Combining 18F-FDG positron emission tomography with Up-to-seven criteria for selecting suitable liver transplant patients with advanced hepatocellular carcinoma

The Up-to-seven (UTS) criteria (sum of tumor size and number not exceeding 7) for indicating liver transplantation (LT) in hepatocellular carcinoma (HCC) were originally based on explant pathology features and absence of microvascular invasion (MVI). 18F-fludeoxyglucose (18F-FDG) positron emission tomography (PET) was shown to indicate the risk of MVI and tumor recurrence. The aim of this study was to analyze the prognostic significance of the clinical UTS criteria when being combined with PET-status of the tumor. Data of 116 liver transplant patients were subject to retrospective analysis. Five-year recurrence-free survival (RFS) rates in patients meeting (n = 85) and exceeding (n = 21) the radiographic UTS criteria were 81% and 55.1%, respectively (p = 0.014). In the UTS In subset, RFS was significantly better in PET-negative (94.9%) than in PET-positive patients (48.3%; p < 0.001). In the UTS Out subset, 5-year RFS rates were 87.1% and 19% in patients with non- 18F-FDG-avid and 18F-FDG-avid tumors (p < 0.001), respectively. Positive PET-status was identified as the only independent clinical predictor of tumor recurrence in beyond UTS patients (Hazard ratio [HR] 19.25; p < 0.001). Combining radiographic UTS criteria with FDG-PET may safely expand the HCC selection criteria for LT.

In 2009, Mazaferro et al. introduced the so-called Up-to-seven (UTS) criteria by simply combining the largest tumor nodule size and the number of HCC nodules, which should not exceed 7 in sum. In a large European multicenter trial including 1556 liver transplant recipients, they demonstrated that patients with HCC exceeding the MC but still meeting the UTS criteria have an excellent prognosis that was not different from patients meeting standard criteria HCC 17 . However, the authors have used post-LT generated histopathologic and not pre-LT radiographic features for their investigation 18 . Apart from that, beneficial outcome of the expanded HCC subset was related to lack of microvascular tumor invasion (MVI) 17 .
Tumor invasion into microscopic vessels is recognized as a major indicator of aggressive tumor behaviour and poor outcome 19,20 . However, it may reliably be assessed only on explant pathology and not by conventional radiographic imaging 21 . Therefore, for a safe application of the UTS criteria, the implementation of appropriate clinical biomarkers of tumor aggressiveness seems to be mandatory 20 .
In recent years, several transplant groups were able to demonstrate that 18 F-fludeoxyglucose ( 18 F-FDG) positron emission tomography (PET) provides useful information on metabolic tumor viability and posttransplant outcome [22][23][24] . PET-positivity was shown to correlate with presence of unfavourable histopathologic features, like MVI and poor grading 23,25,26 .
The primary aim of this retrospective study was to analyze the prognostic value of the UTS criteria when being based on pretransplant imaging. Apart from that, we investigated whether the combination of the radiographic UTS criteria with 18 F-FDG PET may be useful for predicting posttransplant tumor recurrence and, thus, for safely expanding the pool of suitable liver transplant patients.

Methods
Subjects. The study protocol was approved by the local Ethics Committee (Ethical committee of the Medical School, Technical University Munich, Nr. 217/15). Patients' registration, waiting list management and transplant procedures were performed according to national law and re-gulations. Prior to LT, all patients gave informed consent that follow-up data may be used for study purpose and respective publication. This work was supported by the German research Foundation (DFG) and the Technische Universität München within the funding programme Open Access Publishing.
From a prospectively updated data base (1996 to 2012; two-center study under same personal responsibilities and conditions), 116 patients that underwent LT for HCC were identified. Tumor diagnosis was established by clinical staging (radiographic imaging by computed tomography [CT] and/or magnetic resonance tomography [MRI] + alpha-fetoprotein level [AFP] measurement) without tumor biopsy. The minimum cut-off tumor nodule size for diagnostic purpose was 1 cm. The MC were primarily used for justifying patients' listing.
Since December 2007, patients with HCC meeting the MC received exceptional priority status according to the model of end-stage liver disease (MELD) score. Based on HCC topography and remaining liver function, transarterial chemotherapy (TACE) as bridging to LT has been performed. Pretransplant tumor surveillance consisted of liver ultrasound and AFP level determination every 6 weeks, and CT/MRI scan minimum biannually. Additional radiographic imaging was performed when required, such as post-TACE, prior to MELD score upgrading and in the case of tumor-related symptoms.
