Address reprint requests to Albert C. Y. Chan, M.B.B.S., F.R.C.S., Division of Hepatobiliary and Pancreatic Surgery and Liver Transplantation, Department of Surgery, University of Hong Kong, 102 Pokfulam Road, Hong Kong, China. Telephone: (852) 22553639; FAX: (852) 28165284; E-mail: firstname.lastname@example.org
Hepatic resection and radiofrequency ablation (RFA) are important curative treatments for hepatocellular carcinoma (HCC) in patients with preserved liver function reserve. The 2 treatments have been shown to achieve similar efficacy in recent studies.[1, 2] The 5-year survival rate after curative resection is 42% to 52%,[3-6] whereas the corresponding survival rate after RFA is 42% to 70%.[7-10] Despite the improved survival outcomes attained with the primary treatment of HCC over the last decade,[6, 11] tumor recurrence is common. The 5-year tumor recurrence rate after resection is more than 50%, and the corresponding tumor recurrence rate after RFA is up to 80%.[13, 14] Even with the use of an adjuvant treatment such as antiviral or interferon therapy, the 5-year cumulative recurrence rate remains at 50%.[15-17] Of all the recurrence types, intrahepatic recurrence is the most common pattern. However, there is not yet a consensus on a standardized treatment strategy for recurrent HCC. One of the main reasons is that the choice of further treatment is determined by 2 important factors: the function of the remnant liver and macroscopic tumor features. In order to overcome these problems, various surgical strategies have been derived in the form of repeated hepatic resection (RR), repeated radiofrequency ablation (rRFA), and, lately, salvage liver transplantation (SLT). Both RR and rRFA are readily accessible and effective treatments for intrahepatic recurrence. The concept of SLT (ie, resection or ablation of the primary tumor and then transplantation when recurrence develops) has gained popularity in recent years, with its efficacy compared to that of primary transplantation.[20-23] Because transplantation remains viable for up to 80% of patients experiencing intrahepatic recurrence,[24, 25] SLT may offer a good strategy for relieving patients with a good prognosis after primary treatment with liver transplantation and hence optimize liver graft utilization and alleviate the burden on the donor organ pool.
The aim of this study was to review our experience with the survival outcomes of different operative treatments for recurrent HCC so that we could define a logical management algorithm for patients with postoperative tumor recurrences.
PATIENTS AND METHODS
From January 1993 to September 2009, 4611 patients with HCC were managed at the Department of Surgery of Queen Mary Hospital. Only patients who were 65 years old or younger at the time of tumor recurrence were selected for the analysis because this was the accepted upper age limit for organ allocation at our center. In all, 848 patients had HCC within the Milan criteria, and 532 underwent either hepatic resection or RFA as the primary treatment; 288 patients subsequently developed tumor recurrence. 219 of these patients (76.0%) experienced intrahepatic recurrence, 19 (6.7%) experienced extrahepatic recurrence, and 50 (17.4%) experienced concomitant intrahepatic and extrahepatic recurrence. For 160 of the patients with intrahepatic recurrence only (73.1%), the recurrence was within the Milan criteria. Only patients who experienced intrahepatic recurrence after 2001 (n=183) were included in the analysis because SLT began to be performed at our center at that time. Eighty-seven of these patients received further surgical treatments in the form of SLT, RR, or rRFA, and they formed the focus of this study; 19 patients underwent SLT, 24 patients underwent RR, and 44 patients underwent rRFA. The remaining 96 patients underwent transarterial chemoembolization (n=47), chemotherapy (n=7), another form of local ablation (eg, high-intensity focused ultrasound or microwaves; n=13), or conservative treatment (n=29).
Treatment Algorithm for Primary HCC Within the Milan Criteria
Although resection and RFA are regarded as having similar efficacy for small HCCs, hepatic resection remains the preferred curative treatment in the management of primary HCC for patients with preserved liver function at our center. The selection criteria for hepatic resection have been described previously. In brief, anatomical resection was first considered for patients with Child A cirrhosis, an indocyanine green clearance rate ≤15% at 15 minutes, a serum platelet count ≥100 × 109/L, and a future remnant liver volume ≥30% of the estimated total liver volume, regardless of the tumor size. Alternatively, RFA was selected for unresectable tumors ≤5 cm in diameter or fewer than 3 tumor nodules (each <3 cm in diameter) in patients with Child A cirrhosis or select patients with Child B cirrhosis. The preferred mode of delivery was the percutaneous approach, but the laparoscopic or open approach was adopted for tumors located in the vicinity of the right colon, gallbladder, or stomach or for larger tumors requiring multiple overlapping ablations or located in the right posterior section of the liver. For patients with Child C cirrhosis who had primary HCC within the Milan criteria, liver transplantation was the first choice of treatment.
