Introduction

In patients with hepatocellular carcinoma (HCC), internationally validated criteria help to select the relevant candidates for liver transplantation (LT) [1,2,3]. The Milan criteria (solitary tumor ≤ 5 cm or up to three tumors ≤ 3 cm in diameter) have been widely adopted [4]. Although many centers have proposed extended selection criteria for the eligibility of patients with HCC for living donor LT (LDLT) and have reported favorable and comparable outcomes [5,6,7,8], the United Network of Organ Sharing (UNOS) and majority of transplant programs continue to use the Milan criteria as a prioritization tool [9].

In addition to staging, tumor differentiation and microvascular invasion (MVI) are well-known independent predictors of HCC recurrence after LT [10,11,12]. However, these pathologic markers of tumor aggressiveness are evaluated in the explanted liver. Preoperative markers of tumor behavior would help patient selection prior to LT, in addition to tumor size and number. Serum alpha-fetoprotein (AFP), protein induced by vitamin K absence or antagonist-II (PIVKA-II), and positron emission tomography (PET) using 18F-fluorodeoxyglucose have been proposed as predictors of tumor behavior and post-transplant tumor recurrence in LT candidates with HCC [13,14,15].

Magnetic resonance imaging (MRI) with hepatobiliary agents is the most sensitive method for HCC detection [16,17,18]. In addition to accurate diagnosis and morphological staging, preoperative MRI features with hepatobiliary agents have been shown to reflect tumor biology [19,20,21]. Furthermore, a recent single-center study including both treatment-naïve and treated patients suggested that MRI findings could predict HCC recurrence after LDLT [22]. However, little information is available regarding the value of preoperative gadoxetic acid–enhanced MRI as a predictor of HCC recurrence after primary LDLT in treatment-naïve patients. In this multicenter retrospective study, therefore, we investigated the value of preoperative gadoxetic acid–enhanced MRI for evaluating tumor staging and predicting HCC recurrence after primary LDLT.

Materials and methods

This multicenter study was approved by the Institutional Review Board of each participating center and the requirements for informed consent were waived.

Patients

From a prospectively acquired database of LT conducted at two academic referral institutions, we retrieved 812 adult patients (≥ 18 years) who underwent LDLT for HCC without extrahepatic metastasis between January 2009 and December 2013. The inclusion criteria for the study were as follows: (1) treatment-naïve HCCs (i.e., no history of previous HCC treatment); (2) preoperative gadoxetic acid–enhanced liver MRI evaluation within 3 months prior to LT; (3) absence of macrovascular invasion on pre-transplant imaging evaluation; and (4) available tumor marker data, including serum AFP and PIVKA-II levels, measured within 3 months prior to LT. Patients who died of graft failure or complications within 1-month post-transplantation were excluded. Finally, 122 patients met the criteria and were included in this study (Fig. 1). The mean interval between MRI and LT was 43.5 days (range, 0–90 days).

Fig. 1
figure 1

Flowchart of the study population. LDLT, living donor liver transplantation; HCC, hepatocellular carcinoma; MRI, magnetic resonance imaging; LT, liver transplantation; AFP, alpha-fetoprotein; PIVKA-II, protein induced by vitamin K absence or antagonist-II.

Full size image

MRI examination

MRI examinations were performed with a 1.5-T (Magnetom Avanto, Siemens Healthineers; and Intera Achieva, Philips Healthcare) or a 3-T (Magnetom Tim Trio, Siemens Healthineers; and Intera Achieva, Ingenia, or Ingenia CX, Philips Healthcare) scanner. Liver MRI protocol included dual-echo spoiled gradient-echo T1-weighted in-phase and opposed-phase imaging, multi-shot and single-shot turbo spin-echo T2-weighted imaging, and dynamic T1-weighted imaging. For gadoxetic acid (Primovist, Bayer Pharma AG)–enhanced MRI, the following images were obtained using a chemically selective fat-suppressed three-dimensional gradient-echo T1-weighted: unenhanced, enhanced arterial phase (AP, 20–35 s), portal venous phase (PVP, 60 s), delayed phase (DP, 3 min), and 20-min hepatobiliary phase (HBP). The contrast agent was automatically administered intravenously using a power injector at a total dose of 0.025 mmol/kg with a rate of 1 mL/s, followed by a 20-mL saline flush. The detailed parameters of the MRI sequences are provided in Supplementary Table 1.

