• hepatocellular carcinoma
• liver imaging
• liver biopsy

We read with interest the recent work by Ho et al demonstrating mutational hyperactivation of mammalian target of rapamycin signalling in a subgroup of hepatocellular carcinoma (HCC).1 As Berasain and Lechel concluded that the prospect of a positive therapeutic response may outweigh the risk associated with the HCC biopsy procedure2 and histology is essential for confirming a diagnosis of intrahepatic cholangiocarcinoma (ICC),3 we revisited the performance of non-invasive HCC diagnosis as recommended by current guidelines (eg, American Association for the Study of Liver Diseases (AASLD)) in clinical practice.4

We retrospectively analysed all patients (n=182) in which a CT-guided liver biopsy was performed at a tertiary referral centre within a 9-year period due to the clinical differential diagnosis of HCC. After quality control, a total of 94 biopsies (figure 1A, see online supplementary table 1) could be evaluated. HCC development was associated with the presence of cirrhosis (n=51/65, p<0.001), which was also found in 27% (3/11) of ICC and 33% (1/3) of combined hepatocellular–cholangiocarcinoma (HCC–CC) patients. The number of false-positive non-invasive HCC diagnosis was higher in patients without cirrhosis (n=13/19; table 1, figure 1C), while most false-negative diagnoses (n=19/22) occurred in patients with cirrhosis (see online supplementary table 2) Thus, the imaging features of HCC are characteristic but neither specific nor pathognomonic. A particular challenge represents the identification of HCC–CC and ICC (table 1). According to a recent study, about 70% of HCC–CCs were misclassified by both CT and MRI.5

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

(A) Flow chart of 182 liver biopsies obtained from patients with suspected HCC. The central boxes represent patients who were excluded from the study cohort. (B) Comparison of classification systems for non-invasive HCC diagnosis. (Top) Neglecting of arterial phase timing. (Bottom) Considering arterial phase timing. AASLD/EASL-EORTC vs LI-RADS; p<0.05). (C) Liver metastasis of multiple myeloma in a non-cirrhotic liver. Gd-EOB-DTPA-MRI shows an arterial phase hyperenhancement (upper left), washout in portal venous phase (upper centre) and hepatobiliary phase hypointensity (upper right) of a well-defined focal lesion with a diameter of 4.2 cm. Histologically densely packed plasmacytoid cells are seen (lower right), which are negative for the hepatocellular marker HepPar1 (lower centre) and show restriction of the lambda light chain indicating manifestation of a plasma cell neoplasia (lower left). Original magnification, 200-fold. AASLD, American Association for the Study of Liver Diseases; EASL-EORTC, European Association for the Study of the Liver-European Organisation for Research and Treatment of Cancer; Gd-EOB-DTPA, gadolinium-ethoxybenzyl-diethylenetriamine penta-acetic acid; HCC, hepatocellular carcinoma; NPV, negative predictive value; LI-RADS, Liver Imaging Reporting and Data System; PPV, positive predictive value.

To improve standardisation and consensus regarding the imaging diagnosis of HCC, the Liver Imaging Reporting and Data System (LI-RADS) was launched.6 It allows a standardised risk stratification under consideration of multiple, precisely defined parameters.6 LI-RADS recommends late arterial phase imaging, because some HCCs may show hyperenhancement only in the late arterial phase.6 In our cohort 51% (36/70) of CT scans, 50% (14/28) of conventional MRI, and 45% (20/44) of gadolinium-ethoxybenzyl-diethylenetriamine penta-acetic acid-MRI were obtained in late arterial phase. The full diagnostic power of LI-RADS became evident, when applied on late arterial images (figure 1B). Under this condition, the performance of non-invasive HCC diagnosis in daily routine practice was as reported from carefully controlled and supervised studies (sensitivity: 96% vs 64%, p<0.001; accuracy: 80% vs 61%, p=0.017; figure 1B, see online supplementary table 3).4 7 Since modern scanner technology allows the acquisition of a dedicated late arterial phase,8 technical requirements for multiphasic imaging should be included in clinical practice guidelines.

The two main arguments against HCC biopsy are bleeding risk and needle tract seeding. Although we observed biopsy-related bleeding in 3% of patients, all could be managed conservatively without sequela (see online supplementary table 1). These data are in line with a meta-analysis7 and corroborate that liver biopsy is a safe and effective procedure. No needle tract seeding was observed in our series, whereby all biopsies were performed by applying a coaxial 16 G core biopsy needle.9

Our study did not contain tumours <1 cm in diameter, but these lesions are more difficult to biopsy and how to deal with these observations both in the diagnostic as well as in the therapeutic setting is still a matter of debate. Furthermore, it is biased for patients, in which a biopsy was performed. Thus, cases showing the full-blown hallmarks of HCC are likely under-represented.

In summary, we encourage the use of LI-RADS in clinical practice, as the distinction of HCC, ICC and HCC-CC has profound clinical implications and LI-RADS seems to separate these entities better than classical algorithms. To prevent mistreatment a biopsy should be performed whenever there is doubt on the typing of an observation or the patient is low risk for HCC development. Our findings should be validated in a prospective randomised controlled trial.