Introduction

Hepatocellular carcinoma (HCC) is a leading cause of cancer-related death worldwide, and early detection decreases the risk of mortality. Typically, early HCCs are small hypovascular nodules, and advanced HCCs are hypervascular [1]. However, the shared features of hypovascularity during the arterial phase of benign dysplastic nodules (DN) and malignant early HCC make their differentiation challenging on conventional dynamic imaging [2, 3].

Gadoxetic acid has been recently launched as a liver-specific contrast agent for magnetic resonance (MR) imaging [4]. Its utility in the diagnosis of HCC is well studied [5–7], especially in the diagnosis of hypovascular HCC [8–12]. Because HCCs typically do not take up gadoxetic acid, it is used as a contrast agent to enhance hepatocyte phase (HP) imaging to differentiate early HCCs from benign nodules even when features of hypervascularity are not apparent [13]. Thus, the evaluation of its uptake on HP images may improve the detection and characterization of hepatic nodules, particularly borderline early-stage nodules, including DN and early HCC. A recent longitudinal observational study demonstrated that hypovascular hepatic nodules that show hypointensity on gadoxetic acid-enhanced HP images are likely to become hypervascular HCCs within a few years [14–19].

Often detected in clinical practice, hypovascular nodules that show hyperintensity on gadoxetic acid-enhanced HP images have been considered benign and have been shown little attention, and little is known about their outcome. However, a recent report suggested that hypovascular HCCs that show iso- or hyperintensity on HP images [20] may require more intense clinical management than the current practice, including treatment or close follow-up. We therefore undertook the evaluation of the longitudinal risk to patients with cirrhosis of developing HCC from hypovascular hepatic nodules that show positive uptake of gadoxetic acid (hyperintensity) on HP imaging.

Materials and methods

Patients

The ethics committee of our institute approved this retrospective cohort study and waived the need for written informed consent.

We selected subjects from 371 consecutive patients with chronic liver disease who underwent gadoxetic acid-enhanced MR imaging (EOB-MRI) at our institution between 1 January and 31 December 2008. Patients were included for study who demonstrated target nodules in the liver on EOB-MRI (detailed later) and had undergone multiphasic contrast-enhanced computed tomography (CE-CT) and blood testing within one month of the initial MR examination and at least one follow-up EOB-MRI before August 2011 (Fig. 1).

Fig. 1
figure 1

Flow chart of patient inclusion

Full size image

We identified 633 hypovascular nodules that showed hyperintensity on HP images in 69 patients (38 men, mean age, 67.1 ± 10.1 years; 31 women, mean age, 69.2 ± 9.4 years) with chronic liver disease from hepatitis C virus infection (46 patients), hepatitis B virus infection (ten patients), both hepatitis B and C infection (one patient), and other causes (12 patients). Maximum nodule diameters measured 5 to 30 mm (mean ± SD, 8.3 ± 3.6 mm).

Clinical data

We recorded the results of blood tests for serum albumin, total bilirubin, alpha-fetoprotein (AFP), and protein induced by vitamin K absence or antagonist II (PIVKA-II), prothrombin ratio, and platelet counts as well as any history of conventional HCC.

MR imaging parameters

EOB-MRI was performed using a 1.5-tesla superconducting system (Signa™ Excite HD, GE Healthcare, Waukesha, WI, USA) and an eight-channel phased array coil. Sequences obtained for all patients (Table 1) included fast spin-echo T2-weighted imaging (T2WI), dual echo gradient-echo T1-weighted imaging (T1WI), and dynamic gadoxetic acid-enhanced dynamic MR imaging, including hepatic arterial phase and HP images.

Table 1 Magnetic resonance (MR) imaging parameters
Full size table

Computed tomography parameters

Contrast-enhanced CT was performed using a 16-detector row unit (Aquilion™ 16, Toshiba Medical Systems, Tochigi, Japan) with tube voltage of 120 kVp and tube current of 280 to 400 mA automatically adjusted to the patient’s body type. All CE-CT studies utilized nonionic iodine contrast agents (Omnipaque 300, Daiichi-Sankyo, Tokyo, Japan; Iomeron 350, Eisai Global, Tokyo, Japan; andIopamiron 370, Bayer Healthcare Japan, Osaka, Japan) that were rapidly administered intravenously during 30 seconds using a power injector (Auto Enhance A-50; Nemoto Kyorindo, Tokyo, Japan). After initiation of the injection of the contrast agent, images were obtained at 40 seconds during the hepatic arterial phase, 70 seconds at the portal-venous phase, and 180 seconds at the delayed phase.

Computed tomography during hepatic arteriography (CTHA) and during arterial portography (CTAP) were performed using the same scanner. Images with an effective section thickness of 5 mm were reconstructed every 5 mm to provide contiguous sections. Nonionic contrast agent (Omnipaque 300, Daiichi-Sankyo, Tokyo, Japan; Iomeron 350, Eisai Global, Tokyo, Japan; or Iopamiron 370, Bayer Healthcare Japan, Osaka, Japan) was diluted with saline (20 mL of contrast agent diluted with saline to 60 mL for CTHA and 30 mL of contrast agent diluted with saline to 90 mL for CTAP) and injected into the common hepatic artery (CTHA) or superior mesenteric artery (CTAP). Images were obtained 10 and 35 seconds after initiation of the injection of contrast agent for CTHA and 30 to 35 seconds after for CTAP.

