Potential conflict of interest: Nothing to report.
Liver resection (LR) for hepatocellular carcinoma (HCC) as the first-line treatment in transplantable patients followed by “salvage transplantation” (ST) in case of recurrence is an attractive concept. The aim was to identify patients who gain benefit from this approach in an intention-to-treat study. From 1998 to 2008, among 329 potential candidates for liver transplantation (LT) with HCC within the Milan criteria (MC), 138 with good liver function were resected (LR group) from a perspective of ST in case of recurrence, and 191 were listed for LT first (LT group). The two groups were compared on an intention-to-treat basis with special reference to management of recurrences and transplantability after LR. Univariate and multivariate analyses were performed to identify resected patients who developed recurrence beyond MC. Five-year overall and disease-free survival was similar in both groups: LT versus LR group, 60% versus 77% and 56% versus 40%, respectively. Among the 138 patients in the LR group, 20 underwent LT before recurrence, 39 (28%) had ST, and 51 (37%) with recurrence were not transplanted including 21 within MC who were excluded for advanced age, acquired comorbidities, or refusal and 30 (22%) with recurrence beyond MC. Predictive factors for nontransplantability due to recurrence beyond MC included microscopic vascular invasion (hazard ratio [HR] 2.38 [range, 1.10-7.29]), satellite nodules (HR 2.46 [range, 1.01-6.68]), tumor size > 3 cm (HR 1.34 [range, 1.03-3.12]), poorly differentiated tumor (HR 3.18 [range, 1.31-7.70]), and liver cirrhosis (HR 1.90 [range, 1.04-3.12]). Conclusion: The high risk of failure of ST after initial LR for HCC within MC suggests the use of tissue analysis as a selection criterion. The salvage LT strategy should be restricted to patients with favorable oncological factors. (HEPATOLOGY 2012;;55:132–140)
Liver transplantation (LT), which is the most effective treatment for early hepatocellular carcinoma (HCC) in patients with chronic liver disease (CLD), is hampered by an imbalance between the increasing number of candidates and an organ shortage.1–4 While the number of patients with HCC is increasing, the main consequence of organ shortage is an increase in the duration of wait time despite a strict limitation of candidates with HCC within the Milan criteria (MC).5, 6 During the wait time, HCC may progress, with a risk of dropout and extension of vascular invasion, which increases the risk of recurrence after transplant.4, 7–9
Alternative curative treatments, which do not compromise transplantation afterwards if needed, include percutaneous ablation and liver resection (LR), which can be performed with a low operative risk and has good long-term survival rates.8–12 Better liver function assessment, more accurate imaging studies, and refinements in surgical techniques and oncological approach can offer a 5-year survival of approximately 70% in selected patients and do not compromise the possibility to perform LT afterwards in cases of decompensation of cirrhosis and/or tumor recurrence.11, 13, 14 Several studies have shown that initial LR does not impair the technical possibility of performing LT afterwards, does not increase the operative risk, and offers a chance of long-term survival when HCC recurrence is limited.11, 13, 15 Initial LR of HCC as a primary therapy in patients who otherwise could have been transplanted offers a good quality of life and is less demanding than LT. Patients do not need long-term immunosuppression and, in addition, grafts are saved for the community and can be transplanted to other patients who have no other alternative.11, 13, 16
LR as a primary therapy with LT in mind for tumor recurrence or deterioration in liver function, so-called salvage transplantation (ST), was first proposed by Majno et al.17 This attractive concept seems to be applicable in a significant proportion of patients and has a long-term survival similar to that of patients who underwent primary LT.9, 11, 13 However, ST is restricted to patients who develop recurrence within MC and could represent a loss of opportunity for the subgroup of initially transplantable patients who develop recurrence beyond MC. In other words, the decision to delay LT should be based on selecting patients who (1) develop recurrence within MC; (2) will adhere to a careful follow-up with imaging to detect early recurrence; and (3) still have a general condition compatible with LT.
