Alpha-fetoprotein and modified response evaluation criteria in Solid Tumors progression after locoregional therapy as predictors of hepatocellular cancer recurrence and death after transplantation


  • Quirino Lai and Jan Lerut designed the research. Quirino Lai, Alfonso W. Avolio, and Jan Lerut wrote the article. Ivo Graziadei, Gerd Otto, Giuseppe Tisone, Massimo Rossi, Pierre Goffette, Wolfgang Vogel, and Michael B. Pitton participated in the critical evaluation of the article. Quirino Lai performed the data analysis.

  • The members of the European Hepatocellular Cancer Liver Transplant Study Group include Jan Lerut (chair), Quirino Lai, and Pierre Goffette (Starzl Unit of Abdominal Transplantation, St. Luc University Hospital, Catholic University of Louvain, Brussels, Belgium); Ivo Graziadei and Wolfgang Vogel (Gastroenterology and Hepatology, Department of Internal Medicine II, Innsbruck Medical University, Innsbruck, Austria); Gerd Otto and Michael B. Pitton (Department of Transplantation and Hepatobiliary Surgery, University of Mainz, Mainz, Germany); Pasquale B. Berloco and Massimo Rossi (Department of General Surgery and Organ Transplantation, Umberto I Hospital, Sapienza University, Rome, Italy); Giuseppe Tisone and Tommaso M. Manzia (Department of Transplant Surgery, Polyclinic Tor Vergata Foundation, Tor Vergata University, Rome, Italy); and Salvatore Agnes and Alfonso W. Avolio (Liver Unit, Department of Surgery, Agostino Gemelli Hospital, Catholic University, Rome, Italy).

  • None of the authors have any conflict of interest to declare.

  • Quirino Lai was the recipient of a 2012 European Society for Organ Transplantation unrestricted travel grant.

Address reprint requests to Quirino Lai, M.D., Starzl Unit of Abdominal Transplantation, St. Luc University Hospital, Catholic University of Louvain, Avenue Hippocrate 10, 1200 Brussels, Belgium. Telephone: +32027645314; Fax: +32027649039; E-mail: lai.quirino@libero.itor Jan Lerut, M.D., Starzl Unit of Abdominal Transplantation, St. Luc University Hospital, Catholic University of Louvain, Avenue Hippocrate 10, 1200 Brussels, Belgium. Telephone: +32027645314; Fax: +32027649039; E-mail:


Locoregional therapy (LRT) is being increasingly used for the management of hepatocellular cancer (HCC) in patients listed for liver transplantation (LT). Although several selection criteria have been developed, stratifications of survival according to the pathology of explanted livers and pre-LT LRT are lacking. Radiological progression according to the modified Response Evaluation Criteria in Solid Tumors (mRECIST) and alpha-fetoprotein (AFP) behavior was reviewed for 306 patients within the Milan criteria (MC-IN) and 116 patients outside the Milan criteria (MC-OUT) who underwent LRT and LT between January 1999 and March 2010. A prospectively collected database originating from 6 collaborating European centers was used for the study. Sixty-one patients (14.5%) developed HCC recurrence. For both MC-IN and MC-OUT patients, an AFP slope > 15 ng/mL/month and mRECIST progression were unique independent risk factors for HCC recurrence and patient death. When the radiological Milan criteria (MC) status was combined with radiological and biological progression, MC-IN and MC-OUT patients without risk factors had similarly excellent 5-year tumor-free and patient survival rates. MC-IN patients with at least 1 risk factor had worse outcomes, and MC-OUT patients with at least 1 risk factor had the poorest survival (P < 0.001). In conclusion, both radiological and biological modifications permit documentation of the response to LRT in patients waiting for LT. According to these 2 parameters, tumor progression significantly increases the risk of recurrence and patient death not only for MC-OUT patients but also for MC-IN patients. The monitoring of both parameters in combination with the initial radiological MC status is an essential element for further refining the selection criteria for potential liver recipients with HCC. Liver Transpl 19:1108-1118, 2013. © 2013 AASLD.