Progression of tumor load beyond the MC resulted in loss of MELD exceptional priority status. Apart from that, a concise re-evaluation by computed tomography (CT) and/or magnetic resonance tomography (MRI) and AFP level determination every 3 months was initiated. According to an individual decision making process, these patients were primarily scheduled for center-based liver allocation, unless macrovascular tumor infiltration, lymph node infiltration or extrahepatic tumor spread (biological tumor progression) became evident. 18 F-FDG PET was performed in all patients with liver malignancy in order to exclude extrahepatic tumor manifestation. The prospectively collected data were retrospectively used for the assessment of metabolic tumor properties.
As previously described, we distinguished between PET-positive (PET+ status; 18 F-FDG-avid) and PET-negative (PET-status; non-18 F-FDG-avid) tumors. This classification was based on concise visual FDG-uptake assessment of each tumor nodule in very close morphological demarcation to the surrounding non-tumorous liver regions. Any significantly enhanced 18 F-FDG uptake pattern compared to normal adjacent liver tissue (tumor to non-tumor maximum standard uptake value > 1) was indicating PET+ status of HCC 22,23 .
Based on final pretransplant radiographic staging, patients were classified as Milan In (HCC meeting the MC; Milan In) or Milan Out (HCC exceeding the MC; Milan Out), and UTS In (HCC meeting the UTS criteria; UTS In) and UTS Out (HCC exceeding the UTS criteria; UTS Out), respectively. Transplant procedure and posttransplant follow-up. ABO-compatible deceased donor liver grafts were used for transplant procedure in all study patients. Venous reconstruction was performed by using the piggy back technique without veno-venous bypass. In order to avoid the theoretical risk of systemic tumor cell spread, we have not used intraoperative blood salvage autotransfusion. Posttransplant immunosuppression consisted of a calcineurin inhibitor based regimen either by cyclosporine A or tacrolimus augmented by azathioprine or mycophenolate mofetil. Prednisone was withdrawn latest 3 months post-LT with exception of pre-existing autoimmune hepatitis. Tumor surveillance post-LT consisted of AFP-level determination and liver ultrasound at least every three months. Apart from that, CT scans of the chest and abdomen were performed every 6 months during the first posttransplant year and minimum thereafter or in case of suspected HCC relapse.
Statistical analysis. Categorical variables were compared using the χ 2 test. Continuous variables were recorded by median and range, and compared using the Student's t test.
The Kaplan-Meier method was performed to determine overall survival (OS) and recurrence-free survival (RFS) rates. Variables being significant for predicting HCC relapse in univariate analysis (p < 0.05) were entered into a stepwise multivariate logistic Cox regression model in order to identify independent prognostic factors (p < 0.05). Only pretransplant available clinical features were included in the analysis. All statistical analyses were performed by using the software SPSS 23.0 (IBM Inc., Munich, Germany).

Results
Clinicopathologic characteristics. The baseline clinicopathologic characteristics are summarized in  Table 1. Clinicopathologic characteristics of the study cohort (n = 116). *According to pretransplant radiographic imaging. 18  Corresponding RFS rates were 77.2% and 74.4% at 3 and 5 years, respectively. In univariate analysis, AFP level, multiple tumor nodules, maximum tumor diameter, number of HCC nodules, UTS criteria and PET-status were significantly associated with risk of HCC recurrence. Only positive PET-status, AFP level > 400 ng/nl and manifestation of multiple HCC nodules were identified as significant and independent predictors of HCC recurrence on multivariate analysis (Table 2).
Tumor recurrence rate was significantly lower, OS and RFS were both significantly better in Milan In patients as compared to the other two subsets. In contrast, there were no significant outcome differences between Milan Out/UTS In patients and patients with beyond UTS tumors (Table 5).
In the Milan Out/UTS In subgroup, 5-year RFS rates were 80% in PET-positive and 44.4% PET-negative patients (p = 0.078; Fig. 8). There was no significant difference in tumor-specific outcome between Milan In patients and patients with PET-negative Milan Out/UTS In tumors (p = 0.639; Fig. 9).