Pre-transplant imaging and laboratory analysis

Two abdominal radiologists (S.L. and K.W.K., with 7 and 13 years of experience in liver imaging, respectively) retrospectively reviewed the MR images. The reviewers were aware of the presence of HCC but were blinded to the clinical, laboratory, pathologic, and follow-up results. HCC diagnoses were based on the identification of the typical hallmarks of HCC, considering larger than 1 cm with the combination of hyperenhancement in the AP and washout in the PVP on gadoxetic acid–enhanced liver MRI [1]. Using pre-transplant MRI, the sizes and numbers of tumors were assessed and HCC within or beyond the Milan criteria was determined. Additionally, the tumor distribution (unilobar vs. bilobar) was assessed in cases with multiple tumors. The reviewers also evaluated the following MRI features: (1) tumor margin, categorized as smooth (simple nodular tumors with smooth contours) or non-smooth (multinodular confluent or nodular with extranodular growth) [23]; (2) satellite nodule (smaller nodules within 2 cm of the main tumor) [24]; (3) peritumoral enhancement on AP (detectable portion of crescent or polygonal shaped enhancement outside the tumor margin with broad contact to the tumor border on AP, becoming isointense with background liver parenchyma on DP) [25]; and (4) peritumoral hypointensity on HBP (wedge-shaped or flame-like hypointense area of hepatic parenchyma located outside of the tumor margin on HBP) [19]. In cases of multiple tumors, the largest tumor was selected for the evaluation of these MRI features. After independent image review, discordant results between the reviewers were adjusted by consensus review and consensus data were used for analysis.

The neutrophil-to-lymphocyte ratio (NLR) was obtained by dividing the absolute neutrophil count by the absolute lymphocyte count according to pre-transplant laboratory data. An NLR ≥ 5 has been established as an appropriate cutoff according to the published literature [26, 27]. The mean interval between MRI and laboratory data of neutrophil count or lymphocyte count and for Model for End-Stage Liver Disease (MELD) score was 42.5 days (range, 0–89 days).

Based on previously published studies, we used an AFP level of > 200 ng/mL [27, 28] and PIVKA-II level of > 300 mAU/m [29,30,31] as cutoffs indicating marked elevation.

Histopathologic analysis

Pathologic diagnosis was evaluated including tumor size, number, histologic grade, and MVI by two liver pathologists. We also determined HCC within or beyond the Milan criteria according to the pathology of the explanted liver. Histologic grade was classified according to the Edmondson–Steiner grading system [32]. MVI was defined as a tumor within a vascular space lined by the endothelium that was visible only on microscopy [33].

Follow-up after liver transplantation

Every 3–6 months during the first 2 years, the patients were surveyed for HCC recurrence via measurements of serum AFP and PIVKA-II levels and dynamic liver computed tomography (CT) scans. Thereafter, the patients underwent clinical assessments every 3 months and imaging studies once or twice per year. MRI was performed when a follow-up CT suggested recurrence. If an extrahepatic recurrence was suspected based on clinical symptoms or an unexplained elevation of a tumor marker level, the patient was referred to chest CT, whole-body bone scintigraphy, and 18F-fluorodeoxyglucose PET. HCC recurrence was diagnosed when imaging studies revealed evidence of new tumors (n = 14). When new focal lesions in the transplanted liver did not show the typical hallmarks of HCC (the combination of hyperenhancement in the AP and washout in the PVP or DP) on dynamic liver CT or MRI, biopsy was performed for pathologic confirmation (n = 2). The median follow-up period was 69.1 months (interquartile range, 55.2–88.8 months).