area under the curve


area under the receiver operating characteristic curve


corrected Akaike information criterion


confidence interval


diagnostic odds ratio


hepatocellular cancer


hazard ratio


interquartile range


liver resection


locoregional therapy


liver transplantation


Milan criteria


within the Milan criteria


outside the Milan criteria


Model for End-Stage Liver Disease


modified Response Evaluation Criteria in Solid Tumors


percutaneous ethanol injection




transarterial chemoembolization


University of California San Francisco

See Editorial on Page 1055

Liver transplantation (LT) is the widely accepted standard of care for hepatocellular cancer (HCC) in select patients with cirrhosis.[1] The introduction of the Milan criteria (MC) markedly reduced recurrence rates after LT.[1, 2] More recently, the implementation of the less restrictive University of California San Francisco (UCSF) criteria and the up-to-7 criteria has expanded access to LT with only a minimal impact on survival.[3, 4] Despite these extensions, many HCC patients still remain excluded from LT, and many drop from the waiting list because of tumor progression.[5-8] In order to prevent both unfortunate conditions, locoregional therapies (LRTs) have been implemented in the therapeutic algorithm for these potential recipients. These are designed to halt tumor progression (bridging protocols) or even down-stage the tumor burden.[9-11]

Although some centers have reported acceptable tumor-free survival rates after LRT even when it is applied to advanced tumors, the impact of LRT on post-LT survival and tumor recurrence has not been well defined.[12-14] Moreover, no international guidelines exist for the selection of these patients because most studies have examined heterogeneous and small patient cohorts and looked mostly at morphological parameters.[15] There is now increasing evidence that biological behavior expressed by tumor markers such as alpha-fetoprotein (AFP) has to be taken into account as an important selection tool.[16-18] Here also, no universal agreement exists about the best AFP parameter (the absolute value, the initial or last measurement on the waiting list, or the AFP dynamics) or the best cutoff value (a static value of 200 or 400 ng/mL or an AFP slope of 15 ng/mL/month).[16-18]

The aim of the present study was to investigate the impact of preoperative parameters as predictors of HCC recurrence and death after LT in a population of patients previously treated with LRT in 2 separate settings based on the initial radiological status of the tumor [within the Milan criteria (MC-IN) and outside the Milan criteria (MC-OUT)]. Particular attention was paid to AFP modifications during the waiting time with the intention of investigating their prognostic power for the prediction of HCC recurrence.


Data Collection

Prospectively assessed databases from 6 collaborative European transplant centers, which included 722 patients undergoing transplantation from March 1987 to March 2010, led to the creation of the European Hepatocellular Cancer Liver Transplant Study Group database. The study protocol received a priori approval by the appropriate institutional review committees of the different centers. Four hundred twenty-two HCC patients who underwent LRT and then LT between January 1999 and March 2010 and who had posttransplant follow-up for at least 2 years were enrolled in this retrospective study. The patient numbers were as follows: 93 patients from the Catholic University of Louvain in Brussels, 96 patients from the Roman Transplant Consortium (Sapienza University, Tor Vergata University, and Catholic University), 102 patients from the University of Mainz, and 131 patients from Innsbruck Medical University. The demographics and tumor characteristics for the entire cohort and the 2 subcohorts of MC-IN and MC-OUT patients are displayed in Table 1. As of March 31, 2012, the median follow-up for the entire population was 4.9 years [interquartile range (IQR) = 3.1-7.3 years].

Table 1. Clinical and Pathological Factors
 Entire Cohort (n = 422)MC-IN Patients (n = 306)MC-OUT Patients (n = 116)
  1. a

    The data are presented as medians and IQRs.