Determination of target nodules

Two radiologists with 11 and 13 years of experience in abdominal radiology reviewed all the EOB-MRI and CE-CT images to identify hypovascular nodules that showed hyperintensity (uptake of gadoxetic acid) on HP images. They selected nodules (1) delineated from the surrounding liver as round lesions with hyperintensity on both axial and sagittal HP images, (2) 5 mm or larger in maximum diameter, and (3) showing hypovascularity during hepatic arterial phase on EOB-MRI and CE-CT.

We excluded nodules that showed hyperintensity compared to the surrounding liver even on precontrast T1WI regardless of their uptake of gadoxetic acid, because of the difficulty or impossibility of precisely determining a lack of contrast enhancement on hepatic arterial phase images of EOB-MRI and of detecting hypervascularization on follow-up hepatic arterial phase images. Hypovascularity of the nodules was confirmed by the lack of intralesional contrast enhancement on hepatic arterial phase images of both EOB-MRI and CE-CT [21].

Baseline imaging characteristics of the target nodules

We visually classified the signal intensity of nodules on T2WI as high or not compared with the signal of the liver parenchyma, categorized signal on T1WI as hypointense if nodules showed low intensity on both in- and opposed-phase images, and considered decreased signal on opposed-phase T1WI to represent a lipid component. Nodule size was defined as the maximum diameter in the axial plane as measured on HP images.

We performed longitudinal evaluation of the included nodules based on findings of follow-up (second as well as third and fourth, if available) EOB-MRI or CE-CT. The follow-up interval was 3 to 4 months according to the Japanese clinical guideline for the diagnosis and treatment of HCC [22]. Increase in nodule size by 3 mm on follow-up HP images of EOB-MRI indicated increased size during the follow-up period.

Endpoint

We applied the primary endpoint, the development of an arterial supply within the nodules (i.e., hypervascularization), only if HCC was subsequently diagnosed.

We initially confirmed hypervascularization within the nodules using follow-up hepatic arterial phase images of EOB-MRI or CE-CT and added CTHA and CTAP when hypervascularization was ambiguous on hepatic arterial phase with EOB-MRI or CE-CT [23]. If hypervascularization within the nodules could be confirmed with more than one modality, we recorded the earliest date at which hypervascularization within the nodules was confirmed on hepatic arterial phase images as the time of onset of the endpoint.

Statistical analyses

We used the Cox proportional hazard model for univariate analysis to examine the hazard ratios of potential risk factors for the development of HCC from the nodules and for multivariate analysis to test further those patient- and nodule-based variables that showed a significant hazard ratio on univariate analysis. We used Kaplan-Meier time-to-event curves with 95 % confidence interval to estimate the cumulative risk for HCC to develop from the nodules and the log-rank test to compare the curves between those nodules with maximum diameters of 10 mm and larger and those smaller than 10 mm. Data analysis was performed using R version 3.1.2, with a 2-sided p < 0.05 considered to be statistically significant.

Results

Hypervascularization and clinical diagnosis of hepatocellular carcinoma

We observed hypervascularization in six of the 633 nodules (0.9 %) during the follow-up period. Two nodules clearly demonstrated hypervascularization on hepatic arterial phase of CE-CT. Four nodules required additional CTHA and CTAP to confirm hypervascularization that was ambiguous on the hepatic arterial phase of either CE-CT (one case) or EOB-MRI (one case), or both (two cases). All the hypervascularized nodules were clinically diagnosed as HCCs. Three nodules that fulfilled the criteria of wash-in and wash-out on CE-CT proposed by the American Association for the Study of Liver Disease (AASLD) [24] and European Association for the Study of the Liver [25] were treated with transcatheter chemoembolization. Three other nodules that showed hypervascularity on CTHA and decreased portal perfusion on CTAP [26–28] were treated with transcatheter chemoembolization using iodized oil, with confirmation of oil accumulation a week after treatment [29]. The median time to hypervascularization was 679 days (range, 110 to 1150 days).

Risk factors

Univariate analysis showed no patient-based factor as a risk for the development of hypervascular HCC from target nodules (Table 2). Of the 633 nodules, 39 increased in size (mean increase, 4.6 mm; range, 3 to 16 mm). Nodule-based univariate analyses associated a significant risk for HCC to develop with large nodule size at initial examination (hazard ratio [95 % confidence interval [CI], 1.33 [1.17 to 1.51]), increase in nodule size during the follow-up period (hazard ratio [95 % CI], 8.30 [1.68 to 41.14]), and high signal intensity of nodules on T2WI (hazard ratio [95 % CI], 11.50 [2.10 to 62.93]). Multivariate Cox analysis indicated nodule size at first examination (p = 0.009) as the only independent risk factor for the development of hypervascular HCC, with an odds ratio of 1.25 (95 % CI, 1.06 to 1.47) (Table 3). No nodule showed low signal intensity on T1WI or evidence of lipid content.