Statistical analysis

Interobserver agreement between the two reviewers assessing the MRI findings was evaluated by using the Cohen κ coefficient. The κ value (the level of agreement) was defined as follows: κ < 0.21, poor; κ = 0.21–0.40, fair; κ = 0.41–0.60, moderate; κ = 0.61–0.80, good; and κ > 0.80, excellent agreement. Disease-free survival (DFS) was defined as the interval between the date of LT and that of intra- or extrahepatic recurrence. The cumulative DFS rates were estimated using the Kaplan–Meier method, and the differences between curves of four subgroups (all combinations of “within/beyond Milan criteria” and “with/without peritumoral hypointensity on HBP”) were evaluated using the log-rank test. A Cox proportional hazards regression analysis was used to assess the pre-transplant imaging and laboratory findings associated with HCC recurrence. Variables with a p value of < 0.05 in the univariable analysis were included in the multivariable model. Categorical variables were analyzed using Fisher’s exact test. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 23.0 (IBM Corp.). A p value of < 0.05 was considered statistically significant.

Results

Patient characteristics and pre-transplant MRI findings

Baseline characteristics of the patients are presented in Table 1. There were 100/22 patients within/beyond the Milan criteria. Twenty-four patients (19.7%) had bilobar distribution, 52 (42.6%) had non-smooth tumor margin, 12 (9.8%) had satellite nodules, 25 (20.5%) had peritumoral enhancement on AP, and 21 (17.2%) had peritumoral hypointensity on HBP.

Table 1 Clinical characteristics of the study populations
Full size table

Interobserver agreement for MRI findings

Interobserver agreement MRI findings were good or excellent (κ = 0.809 for beyond the Milan criteria, κ = 0.849 for bilobar distribution, κ = 0.783 for non-smooth tumor margin, κ = 0.739 for satellite nodule, κ = 0.770 for peritumoral enhancement on AP, and 0.755 for peritumoral hypointensity on HBP).

Predictors for HCC recurrence from pre-transplant MRI

HCC recurrence was detected in 13.1% (16 of 122). The initial sites of recurrence were as follows: liver only, 6 (37.5%); lung only, 3 (18.6%); adrenal only, 2 (12.5%); lymph node only, 2 (12.5%); bone only, 1 (6.3%); peritoneum only, 1 (6.3%); and liver and lung, 1 (6.3%). In the multivariable analysis, beyond the Milan criteria (HR, 3.54; 95% CI, 1.13–11.12; p = 0.030) and peritumoral hypointensity on HBP (HR, 18.30; 95% CI, 4.33–77.34; p < 0.001) were significantly independent predictors of HCC recurrence (Table 2).

Table 2 Cox proportional hazard models of pre-transplant data in predicting HCC recurrence
Full size table

Comparison of pre-transplant MRI with pathology

Compared with pathology, which was the reference standard, 10 of 100 patients who were preoperatively categorized as within the Milan criteria were underestimated. In contrast, two of 22 patients who were preoperatively diagnosed as beyond the Milan criteria were overestimated. The use of pre-transplant MRI to categorize within or beyond the Milan criteria yielded an accuracy of 90.2%.

Peritumoral hypointensity on HBP was significantly associated with a worse tumor grade (p = 0.010) and MVI (p < 0.001) (Table 3).

Table 3 Relationship between peritumoral hypointensity on HBP and explanted pathologic features of the tumor
Full size table

Disease-free survival

The overall 1-, 3-, 5-, and 7-year cumulative DFS rates were 93.3%, 90.7%, 88.9%, and 86.1%, respectively. DFS according to the Milan criteria and peritumoral hypointensity on HBP of pre-transplant MRI is shown in Fig. 2. Patients within the Milan criteria had a significantly higher DFS than those beyond the Milan criteria (93.7% vs. 67.9% at 5 years; p < 0.001; Fig. 2a). Patients without peritumoral hypointensity on HBP demonstrated significantly better DFS compared with those with peritumoral hypointensity on HBP (5-year DFS rate, 96.5% vs. 45.0%; p < 0.001; Fig. 2b).

