Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Liver transplantation for hepatocellular carcinoma (HCC) and cirrhosis is complicated by the prognosis of both diseases. Although there is reasonable prognostic evidence for the likelihood of surviving cirrhosis with and without liver transplantation, the issue with HCC is the likelihood of its recurrence as well as its consequences. For patients with HCC, there is currently only 1 drug (sorafenib) that has been shown to increase survival (in the pretransplant setting), and the median increase is only 3 months.1
No therapies for recurrent HCC after transplantation have been shown to improve survival. Therefore, the prevention of recurrence or recurrence-free survival could be considered paramount, but there is no consensus. Overall survival may not be a sufficiently sensitive outcome measure for the assessment of selection criteria for liver transplantation for HCC.
Recurrence occurs at a rate of 50% in the liver and at a rate of 50% in extrahepatic sites (mainly the lungs). The first issue is the acceptable rate of recurrence because recurrence is one of the major causes of death in patients with HCC (along with the recurrence of progressive hepatitis C). In part, the acceptable recurrence rate could be determined by oncological principles, but it must also be dictated by the availability of donor organs and by comparisons with other indications for liver transplantation via models of utility and transplant benefits. In other words, should the selection for liver transplantation be based on an estimate of no recurrence or a rate of recurrence of 15% (according to the current explant-based Milan criteria)?
The second issue is the predictability of recurrence (and thus the confidence limits for estimates) and the usefulness of this predictability in selection. In other words, are there tumor characteristics (eg, molecular profiling) that make recurrence more likely?
In this article, we assess the size and number of diagnosed HCC nodules in a cirrhotic liver as prognostic factors for recurrence and survival after transplantation because clinical experience has generally shown that both of these features have prognostic value. The number and particularly the size are related to the degree of microvascular invasion,2 which is a very strong prognostic factor. Indeed, Mazzaferro et al.3 showed that in a worst-case scenario (ie, microvascular invasion), there is a 50% chance of 5-year survival (which could be considered too low) with a single nodule up to 5 cm in diameter or with 3 or fewer nodules with individual diameters no greater than 3 cm. The nomogram in their article does not provide data for a 5-year survival rate of 70% (which is similar to the rates associated with other indications for transplantation).
With respect to the staging of HCC in the transplant setting, the following factors are important for the evaluation and interpretation of the literature data: (1) whether the staging is based on pretransplant imaging or explanted materials and what the differences are between them (with consideration also given to the waiting time on the list); (2) how the staging is classified [ie, whether the number of nodules and their sizes or the total tumor volume (ie, the total burden) is documented]; (3) whether “beyond the criteria” status is based on the size of the largest nodule, the number of nodules, or both when specific systems such as the Milan criteria and the University of California San Francisco (UCSF) criteria are being used; and (4) whether the outcomes include posttransplant survival, recurrence, or both.
CI, confidence interval; HCC, hepatocellular carcinoma; HR, hazard ratio; NS, not significant; UCSF, University of California San Francisco.
MATERIALS AND METHODS
Identification of Trials
The MEDLINE, Cochrane Central Register of Controlled Trials, Embase, and Science Citation Index databases were searched until April 2010. Comparative studies were identified with the following key words: carcinoma, hepatocellular, and liver transplantation. Validated search filters were used to identify only prognostic studies. Equivalent free-text terms were used. There were no language restrictions. We included only studies reporting hazard ratios (HRs) or Kaplan-Meier curves for death, HCC recurrence, or both according to (1) selection criteria (ie, the Milan criteria, the UCSF criteria, or the tumor-node-metastasis system), (2) the tumor size and volume, and/or (3) the number of tumors. For studies not reporting HRs, we performed calculations with the Excel sheet published by Tierney et al.,4 which is based on the Parmar method of extracting data from survival curves.5 This is a method suggested by the Cochrane Collaboration for time-to-event outcomes.6 When both pretransplant and posttransplant HRs based on staging were available, the pretransplant criteria were used for the meta-analysis.
The studies that were identified in the search were screened independently by G.G. and M.G. and then were verified reciprocally. Any disagreements were arbitrated by A.K.B.
The primary outcomes evaluated in this meta-analysis were survival (HRs for overall survival and disease-free survival) and recurrence (HRs for recurrence). We considered either distant or local recurrence of the disease to be disease recurrence. In addition, any deaths for which another cause was not given were considered to be due to recurrence of the disease.
