Objectives:

Hepatocellular carcinoma (HCC) is the most common primary malignant neoplasm of the liver, and accurate diagnosis and staging of HCC are important for guiding treatment and other clinical decisions. A number of imaging modalities are available for detection of HCC in surveillance and nonsurveillance settings, evaluation of focal liver lesions to identify HCC, and staging of HCC. The purpose of this review is to compare the effectiveness of imaging techniques for HCC on test performance, clinical decisionmaking, clinical outcomes, and harms.

Data sources:

Articles were identified from searches (from 1998 to 2013) of electronic databases, including Ovid MEDLINE^®, Scopus, and the Cochrane Libraries. The searches were supplemented by reviewing reference lists and searching clinical trials registries.

Review methods:

We used predefined criteria to determine study eligibility. We selected studies of ultrasound (US), computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) that evaluated test performance for detection of HCC lesions, evaluation of focal liver lesions, or staging of HCC. We also included randomized trials and comparative observational studies on effects of imaging on clinical decisionmaking, clinical outcomes, and harms. The risk of bias (quality) of included studies was assessed, data were extracted, and results were summarized quantitatively (through meta-analysis) and qualitatively. Analyses were stratified by imaging type and unit of analysis (patient or HCC lesion). Additional analyses were conducted to evaluate the effects of the reference standard used and study, patient, tumor, and technical characteristics on estimates of test performance.

Results:

Of the 4,846 citations identified at the title and abstract level, we screened and reviewed 851 full-length articles. A total of 281 studies were included, 274 of which evaluated test performance. No body of evidence was rated high strength of evidence due to methodological shortcomings, imprecision, and/or inconsistency. Moderate strength-of-evidence ratings were primarily limited to estimates of diagnostic accuracy for CT and MRI and to some direct comparisons involving US versus CT or MRI. Few studies evaluated diagnostic accuracy in surveillance settings, and the only two studies that directly compared imaging modalities found US without contrast associated with lower sensitivity and specificity than CT for detection of patients with HCC (low strength of evidence). For detection of HCC in nonsurveillance settings, based on studies that directly compared imaging modalities and using HCC lesions as the unit of analysis, sensitivity was lower for US without contrast than for CT or MRI (difference in sensitivity based on within-study comparisons of 0.11 to 0.22) (moderate strength of evidence) and sensitivity was higher for MRI than CT (pooled difference 0.09; 95% confidence interval [CI], 0.07 to 12) (moderate strength of evidence). For evaluation of detected focal liver lesions, we found no clear differences in sensitivity for identifying HCC between US with contrast, CT, and MRI (moderate strength of evidence). Across imaging modalities and indications for imaging, specificity was generally 0.85 or higher, but specificity was not reported in a number of studies. Sensitivity of ¹⁸F-fluorodeoxyglucose (FDG) PET for identification of metastatic HCC lesions was 0.82 (95% CI, 0.72 to 0.90) (low strength of evidence), but sensitivity of FDG PET for intrahepatic lesions was poor (moderate strength of evidence). Evidence suggests that imaging strategies involving more than one imaging modality, in which a positive test is defined as typical imaging findings on one or more imaging modalities, is associated with higher sensitivity than a single test, with little effect on specificity (moderate strength of evidence).

Across imaging modalities, factors associated with lower estimates of sensitivity included use of explanted liver as the reference standard, use of HCC lesions as the unit of analysis, smaller HCC lesion size, and more well-differentiated HCC lesions. For MRI, hepatic-specific contrast agents were associated with slightly higher sensitivity than nonspecific contrast agents. For PET, evidence suggested higher sensitivity with use of PET/CT than with PET alone and with ¹¹C-acetate than with FDG.

Evidence on the comparative effects of imaging for HCC on clinical decisionmaking was extremely limited. The proportion of patients correctly assessed with CT for transplant eligibility based on Milan criteria ranged from 40 to 96 percent (moderate strength of evidence). Evidence on the effects of surveillance with imaging versus no surveillance on clinical outcomes was limited to a single randomized trial (low strength of evidence). Although it found an association between surveillance with US and alpha-fetoprotein (AFP) and decreased liver-specific mortality, the trial was conducted in China, potentially limiting applicability to screening in the United States, and there were important methodological shortcomings. Evidence on comparative harms associated with imaging was also extremely limited but indicates low rates of serious direct harms.

Conclusions:

Several imaging modalities have relatively high sensitivity and specificity for diagnosis or staging of HCC, although test performance is suboptimal for small or well-differentiated HCC. Although there are some potential differences in test performance between different imaging modalities and techniques, more research is needed to understand the effects of such differences on clinical decisionmaking and clinical outcomes.