Tumor Surveillance After Primary Treatment of HCC
All patients underwent computed tomography with measurements of serum alpha-fetoprotein (AFP) and liver biochemistry 1 month after treatment to confirm complete tumor clearance. The same set of investigations was repeated at 3-month intervals for tumor surveillance. Intrahepatic recurrence was defined as a new lesion with arterial contrast enhancement and portal venous washout according to the latest American Association for the Study of Liver Diseases guidelines.
Treatment Selection for Recurrent HCC
As for the treatment strategy for recurrent HCC, our center adopted a re-resection first policy whenever possible. RR was considered for intrahepatic recurrence when anatomical monosegmental or bisegmental resection was feasible, regardless of the tumor size. With respect to the liver function reserve, re-resection was considered for patients with Child A cirrhosis, an indocyanine green clearance rate ≤15% at 15 minutes, an absence of gross ascites, and a platelet count ≥100 × 109/L. When RR was deemed not feasible, rRFA was considered. The indications for rRFA for recurrent HCC were similar to those for the primary tumor. It was selected for unresectable tumors ≤5 cm in diameter or fewer than 3 tumor nodules (each <3 cm in diameter) in patients with Child A cirrhosis or select patients with Child B cirrhosis. It was avoided for periductal tumors or tumors adjacent to major portal pedicles for fear of bile duct injury. SLT was considered otherwise when tumor recurrence was deemed unresectable or nonablatable because of the tumor location and/or poor liver function reserve. For tumor recurrence within the Milan criteria, both deceased donor liver transplantation and living donor liver transplantation (LDLT) were considered. For recurrent tumors beyond the Milan criteria, only LDLT was considered because of the local organ allocation policy on deceased liver grafts. The preoperative workup for both donors and recipients in LDLT has been described previously. In brief, the donor left remnant liver should be no less than 30% of the total liver volume, and the right liver graft should be more than 35% of the recipient's estimated standard liver volume. The technical details of both donor and recipient operations have also been described previously. Preoperative tumor staging was performed via contrast computed tomography scanning of the abdomen and thorax, bone scintigraphic scanning, and/or positron emission tomography with dual tracers using 18-fludeoxyglucose and C-acetate.
Immunosuppression and Antiviral Prophylaxis
Both basiliximab (20 mg) and hydrocortisone (1000 mg) were given intravenously and on postoperative day 4. Long-term immunosuppression was maintained with oral tacrolimus (0.15 mg/kg/day) and a 3-month course of mycophenolate mofetil (1.5 g/day). An oral nucleoside analogue was given as monoprophylaxis for patients with a chronic hepatitis B virus (HBV) infection, and none of the patients received hepatitis B immunoglobulin. Surveillance for HBV recurrence was performed through the monitoring of liver biochemistry, hepatitis B surface antigen, hepatitis B e antigen, and HBV DNA levels initially at each follow-up at monthly intervals after liver transplantation and at 3-month intervals thereafter.
Continuous variables were expressed as medians and interquartile ranges and were compared among the 3 groups with the Kruskal-Wallis test. Categorical variables were compared with the χ2 test. The disease-free survival of our cohort was estimated with the Kaplan-Meier method and was compared between different subgroups with the log-rank test. Disease-free survival was selected as the primary endpoint for the data analysis and was defined as the period from the time of treatment for the first postoperative recurrence to the time of treatment for the second recurrence or death related to any cause. Nine clinicopathological factors [age, sex, HBV serology positivity, serum AFP level, Model for End-Stage Liver Disease (MELD) score, time to tumor recurrence after primary treatment, recurrent tumor status according to the Milan criteria, recurrent tumor number, and choice of treatment for HCC recurrence] were selected for the multivariate analysis with a proportional hazards model to identify independent prognostic factors for disease-free survival after the treatment of recurrent HCC. In the multivariate analysis, the median value was selected as the cutoff point for continuous variables except for the serum AFP level and the time to recurrence, for which cutoff values of 400 ng/mL and 1 year, respectively, were chosen. Categorical values were dichotomized for the multivariate analysis. A P value<0.05 was considered to be statistically significant. Statistical analysis was performed with the SPSS 18.0 computer software program.
