The experiments were conceived and designed by Zhenhua Hu, Wei Wang, and Shu-Sen Zheng. Zhenhua Hu, Wei Wang, Zhiwei Li, and Sunyi Ye performed the experiments, analyzed the data, and contributed reagents, materials, and analytical tools. Zhenhua Hu, Wei Wang, and Shu-Sen Zheng wrote the article.
This work was supported by a grant from the National Basic Research Program of China through the 973 Program (2009CB522403).
Hepatocellular carcinoma (HCC) is the seventh most common cancer and the third most common cause of cancer-related deaths worldwide: it accounts for more than 90% of primary liver neoplasms.1 Liver transplantation (LT) is recommended as the best option for HCC treatment because it resects the tumor, removes the underlying cirrhotic tissue, and restores well-balanced liver function.2, 3 However, because of the shortage of donated organs from both deceased and living donors and the high incidence of HCC, the applicability of LT as the prevailing curative therapy for HCC worldwide is undoubtedly limited.
Recently, salvage liver transplantation (SLT), which involves LT after primary liver resection (LR), has been performed to meet the rising challenge of organ shortages.4 Moreover, SLT can theoretically solve the problem of tumor progression during the period of waiting for primary liver transplantation (PLT) and thus greatly extend the potential waiting time.5, 6 Nevertheless, little is known about SLT and especially about its outcomes in comparison with PLT, and this is vital information for the application of this treatment.
To our knowledge, no systematic evaluation of the survival rate of SLT recipients or the incidence of complications has previously been performed. The primary aim of this study was, therefore, to conduct a comprehensive meta-analysis comparing the short- and long-term survival rates and incidence of complications for SLT and PLT patients. In addition to aiding in the assessment of the feasibility of SLT and its indications, the findings of this study may inform clinical practice and help in the development of reasonable guidelines for the management of patients with HCC.
MATERIALS AND METHODS
Data Sources and Search Strategy
To identify relevant studies and reports, we searched PubMed, Embase, and the Cochrane Library with combinations of the following terms: salvage*; secondary*; transplantation, liver; liver transplantations; transplantations, liver; transplantation, hepatic; grafting, liver; liver grafting; hepatic grafting; hepatic transplantation; hepatic transplantations; transplantations, hepatic; and liver transplantation. Studies were limited to those involving humans and published in English between 1990 and the first week of June 2011. We checked the reference lists of identified review articles, meta-analyses, and original studies to obtain further eligible data. We also searched for clinical trials in the databases of the National Institutes of Health, the European Liver Transplant Registry, and the Scientific Registry of Transplant Recipients. When articles did not report all the types of data necessary for this analysis, we contacted the authors via e-mail for additional information.
A priori, we established the following inclusion criteria for eligible trials:
1Trials should include recipients with HCC who underwent SLT (ie, LR followed by secondary LT).
2Trials should include comparisons of PLT and SLT groups with respect to survival rates or the incidence of complications with follow-up exceeding 1 year.
3Recipients must have been selected on clinical grounds with no restriction on the number of enrolled patients.
4If multiple articles reported the same trial, the most informative report with respect to the study's aims was to be chosen.
We excluded patients with extrahepatic metastases and major vascular invasion and patients who underwent SLT or PLT as part of a multiorgan transplant (eg, liver-kidney transplantation) or whose age was greater than 70 years at the time of transplantation.
Data Extraction and Quality Assessment
Two investigators (W.W. and S.Y.) independently extracted data from all selected articles and stored them in a customized database. Disagreements were adjudicated by re-evaluation and consensus. Another reviewer (Z.H.) independently confirmed all the extracted data and was responsible for their accuracy. Additionally, all extracted data and calculations were double-checked. For all eligible articles, the following data were extracted from the original publications: the first author; the country of origin; the year of publication; the trial scales; the mean age of the recipients; the duration of follow-up; the overall 1-, 3-, and 5-year survival rates; and the incidence of postoperative complications (eg, bleeding, sepsis, and biliary complications). Data that were not directly reported were calculated or obtained indirectly from the original study when they were available.
We assessed the risk of bias in trials according to the recommendations of the Cochrane Collaboration.7 All selected studies came from the published literature, and no further ethics approval was required.
Statistical Analysis and Quantitative Data Synthesis
Our systematic review and meta-analysis were performed according to a recommended protocol that was consistent with the Meta-Analysis of Observational Studies in Epidemiology consensus statement and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.8, 9 All analyses were conducted with Stata 10.0 (Stata, College Station, TX).
Heterogeneity across studies was assessed with the I2 statistic, which is the proportion of the total variation across studies attributable to heterogeneity rather than sampling error.10, 11 We chose a fixed effects model (the Mantel-Haenszel method) to calculate estimates with low heterogeneity and a random effects model (the DerSimonian-Laird method) when there was high heterogeneity (I2 > 50%). Relative risk (RR) was used for the analysis of dichotomous variables. A sensitivity analysis was performed to assess the reliability of the results of both the random effects and fixed effects models. When it was possible, a meta-regression was performed to further evaluate heterogeneity between study results.12 Egger's test (Stata 10.0) was used to evaluate the potential risk of small study effects (eg, publication bias).13 A 2-tailed P value < 0.05 was considered to indicate a statistically significant difference.