General variable   
Patient age (years)a60 (53-64)60 (53-64)59 (54-64)
Male sex [n (%)]348 (82.5)249 (81.4)99 (85.3)
Hepatitis B virus positivity [n (%)]67 (15.9)45 (14.7)22 (19.0)
Hepatitis C virus positivity [n (%)]192 (45.5)139 (45.4)53 (45.7)
Waiting period (months)a6 (3-10)6 (3-10)5 (3-9)
Laboratory MELD scorea12 (8-17)7 (7-14)6 (6-11)
Child-Turcotte-Pugh class C [n (%)]54 (12.8)40 (13.1)14 (12.1)
Initial radiological examination   
Largest lesion (cm)a2.7 (2.0-3.8)2.3 (1.6-3.0)4.0 (3.0-5.4)
Lesion > 5 cm [n (%)]37 (8.8)0 (0)37 (31.9)
Number of lesionsa2 (1-3)1 (1-2)4 (2-9)
>3 lesions [n (%)]62 (14.7)0 (0)62 (53.4)
MC-OUT [n (%)]116 (27.5)0 (0)116 (100)
Outside UCSF criteria [n (%)]77 (18.2)0 (0)77 (66.4)
Last pre-LT radiological examination   
Largest lesion (cm)a1.4 (0-2.4)1.0 (0-2.0)2.0 (0.8-3.5)
Lesion > 5 cm [n (%)]9 (2.1)0 (0)9 (7.8)
Number of lesionsa1 (0-2)1 (0-2)2 (1-3)
>3 lesions [n (%)]0 (0)0 (0)0 (0)
mRECIST complete response [n (%)]121 (28.7)95 (31.0)26 (22.4)
mRECIST progression [n (%)]30 (7.1)16 (5.2)14 (12.1)
Explant examination   
Largest lesion (cm)a2.0 (1.0-3.0)1.5 (0.7-2.6)2.7 (1.5-4.0)
Lesion > 5 cm [n (%)]32 (7.6)14 (4.6)18 (15.5)
Number of lesionsa1 (1-2)1 (1-2)2 (1-3)
>3 lesions [n (%)]44 (10.4)20 (6.5)24 (20.7)
MC-OUT [n (%)]97 (23.0)44 (14.4)53 (45.7)
Outside UCSF criteria [n (%)]64 (15.2)30 (9.8)34 (29.3)
Tumor understaging [n (%)]70 (16.6)53 (17.3)17 (14.7)
Poor tumor grading [n (%)]43 (10.2)26 (8.5)17 (14.7)
Multifocality [n (%)]168 (39.9)126 (41.2)42 (36.2)
Bilobarity [n (%)]95 (22.5)66 (21.6)29 (25.0)
Microvascular invasion [n (%)]44 (10.4)19 (6.2)25 (21.6)
Initial AFP (ng/mL)a9.7 (5.0-31.4)8.8 (4.7-23.8)13.1 (6.1-102.5)
Last AFP (ng/mL)a6.9 (4.0-17.0)6.6 (3.8-13.9)8.1 (4.7-55.3)
AFP modification (ng/mL)a−1.1 (−8.3 to 1)−1.1 (−6.0 to 0.6)−1.3 (−29.1 to 2.9)
Time from initial AFP to last AFP (months)a6 (4-10)6 (4-11)6 (3-8)
Last AFP > 200 ng/mL [n (%)]24 (5.7)12 (3.9)12 (10.3)
AFP progression > 15 ng/mL/month [n (%)]26 (6.2)12 (3.9)14 (12.1)
Number of treatmentsa3 (1-4)2 (1-4)4 (2-5)
Time from successful down-staging to listing (months)a3 (1-6)
Time from last LRT to LT (months)a3 (2-5)3 (1-6)4 (3-4)
Time from last radiology to LT (months)a2 (0-4)2 (0-4)2 (1-4)
Multiple types of LRT [n (%)]81 (19.2)55 (18.0)26 (22.4)
100% necrosis at pathology [n (%)]127 (30.1)98 (32.0)29 (25.0)

Diagnosis and HCC Staging

HCC was diagnosed on the basis of different guidelines according to the period in which LT was performed.[19, 20] The radiological assessment was performed at the time of registration on the waiting list. Initially, the MC were adopted as the selection criteria for registration on the waiting list. In the case of MC-OUT patients, LRT was used as the down-staging procedure. Only patients down-staged within the MC were considered eligible for LT. From 2001 onward, the acceptable upper limit for the tumor burden after down-staging was extended to meet the UCSF criteria.

Liver Allograft Allocation

Graft allocation was based on the laboratory Model for End-Stage Liver Disease (MELD) score. Patients with a stage II tumor according to the International Union Against Cancer and a laboratory MELD score less than 22 were raised to the 22-point level at registration on the waiting list.[21]

Treatment and Follow-Up of HCC on the Liver Waiting List

LRT was performed according to the guidelines of the European Association for the Study of the Liver.[19] LRT was used in 2 different contexts: for MC-IN patients, neoadjuvant approaches were adopted with the intent of bridging the patients to LT, whereas for MC-OUT patients, LRT was used with the intent of down-staging the tumor burden. All the different LRT modalities that were applied are listed in Table 2. LRT consisted of transarterial chemoembolization (TACE; n = 347 or 82.2%), percutaneous ethanol injection (PEI; n = 72 or 17.1%), radiofrequency (RF; n = 64 or 15.2%), and/or liver resection (LR; n = 27 or 6.4%). Eighty-one patients (19.2%) underwent 2 or more types of LRT: 74 patients (17.5%) had 2 different types of LRT, and 7 patients (1.7%) had 3 different types of LRT.