The Milan criteria (a single nodule ≤ 5 cm in diameter or 3 or fewer nodules with individual diameters ≤ 3 cm) were derived from explanted specimens, and when they are used with pretransplant imaging, a recurrence rate less than or equal to 15% results.7 On the basis of 15 studies using Mazzaferro et al.'s criteria3 before transplantation and reporting the mean survival rates (and not the median rates), we found the average survival rate 5 years after transplantation to be 64.9%; on the basis of 13 studies using explant staging, we found the rate to be 73.4%. These figures can be used with the HRs presented in this article to estimate the increases or decreases in survival when different staging systems are used. However, when specific comparisons are made (ie, multiple nodules versus a single nodule), the baseline reference is the lesser tumor load.
The data were extracted with a predefined form independently by G.G. and M.G. Discrepancies were resolved by consensus and by arbitration by K.G. Data related to the following were extracted: the outcomes, the methodological quality, the country of origin, the year, the study design, the number of patients, the patient characteristics, the etiology of the liver disease, the size and number of HCC nodules, the presence of vascular invasion, the time on the waiting list, the use of bridge therapy before liver transplantation, the use of immunosuppressive therapy, and the median and minimum follow-up periods.
The results of the studies were combined by the generic inverse variance method with a random effects model. RevMan 5 from the Cochrane Collaboration was used for the meta-analysis.
Heterogeneity, Subgroup, and Sensitivity Analyses
Heterogeneity was identified by a visual inspection of Forest plots and with a Higgins I2 value of 30 or higher. A chi-square value for heterogeneity < 0.10 was considered to indicate statistically significant heterogeneity. We used a test for interactions8 to identify subgroup differences. A P value < 0.05 was considered to indicate statistically significant subgroup differences.
For the risk of bias, the main factors that we assessed were the attrition bias (due to the omission of some patients from the analysis), the baseline imbalance in factors other than the tumor size or number of nodules, and the adjustments for confounding variables.
A funnel plot was used to explore the publication bias. A visual inspection of the funnel plot was used to assess this bias. In addition, Egger's test was used to test the publication bias: a P value of 0.10 was considered statistically significant.
Description of the Studies
One hundred one studies were identified, and data from 74 of these studies were used for the quantitative analysis (Fig. 1). Thirty-four of the 74 studies were performed in Europe [France (2), Germany (7), Ireland (1), Italy (12), Spain (7), Sweden (1), and United Kingdom (4)], 20 were performed in North America [Canada (4) and the United States (16)], 19 were performed in Asia [China (7), Japan (7), and Korea (5)], and 1 was performed in Australia; there were 22,392 patients in all. Fifty-four of the 74 studies (73%) reported data on vascular invasion, 52 (70%) reported data on the tumor size, and 51 (69%) reported data on the number of tumors. Only 23 (31%) reported the length of stay on the waiting list, and 50 (68%) reported the number of patients who underwent preoperative treatments. Immunosuppressive therapy was reported in 41 of the 74 studies (55%), and the length of follow-up was included in 56 (76%). Most of the studies (65%) included in our meta-analysis were based on explant findings; only 35% assessed the impact of pretransplant tumor characteristics on survival and recurrence.
Overall Survival and Disease-Free Survival According to Different Transplant Criteria
Thirty-nine studies were included in the meta-analysis (8981 patients).7, 9–46 The HR for overall survival was reduced by 1.63 [95% confidence interval (CI) = 1.31-2.03] in patients beyond the Milan criteria versus patients within the Milan criteria. The HR for disease-free survival was reduced by 3.27 (95% CI = 2.12-5.05) in patients beyond the Milan criteria versus patients within the Milan criteria. The HR for recurrence was increased by 2.79 (95% CI = 1.71-4.54) in patients beyond the Milan criteria versus patients within the Milan criteria (Table 1).
Table 1. Summary of the Outcomes of Liver Transplantation for HCC and Cirrhosis According to the Selection Criteria, the Tumor Size, and the Number of Tumors
NOTE: This table is based on the studies included in the meta-analysis.
Thirteen studies were included in the meta-analysis (1987 patients).10, 26, 29, 31, 39, 40, 43, 47–52 The HR for ovreall survival was reduced by 1.79 (95% CI = 1-3.21) in patients outside the UCSF criteria versus patients within the UCSF criteria. The HR for disease-free survival was reduced by 3.41 (95% CI = 1.21-9.60) in patients outside the UCSF criteria versus patients within the UCSF criteria. The HR for recurrence was increased by 6.11 (95% CI = 2.45-15.23) in patients outside the UCSF criteria versus patients within the UCSF criteria (Table 1).