PET correlations with tumor characteristics and outcome. Table 6    Positive PET-status correlated significantly with tumor nodule size and total tumor diameter, whereas it tended to be associated with elevated AFP levels. Apart from that, enhanced 18 F-FDG uptake on PET was significantly correlated with presence of aggressive histopathologic features (Table 6).

Discussion
At first sight, our data seem to implicate that expansion to the clinical UTS criteria does not dramatically increase the risk of tumor recurrence. We found considerable 5year RFS rates in patients meeting the radiographic UTS criteria (81%), which was only slightly inferior to the Milan In cohort (86.2%). Apart from that, an additional of 19 patients beyond the MC could, thereby, be declared as being suitable for LT (Table 5), which corresponds to an increase of 19.6%. Insofar, our study seems to validate the UTS criteria on basis of clinical staging. However, in the additionally recruited patients with tumors exceeding the MC but still meeting the UTS criteria (n = 19), the tumor recurrence rate was significantly higher (36.8% versus 13.6%; p = 0.023), and OS (62.2% versus 81.7%; p = 0.001) and RFS rates (63.2% versus 86.2%; p = 0.017) were both significantly lower compared to the Milan In subset (Table 5). Probably, just because of the small sample size, this has not yet become noticeable in the overall survival. Therefore, on closer examination, our data rather indicate that the undifferentiated application of the UTS criteria enhances the oncological risk.
Whether our results finally justify the implementation of the UTS criteria is a matter of critical discussion. Liberalizing the transplant criteria without expanding the pool of available donor organs results in growing waiting lists and prolongation of pre-LT waiting times. This may in turn enhance the drop out risk following tumor    27 . In our Milan Out/UTS In cohort, 5-year OS and RFS rates were 61% and 63.2% (Table 5), respectively, which exceeded this cut-off value. However, a tumor relapse risk about 40% in an additionally selected expanded criteria HCC subset may be unacceptable, particularly in view of an escalating donor organ shortage 28 .
In recent years, several trials emphasized on the predictive power of the UTS criteria 17,[29][30][31] . Nevertheless, they have not yet been implemented as standard for patients' selection. It was a major limitation of these studies that pathomorphometric and not radiographic UTS features were used. Well-known discrepancies between pre-and posttransplant tumor staging may have limited clinical applicability 32 . Apart from that, absence of MVI contributed substantially to the prognostic value of the UTS criteria, which additionally hampered their implementation into clinical routine 17,[29][30][31] . In order to create a realistic clinical scenario, we were using only pretransplant available variables in our analysis. Nevertheless, our data indirectly confirmed the specific significance of MVI in this setting, as tumor recurrence rate was significantly higher in the Milan Out/UTS In patients compared to the Milan In subset when MVI was not considered (Table 5).
In fact, vascular tumor invasion is one of the most important prognostic features in LT for HCC 19,20 . While macrovascular infiltration may be appropriately detected by modern radiographic imaging and generally excludes patients from LT, MVI may reliably be confirmed only post-LT 33,34 . Pretransplant tumor biopsy is inappropriate due to high risk of sample errors caused by intratumoral heterogeneity 21 . Apart from that, there is a theoretical risk of tumor cell seeding that might affect posttransplant prognosis 35 .
Since tumor load correlates with risk of MVI and poor grading 36,37 , clinical surrogate markers of tumor aggressiveness are essential for safely expanding macromorphometric selection limits. In the past, response to   Table 4. Prognostic variables for HCC recurrence in UTS Out patients (n = 31). *According to pretransplant radiographic staging. AFP -alpha-fetoprotein. CI -confidence interval. HCC -hepatocellular carcinoma. LT -liver transplantation. MELD -model for end-stage liver disease. PET -positron emission tomography. TACE -transarterial chemoembolization. UTS -Up-to-seven criteria. TACE 38 , AFP-level 39 , C-reactive protein 40 , PIVKA II 41 , yglutamyltransferase 42 and 18 F-FDG PET 22,24 were applied for biological tumor evaluation. Among them, increased 18 F-FDG uptake on PET was shown to be highly associated with unfavorable histopathology and risk of tumor recurrence [43][44][45] . 18 F-FDG PET is a well-established non-invasive tool for metabolic imaging of different malignancies 46 . Like glucose, 18 F-FDG is uploaded by the tumor cells via several overexpressed glucose transporters. Well differentiated HCC nodules exhibit an enzyme activity that is not different to that of normal liver tissue, resulting in a similar FDG uptake pattern. On contrary, poorly differentiated HCCs are characterized by enhanced FDG uptake pattern as compared to the surrounding healthy liver regions. In contrast to several other cancers, 18 F-FDG-PET is therefore not appropriate for detection of HCC, but rather for evaluation of metabolic tumor viability 47 .