The median follow-up for our patients was 38.5 months. Table 1 illustrates the patient demographics for the present study. HBV infection was the most common viral etiology among the 3 groups of patients. As expected, the SLT group had a much worse liver function status, as indicated by the higher MELD score and the lower serum platelet count, which reflected the severity of the underlying portal hypertension.
Table 1. Patient Demographics
The data are presented as medians and interquartile ranges.
As for the primary tumor characteristics, there were no significant differences in the tumor size, tumor number, or serum AFP level among the 3 groups (Table 2). The median tumor sizes for the SLT, RR, and rRFA group were 2.6, 3.0, and 2.9 cm, respectively. More than 80% of the patients in all 3 groups had a solitary tumor on the initial diagnosis. Hepatic resection was performed for 83.3% of the patients in the RR group, whereas only 31.6% of the patients in the SLT group and 29.5% of the patients in the rRFA group underwent hepatic resection as the primary treatment. Forty-five percent of the patients undergoing primary resection and then RR for recurrence had a cirrhotic liver at the time of the first treatment, and all patients who underwent primary resection and then SLT had liver cirrhosis at the time of the first treatment. The median disease-free survival times after primary treatment for the SLT, RR, and rRFA groups were 17.6, 22.6, and 8.8 months, respectively (P<0.01). There was no significant difference between the SLT and RR groups (P=0.45), but both groups had significantly longer disease-free survival after primary treatment than the rRFA group (P<0.01 for SLT versus rRFA and P=0.01 for RR versus rRFA).
Table 2. Primary Tumor Characteristics
The data are presented as medians and interquartile ranges.
Patients with hepatectomy as first treatment [n (%)]
With respect to the recurrent tumor status (Table 3), solitary intrahepatic recurrences accounted for 31.6%, 66.7%, and 86.4% of the patients in the SLT, RR, and rRFA groups, respectively. Ten of the 19 SLT patients experienced tumor recurrence within the Milan criteria, whereas the majority of tumor recurrences were within the Milan criteria for the RR (87.5%) and rRFA groups (97.7%). The recurrent tumor sizes for the SLT, RR, and rRFA groups were 3.8, 2.1, and 1.8 cm, respectively. The 1-, 3-, and 5-year disease-free survival rates were 68.4%, 57.9%, and 57.9%, respectively, after SLT; the corresponding survival rates were 69.7%, 49.3%, and 49.3% after RR and 40.0%, 19.8%, and 10.6% after rRFA (Fig. 1). The 5-year cumulative tumor recurrence rates (Table 4) after SLT, RR, and rRFA were 42.1%, 50.7%, and 89.4%, respectively (P<0.001). There was no significant difference in survival between the SLT and RR groups (P=0.81), but both groups had significantly better survival outcomes than the rRFA group (P<0.001).
Table 3. Recurrent Tumor Characteristics
The data are presented as medians and interquartile ranges.
Patients with recurrence within Milan criteria [n (%)]
Table 4. Cumulative Tumor Recurrence Rates
1 year (%)
3 years (%)
5 years (%)
According to a subgroup analysis confined to recurrent tumors within the Milan criteria (Fig. 2), the 1-, 3-, and 5-year disease-free survival rates after SLT were all 60%; the corresponding survival rates were 70.2%, 48.0%, and 48.0% after RR and 41.0%, 20.3%, and 10.9% after rRFA. Likewise, when the survival analysis was performed after adjustment of the MELD score in order to eliminate the confounding effect of liver function on oncological outcomes regardless of the Milan status (Fig. 3), the 1-, 3-, and 5-year disease-free survival rates were all 50% after SLT; 70.2%, 48.0%, and 48.0%, respectively, after RR; and 41.3%, 21.4%, and 11.4%, respectively, after rRFA (P<0.003).
For those patients who experienced recurrence again after the treatment of first recurrences, intrahepatic recurrence remained the most common pattern of recurrence in the RR and rRFA groups, whereas multiple-site recurrence was most common in the SLT group (Table 5). This distinctive difference in the patterns of recurrence implies that although locoregional treatments in the form of transarterial oily chemoembolization or local ablation remain feasible for subsequent recurrences after RR and rRFA, the choice of locoregional treatments for tumor recurrence after SLT is limited.
