viral hepatitis. The accuracy of non-contrast MRI has been investigated with varying results. We performed this meta-analysis to consolidate the evidence on the accuracy of non-contrast MRI for the detection of HCC. We conducted a systematic search in the databases of PubMed Central, SCOPUS, MEDLINE, EMBASE and Cochrane from inception till November 2020. We used the STATA software "Midas" package for meta-analysis. We included 15 studies with 3,756 patients. The pooled sensitivity and specificity of non-contrast MRI for HCC detection were 84% (95%CI, 78%-88%) and 94% (95%CI, 91%-97%). The positive likelihood ratio was 14.9 (95% CI, 9.0-24.7) and the negative one 0.17 (0.12-0.23). The overall quality of the studies was high. We found significant heterogeneity based on chi-square test results and I(2) statistic > 75%. Deek's test showed the absence of publication bias. We found that non-contrast MRI has high sensitivity and specificity as a tool for detecting HCC. Studies exploring its accuracy in different ethnic populations are required to strengthen the evidence.