In fact, the application of 18 F-FDG PET was recently demonstrated to select suitable liver transplant patients with HCC beyond MC and UCSF criteria [48][49][50][51] . Although MVI plays an essential role for the prognostic reliability of the UTS criteria 17,[29][30][31] , the are no comparable investigations in this context.
Our study impressively confirmed that enhanced 18 F-FDG uptake on pretransplant PET is a valuable indicator of biological tumor aggressiveness and poor outcome ( Table 6). Positive PET-status was even identified as most powerful independent clinical predictor of HCC recurrence in our series (Table 2). Apart from that, we were able to demonstrate that combining the clinical UTS criteria with FDG-PET leads to an extremely low tumor relapse  However, by strictly adhering to this selection concept, more than 30% of our UTS In patients would have been excluded from LT, with almost half of them still being tumor-free alive after 5 years (Fig. 5). The number of liver transplants had thereby been reduced from originally 66 meeting standard criteria (Milan In) to 59 fulfilling the novel hybrid criteria set (UTS In + PET-negative). Rather, our data pointed out that the MC are excellent for selecting suitable liver transplant patients (Figs 1 and 2), whereas further biological tumor evaluation is necessary beyond the Milan boundaries ( Table 5).
As shown in Fig. 8, 18 F-FDG PET identifies those Milan Out/UTS In tumors that have an aggressive biological potential. Probably due to the small sample size (n = 19), the survival difference between PET-positive and PET-negative patients was just not significant (Fig. 8). However, non-18 F-FDG-avid Milan Out/UTS In patients were able to achieve a 5-year RFS rate of 80%, which was comparable to Milan In patients (86.2%, Fig. 9). This needs to be validated in a larger study cohort.
Noteworthy, we did not find significant outcome differences between Milan Out/UTS In and beyond UTS patients, although tumor load was significantly higher in the UTS Out cohort. One explanation might be that   both subgroups did not differ regarding aggressive histopathologic variables (Table 5). According to this finding, the prognostic significance of tumor load may reach a plateau beyond the UTS limit. In fact, none of macromorphometric features but only positive PET-status was identified as an independent clinical promoter of HCC relapse in beyond UTS patients (Table 4). Furthermore, outcome was not significantly different between patients meeting the UTS criteria and PET-negative patients exceeding them (Fig. 7). This interesting result indicates that biological tumor activity is the major prognostic determinant in advanced HCC stages. D' Amico et al. recently identified poor grading and MVI as the only independent predictors of tumor relapse in a series of 124 liver recipients with HCC beyond the UTS criteria, which supports this assumption on a histopathologic basis 31 . If being confirmed in large prospective trials, our data support the implementation of a purely tumor-biology based selection approach, as was already favored by others 52 .
There are several limitations of our study. First, it was a retrospective investigation with all the possible disadvantages of this study design. Second, the number of HCC patients beyond MC but still meeting the UTS criteria was rather low. Another critical point is that we have stratified our data according to semiquantitative and not quantitative PET results, which did not allow for further risk stratification in the 18 F-FDG-avid subset. In contrast,   Table 6. Correlation of FDG-PET with tumor-specific features and outcome. *According to pretransplant radiographic staging. AFP -alpha-fetoprotein. HCC -hepatocellular carcinoma. LT -liver transplantation. PET -positron emission tomography.
our study was powered by a large PET data collection and a well-documented long-term follow-up. Apart from that, only preoperatively available tumor characteristics were used for risk analysis.
In conclusion, our study demonstrated that expansion to the clinical UTS criteria carries a considerable risk of selecting tumors with aggressive potential. The implementation of 18 F-FDG PET improves the prognostic power of the UTS criteria, since it identifies patients with beneficial tumor biology. Expansion to the UTS criteria may, thereby, be safely realized.