Within the Milan Criteria Versus Within the UCSF Criteria but Beyond the Milan Criteria
Three studies were included in the meta-analysis (7280 patients).47, 53, 54 The HR for overall survival was 1.37 (95% CI = 1.04-1.80) in patients outside the Milan criteria but within the UCSF criteria versus patients within the Milan criteria (Table 1). The HR for disease-free survival was 1.66 (95% CI = 1.17-2.34) in patients outside the Milan criteria but within the UCSF criteria versus patients within the Milan criteria (Table 1). No data were available on recurrence.
Overall Survival and Disease-Free Survival According to the Tumor Size and Volume
Total Tumor Diameter
Seven studies were included in the meta-analysis (1502 patients).20, 40, 41, 55–58
Two studies treated the total tumor diameter as a continuous variable (the summation of the diameters of all nodules), and both assessed recurrence (HR = 1.19, 95% CI = 0.95-1.49; Table 1).40, 58
Three studies used a cutoff of 10 cm, and they all assessed overall survival.20, 55, 57 The HR for overall survival was reduced by 4.59 (95% CI = 1.26-16.79) in patients with a total tumor diameter ≥ 10 cm versus patients with a total tumor diameter < 10 cm (Table 1).
Two studies used a cutoff of 9 cm, and both assessed disease-free survival.41, 56 The HR for disease-free survival was reduced by 1.98 (95% CI = 1.49-2.64) in patients with a total tumor diameter ≥ 9 cm versus patients with a total tumor diameter < 9 cm (Table 1).
Diameter of the Largest Tumor Nodule
Nine studies were included in the meta-analysis (2743 patients).2, 3, 16, 17, 37, 40, 48, 50, 59
Three studies treated the diameter of the largest tumor nodule as a continuous variable, and they all assessed recurrence (HR = 1.03, 95% CI = 0.99-1.07; Table 1).16, 37, 40
Six studies treated the diameter of the largest tumor nodule as a cutoff (≥3 versus <3 cm).2, 3, 17, 48, 50, 59 The HR for overall survival was reduced by 1.55 (95% CI = 1.29-1.86) in patients whose largest nodule had a diameter ≥ 3 cm versus patients whose largest nodule had a diameter < 3 cm. The HR for recurrence was increased by 6.69 (95% CI = 2.34-19.12) in patients whose largest nodule had a diameter ≥ 3 cm versus patients whose largest nodule had a diameter < 3 cm (Table 1).
Tumor Size Without Specifications
Nineteen studies were included in the meta-analysis (2497 patients).13, 27, 31, 44, 60–74
Four studies treated the tumor size as a continuous variable, and they all assessed overall survival (HR = 1.14, 95% CI = 1-1.30; Table 1).67, 69, 71, 74
Sixteen studies treated the tumor size as a cutoff.13, 27, 31, 44, 60–68, 70, 72, 73 The HR for overall survival was reduced by 1.92 (95% CI = 1.48-2.50) in patients with a tumor size ≥ 5 cm versus patients with a tumor size < 5 cm. The HR for disease-free survival was reduced by 4.30 (95% CI = 2.48-7.49) in patients with a tumor size ≥ 5 cm versus patients with a tumor size < 5 cm. The HR for recurrence was increased by 2.56 (95% CI = 1.53-3.34) in patients with a tumor size ≥ 5 cm versus patients with a tumor size < 5 cm (Table 1).
In the meta-analysis, only 1 study (reporting 2 case series) evaluated the overall survival rate with respect to the tumor size with a cutoff of 115 cm3.75 The HR for death was reduced by 8.40 (95% CI = 3.03-23.25) in patients with a total tumor volume ≥ 115 cm3 versus patients with a total tumor volume < 115 cm3 (Table 1). For a single lesion, this corresponded to a diameter of 6 cm. These 2 case series used a rough estimate of the total tumor volume that was based on the largest radius rather than the true radius (the first, second, and third radii of the tumor) or a volumetric analysis.