Table 2 Univariate analysis: patient-based results
Full size table
Table 3 Univariate and multivariate analyses: nodule-based results
Full size table

The cutoff size of 10 mm corresponded with 100 % (6/6) sensitivity and 73.5 % (461/627) specificity to discriminate nodules with and without development of hypervascular HCC.

Cumulative risk of the development of hepatocellular carcinoma from the target nodule

At one year, the patient-based risk was 4.08 % (95 % CI, 1.74 to 9.55), and the nodule-based risk was 0.44 % (95 % CI, 0.20 to 0.95). The Kaplan-Meier curves differed significantly for nodules 10 mm and larger and those smaller than 10 mm (p = 0.0022; Fig. 2). The one-year risk of developing hypervascular HCC was 1.31 % [0.56 to 3.07] for the larger nodules and only 0.10 % [0.02 to 0.57] for those smaller than 10 mm (Figs. 2 and 3).

Fig. 2
figure 2

Longitudinal rates of hypervascular hepatocellular carcinoma (HCC) developing from target nodules (hypovascular nodules that showed uptake of gadoxetic acid on hepatocyte-phase images). Nodules with maximum diameters of 10 mm and larger showed significantly higher incidence of HCC development over time than those smaller than 10 mm (p = 0.002). However, even the larger nodules developed no HCC for the first year, which suggests no need to intensify follow-up for these nodules

Full size image
Fig. 3
figure 3

A 67-year-old man with a hypervascularized nodule after 2 years of follow-up. Top images are baseline studies (initial gadoxetic acid-enhanced magnetic resonance imaging [EOB-MRI] and computed tomography during hepatic arteriography and arterial portography [CTHA/CTAP] around the same time). The nodule with 24-mm diameter is hypovascular on hepatic arterial phase images of EOB-MRI and CTHA (arrow) and showed hyperintensity on hepatocyte phase (HP) images. Two years later (bottom images); hypervascularization was apparent within the nodule (arrow), which was confirmed on CTHA. Decreased portal perfusion in the nodule, i.e. hypodensity on CTAP, suggests the development of hypervascular hepatocellular carcinoma

Full size image

Discussion

This retrospective longitudinal study revealed that hypovascular hepatic nodules that showed hyperintensity on gadoxetic acid-enhanced HP images rarely developed into hypervascular HCC. Though the risk was higher for larger nodules, we observed a low risk even for those measuring 10 mm and larger.

Arterial-phase imaging of the cirrhotic liver may demonstrate numerous hypovascular nodules, including regenerative and dysplastic nodules and early HCC [1]. Hypointensity on gadoxetic-acid HP images has been well reported as an indicator of malignancy or HCC [12, 13, 30] or at least high-grade dysplastic nodule [31]. The one-year risk of hypervascular HCC developing within a hypovascular nodule that shows hypointensity on initial HP images ranges from 15.6 to 77.3 % [14–17, 19]. We found the risk minimal for hypovascular nodules that showed hyperintensity on HP images compared with those showing hypointensity. Further, the reported risk for the development of HCC in the cirrhotic liver (even without nodules) is around 8 % at one year [32] and 13 % at 3 years [33] in patients with no previous HCC. The AASLD guidelines for the management of HCC recommend ultrasonographic screening of cirrhotic liver every 6 months [24]. Considering the low risk we observed for the development of HCC from the target nodule, we believe these hypovascular, hyperintense nodules require neither treatment nor extensive follow-up beyond the AASLD recommendations.

HCCs typically show hypointensity on HP images [34], but this is not always the case [35]. The signal intensity of HCC during HP is well correlated with the expression of organic anion transporter (OATP) [36]. Recent publications revealed the favorable prognoses of some typical (hypervascular) HCCs that expressed OATP on the surface of tumour cells and showed hyperintensity during HP imaging [37, 38]. Although hypovascular HCCs can show increased uptake of gadoxetic acid during HP [20], our results suggest its uptake indicates favorable outcomes.

Our univariate analysis showed no patient-based factor as a risk for the development of hypervascular HCC from target nodules, though others have associated patient-based factors with both occurrence or recurrence of HCC [39–43]. Indeed, previous local therapy for HCC, Child-Pugh Class B cirrhosis, and coexistence of hypervascular HCC have been associated with an increased risk of hypervascularization in hypovascular nodules that demonstrated hypointensity on HP images [16]. Thus, we supposed that hypervascularization might be less aggressive in those hypovascular nodules with uptake of gadoxetic acid than those without. This point requires further investigation.

Our study has several limitations, including its retrospective design, which yielded varying intervals of imaging-based follow-up, and our relatively small number of patients, some with multiple nodules that produced a clustering effect in nodule-based results. Nevertheless, our consistent results in patient- and nodule-based analyses may indicate that any clustering effect was minimal.

In conclusion, hypovascular, hyperintense hepatic nodules rarely develop into hypervascular hepatocellular carcinomas. Though that risk is higher for larger nodules, even those 10 mm and larger require no altered management for patients with cirrhosis.