None of the previous series in the field have addressed this issue and only an intention-to-treat study could analyze these endpoints. In 1998, we adopted the strategy of performing LR first for HCC in patients with good liver function and considering LT in case of recurrence within MC. With special reference to survival and management of recurrences after LR as a primary therapy, our aim in this intention-to-treat study was to evaluate the rate of resected patients who could not be transplanted at the time of recurrence and the risk factors for the failure of the ST strategy.
CLD, chronic liver disease; HCC, hepatocellular carcinoma; LR, liver resection; LT, liver transplantation; MC, Milan criteria; MELD, Model for End-Stage Liver Disease; MRI, magnetic resonance imaging; RFA, radiofrequency ablation; TACE, transarterial chemoembolization.
Patients and Methods
The study was performed from January 1998 to December 2008 and included 329 patients younger than 65 years with HCC within MC on imaging, occurring with chronic liver disease (CLD; F3 or F4 stage according to the Metavir score). The diagnosis of HCC was based on at least two concordant imaging studies including triple-phase computed tomography (CT) and/or magnetic resonance imaging (MRI) showing both early hyperenhancement and delayed hypoenhancement (washout) in accordance with the American Association for the Study of Liver Diseases (AASLD) Practice Guideline for Management of Hepatocellular Carcinoma.4
LR as a primary therapy was considered in 138 patients with excellent liver function. All were Child-Pugh class A, without clinically significant portal hypertension and with a platelet count over 100,000/mm3. Anatomic resection, with complete removal of at least one Couinaud's segment including the tumor area fed by portal branches, was considered.18 If anatomic resection was not technically possible, we tried to obtain an appropriate margin, greater than 2 cm.19 After LR, follow-up included liver function tests, α-fetoprotein (AFP) level, ultrasound (US), and triple-phase CT scan or MRI every 3 months during the first year and then every 6 months indefinitely. The period of inclusion allowed for extension of follow-up for a minimum of 24 months in order to evaluate the rate of recurrence. Patients with chronic hepatitis B were all treated by appropriate antiviral therapy before and after surgery. Treatment of hepatitis C virus (HCV) infection was considered case by case.
Recurrence was defined as the appearance of a new lesion with features of HCC on imaging. As for the initial tumor, all recurrences were discussed at the multidisciplinary meeting and were classified as transplantable or nontransplantable using the same criteria as those of primary LT for HCC. The policy was to consider salvage LT for patients resected previously and who developed documented HCC recurrence within MC during follow-up. During the study period, it was our policy not to consider specimen analysis as a selection tool to orient patients considered at high risk of recurrence to LT.20 Patients with recurrence outside the criteria for transplantation were treated with a second LR, radiofrequency ablation (RFA), and/or transarterial chemoembolization (TACE).
According to the same criteria, 191 patients not eligible for LR due to multiple tumors, clinically significant portal hypertension, or insufficient liver function (Child-Pugh grade B or C) were listed for LT during the study period. From 1998 to 2007, liver grafts were allocated primarily to transplant centers, not to patients. In March 2007, the Model for End-Stage Liver Disease (MELD) score was adopted as the basis of the French liver allocation system. Patients with HCC are given a score (extra points) corresponding to their risk of wait list mortality or dropout. Although we observed a higher rate of LT for HCC during the second period,21 the mean waiting time for patients with HCC was similar before and after MELD was adopted (5.8 ± 8 versus 7.1 ± 5 months, respectively). For all patients, collection of data ended in December 2010. In patients awaiting LT, adjuvant therapies including TACE and RFA were used whenever possible in order to slow tumor growth and achieve extended or complete tumor necrosis. Patients who died while on the waiting list and/or patients with HCC progression beyond MC or with extrahepatic metastases were removed from the waiting list.