Table 2. Different LRT Modalities Performed in the Study Population
LRT ProcedureEntire Cohort (n = 422)MC-IN Patients (n = 306)MC-OUT Patients (n = 116)
TACE alone [n (%)]273 (64.7)194 (63.4)79 (68.1)
PEI alone [n (%)]28 (6.6)22 (7.2)6 (5.2)
RF alone [n (%)]26 (6.2)23 (7.5)3 (2.6)
LR alone [n (%)]14 (3.3)12 (3.9)2 (1.7)
TACE + PEI [n (%)]34 (8.1)19 (6.2)15 (12.9)
TACE + RF [n (%)]27 (6.4)24 (7.8)3 (2.6)
TACE + LR [n (%)]6 (1.4)3 (1.0)3 (2.6)
PEI + RF [n (%)]4 (0.9)2 (0.7)2 (1.7)
PEI + LR [n (%)]2 (0.5)1 (0.3)1 (0.9)
RF + LR [n (%)]1 (0.2)1 (0.3)
TACE + PEI + RF [n (%)]3 (0.7)2 (0.7)1 (0.9)
TACE + PEI + LR [n (%)]1 (0.2)1 (0.3)
TACE + RF + LR [n (%)]3 (0.7)2 (0.7)1 (0.9)

Follow-up on the transplant waiting list after LRT consisted of computed tomography or magnetic resonance scanning every 3 months. The tumor response after LRT was evaluated with the initial and last imaging before LT according to the modified Response Evaluation Criteria in Solid Tumors (mRECIST).[22] At each center, independent readers retrospectively reviewed the imaging in order to assess the tumor response to LRT. Radiological progression was defined as an at least 20% increase in the sum of the largest diameters of the target lesion (with an absolute increase of at least 5 mm) or the appearance of new lesions. In accordance with this, the indication for LT remained valid with tumor progression confined to transplant criteria (MC or UCSF criteria). Sometimes, the paradoxical situation of a successfully down-staged target lesion in the presence of a rise in the total tumor burden was encountered. Two examples of this condition are given here.

A patient who presented with a 50% reduction in a solitary, large (6-cm) lesion (MC-OUT status) after TACE but developed a 2-cm lesion according to the last pre-LT imaging improved his oncological transplant condition from MC-OUT status to MC-IN status, but at the same time, his oncological status deteriorated when his mRECIST classification was taken into account. Similarly, an MC-IN patient who presented with a 75% reduction in a solitary, large (4-cm) lesion but simultaneously developed a 2-cm lesion remained MC-IN but presented with radiological progression.

Pathological underestimation of the tumor was defined, and this took into account the last pre-LT radiological assessment and the explant examination. The criteria for its definition were arbitrarily established for single lesions as a tumor diameter >20% at pathology and for multiple lesions as the presence of lesions > 1 cm in diameter that were undiagnosed at radiology.

Assessment of the Evolution of the AFP Level

AFP determinations at 2 well-defined time points—registration on the waiting list and LT—were considered. Two different variables were investigated: the last static AFP value before LT and the slope of AFP evolution before LT. The slope of AFP progression was calculated according to Vibert et al.[18] through the consideration of the difference between the AFP values divided by the time lapse between the 2 referral points. Different cutoff values were tested for both variables (Table 3). Threshold values of 200 and 400 ng/mL were considered for the last static AFP examination. A threshold value of 15 ng/mL/month was considered for the AFP slope.[18] All 3 cutoff values were validated in previous studies.[16-18]

Table 3. Receiver Operating Characteristic Analysis and Cutoff Values for the 2 Subgroups of MC-IN and MC-OUT Patients
VariableAUC (95% CI)P ValueCutoffSensitivity (%)Specificity (%)DOR
  1. NOTE: AFP cutoff values were previously log-transformed (log10).

MC-IN patients      
Last static AFP value57.0 (44.9-69.1)0.18200 ng/mL17.197.47.7
400 ng/mL14.398.18.6
AFP slope51.6 (39.8-63.4)0.5915 ng/mL/month14.397.46.5
MC-OUT patients      
Last static AFP value59.6 (46.3-73.0)0.14200 ng/mL26.993.45.2
400 ng/mL23.195.66.3
AFP slope66.6 (54.5-78.7)0.0115 ng/mL/month30.892.35.3


Immunosuppressive therapy consisted of a calcineurin inhibitor–based regimen in combination with steroids and either azathioprine or mycophenolic acid. Steroids were gradually tapered and discontinued after 3 to 6 months in nearly all patients. Mammalian target of rapamycin inhibitors were used in 108 patients (25.6%) because of renal dysfunction or HCC recurrence and, more recently, as a main immunosuppressant for HCC patients during follow-up.

