Hepatocellular carcinoma: diagnostic performance of multidetector CT and MR imaging-a systematic review and meta-analysis.

Abstract

To perform a systematic review and meta-analysis of the diagnostic performance of computed tomography (CT) and magnetic resonance (MR) imaging as noninvasive modalities for evaluating hepatocellular carcinoma (HCC) in patients with chronic liver disease. A search of the MEDLINE, EMBASE, and Cochrane Library databases was performed to identify studies providing per-patient or per-lesion diagnostic accuracies of multidetector CT and MR imaging for HCCs in patients with chronic liver disease. Studies published from January 2000 to December 2012 that used a reference standard based on histopathologic findings and/or findings at follow-up were included. Summary estimates of diagnostic accuracy were obtained by using a random-effects model with further exploration with meta-regression and subgroup analyses. Forty studies (six on multidetector CT, 22 on MR imaging, and 12 on both CT and MR imaging) were included. The studies evaluated a total of 1135 patients with multidetector CT and 2489 patients with MR imaging. The overall per-patient sensitivity of MR imaging was 88% (95% confidence interval [CI]: 83%, 92%), with a specificity of 94% (95% CI: 85%, 98%). The overall per-lesion sensitivity of MR imaging was higher than that of multidetector CT when the paired data of the 11 available studies were pooled (80% vs 68%, P = .0023). Gadoxetic acid-enhanced MR imaging showed significantly higher per-lesion sensitivity than MR imaging performed with other contrast agents (87% vs 74%, P = .03). Per-lesion sensitivity was significantly lower for HCCs smaller than 1 cm than that for HCCs 1 cm or larger (P < .001 for CT, P = .02 for MR imaging) and for those in explanted livers (P = .04 for CT, P < .001 for MR imaging). MR imaging showed higher per-lesion sensitivity than multidetector CT and should be the preferred imaging modality for the diagnosis of HCCs in patients with chronic liver disease.