The research was approved by local institutional review board. An intention-to-treat analysis was performed for all potential candidates for LT (within MC) who underwent LR (LR group) and in all those listed for LT first (LT group) during the study period, whatever the histological findings on the specimen. Intention-to-treat analysis was restricted to patients younger than 65 years old. Data from all patients in the LR group (n = 138) were compared with data from patients who were listed for LT (n = 191) in order to assess the outcome following each treatment strategy, with special emphasis on management of tumor recurrence in the LR group. Resection specimens and liver explants were routinely cut into thin serial slices 3-4 mm thick and analyzed by highly experienced pathologists. Data for pathological variables including tumor size, differentiation grade, vascular invasion (macroscopic or microscopic), surgical margins (R1 = 0 mm versus R0 > 0 mm)10 as well as satellite nodules (defined as tumors ≤2 cm in size and located ≤2 cm from the main tumor),22 and fibrosis stage (advanced fibrosis [F3] or cirrhosis [F4]) were systematically collected in the analysis. Early postoperative mortality was defined as death within the first 90 postoperative days.
The primary endpoint was to identify patients who would recur beyond MC, in the intention-to-treat population. Secondary endpoints included the results of the ST strategy compared with patients listed for LT first. Patient baseline characteristics are expressed as mean ± standard deviation (SD) and median with interquartile range for continuous data, and as frequency for categorical data. All data were analyzed according to the intention-to-treat principle. Survival time in each group started at the time of the procedure for those who underwent LR and at the time of listing for those who were considered for LT first. Disease-free survival was calculated considering patients who developed recurrence of HCC after LT or LR and patients who died in the group of those listed for LT. Because of risks for death and recurrence as competing events, predictive factors for time to recurrence were analyzed using the Fine and Gray method.23 All variables were dichotomized for analysis. A P value of less than 0.05 was considered to indicate statistical significance. Multivariate regression analysis was performed using the Cox proportional hazards model, and associations with a P value of less than 0.10 were entered into the final model that was adopted. All statistical tests were carried out using R software (version 2.12.0).
According to the intention-to-treat analysis of all transplantable HCC patients, the 5-year overall survival of the patients who were resected first (n = 138) was 77% compared with 60% (P = 0.12) in the group listed for LT (n = 191). In resected patients, 5-year disease-free survival in intention-to-treat analysis was lower than in patients listed for LT (42% versus 56%) but not significantly (P = 0.3).
The 191 patients listed for LT (median MELD score of 19.8 [range, 8-37]) had a mean waiting time of 6.6 ± 5.4 months (range, 2-17). During this period, 28 patients (15%) died while on the waiting list or were removed from the waiting list due to tumor progression. Among the 163 patients who underwent LT, six (4%) died in the early postoperative period. Eleven patients were found to have nodules other than HCC on explant pathology (hepatocholangiocarcinoma in nine and benign macroscopic regeneration nodule in two (Fig. 1).
Among the 138 resected patients, anatomic LR was performed in 102 (74%), including 29 (21%) patients who underwent major hepatectomy (preceded by portal vein embolization in 15 patients). Two (1.4%) patients died postoperatively, and the overall morbidity rate was 41% (n = 56) with a median length of in-hospital stay of 13 ± 2 days (range, 4-49). Analysis of the resected specimen after LR revealed that one patient with a 2-cm nodule and HCV-related cirrhosis had a benign regeneration nodule, two patients had a hepatocholangiocarcinoma, and one had a carcinosarcoma. Two patients experienced postoperative deterioration in liver function during follow-up and underwent transplantation before recurrence, 12 and 23 months after LR. Eighteen (13%) patients underwent LR as a bridge, including laparoscopic approach in nine and transthoracic approach in five, and underwent transplantation before evidence of recurrence.
Clinical and pathological data for the LR (n = 138) and LT groups (n = 191) are shown in Table 1. Preoperative nonsurgical treatments (TACE, RFA) were similar in both groups (38% versus 37%, P = 0.50).
Table 1. Comparison of 138 Resected Patients for Transplantable HCC (LR Group) and 191 Patients Listed for LT (LT Group) Within MC
LR Group (n = 138)
LT Group (n = 191)
Abbreviations: AFP, α-fetoprotein; MELD, Model for End-Stage Liver Disease; VHB, viral hepatitis B; VHC, viral hepatitis C.
In 39 patients, the specimen was not available because patients died or were removed from the waiting list.
