Introduction

Hepatocellular carcinoma (HCC) develops via multistep progression from a dysplastic nodule to early HCC and, finally, to advanced HCC [1]. The 5-year survival rate is higher among patients with early-stage HCC than among patients with advanced HCC. Thus, the assessment of risk for HCC is essential for the management of patients with chronic liver diseases. Cirrhosis is one of the important risk factors for HCC development, especially in patients with viral and alcoholic hepatitis. Not only the presence of cirrhosis but also the degree of fibrosis in non-cirrhotic liver is known to be correlated with the risk of HCC [2]. Ultrasound transient elastography has been developed to non-invasively measure liver stiffness, and the results are reported to be well correlated with histologically assessed liver fibrosis stage [3]. Recently, magnetic resonance elastography (MRE) has emerged as a non-invasive method of assessing liver elasticity. Excellent correlation between liver stiffness measured by MRE and liver fibrosis stage was reported by several researchers [4–9]. Because progressed liver fibrosis poses a greater risk of HCC development than less fibrotic livers, liver stiffness can be a risk factor for HCC development. One paper using ultrasound transient elastography revealed that the liver stiffness may well be a possible risk factor for HCC development [10].

The purpose of this study was to examine if liver stiffness measured by MRE is a risk factor for HCC in patients with chronic liver disease.

Patients and methods

Patients

This retrospective case–control study was performed in accordance with the principles of the Declaration of Helsinki. The ethics committee at our institute approved this study. The need for written informed consent was waived by the committee.

First, we reviewed the records of MR examinations performed at our institute from January 2010 to January 2011 to identify patients with chronic liver disease who had undergone MRE and gadoxetic acid-enhanced MR imaging. All patients underwent abdominal ultrasound within 2 weeks of the MR examination. The patients with a known previous medical history of HCC were excluded. In all, 301 patients matched the criteria, all of whom had undergone MR for liver cancer screening. HCC was observed on MR images in the case of 66 of the 301 patients. No tumours were found in the liver of the remaining 229 patients, and this was also confirmed by abdominal ultrasound performed 3 months after the MR examination. These patients were assigned to the “without HCC” group. Four patients were excluded, because liver tumours other than HCC were found in the liver. Two patients were also excluded because of inconclusive imaging findings. The 66 patients with HCC were assigned to the “with HCC” group; 66 age-matched controls were selected from the group of patients without HCC (Table 1; Fig. 1). The median (range) size of the HCCs was 24 (9–150) mm. The numbers of HCCs were 1 in 52 patients, 2 in 7 patients, 3 in 4 patients, and 4 in 3 patients. Seventy-two patients had Child–Pugh class A disease, whereas 58 and 2 patients had Child–Pugh class B and C disease, respectively.

Table 1 Patients’ demographics
Full size table
Fig. 1
figure 1

Inclusion criteria for the study

Full size image

MRI examination

Gadoxetic acid (Primovist, Bayer Health Care, Osaka, Japan)-enhanced MRI (EOB-MRI) was performed in all patients with a 1.5-T superconducting magnetic resonance system (Signa Excite HD; GE Healthcare, Milwaukee, WI, USA) and an eight-channel phased-array coil. First, respiratory-triggered fat-saturated T2-weighted fast spin-echo images (T2WI) were acquired, followed by dual-echo fast-spoiled gradient-echo T1-weighted images (T1WI). Dynamic images using fat-saturated T1-weighted gradient-echo images with a three-dimensional acquisition sequence (liver acquisition with volume acceleration) were obtained before (precontrast) and 20–30 s (arterial-phase scan triggered by fluoroscope technique), 60 s (portal venous phase), 2 min (late phase), 5 min, 10 min and 20 min (hepatocyte phase) after the administration of gadoxetic acid. The contrast agent was administered at a rate of 1 mL/s as a bolus dose (0.025 mmol/kg body weight) through an intravenous cubital line (22-gauge) flushed with 20 mL saline by using a power injector (Sonic Shot 50; Nemoto Kyorindo, Tokyo, Japan). The hepatocyte-phase images obtained 20 min after the injections were used for evaluation in this study [11]. The images were acquired in the transverse plane and had a section thickness of 5 mm and a 2.5-mm overlap (i.e. 2.5-mm interval). Sagittal hepatocyte-phase images were also obtained once just before axial hepatocyte-phase imaging at 20 min.