Fig. 2
figure 2

Disease-free survival curves comparing liver transplantation (LT) recipients for hepatocellular carcinoma (HCC) within and beyond the Milan criteria (a) and with and without peritumoral hypointensity on hepatobiliary phase (HBP) (b) of pre-transplant gadoxetic acid–enhanced magnetic resonance imaging (MRI)

Full size image

Stratification of patients according to the Milan criteria (within or beyond) and peritumoral hypointensity on HBP (hypointensity or non-hypointensity) was as follows: in group 1, 87 patients were within/non-hypointensity; in group 2, 14 were beyond/non-hypointensity; in group 3, 13 were within/hypointensity; and in group 4, 8 were beyond/hypointensity. Interestingly, patients within the Milan criteria and peritumoral hypointensity on HBP had a higher risk for recurrence than those beyond the Milan criteria who did not have peritumoral hypointensity on HBP, although this difference was not statistically significant (p = 0.086; Fig. 3a). DFS of patients with HCC beyond the Milan criteria and non-peritumoral hypointensity on HBP was not different in patients within the Milan criteria (p = 0.438) and was significantly better than that of patients with HCC beyond the Milan criteria and peritumoral hypointensity on HBP (p < 0.001; Fig. 3b).

Fig. 3
figure 3

Disease-free survival curves comparing LT recipients stratified into four groups using a combination of independent factors (Milan criteria and peritumoral hypointensity on HBP of pre-transplant gadoxetic acid-enhanced MRI) (a) and within the Milan criteria, beyond the Milan criteria without peritumoral hypointensity on HBP, and beyond the Milan criteria with peritumoral hypointensity on HBP (b)

Full size image

Discussion

Among the available pre-transplant data, we identified MRI findings, including beyond the Milan criteria and peritumoral hypointensity on HBP, as independent predictors of HCC recurrence after LT. Categorization of the Milan criteria using pre-transplant gadoxetic acid–enhanced MRI yielded a 90.2% accuracy compared with explanted liver. In addition, this study showed that peritumoral hypointensity on HBP was associated with an increased risk of recurrence, which seems to correlate with a worse tumor grade and the presence of MVI in the explanted pathology.

According to previously published data, dynamic CT or MRI with extracellular agents yields sensitivity rates of approximately 50–80% for the pre-transplant detection of HCC in correlation with the whole-explant liver [34, 35]. One prior study indicated that pre-transplant radiological staging based on dynamic CT or MRI with extracellular agents may underestimate or overestimate the tumor burden in up to 25% of cases compared with pathologic staging of the explanted liver [36]. In contrast, accumulating evidence suggests that MRI with hepatobiliary agents is the most sensitive method for detecting HCCs and the addition of HBP images improves the sensitivity for HCC by 6–15% on a per-lesion basis [24, 37,38,39]. Recent incremental advances in MRI techniques have enabled the detection of small lesions, which could not be easily identified via imaging even a decade earlier [40]. Gadoxetic acid–enhanced MRI is very effective for the detection and diagnosis of HCC, and a recent study of patients with treatment-naïve HCCs reported an accuracy rate of 92.1% for patient allocation based on the Milan criteria and UNOS guidelines [41]. Our rate of 90.2% was consistent with this study.