Table 5. Patterns of Recurrence After Treatment of Recurrent HCC
Intrahepatic [n (%)]
Extrahepatic [n (%)]
Intrahepatic and extrahepatic [n (%)]
Impact of Primary Treatment on Outcomes of SLT
Six patients underwent hepatic resection as the primary treatment [left lateral sectionectomy (n=2), segmentectomy (n=2), or wedge resection (n=2)], and 13 patients underwent RFA for the primary tumor. When the survival outcomes were stratified according to the choice of primary treatment, the patients undergoing hepatic resection as the primary treatment attained significantly better survival outcomes than the patients undergoing RFA as the primary treatment (P=0.02). The 1-, 3-, and 5-year disease-free survival rates for those undergoing hepatic resection as the primary treatment were 83.3%, 66.7%, and 66.7%, respectively, and the corresponding survival rates for RFA were 61.5%, 30.8%, and 0%.
Prognostic Factor for Disease-Free Survival After the Treatment of Recurrent HCC
When the 9 clinicopathological parameters were entered into the multivariate analysis, a serum AFP level>400 ng/mL, tumor recurrence within 1 year of treatment, and rRFA as the secondary treatment were identified as risk factors for shorter disease-free survival (Table 6). Because tumor recurrence within 1 year could have a major impact on the decision for further treatment, a survival analysis stratified according to this prognostic factor was performed. For those who experienced recurrence within 1 year of the first treatment, the 1-, 3-, and 5-year disease-free survival rates after the second treatment was 23.3%, 10.0%, and 6.7%, respectively; for those who experienced recurrence more than 1 year after the first treatment, the corresponding survival rates were 76.1%, 53.5%, and 47.5%, respectively (Fig. 4).
Table 6. Independent Prognostic Factors for Disease-Free Survival in the Multivariate Analysis
95% Confidence Interval
Serum AFP level>400 ng/mL
Tumor recurrence≤12 months after treatment of primary HCC
RFA as treatment for recurrent HCC
The findings from the present study show that long-term survival is achievable with an aggressive treatment strategy for recurrent HCC, which leads to a 5-year disease-free survival rate of approximately 50% after SLT and RR. The survival benefits of SLT and RR appear to be unchanged even for tumor recurrences within the Milan criteria. Taking this finding into consideration, we believe that a policy of RR first and then SLT for unresectable disease would likely alleviate the burden on the donor organ pool without compromising the chance for long-term survival. A salvage transplantation rate of 8.7% for all intrahepatic recurrences occurring after primary treatment is somewhat lower than that reported in other series.[21, 22, 33] A recent intention-to-treat analysis from a French center showed that up to 28% of patients were able to undergo salvage transplantation when recurrence developed after liver resection. Such a drastic difference in the salvage transplantation rates could be explained by the differences in graft types between the 2 centers. The relatively low preoperative MELD score of our cohort implies a low priority for organ allocation in our locality, where deceased organs are in short supply. This implies that patients with recurrent HCC are less likely to receive a liver graft in time before the tumor progresses, so a more practical treatment strategy would be the initiation of LDLT. However, there are currently limited data on the efficacy of SLT with living donor partial grafts for recurrent HCC. A retrospective study of 17 patients by Hwang et al. showed a 5-year survival rate of 54% after SLT for all intrahepatic recurrences and an improvement to 80% for recurrences within the Milan criteria. A major concern of living related SLT is being able to balance the perioperative risk of donor hepatectomy against the long-term oncological benefit of transplantation. Whether a 5-year disease-free survival rate in the range of 50% to 60% attainable by SLT is acceptable to the patient and the potential donor remains debatable, but the decision for donation should be individualized on a case-by-case basis. Nonetheless, the findings from the present study provide insight into the long-term outcomes of SLT and can facilitate the conveyance of prognostic information when informed consent is being sought from both the patient and the living donor.