Diagnosis of HCC

All HCCs detected by MRI were confirmed pathologically or by dynamic contrast-enhanced CT which was performed within 1 month of MR examination, whereas the absence of HCC was confirmed by abdominal ultrasound performed 3 months after the MR examination [12].

The HCCs were pathologically confirmed in 40 of the 66 patients, of whom 28 underwent surgical partial hepatectomy, whereas the other 12 patients underwent percutaneous needle biopsy (Bard Monopty [16G], Medicon, Osaka, Japan). Imaging-based diagnosis was applied using criteria established by the American Association for the Study of Liver Disease in the other 26 patients [12]. For the imaging-based diagnosis, dynamic contrast-enhanced multidetector CT (Aquilion 64 or Aquilion one; Toshiba Medical Systems Corp., Tochigi, Japan) was used in helical mode with a tube voltage of 120 kV and a tube current of 280–400 mA (automatically adjusted to the patient’s body type). Unenhanced CT images were obtained first, and subsequently, a non-ionic contrast medium with 600 mg iodine/kg body weight (Omnipaque 300, Daiichi-Sankyo, Tokyo, Japan; Iomeron 350, Eisai Global, Tokyo, Japan; Iopamiron 370, Bayer Healthcare, Osaka, Japan) was administered intravenously as a bolus up to a maximum of 45 g in 30 s by using a power injector (Auto Enhance A-50; Nemoto Kyorindo, Tokyo, Japan). Image acquisition in the arterial, portal venous and delayed phases began at 40, 70 and 180 s, respectively, after initiation of contrast medium injection [13, 14]. The diagnosis of HCC was performed on the basis of typical findings on CT, i.e. hyperattenuation on the arterial-phase and subsequent washout on the portal- or delayed-phase images as proposed by the American Association for the Study of Liver Disease [12]. In the gadoxetic acid-enhanced MR imaging, we used the following criteria for identifying HCCs in the liver: (i) HCCs are suggested by hypervascularity on arterial-phase image and subsequent hypointensity on late-phase and hepatocyte-phase images , (ii) partial uptake of gadoxetic acid cannot exclude HCCs, (iii) hypointense lesion on hepatocyte-phase without arterial hypervascularity is not sufficient to be confirmed as HCC. Pathological confirmation and/or arterial enhancement in dynamic CT is required.

MR elastography (MRE)

Magnetic resonance elastography was performed before the administration of contrast agent. We used a cylindrical passive driver placed across the right chest wall to deliver vibrations via a transcostal approach [5, 15]. The generator placed outside of the MR examination room produced a pneumatic vibration that was delivered to the passive driver via a plastic cylinder. The passive driver was attached using an elastic (rubber) belt to deliver the vibration to the patients’ chest wall and liver. The MRE system including the generator of the vibration and passive driver was developed in the Mayo Clinic (Rochester, MN, USA) and provided to our institute with a service agreement.

A two-dimensional (2-D) gradient-echo MRE sequence was used to acquire axial wave images. The imaging position was set above the gall bladder and below the subphrenic region of the liver. For acquiring images of the liver at consistent positions at each phase offset, the patients were asked to hold their breath after expiration [4]. The imaging parameters for MRE were as follows: repetition time/echo time (TR/TE), 100/27 ms; continuous sinusoidal vibration, 60 Hz [4]; field of view, 30–34 × 40–45 cm; matrix size, 256 × 64; flip angle, 30°; slice thickness, 10 mm; number of slices obtained, 2; evenly spaced phase offsets, 4; and a single cycle of a 60-Hz trapezoidal motion-encoding gradient with zeroth- and first-moment nulling along the through-plane direction. A parallel imaging technique was not used. Two spatial presaturation bands were applied to each side of the selected slice to reduce motion artefacts due to blood flow. The total acquisition time was 64 s (four 16-s breath holds) (Table 2).