Patient Follow-Up

All patients were followed at their respective transplant outpatient clinics. Screening for tumor recurrence was done via the measurement of AFP levels and via ultrasound examination every 3 months. A thoracoabdominal computed tomography scan was performed yearly; additional imaging techniques such as bone scanning and cerebral magnetic resonance were used for cases of suspected HCC recurrence. No patient received adjuvant chemotherapy.

Statistical Analysis

Categorical variables were reported as numbers of cases and percentages, and continuous variables were reported as medians and IQRs. Because of the poor homogeneity between the groups of patients treated with LRT for bridging or down-staging, an analysis of comparisons between them was not performed. The Kolmogorov-Smirnov test was used for evaluating the parametric distribution of serum AFP values. Because of their highly skewed distribution (P < 0.001), static AFP values were log-transformed (log10).

The last log-transformed static AFP determination and the AFP slope were tested for MC-IN patients and MC-OUT patients with c-statistic analysis. The best HCC predictor was established on the basis of the higher area under the receiver operating characteristic curve (AUROC). Different cutoffs for these 2 variables were also investigated. Threshold values previously described in the literature—200 and 400 ng/mL for the static AFP value and 15 ng/mL/month for the AFP slope—were analyzed.[16-18] The diagnostic odds ratio (DOR), which measures the overall accuracy of a diagnostic test, was calculated for each cutoff value. A higher DOR indicates higher accuracy for the test.

Separate univariate Cox regression analyses were performed for the entire population and for the subgroups of MC-IN and MC-OUT patients with the intention of investigating risk factors for HCC recurrence in the different settings of LRT used for bridging or down-staging. Only variables detectable before LT were analyzed: all continuous variables were transformed into dummy variables. Multivariate logistic regression analyses were performed: only variables with a P value ≤ 0.05 according to the univariate analysis were included to avoid the risk of overfitting the data.

The goodness of fit of the models was tested with the corrected Akaike information criterion (cAIC) test. Predictions of risk were reported as hazard ratios (HRs), 95% confidence intervals (CIs), and P values. A P value < 0.05 indicated statistical significance.

Tumor-free and patient survival rates were analyzed with the Kaplan-Meier method and compared with the log-rank test. Statistical analyses and plots were performed with SPSS 19.0 (SPSS, Chicago, IL).


Analyses of Tumor Lesions and Outcomes for Patients

At the initial radiological evaluation, 306 patients (72.5%) were MC-IN and 116 (27.5%) were MC-OUT; 77 of the latter patients (18.2%) were also outside the UCSF criteria. Several differences were observed between the 2 subgroups with respect to tumor- and management-related characteristics. According to mRECIST, at the time of LT, 95 MC-IN patients (31.0%) and 26 MC-OUT patients (22.4%) had a complete radiological tumor response. Sixteen MC-IN patients (5.2%) and 14 MC-OUT patients (12.1%) showed tumor progression.

At the pathological examination of the explanted livers, tumor underestimation was observed for 53 MC-IN patients (17.3%) and 17 MC-OUT patients (14.7%). For 9 of these MC-IN patients (17.0%) and 3 of these MC-OUT patients (17.6%), the underestimation was determined by the presence of lesions > 1 cm in diameter that were undiagnosed at the last pre-LT radiological assessment. Complete necrosis at the site of LRT was seen for 98 MC-IN patients (32.0%) and 29 MC-OUT patients (25.0%).

MC-OUT patients had a higher median number of LRT treatments (4 versus 2 for MC-IN patients) with a median time of 4 months between last LRT and LT.

Sixty-one patients (14.5%) in the entire cohort developed HCC recurrence after a median period of 16.6 months (IQR = 10.4-33.0 months). Thirty-six recurrences (11.8%) were observed among MC-IN patients, and 25 (21.6%) were observed among MC-OUT patients; the median times from LT were 13.2 and 22.2 months, respectively.

AFP Modification After LRT

For MC-IN patients, the median AFP values were 8.8 and 6.6 ng/mL at the time of waiting list registration and immediately before LT, respectively. The last static AFP value and the AFP slope had AUROCs of 57.0 (95% CI = 44.9-69.1) and 51.6 (95% CI = 39.8-63.4), respectively, for the risk of HCC recurrence. An AFP slope value of 15 ng/mL/month had a good DOR (6.5) with 14.3% sensitivity and 97.4% specificity. Half of the 12 patients exceeding this latter value (3.9%) developed recurrence (Fig. 1A).