Overall Survival and Disease-Free Survival According to the Number of Tumor Nodules
Fifteen studies (4575 patients) evaluating the impact of the number of tumor nodules on overall survival and disease-free survival were included in the meta-analysis.3, 15, 16, 18, 31, 40, 41, 50, 62, 64, 67, 71, 72, 74, 76
Five studies that evaluated the tumor number as a continuous variable were included in the meta-analysis (719 patients).15, 16, 40, 71, 74 The HR for overall survival was 1.09 (95% CI = 0.88-1.34). The HR for recurrence was 1.07 (95% CI = 0.93-1.23; Table 1).
Three studies (468 patients) that evaluated the tumor number in terms of multiple nodules versus a single nodule were included in the meta-analysis.67, 72, 76 The HR for overall survival was reduced by 1.23 (95% CI = 1-1.53) in patients with multiple nodules versus patients with a single nodule (Table 1).
Seven studies (3289 patients) that evaluated the tumor number as a specific cutoff were included in the meta-analysis, and they all used <3 nodules and ≥3 nodules.3, 18, 31, 41, 50, 62, 64 The HR for overall survival was 1.29 (95% CI = 1.14-1.46) in patients with ≥3 nodules versus patients with <3 nodules. However, the HR for disease-free survival was not reduced significantly in patients with ≥3 nodules versus patients with <3 nodules (HR = 1.24, 95% CI = 0.77-2.01). The HR for recurrence was not significantly different in patients with ≥3 nodules versus patients with <3 nodules (HR = 1.02, 95% CI = 0.25-4.24; Table 1).
The purpose of this meta-analysis is a systematic review of all articles dealing with HCC and liver transplantation and reporting the staging of HCC in terms of the size and number of the nodules with respect to posttransplant recurrence and survival.
In comparison with patients meeting the Milan criteria, patients beyond the Milan criteria have worse overall and disease-free survival and experience more recurrence after liver transplantation for HCC. In comparison with patients beyond the UCSF criteria, patients within the UCSF criteria do marginally better in terms of overall survival after liver transplantation and have a higher rate of disease-free survival and a lower rate of recurrence. Transplant patients outside the Milan criteria but within the UCSF criteria have an increased risk of death in comparison with transplant patients within the Milan criteria, but this finding is based on only 3 studies. Moreover, when the UCSF criteria are extended from the Milan criteria, it is not clear whether the size of the largest nodule, the number of nodules, or both are critical permissive features.
Patients with a total tumor diameter ≥ 10 cm have an overall survival probability that is nearly 5 times lower than the overall survival probability of patients with a total tumor diameter < 10 cm. Patients with a total tumor diameter ≥ 9 cm have a probability of disease-free survival that is nearly 2 times lower than that of patients with a total tumor diameter < 9 cm. The UCSF criteria include a total tumor diameter up to 8 cm, and the Milan criteria include a total tumor diameter up to 9 cm.
When the diameter of the largest tumor nodule is considered, patients whose largest tumor nodule has a diameter ≥ 3 cm have a probability of recurrence after liver transplantation more than 6 times higher than that of patients whose largest tumor nodule has a diameter < 3 cm. Studies evaluating the impact of the tumor diameter with no other specification (ie, the total tumor diameter or the largest tumor diameter) have found that patients with a tumor diameter ≥ 5 cm have a probability of disease-free survival after liver transplantation approximately 4 times lower than that of patients with a tumor diameter < 5 cm. Moreover, patients with a tumor diameter ≥ 5 cm have a probability of recurrence after liver transplantation 2.56 times higher than that of patients with a tumor diameter < 5 cm. The impact on overall survival is not as large as the impact on disease-free survival.
When the number of tumors is evaluated as a continuous variable, it does not have a clear impact on overall survival or recurrence. However, when it is considered as a categorical variable (ie, multiple tumors versus a single tumor), the risk of recurrence after liver transplantation is doubled for patients with multiple tumors versus patients with a single tumor. The impact of multiple tumors on overall survival is borderline because the lower limit of confidence for the HR is 1. When the number of tumors is considered as a cutoff, patients with ≥3 tumors have an increased risk of death in comparison with patients with <3 tumors. However, the impact of 3 or more tumors on disease-free survival and recurrence after liver transplantation is not significant (NS), possibly because of the inclusion of only a small number of studies (2 and 3, respectively).
