Purpose: To perform a systematic review to evaluate the diagnostic accuracy of hepatobiliary (HPB) phase gadoxetic acid-enhanced MR imaging of the liver in the diagnosis of focal nodular hyperplasia (FNH) versus hepatocellular adenoma (HCA) and to identify the rate of (a) reported HCAs that are iso- or hyperintense to liver and (b) reported FNHs that are hypointense to liver on HPB phase MR images.
Materials and methods: The institutional review board granted a waiver for this study type, and multiple databases were searched for studies in which researchers distinguished between FNH and HCA with gadoxetic acid-enhanced MR imaging. Studies to evaluate diagnostic accuracy were included; case reports and series were included to analyze the rate of iso- or hyperintense HCAs on HPB phase MR images. Risk of bias was assessed by using the Quality Assessment Tool for Diagnostic Accuracy Studies-2. Sensitivity and specificity were plotted with a forest plot; pooling was not performed because a small number of heterogeneous studies were included. Rate of iso- or hyperintense HCA on HPB phase gadoxetic acid-enhanced MR images was evaluated.
Results: Six studies (309 patients; 164 with HCA, 233 with FNH) were included for diagnostic accuracy assessment. Twelve case series (129 patients; 81 with HCA, 70 with FNH) were included (studies with insufficient 2 × 2 table data for diagnostic accuracy assessment). Sensitivity was high (range, 0.91-1.00; lower margin of the 95% confidence interval: 0.77). Specificity was high (range, 0.87-1.00; lower margin of the 95% confidence interval: 0.54). Specificity was lowest among studies in which molecular subtyping of HCA was performed. Rate of iso-or hyperintensity of HCA on HPB phase MR images was variable (range, 0%-67%) and occurred more frequently in the inflammatory subtype. High risk of bias was identified in the domains of patient selection and reference standard.
Conclusion: The reported diagnostic accuracy of HPB phase gadoxetic acid-enhanced MR imaging in the diagnosis of HCA versus FNH is high; however, studies are few, heterogeneous, and at high risk for bias, indicating that diagnostic accuracy may be overestimated.
© RSNA, 2015