One interesting finding of the current study is that survival outcomes after SLT were only marginally better after the exclusion of patients with recurrence beyond the Milan criteria. This is in agreement with the study by Scatton et al., in that adopting the Milan criteria for recurrent tumors failed to differentiate the survival outcomes after SLT for tumors within the Milan criteria and tumors beyond the Milan criteria, and that could be attributed to the possibility of tumor understaging on preoperative imaging. A more precise selection criterion for SLT is, therefore, needed. Our study indicates that the preoperative serum AFP levels and the time to recurrence are associated with poor disease-free survival and hence are good clinical parameters to be adopted for patient selection. Serum AFP levels could reliably predict tumor biological behavior,[36, 37] but using an AFP level>400 ng/mL as a cutoff point for selection would not be useful for non–AFP-secreting recurrences. The time to recurrence, on the other hand, is readily applicable to all recurrences. It is a useful marker for distinguishing recurrence due to intrahepatic metastasis from de novo tumor formation, both of which are associated with different tumor behaviors and prognoses. Intrahepatic metastasis, signified by early tumor recurrence, is associated with more aggressive tumor behavior, whereas late recurrence is more likely to be associated with de novo tumor formation.[38, 39] Because of the organ shortage issue and the natural history of intrahepatic metastasis, SLT may be more appropriately reserved for patients with late recurrences and low serum AFP levels. Alternatively, the issue of preemptive liver transplantation has been proposed for patients with a potential risk for early recurrence.[20, 35] However, such a policy may not be a practical treatment strategy because of the low priority of our cohort for deceased organ allocation. Whether LDLT is justified for this group of patients in the absence of tumor recurrence remains controversial. A more feasible alternative would be to explore new avenues via adjuvant therapies for HCC to reduce the risk of intrahepatic metastasis and hence the chance of early recurrence.
Another original finding of this study is that RFA as a primary treatment impairs survival outcomes after SLT. Our explanation for this phenomenon is an underestimation of tumor necrosis on postablation imaging. Recent histological studies of tumor viability after the application of RFA as a bridging treatment showed that more than 40% of patients had viable tumors despite a complete radiological response, and more than 30% of patients had tumor progression in the explant livers.[40, 41] It is known that partial necrosis promotes tumor recurrence after transarterial chemoembolization, and it is possible that the same phenomenon might also occur with RFA.
RR and rRFA
The 5-year disease-free survival rate of 49.3% corroborates the findings of other series.[43-45] The advantages of RR over SLT are 2-fold: it is a readily accessible treatment option and a less complicated procedure. Besides, the potential hazards of immunosuppression and the risk of opportunistic infection are omitted. The safety profile of RR has been well established in previous studies, with a mortality rate ranging from 0% to 11%.[43, 44, 46, 47] The overall RR rate ranges from 4.3% to 31%.[43-45] Our relatively low RR rate (11.0%) is partly attributable to the fact that the initial study period coincided with the introduction of RFA at our center, and the early enthusiasm for embarking on this local ablative therapy has shifted the treatment paradigm to selecting RR for tumor recurrences. Repeatability and minimal invasiveness are intrinsic advantages of RFA for the treatment of intrahepatic recurrences. Nonetheless, the survival outcomes of rRFA in the present study are rather disappointing, and further recurrences were frequent. Tumor seeding along the needle tract is one possibility promoting tumor recurrence, and increased intratumoral pressure during ablation might also induce tumor dissemination into the adjacent portal vessels.
Selection bias was a major pitfall of this retrospective analysis. Patients with relatively well-preserved liver function and solitary recurrences were more likely to be selected for re-resection. Likewise, patients with multiple recurrences were more likely to undergo SLT. The choice of SLT was also determined by the availability of living donors for patients with low MELD scores, and rRFA may be the only option in this situation if no donor is available. Nonetheless, our survival analysis appears to show that SLT provides a long-term survival advantage over RR because of the selection bias, and rRFA was not as effective as the other 2 treatment strategies for controlling tumor recurrence. The relatively small sample size did not permit any robust statistical analysis for differentiating the efficacy of SLT from that of RR. Hence, the 3 treatment options should be regarded as complementary rather than competing therapies.
In conclusion, to the best of our knowledge, this is the first study evaluating the efficacy of SLT, mainly with living donors, to that of other commonly employed salvage treatment strategies. Although the findings of this study might have been influenced by a potential selection bias and the small sample size, they highlight the importance of a multidisciplinary team approach to the management of patients with postoperative tumor recurrences. Nonetheless, future prospective studies with larger sample sizes are warranted to ascertain the role of SLT versus RR in the management of recurrent HCC.