Several studies have reported that parameters of tumor biology, such as the tumor grade and MVI, can significantly influence patient outcomes after LT [10,11,12]. Instead, several studies have investigated preoperative biological markers that are closely associated with the abovementioned pathologic features, including AFP, PIVKA-II, and PET [13, 14, 42]. In addition to its roles in diagnosis and accurate morphological staging, gadoxetic acid–enhanced MRI may be used to characterize tumor biology. Prior studies have demonstrated that peritumoral hypointensity on HBP was a significant predictor of MVI after curative resection or LT for HCC [19, 21, 22]. Similarly, the present study also found that peritumoral hypointensity on HBP correlated significantly with the tumor grade and MVI and thus reflected tumor biology. Potentially, the presence of peritumoral hypointensity on HBP could be attributed to the altered expression of organic anion-transporting polypeptide transporters of hepatocytes around the tumor, which are a consequence of hemodynamic changes associated with the tumor-mediated obstruction of minute portal veins [19, 21]. This result suggests that peritumoral hypointensity on HBP could serve as a preoperative parameter of biological tumor behavior in LT candidates with HCC.

In the present study, the combined use of the Milan criteria (as part of morphological information) and peritumoral hypointensity on HBP (a predictor of aggressiveness) improved the stratification of patients in terms of the risk for HCC recurrence after primary LT. Our study shows that pre-transplant gadoxetic acid–enhanced MRI has the potential to optimize the selection process for LT candidates with HCC. This might be particularly relevant for patients with tumors beyond the Milan criteria as the 5-year DFS rate of patients beyond the Milan criteria and without peritumoral hypointensity were comparable to those of patients within the Milan criteria and with peritumoral hypointensity on HBP.

Our study had several limitations. We included only patients who had undergone LT for treatment-naïve HCCs. In patients with previous treatments, it is not easy to evaluate the size and number of tumors (i.e., essential components of the Milan criteria) after treatment, and it is difficult to determine whether the Milan criteria are met. Previous treatment can also alter imaging features of the tumor and peritumoral parenchyma and make the accurate analysis of MRI findings impractical [43]. We included only LT recipients who underwent pre-transplant gadoxetic acid–enhanced MRI. This may have led to a selection bias because the specific reason for performing MRI prior to LT could not be ascertained owing to the retrospective nature of the study. Our cohort comprised mostly patients with hepatitis B (Figs. 4, 5, and 6), and therefore, our results might not be generalizable to other liver diseases.

Fig. 4
figure 4

A 50-year-old male with cirrhosis related to hepatitis B and who underwent liver transplantation (LT) for HCC. a In the arterial phase image after administration of gadoxetic acid, a 57-mm liver mass (arrow) showed hyperenhancement and was diagnosed as HCC beyond the Milan criteria on pre-transplant MRI. b In the hepatobiliary phase (HBP), the mass showed peritumoral hypointensity (arrowheads). On histopathology after LT, the tumor was classified as Edmondson grade III HCC with microvascular invasion. Tumor recurrence occurred 18 months after LT

Full size image
Fig. 5
figure 5

A 49-year-old female with cirrhosis related to hepatitis B and who underwent LT for HCC. a In the arterial phase image after administration of gadoxetic acid, a 47-mm liver mass (arrow) showed hyperenhancement and was diagnosed as HCC within the Milan criteria on pre-transplant MRI. b In the HBP, the mass showed peritumoral hypointensity (arrowheads). On histopathology after LT, the tumor was classified as Edmondson grade IV HCC with microvascular invasion. Tumor recurrence occurred 22 months after LT

Full size image
Fig. 6
figure 6

A 53-year-old male with cirrhosis related to hepatitis B and who underwent LT for HCC. a In the arterial phase image after administration of gadoxetic acid, 50-mm and 15-mm liver masses showed hyperenhancement and were diagnosed as HCCs beyond the Milan criteria on pre-transplant MRI. b In the HBP, the masses did not show peritumoral hypointensity. On histopathology after LT, the tumors were classified as Edmondson grade II HCCs without microvascular invasion. Tumor recurrence did not occur during 65 months of follow-up period after LT

Full size image

In conclusion, our study suggests that pre-transplant gadoxetic acid–enhanced MRI can assist in the HCC recurrence risk prediction, based on for the combination of morphological tumor staging and evaluation of tumor aggressiveness.