Patients in the LR group were characterized by better liver function with a lower MELD score (6.5 versus.19.8), less chronic alcoholism (12% versus 35%), and more cryptogenic CLD (19% versus 4%). In the LR group, tumors were significantly larger, less often multiple, and less often developed on cirrhosis (75% versus 100%). In the LR group, the relationship between the degree of fibrosis and the cause of CLD showed that F3 was present in nine (16%) with HCV infection, 13 (37%) with hepatitis B virus (HBV) infection, and 14 (54%) with cryptogenic CLD. After LR, the mean surgical margin was 2.8 ± 1.1 cm, including 117 (85%) R0.
Outcomes Among Patients Resected for Documented HCC With a View to Salvage LT
This group of patients comprised 112 patients in all (Table 2).
Table 2. Pattern of Recurrence of Patients Resected for Documented HCC Planned for Salvage LT (n = 112)*
No Recurrence (n = 22)
Recurrence Within MC (n = 60)
Recurrence Beyond MC (n = 30)
Abbreviations: AFP, α-fetoprotein; HCC, hepatocellular carcinoma; LT, liver transplantation; MC, Milan criteria; VHB, viral hepatitis B; VHC, viral hepatitis C.
Patients who underwent liver transplantation before recurrence are not reported in this table.
Median age, years
>60 years, no (%)
AFP > 400 ng/mL, no (%)
Tumor, no (%)
Diameter > 3 cm
Poorly differentiated tumor
Macroscopic vascular invasion
Microscopic vascular invasion
Underlying liver disease, no (%)
R0 resection, no (%)
Time to recurrence, months
Patients With No Recurrence.
With a median follow-up of 68 months (range, 24-168), 22 (20%) patients were still alive without recurrence. This subgroup was characterized by curative resection with R0 in all patients and good oncological characteristics including a single nodule in 86%, tumor < 3 cm in 50%, including 10 (45%) < 2 cm, no poorly differentiated tumor, and no macroscopic vascular invasion. Underlying liver disease was characterized by the presence of extensive fibrosis (F3) in 64%.
Patients With Recurrence.
Recurrence was observed in 90 (80%) patients, including 60 (67%) within MC and 30 beyond MC. The median time to recurrence was 15.9 ± 15 months (range, 6-89) with no significant difference between recurrence within MC and beyond MC (15.5 ± 15 versus 16.2 ± 14 months, P = 0.87).
Recurrence Within MC.
Among the 60 patients with HCC recurrence within MC, 39 (65%) underwent an ST with a median time on the waiting list of 6.9 ± 10 months. During the waiting time, 12 underwent treatment of recurrence, including RFA in five and TACE in seven with no dropout before LT. However, in two cases recurrence was beyond MC on the specimen. Mortality after LT was 5% (n = 2) and during a median follow-up of 68 months, six patients (15%) developed HCC recurrence. After ST, overall survival at 1, 3, and 5 years was 94%, 81%, and 71%, respectively.
Analysis of the 21 (35%) patients with recurrence within MC not considered for LT showed that reasons for discarding LT included advanced age (range, 67-72 years) at the time of HCC recurrence in nine patients and the occurrence of significant comorbidities during follow-up representing contraindications for LT (severe ischemic heart disease) in two patients. In addition, 10 patients were denied LT (n = 7) at the time of recurrence or were lost to follow-up (n = 3).
Recurrence Beyond MC.
Among the 90 patients with recurrence after LR, 30 (33%) patients were beyond MC, with tumor size > 5 cm in eight cases; multifocal (more than three locations) nodules in eight cases; macroscopic vascular invasion in eight cases; and distant metastases in six cases (lungs, n = 5; adrenal, n = 1; peritoneum, n = 2). Among these 30 patients, six (20%) were lost to follow-up over several months (range, 6-24).
In intention-to-treat analysis, the risk analysis of recurrence beyond MC identified seven factors in univariate analysis (Table 3). On multivariate analysis, independent predictive factors were microscopic vascular invasion (hazard ratio [HR] 2.83 [range, 1.10-7.29]), satellite nodules (HR 2.46 [1.01-6.68]), tumor size >3 cm (HR 1.34 [1.03-3.12]), poorly differentiated tumor (HR 3.18 [1.31-7.70]), and liver cirrhosis (HR 1.90 [1.04-3.12]) (Table 3). All patients without recurrence or recurrence within MC had fewer than three factors of the five pejorative histological factors found on multivariate analysis (Table 4). Although some patients with one or two pejorative histological factors did have recurrence beyond MC, all those with three or more pejorative histological factors developed recurrence beyond MC.