Table 2 Parameters of MR elastography
Full size table

The MRI system automatically generates elastograms by processing the acquired images of propagating shear waves with a previously described inversion algorithm [16]. The shear stiffness of the tissue is determined as a pixel value (kPa). One of the authors (T.K.) with 6 years’ experience in radiology selected the region of interest (ROI) in the right lobe of the liver on the elastogram, in which one of the two slices of the MRE was chosen for evaluation (the one that was nearer the centre of the passive driver placed on the patients’ chest wall). During the ROI measurements, the left lobe was not used because cardiac motion is considered to affect the phase image of MRE. As a rule, the ROIs larger than 1.5 cm2 in size were placed in the anterior lobe, segment 8, of the liver on the phase images and copied and pasted onto the elastogram in which liver stiffness was shown in kPa (Fig. 2). The distance from the surface of the liver (capsule) was 5 to 10 mm. The magnitude image was also referred to in order to confirm that the ROI was in the liver. Careful attention was paid to place ROIs in a part where the penetrating wave was well visualised and no interference was observed on the phase image. Intrahepatic vessels or bile ducts were also avoided by referring to the magnitude image. The average value in the ROI was used as a measure of liver stiffness by the MRE.

Fig. 2
figure 2

Rule of region of interest (ROI) measurement. ROIs were placed in the anterior lobe of the liver. During the ROI measurement, the phase images of MRE were referred to select where the wave is propagating without interference (top left). Magnitude image was also referred to in order to confirm that the ROI is in the liver (bottom left). The ROI placed on the phase image was copied and pasted onto the elastogram in which the stiffness of the liver is provided in kPa (right)

Full size image

Statistical analysis

We considered the following variables obtained by MR examination and laboratory analysis in the prediction of risk of HCC development: albumin, total bilirubin, aspartate transferase (AST) level, alanine transferase (ALT) level, per cent prothrombin time, platelet count, alpha-fetoprotein, and protein induced by vitamin K absence-II (PIVKA-II). Categorical variables were compared using the chi-squared test, whereas continuous variables were compared using the Wilcoxon test. For multivariate analysis, the odds ratio was estimated by logistic regression analysis using age, gender, and the variables which were identified as significant predictors for HCC development by univariate analysis. Although the age-matched control was selected, patients’ age was also added as a variable to ensure the exclusion of effect of age to enhance the adjustment of age between the two groups. Data analysis was performed using JMP version 8 (SAS Institute Japan, Tokyo, Japan). A two-sided P value of less than 0.05 was considered statistically significant.

Results

Risk of HCC development: univariate analysis

Liver stiffness (kPa) measured by MRE in patients with HCCs (median [range], 5.0 [2.3–9.3]) was significantly greater than that in patients without HCCs (3.9 [1.8–8.8]; P = 0.0025). A significant difference between the two groups was also observed in serum AST (IU/L) (patients with HCC vs. without HCC, 48 [11–194] vs. 36 [15–121]; P = 0.0064) and ALT (IU/L) (48 [10–310] vs. 36 [7–123]; P = 0.0447) as well as in the tumour markers, alpha-fetoprotein (ng/mL) (10 [1.6–8,478] vs. 3.6 [1.1–808]; P = 0.0012) and PIVKA-II (mAU/mL) (24 [6–22,153] vs. 16 [7–157]; P < 0.0001) (Table 3 and Fig. 3).

Table 3 Results of univariate analysis
Full size table
Fig. 3
figure 3

Scatter plots of liver stiffness measured by MRE, AST, ALT, alpha-fetoprotein and PIVCA-II between “with HCC group” and control

Full size image

Risk of HCC development: multivariate analysis

Multivariate analysis revealed that only liver stiffness measured by MRE was a significant predictor of HCC development, with an odds ratio (95 % CI) of 1.38 (1.05–1.84) (Table 4, Fig. 4).