The database searches identified 2779 references for evaluation, and we obtained 185 references from reference lists for further assessment. Nineteen articles were included, but 10 were subsequently excluded because of a lack of available data or information for analysis. Two of the remaining 9 articles14, 15 were subsequently excluded because they did not meet the inclusion criteria. These studies had transplant survival data only for patients with recurrent HCC and definitely excluded LT for patients with other indications such as liver failure. The remaining 7 articles,16–22 which were all observational, retrospective cohort studies, were included in the meta-analysis (Fig. 1). No randomized controlled trials were found, and this was likely a result of ethical limitations; this means that they are unlikely to be performed.
The baseline characteristics of the 7 eligible studies (which were published between 2003 and 2009)16–22 are provided in Table 1. Three of these studies were performed in France, 2 were performed in Korea, 1 was performed in Italy, and 1 was performed in Spain. The ratio of SLT and PLT patients varied from 0.085 to 0.483.
Table 1. Characteristics of the Eligible Studies Included in This Meta-Analysis
The 1-, 3-, and 5-year overall survival rates were calculated from the transplant data and were compared for the SLT and PLT recipients. A sensitivity analysis was conducted to confirm the reliability of the calculated outcomes, and Egger's test was used to indicate any potential publication bias. The results of the meta-analysis of the overall survival rates are presented in Table 2 and Fig. 2.
Table 2. Summary of the Meta-Analysis Results
RR (95% CI)
Included biliary strictures or fistulas that required intervention or reoperation.
Six studies16, 17, 19–22 compared the 1-year overall survival rates of SLT and PLT recipients. Homogeneity tests with a fixed effects model revealed no evidence of heterogeneity across studies (I2 = 0.0%, P = 0.75). The meta-analysis with a fixed effects model showed that the pooled RR was 0.99 [95% confidence interval (CI) = 0.90-1.09], which was not statistically significant (P = 0.87). A low level of sensitivity was revealed by a sensitivity analysis, which produced the same result (I2 = 0.0%, P = 0.75) as the random effects model. Egger's test suggested that there was no evidence of publication bias (P = 0.66).
Four studies19–22 compared the 3-year overall survival rates of SLT and PLT recipients. Homogeneity tests with a fixed effects model showed no evidence of heterogeneity across studies (I2 = 0.0%, P = 0.78). A sensitivity analysis revealed a low level of sensitivity (I2 = 0.0%, P = 0.78) with a random effects model. It was reasonable to perform the meta-analysis with a fixed effects model, from which a statistically nonsignificant pooled RR of 0.97 (95% CI = 0.83-1.13, P = 0.68) was obtained. Egger's test suggested that there was no evidence of publication bias (P = 0.61).
Six studies17–22 compared the 5-year overall survival rates of SLT and PLT recipients. Homogeneity tests with a fixed effects model showed no evidence of heterogeneity across studies (I2 = 0.0%, P = 0.45). We also recalculated the meta-analysis with a random effects model and obtained the same result (I2 = 0.0%, P = 0.45), and we thus concluded that the sensitivity was low. It was reasonable to perform the meta-analysis with a fixed effects model, from which a pooled RR of 0.96 (95% CI = 0.81-1.13, P = 0.61) was calculated. Egger's test suggested no evidence of publication bias (P = 0.99). However, because of the insufficient amount of data and number of studies, it was not possible to perform a meta-regression analysis.23
We also performed a meta-analysis of the overall survival rates with all 9 articles (ie, we included those articles that were excluded because of the inclusion criteria as mentioned previously). The results of this analysis (data not shown) were similar to the results of the analysis of the 7 included studies.
Incidence of Postoperative Complications in Recipients
We collected reports regarding several primary complications after LT for the meta-analysis. Bleeding, sepsis, and biliary complications were chosen as parameters for the subgroup analysis because sufficient data were available for them. Homogeneity testing of the bleeding, sepsis, and biliary complication subgroups showed coherent results (see Table 2 and Fig. 3 for more details). All subgroup analyses were performed with a fixed effects model.
The incidence of bleeding was reported by 3 articles,16, 20, 21 for which the pooled RR was 2.84 (95% CI = 1.57-5.13); it was statistically significant (P = 0.001). Two studies20, 21 reported the incidence of sepsis (pooled RR = 0.98, 95% CI = 0.53-1.81), and 4 studies16, 20–22 reported biliary complications (pooled RR = 1.28, 95% CI = 0.77-2.13). The pooled RR values for the sepsis and biliary complication subgroups were not statistically significant (P > 0.05; Table 2). The sensitivity analysis for all complication subgroups revealed a low level of sensitivity.