Table 3. Risk Factors for Recurrence Beyond MC Among All the 138 Resected Patients for Transplantable HCC (Univariate and Multivariate Analysis)
Abbreviations: AFP, α-fetoprotein; HCC, hepatocellular carcinoma; HR, hazard ratio; MC, Milan criteria; VHB, viral hepatitis B; VHC, viral hepatitis C.
Age > 60 years
AFP > 400 ng/mL
Presence of cirrhosis
Diameter > 3 cm
Microscopic vascular invasion
Macroscopic vascular invasion
Table 4. Outcome of Patients Resected and Planned for Salvage LT (n = 112) According to the Number of Histological Risk Factors
Abbreviations: LT, liver transplantation; MC, Milan criteria.
Factors included: microscopic vascular invasion; presence of satellite nodules; tumor size > 3 cm; poorly differentiated tumor; and cirrhosis.
Benefit of the ST Strategy
Among the 138 patients who underwent initial LR, this strategy was beneficial in 81 (59%) patients including 22 with long-term survival without recurrence, 39 transplanted for recurrence, and 20 who underwent LT without recurrence.
This strategy saved at least 22 grafts that would have been transplanted in patients with no recurrence and four without benefit from LT (absence of HCC or HCC with poor prognosis). However, the transplantability rate for recurrence was 28% of the intention-to-treat population, 39% of patients with recurrence, and 65% of patients with recurrence within MC.
This study, evaluating the outcome of potential candidates for LT with HCC strictly within the MC, showed that initial LR in patients with good liver function is a valid treatment because 5- and 10-year overall survivals were similar to the group who underwent LT first. We confirmed that initial LR as a primary therapy, considering LT for tumor recurrence, is an effective treatment for those who developed recurrence within MC. However, this intention-to-treat analysis has shown that salvage LT for recurrence was successful in less than a half of our patients. Therefore, we propose to restrict this option to patients with favorable oncological factors on specimen analysis, those who are less likely to develop recurrence outside MC.
Patients considered for initial LT were characterized by more severe liver dysfunction, as shown by a significantly higher MELD score.9, 16 Despite a low (6%) rate of recurrence, the long-term survival of this group was not significantly better than that of the LR group, suggesting that in transplanted patients, survival also depends on early and delayed technical complications, recurrence of underlying liver disease, and adverse effects of immunosuppression.24, 25 Although patients with multiple nodules and cirrhosis were more likely to be oriented to LT, our low rate of HCC recurrence after LT could be explained by better oncological factors in the LT group compared with the LR group. This finding, which was also observed by others,19, 15 could result from a stringent selection policy for primary LT, whereas partial LR was more easily considered in patients with good liver function.
The results of the present study confirmed that partial LR in patients with good liver function can be performed with a low operative risk (<2%).10, 26 The present study also confirmed that partial LR with an oncologic intent can offer an excellent outcome without recurrence in some patients characterized by favorable histological factors.27, 28 Interestingly, nearly half of these patients had tumors ≤2 cm, confirming the size criteria of the Barcelona Clinic Liver Cancer (BCLC) prognosis staging,3 and 60% had advanced fibrosis. The high rate of stage F3 patients in this study could be related to the high number of patients considered with cryptogenic CLD, including patients with metabolic syndrome.29 The possibility of selecting such a favorable subgroup that could be cured by partial LR presents a solid argument in favor of preoperative biopsy of the nontumorous parenchyma, considering partial LR as the first-line option in stage F3 patients with HCC.
The 5-year rate of tumor recurrence after LR was of almost 60%, similar to that reported in other studies focusing on HCC within MC.10, 11, 16 This high risk of recurrence after LR was integrated into the ST strategy, which offered LT in case of recurrence within MC. In such cases, our study confirmed the feasibility of LT afterwards, with a low (5%) operative risk and a favorable outcome similar to that of the group that underwent primary LT.9, 11, 16 These results illustrate why the vast majority of transplant surgeons do not contraindicate LR in a potential candidate for LT.