Figure 1.

Distributions of (A) MC-IN and (B) MC-OUT patients according to the AFP slope and the final static AFP value. Both AFP values are reported on a logarithmic scale. Horizontal, dashed lines correspond to the cutoff value of 15 ng/mL/month.

MC-OUT patients had median AFP values of 13.1 and 8.1 ng/mL at waiting list registration and before LT, respectively. The AFP slope was the best predictor of recurrence with an AUROC of 66.6 (95% CI = 54.5-78.7). An AFP slope > 15 ng/dL/month had a DOR of 5.3 with a sensitivity of 30.8% and a specificity of 92.3%. Fourteen patients (12.1%) exceeded this latter cutoff value, and 8 of these patients (57.1%) developed recurrence (Fig. 1B).

Multivariate Analyses

In the multivariate Cox regression analysis of the entire population, an AFP slope > 15 ng/mL/month (HR = 5.4, P < 0.001) and mRECIST progression (HR = 3.5, P value < 0.001) were the strongest independent risk factors for recurrence, and they were followed by a largest lesion diameter > 5 cm (HR = 2.0, P = 0.03).

Among MC-IN patients, 2 variables were significant risk factors for recurrence in the multivariate Cox regression analysis: increased AFP (HR = 4.9, P = 0.001) and mRECIST progression (HR = 2.2, P = 0.02).

The multivariate analysis of MC-OUT patients confirmed both variables as independent risk factors for recurrence, with 3.8- and 3.9-fold increased risks of recurrence in patients with radiological progression and increased AFP before LT, respectively (Table 4).

Table 4. Multivariate Analyses of the Risk of Post-LT HCC Recurrence and Patient Death in the Entire Population and in the MC-IN and MC-OUT Groups
VariableHCC RecurrencePatient Death
P ValueHR95% CIP ValueHR95% CI
  1. a

    cAIC = 714.4 for HCC recurrence; cAIC = 818.3 for patient death.

  2. b

    cAIC = 386.1 for HCC recurrence; cAIC = 476.4 for patient death.

  3. c

    cAIC = 238.1 for HCC recurrence; cAIC = 268.6 for patient death.

Entire populationa
AFP slope > 15 ng/mL/month<0.0015.42.8-10.1<0.0013.82.0-6.9
mRECIST progression<0.0013.51.9-
Major lesion diameter > 5 cm0.032.01.1-
mRECIST complete response0.20.60.3-
Total number of LRT treatments >
MC-IN patientsb
AFP slope > 15 ng/mL/month0.0014.91.9-12.60.0073.61.4-9.2
mRECIST progression0.022.21.1-
Total number of LRT treatments >
MC-OUT patientsc
mRECIST progression0.0023.81.6-
AFP slope (15 ng/mL/month)0.0083.91.4-
Male sex0.40.60.2-

The multivariate Cox regression analyses that were designed to investigate risk factors for death after LT obtained similar results for the different populations. In the whole patient cohort, an AFP slope > 15 ng/mL/month (HR = 3.8, P < 0.001) and mRECIST progression (HR = 1.6, P = 0.04) were independent risk factors. The same risk factors were observed for MC-IN patients: increased AFP (HR = 3.6, P = 0.007) and mRECIST progression (HR = 2.0, P = 0.02). Among MC-OUT patients, only increased AFP (HR = 3.0, P = 0.02) was an independent risk factor for patient death.

Cohort Stratification According to the Number of Risk Factors Predictive of Recurrence and Survival

When the patient population was stratified according to the initial radiological MC status and the presence of risk factors detected in the inferential analyses, MC-IN patients with neither risk factor had the best 5-year tumor-free survival (90.0%), whereas MC-IN patients with at least 1 risk factor had clearly worse results (67.7%, P < 0.001). Interestingly, patients who were initially MC-OUT without risk factors had an excellent 5-year disease-free survival rate of 87.0%, which was significantly better than the rate observed for MC-IN patients with risk factors (P = 0.01). MC-OUT patients with at least 1 risk factor had the worst results (42.0%), which were significantly inferior to the results for all the other groups (Fig. 2).

Figure 2.

Patient survival curves (right) and tumor-free survival curves (left) based on the initial MC status and the presence of risk factors (ie, AFP slope > 15 ng/mL/month and radiological tumor progression).

