Background and aim: In patients with chronic liver disease, the accuracy of ultrasound scan (US), spiral computed tomography (CT), magnetic resonance imaging (MRI), and alpha-fetoprotein (AFP) in diagnosing hepatocellular carcinoma (HCC) has never been systematically assessed, and present systematic review was aimed at this issue.
Methods: Pertinent cross-sectional studies having as a reference standard pathological examinations of the explanted liver or resected segment(s), biopsies of focal lesion(s), and/or a period of follow-up, were identified using MEDLINE, EMBASE, Cochrane Library, and CancerLit. Pooled sensitivity, specificity, and likelihood ratios (LR) were calculated using the random effect model. Summary receiver operating characteristic (SROC) curve and predefined subgroup analyses were made when indicated.
Results: The pooled estimates of the 14 US studies were 60% (95% CI 44-76) for sensitivity, 97% (95% CI 95-98) for specificity, 18 (95% CI 8-37) for LR+, and 0.5 (95% CI 0.4-0.6) for LR-; for the 10 CT studies sensitivity was 68% (95% CI 55-80), specificity 93% (95% CI 89-96), LR+ 6 (95% CI 3-12),and LR- 0.4 (95% CI 0.3-0.6); for the nine MRI studies sensitivity was 81% (95% CI 70-91), specificity 85% (95%CI 77-93), LR+ 3.9 (95%CI 2-7), and LR- 0.3 (95% CI 0.2-0.5). The sensitivity and specificity of AFP varied widely, and this could not be entirely attributed to the threshold effect of the different cutoff levels used.
Conclusions: US is highly specific but insufficiently sensitive to detect HCC in many cirrhotics or to support an effective surveillance program. The operative characteristics of CT are comparable, whereas MRI is more sensitive. High-quality prospective studies are needed to define the actual diagnostic role of AFP.