The most important result of the present study is that nearly half of our patients with HCC recurrence following LR did not undergo transplantation, including one-third due to recurrence beyond MC. There is debate concerning the respective roles of LR and LT in the treatment of HCC.9, 11, 16 Analysis of the subgroup of patients who could not undergo transplantation at the time of recurrence showed that three factors can result in failure. These factors were respectively related to the patient, to the tumor, and to the underlying parenchyma.
The first striking factor was related to patient adherence to regular follow-up for prolonged periods after LR. Nearly 20% of our patients who developed recurrence beyond MC had suboptimal follow-up. This failure illustrates the need for a more stringent screening program with repeated abdominal CT scans or MRI, similar to that used in patients on the waiting list before LT.2, 3, 8 Therefore, we would be reluctant to propose this strategy to patients who are geographically distant from the tertiary center. Another important result of this intention-to-treat analysis was to show that nearly 15% of patients with recurrence within MC did not undergo transplantation because they had become too old or had developed severe comorbidities at the time of recurrence. This appears to be an unavoidable consequence of the improvement of life expectancy after LR.9, 11, 16
Prediction of recurrence after LR depends on tumor aggressiveness, which can be assessed by both pathologic and genomic data.30 Although the present study did not explore gene expression, we confirm these findings from clinical and pathological analysis for the risk of recurrence. Sala et al. were the first to recommend performing LT after LR in patients with a high histological risk of recurrence (vascular invasion, satellites nodules).20 We considered these selection criteria not restrictive enough because the presence of an unique pejorative histological factor (such as microvascular invasion) would lead to transplant some patients who would not experience recurrence.22 We focused our results on recurrence beyond MC, which represents a major drawback for those patients who lose their chance for transplantation. The present study presents substantial arguments for a selection process based on specimen analysis. Highly pejorative factors, including macrovascular invasion in subsegmental portal veins, not detected on standard imaging, and specific cholangiocarcinoma components, which are more often discovered,31 should rule out patients for LT. The results of the present study showed that patients with three or more pejorative histological factors were at very high risk of recurrence beyond MC. This selection analysis, which takes into account the addition of several pejorative histological factors, clearly requires external validation but gives solid arguments for considering LT before recurrence.
Prediction of recurrence after LR depends on the liver carcinogenic field. Cirrhosis was present in the vast majority of our patients with recurrence beyond MC, emphasizing the preneoplasic status of the underlying liver disease.32, 33 These recurrences, which could be attributed to de novo cancers, may be prevented by treating the cause of the underlying liver disease.34, 35 However, to what extent antiviral therapy had a significant impact in reducing recurrence following resection is still a matter of debate.36, 37 HCV infection was not associated with higher risk of recurrence beyond MC, suggesting that the presence of cirrhosis in this etiology has the main pejorative impact.38, 39 Interestingly, time to recurrence was not an unfavorable prognostic factor in this study. As a matter of fact, we do not share the view that patients with late versus early recurrence should be considered differently.40
In this intention-to-treat study, the methodological inclusion of the group that underwent transplantation before recurrence and the low number of patients who recurred beyond MC may limit the power of our message, but this method was less subject to selection bias than previously reported series.9, 11, 14 The strengths of this study are the large number of transplantable patients resected in a single center with a long follow-up, allowing analysis of the largest number of published patients transplanted for recurrence with an acceptable waiting time. It is not clear whether the salvage policy would be appropriate for areas with especially long waiting times or whether it would fit into the current MELD scoring system in the United States.
In conclusion, this study confirms that initial LR of transplantable patients within MC with good liver function is a valid option allowing selection of the subgroup of patients who could benefit from follow-up with LT in case of recurrence. However, the high risk of nontransplantability after initial LR critically emphasizes the utility of specimen analysis as a selection tool to assess both the underlying parenchyma and histological factors. The presence of pejorative factors led us to consider LT before recurrence (Fig. 2).