When we analyzed patient survival, similar results were obtained. MC-IN patients without risk factors had the best 5-year patient survival (88.0%); MC-OUT patients without risk factors had better results than MC-IN patients with risk factors (83.5% versus 67.3%); and finally, MC-OUT patients with at least 1 risk factor had the worst results (55.4%), which again were significantly inferior to the results for all the other groups (P < 0.001; Fig. 2).


Efforts to identify a metric system based on the size and number of nodules in HCC recipients led to the development of the MC in 1996.[1] More recently, the spectrum of the original metric system was widened, and the result was the Metroticket concept.[4] Although some authors have demonstrated that priority could be based on survival benefit, research focused on a ranking system for HCC candidates remains a major point of interest in the field of LT.[23, 24] Today, several systems based on tumor burden have been validated in large LT series.[1, 3, 4] Nevertheless, all of them lack continuity, and no agreement exists about the variables to be hypothetically considered or about whether there is sufficient correction for waiting time in these patients.[15] In this scenario, LRT may have 2 different functions. Not only is LRT a valid tool for stabilizing or down-staging tumors surpassing the commonly accepted selection criteria, but it also can serve as a tool for the documentation of tumor behavior and/or aggressiveness.[12, 13, 25] In fact, besides the morphological variables that are currently used as parameters for patient listing, great interest has recently grown in modifications of the radiological tumor burden and serological markers after LRT.[16-18, 26-28] Tumor behavior indeed plays an important role in (surgical) oncology.[29] The integration of dynamic biological and morphological variables into the classic static morphological presentation of the tumor (translated mainly by the tumor number and volume) seems to be a logical approach to identifying patients with the best or worst prognoses.[30-32] In the recently published European Association for the Study of the Liver guidelines for the treatment of HCC, changes in serum AFP levels and radiological modification according to mRECIST have been proposed to assess the post-LRT response.[33] In the present study, these 2 dynamic variables were confirmed to be the strongest predictors of recurrence after LT. Our findings are strongly in agreement with several publications in which AFP modification and radiological progression after LRT correlate with the risk of recurrence.[18, 34-36]

In a large multicenter French study (n = 537), a prognostic model for predicting recurrence was generated and validated, and the incorporation of AFP significantly improved the prediction of tumor recurrence.[34] Another French experience (n = 153) clearly showed that AFP progression > 15 ng/mL/month was more relevant than static AFP levels in predicting LT outcomes.[18] Similarly, a Canadian study (n = 144) revealed that a rising AFP slope (>0.1 μg/L/day) was the best predictor of microvascular invasion and HCC recurrence after OLT.[35]

Looking at the role of radiological progression, a study from Germany (n = 136) demonstrated that changes in tumor features resulting from pre-LT TACE were better criteria for predicting tumor recurrence than the MC, mainly because of the imprecise assessment of the tumor size and the number of lesions at initial imaging.[36]

In our study, the impact of these dynamic variables has been analyzed for the entire patient cohort and for 2 different subgroups: bridged MC-IN patients and down-staged MC-OUT patients. We decided to perform both analyses even though the reduced number of cases in each subgroup weakened their statistical power. In fact, our intention was also to investigate the performance of LRT by an examination of 2 different endpoints: stabilization and destruction of the tumor burden.

Interestingly, for both MC-IN and MC-OUT patients, an AFP slope > 15 ng/mL/month and radiological progression according to the mRECIST classification were unique independent risk factors for HCC recurrence and death after LT. Stratifying the entire cohort according to these 2 variables and the initial radiological MC status, we observed that MC-OUT patients without risk factors at the time of LT had excellent tumor-free survival and patient survival. These survival rates not only were similar to those observed for MC-IN patients without risk factors but also were significantly superior to those reported for MC-IN patients with risk factors.

These findings are important because they may profoundly change the outlook of patients who are initially MC-OUT. Not only can such patients be reconsidered for LT, but they may also expect a good survival rate. Conversely, patients who are initially MC-IN but present with one or both risk factors experience less good results. These observations confirm not only that morphological data are insufficient (surrogate) markers of tumor aggressiveness but also that they are insufficient predictors of recurrence.