Table 4 Results of multivariate analysis
Full size table
Fig. 4
figure 4

a A 70-year-old woman with type C hepatitis and moderately differentiated hepatocellular carcinoma in hepatic segment 6. The liver shows as a red area on the fusion image of MRE and the T1-weighted image. The liver shows a stiffness value of 7.2 kPa. b A 73-year-old man with type C hepatitis. No hepatocellular carcinoma was observed in the liver. The stiffness value of the liver was 2.2 kPa

Full size image

Discussion

Magnetic resonance elastography is a recently developed and promising tool for measuring liver stiffness. The further the liver fibrosis progression, the stiffer the liver. Indeed, preliminary results showed excellent correlation between fibrosis stages and liver stiffness measured by MRE [4, 7, 8]. One of these studies also revealed that MRE has a better diagnostic accuracy than ultrasound transient elastography [17]. Several other recent studies also confirmed sufficient repeatability and reproducibility of this new technique [6, 18–20].

Various risk factors other than liver fibrosis have been reported for HCC development: older age, male gender, heavy alcohol intake, cirrhosis, lower platelet count, high serum AFP level, low serum albumin level and high serum ALT level [2, 21–23]. Of these, many researchers consistently agree that patients’ age is one of the strongest risk factors for HCC development [24–27]. A previous study using ultrasound transient elastography revealed that liver stiffness is an independent indicator for HCC development compared with age-matched controls [10]. A previous case–control study using MRE, however, suggested that liver stiffness measured by MRE was not significantly different between cirrhotic patients with and without HCC [28]. What made that difference? We suppose that the most important cause of that difference was the patients’ population included in those studies. In general, MR examination for screening for HCC is planned for the relatively advanced stages of fibrosis, whereas ultrasound is a primary examination for patients with chronic liver disease [29]. This may imply that the patient population of a study using MRI tends to have only a progressed stage of fibrosis, which can be confirmed by comparing the scattered plot graphs of those studies [27]. If the fibrosis stage were not so varied, the effect of liver stiffness on HCC development might be obscured by other known risk factors including age and sex. In our study, we tried to eliminate the noise of the other risk factor (age) to clarify the effect of liver stiffness by using an age-matched control group.

Radiologists should consider risk factors for cancer development in daily practice, because they are expected to evaluate images that serve as a basis for patient management, e.g. to suggest how often imaging should be performed. On the other hand, risk factors represent good opportunities for patient education and information. Screening the liver with imaging techniques is mandatory to identify HCC in its early stages and improve the prognosis of patients with chronic liver disease. The guidelines proposed by the American Association for the Study of Liver Disease recommend surveillance using ultrasound with an interval of 6 months [14]. According to the guidelines of the Japanese Liver Cancer Study Group, surveillance combined with dynamic enhanced computed tomography or MR imaging is recommended for patients with liver cirrhosis with type B or C hepatitis, which is considered a very high risk factor for HCC development [29]. Furthermore, several recent studies consistently revealed that gadoxetic acid-enhanced MR imaging is the best tool for detecting early-stage HCC [30–33]. It is mandatory to appropriately select the patients at very high risk of HCC or cirrhosis for screening using MR imaging. From our results, liver stiffness on MRE is an independent risk factor for HCC development. This suggests that MRE might have a potential utility for selecting patients at high risk of HCC for a subsequent MR or CT examination within a short interval. That combined with the usefulness of identifying early-stage HCC with gadoxetic acid places MRE imaging based-surveillance as being of good value for improving prognosis of patients with a risk of HCC or chronic liver disease.

The major limitation of our study was its retrospective design. The most appropriate method for assessing risk factors is a prospective study with longitudinal observation of a sufficient number of patients, in which incidence of HCC would be set as an endpoint. A future study with uniform subjects and a prospective study design would be necessary to preserve evidence of the usefulness of the liver stiffness measured by MRE as a risk factor for HCC. We did not think it worthwhile for radiologists to include various clinical variables, which have been previously suggested as risk factors for HCC development, e.g. body mass index, diabetes, heavy alcohol intake and fatty liver disease. We believe, however, that evaluation of liver function tests (blood test) and alpha-fetoprotein, easily available for the radiologist consultation through the patients’ charts, are of clinical relevance in determining the pretest likelihood of the patient being positive for HCCs. Another possible limitation is that we did not assess inflammation in the liver. Previous results suggested that only inflammation can increase the stiffness value of the liver even if no fibrosis develops [34]. We should be aware that MRE does not measure fibrosis but stiffness instead.

In conclusion, our retrospective case–control study revealed that liver stiffness measured by MRE is an risk factor for HCC in patients with chronic liver disease.