LR and LT are now widely considered to be complementary treatments (rather than competitive ones) for HCC in patients with cirrhosis and well-preserved liver function.24–26 Salvage transplantation after primary LR is regarded as a reasonable strategy for patients with HCC.6 As a result of the uncertainty about the outcomes of SLT and PLT recipients, there is a debate concerning the optimal treatment of HCC patients with cirrhosis. This is the first meta-analysis to comprehensively review the relevant literature in order to specifically compare the survival rates and postoperative complications of SLT and PLT recipients. We have found that SLT is feasible and safe in terms of overall survival rates after transplantation, and the rates of postoperative complications are comparable to those associated with PLT, except that the bleeding rate is higher with SLT.
The first consideration regarding SLT is its impact on the posttransplant survival rate in comparison with PLT. Since SLT was first proposed by Majno et al.27 with encouraging results, series of trials have been performed to explore the validity of this approach. Two early series of clinical case studies by Adam et al.14 and Belghiti et al.21 reported conflicting results. The first study showed that the rates of survival after SLT were inferior to those after PLT, whereas the latter study indicated that the survival rates of SLT and PLT were similar. In the report by the Paul-Brousse group,14 the poor outcomes after SLT were primarily due to high operative mortality rates and intraoperative bleeding, which may have been related to technical difficulties during LT. Furthermore, all patients in Adam et al.'s study underwent LT because of HCC recurrence, whereas only 61% of the 18 patients in the study by Belghiti et al. had morphological evidence of recurrence. This may have also resulted in the different comparative results. According to the combined results of the 7 studies selected for our meta-analysis, primary LR before LT did not impair overall survival rates in comparison with PLT alone at 1, 3, and 5 years after transplantation.
In addition, these outcomes should be reconsidered on the basis of the Milan criteria for the threshold of resection; the Milan criteria state that transplantation is indicated if there is a single lesion < 5 cm in size or there are no more than 3 lesions with the largest ≤ 3 cm in size.24 Traditionally, it has been accepted that SLT (including PLT) should be proposed for recipients who meet the Milan criteria.28 Recently, some promising results after PLT and SLT have been reported for HCC patients who exceed the Milan criteria.19, 29–32 Kaido et al.5 claimed that the Kyoto criteria (≤10 tumors with all ≤ 5 cm in diameter and serum des-gamma-carboxy prothrombin levels ≤ 400 mAU/mL) were also effective for the selection of patients for SLT because the same results were found for SLT and PLT recipients in terms of overall survival rates. This might indicate that the criteria for the selection of patients for SLT require redefinition. Indeed, the Milan criteria, which appear to be the first-line selection standard advocated and validated by many transplant centers,33 are inevitably restricted by the low sensitivity and accuracy of morphological assessments.
However, some patients dropped from the waiting list because their tumors progressed during this time. For example, recipients with HCC that met the Milan criteria had a decreased 3-year survival rate (from 80% to 60%) as a result of this dropout. It has been reported that almost one-third of patients with advanced HCC beyond the Milan criteria who are waiting for PLT will drop out.34 Our analysis suggests that whatever the HCC stage is, the overall survival rates of SLT and PLT patients are similar. It is unreasonable to exclude such patients from the waiting list on the basis of the Milan criteria at present. Furthermore, some authors have reported that LR can help to downgrade HCC, and this could reduce the existing waiting-list time and provide more opportunities for increasing the number of patients on the waiting list. Therefore, patients who have undergone LR and patients who have not undergone LR should have an equal right to organ allocation. Besides the question of organ allocation, SLT may offer a potential advantage as an alternative approach by enabling the better selection of patients who will benefit more from LT.15
The second concern regarding SLT is the rate of postoperative complications. Adhesions between the cut surface from a previous LR and the omentum or intestine are the main technical difficulties encountered. However, a recent study suggested that sharp dissection and meticulous hemostatic control would reduce the difficulty of repeated surgery and decrease the incidence of postoperative complications.16 Our meta-analysis shows that the risks of sepsis and biliary complications are similar for SLT and PLT patients, but there is an increased risk of bleeding with SLT. One explanation for this finding is that transplantation procedures after previous upper abdominal surgery generally involve more bleeding because of vascular adhesions and a degree of portal hypertension. Another possible reason is a publication bias: authors might be more willing to share their successful experiences than their experiences with high levels of complications. Laurent et al.35 showed that adhesions are reduced if the initial LR is performed laparoscopically, and this suggests a new way to decrease the incidence of complications in SLT.
Several limitations of our meta-analysis should be considered. The first limitation is the lack of randomized controlled trials; this could not be avoided because of limited resources and ethical restrictions. Second, the number of high-quality studies was insufficient for more detailed subgroup analyses. For example, only 1 article included information on disease-free survival. Finally, fewer than 10 trials were involved in our meta-analysis, and thus we failed in our intention to investigate methodological differences as a source of heterogeneity.
In conclusion, our analysis demonstrates that SLT after primary LR is a safe procedure with overall survival rates similar to those associated with PLT. Because of the shortage of donor organs, SLT might be accepted as a good alternative for the treatment of patients with HCC.
The authors gratefully acknowledge Mrs. Mingjuan Jin (Department of Statistics, Zhejiang University) for her assistance with the statistical analysis and Dr. Jie Zhou (Department of Hepatobiliary and Pancreatic Surgery, First Affiliated Hospital, Zhejiang University School of Medicine) for his help with the data collection.