A recent study based on US data (n = 45,267) confirmed our findings and showed that MC-OUT patients with low AFP values (0-15 ng/mL) experienced excellent post-LT survival, whereas MC-IN patients had poor survival if their serum AFP level was elevated.[37]

Finally, patients who were initially MC-OUT with at least 1 risk factor had the worst results. Poor responders to LRT, therefore, need a more complete evaluation using novel tools such as positron emission tomography scanning in order to improve the assessment of tumor aggressiveness and to detect distant metastases.[38]

Patients presenting with increased AFP levels and/or mRECIST progression and patients presenting with tumor recurrence represented a small percentage of the studied cohort. There was an elevated risk that both parameters were reported contemporaneously for the same patients, and this could potentially reduce the statistical power of the multivariate models because of collinearity. However, it is interesting that very few patients presented with both increased AFP levels and radiological progression [1 MC-IN patient (0.3%) and 2 MC-OUT patients (1.7%)]. The absence of collinearity in the multivariate models was tested with the cAIC test, and this confirmed the validity of the results.

The rationale for determining the AFP slope with 2 time points only was based on the report by Vibert et al.[18]: their slope of AFP progression was calculated through the consideration of the difference between the lowest and highest AFP values divided by the time span between the 2 values. We designed an AFP slope determined by the initial and last pre-LT values, although we realized the potential loss of information generated by the fluctuations in AFP between the 2 time points. Although it has been established that the last pre-LT AFP value has prognostic value, it is unclear which other value (the initial value, the highest value during the waiting time, or the next-to-last value) is best to analyze.[17] We decided to use the initial value with the intent of investigating the slope between the baseline value obtained at the time of waiting-list registration and the value immediately before LT. Further research is needed in order to investigate this controversial matter.

Another important aspect that should be emphasized with respect to AFP is the low area under the curve (AUC) observed in the present study, mainly in the subgroup of MC-IN patients. Unfortunately, AFP is not a perfect predictive parameter for the risk of recurrence; as a result, its exclusive use without combination with other selective parameters (eg, radiological progression as in the present case) could be associated with a risk of mistakes in selection. Our data are in agreement with other studies in which static and dynamic AFP values had low-to-intermediate AUCs: in the previously reported study by Vibert et al.,[18] the AUC was 55.7, whereas in a study from Italy, it was 75.0.[16]

We considered only patients who underwent LRT before LT. The high rate of complete tumor necrosis observed during the pathological examination of total hepatectomy specimens could explain the low incidence (10.4%) of microvascular invasion in comparison with other studies such as the Metroticket study (33.8%). The fact that 72.5% of our patients were MC-IN and only 28.5% were in the Metroticket study could be another explanation for this observation.[4]

Our study has some of the limitations of a retrospective, multicenter study. The less restrictive selection criteria used for acceptance on the waiting list in the present patient cohort in comparison with the United Network for Organ Sharing data gave us, however, the opportunity not only to study 116 MC-OUT patients at first radiological examination[17, 39, 40] but also to obtain 2 different cohorts of bridged and down-staged patients and set up different complex prognostic models based on morphological and biological parameters.

The heterogeneity and different timing of the LRT treatments between the participating centers represent a possible second bias. However, the European Hepatocellular Cancer Liver Transplant Study Group was set up to look at a few centers with similar pre-LT approaches to potential HCC recipients. Clinical experience has, moreover, clearly shown that a combination of several LRT types for the same patients is unavoidable because of the technical limitations of the different LRT modalities. Although the analysis of a more homogeneous patient cohort treated with 1 type of LRT (eg, TACE) only could have some statistical benefit, we accepted the bias from analyzing a larger patient cohort treated with different approaches. It should also be stressed that this study did not aim to look at the performance status of the different types of LRT. Because the analysis looked only at patients undergoing LRT, the artificial bias in radiological and AFP determinations in comparison with HCC recipients untreated before LT was avoided.

Consequently, an intent-to-treat analysis of the waiting-list population and the survival benefit in comparison with the possible harm to the population of potential non-HCC recipients was not performed.[41]

In conclusion, integrating radiological and biological progression into classic morphological HCC staging could represent an effective means of stratifying patients before LT. Patients who initially are MC-OUT but do not experience radiological (morphological) and biological progression after LRT can achieve very good post-LT tumor-free survival and patient survival. Conversely, initially MC-OUT patients with a poor morphological and biochemical response after LRT should be re-evaluated and finally withdrawn from the waiting list to prevent the waste of a scarce organ resource. MC-IN patients experiencing the progression of at least 1 of these variables after LRT have an increased risk of tumor recurrence and death after LT in comparison with MC-IN patients and even MC-OUT patients presenting without these risk factors. LT still represents the best therapy for these patients, but the preoperative selection of this subgroup of patients is crucial, and transplant physicians should be urged to adapt their selection procedures and eventually the posttransplant follow-up and (immunosuppressive) care.