Purpose

To compare the per-lesion sensitivity and positive predictive value (PPV) of ultrasonography (US), computed tomography (CT), and magnetic resonance imaging (MRI) for the diagnosis of hepatocellular carcinoma (HCC).

Materials and methods

The meta-analysis of sensitivity included 242 studies (15,713 patients); 116 studies (7492 patients) allowed calculation of PPV. Pooled per-lesion sensitivity and PPV for HCC detection were compared using empirical Bayes estimates of a beta-binomial model.

Results

The pooled per-lesion sensitivity and PPV of contrast-enhanced CT (73.6%, 85.8%) and gadolinium-enhanced MRI (77.5%, 83.6%) are not significantly different (P = 0.08, P = 0.2). However, if the hepatobiliary agent gadoxetate is used, MRI has significantly higher pooled per-lesion sensitivity and PPV (85.6%, 94.2%) than CT (P < 0.0001) or than MRI with other agents (P < 0.0001). Non-contrast-enhanced US has the lowest overall sensitivity and PPV (59.3%, 77.4%). Pooled per-lesion sensitivity and PPV of contrast-enhanced US (84.4%, 89.3%) are relatively high, but no contrast-enhanced US study used the most rigorous reference standards.

Conclusion

MRI utilizing the hepatobiliary agent gadoxetate has the highest overall sensitivity and PPV, and may be the single optimal method for diagnosis of HCC. Non-contrast-enhanced US has the lowest sensitivity and PPV. More rigorous reference standards are needed to compare the performance of contrast-enhanced US with CT and MRI. Differences in sensitivity and PPV between CT and conventional gadolinium-enhanced MRI are not statistically significant overall.