These findings and this analysis are extremely important: imaging has not particularly changed for the detection of nodules that are 2 cm in diameter or larger, but it has improved for the detection of smaller nodules (0.5-1.5 cm). If we assume that these nodules were missed on pretransplant imaging in the past (and were found during explant examinations only when specimens were carefully examined), we may already have a case for the number of nodules (particularly small nodules 1-2 cm in diameter) being less important for the risk of recurrence. Moreover, these small nodules are far less likely to be associated with microvascular invasion.2
We believe that selection criteria for transplantation in patients with cirrhosis and HCC, whether they are based on staging or other criteria, should have the following ground rules.
Newly selected patients should have intent-to-treat survival rates (from the time of listing) and posttransplant survival rates similar to those of patients without HCC. The criteria should be based on pretransplant imaging with accepted diagnostic criteria. The Milan criteria should be used as a reference, and the assessment should compare patients within the Milan criteria to patients beyond the Milan criteria but within any newly proposed criteria. Although the Milan criteria are based on explant materials, they are currently used before transplantation, and this should be noted. There is a survival difference of approximately 10%, and the survival rate based on pretransplant criteria is reduced. Many studies compare only patients within the Milan criteria to patients within the new criteria. The latter group thus also includes patients within the Milan criteria; this introduces a major bias, artificially improves the outcomes of this group, and prevents any meaningful assessment. Ideal selection criteria should lead to posttransplant survival rates similar to those of patients without HCC. This will avoid unfair competition on a common waiting list. A Markov model has been developed that compares the survival benefit of transplantation for a patient beyond the Milan criteria and the harm caused to other patients on the waiting list. It appears that a 5-year patient survival rate of at least 61% is required for selection criteria to be seen as ethically acceptable,77 but this may depend on donor availability. Interestingly, this is similar to the survival rate expected with pretransplant Milan criteria.
Intent-to-treat survival is a secondary outcome in comparisons of patients with HCC and patients without HCC. Any increase in the number of potential HCC candidates will lead to decreased intent-to-treat survival from the time of listing for patients without HCC (ie, there will be fewer grafts available for them), and this can be acceptable only if similar posttransplant outcomes are achieved.
Because major differences in staging policies have been observed between centers, selection criteria should be based on a standardized radiological assessment (including the type and the timing) so that the clinical applicability of a score can be tested and validated.78 The use of accurate radiology protocols is especially important when we are considering scores that include a cutoff for the number of HCC nodules because a small additional lesion can tilt a patient away from transplantation and toward palliative therapy.
However, there will always be a discrepancy between pretransplant staging and explant staging of HCC. Most often, pretransplant imaging leads to understaging because of the inherent limitations of the accuracy of imaging and because of lesion growth while the patient is waiting. This emphasizes the need for using pretransplant imaging as the baseline assessment for fulfilling the listing criteria.
We can surmise that using a strict tumor nodule number as a cutoff should be avoided because it might partially obviate the better resolution of current imaging (ie, the diagnosis of more nodules) and the use of pretransplant imaging versus explant imaging (which usually shows more nodules and nodule growth). A system taking into account larger lesions (eg, a diameter ≥ 1.5 cm) might be better, but this requires a formal evaluation.
Limitations of the Meta-Analysis
The studies included in the quantitative analysis have a high level of heterogeneity. Most of the studies do not report tumor characteristics (size, number, and, when available, volume) in a uniform way, and they use different cutoff values and continuous/dichotomous classifications for the total tumor diameter, the diameter of the largest tumor nodule, and the tumor size (with no other specification). Moreover, several parameters are missing, especially with respect to the time on the waiting list and the use of immunosuppression and bridge therapy.
Most of the studies included in our meta-analysis (65%) report findings based on explants, whereas only 35% of the studies assess the impact of pretransplant tumor characteristics on survival and recurrence. Moreover, only a few of the studies compare pretransplant imaging with explant findings for the prediction of survival and recurrence. This can influence the interpretation of the meta-analysis.
In conclusion, this meta-analysis demonstrates consistent evidence supporting the importance of the largest tumor nodule diameter or the total summation of the tumor nodule diameters with respect to both overall survival and disease-free survival after liver transplantation. This is clear especially for patients with a total tumor diameter ≥ 10 cm, whose risk of death or recurrence is more than 4 times higher than the risk for patients with a total tumor diameter < 10 cm (ie, patients within the maximal Milan criteria).
The current analysis does not provide clear evidence for the extension of the Milan criteria based on pretransplant imaging. We assume that small nodules (probably ≤1-1.5 cm) that are not seen on pretransplant imaging (but are diagnosed by a careful examination of the explant) are not influencing outcomes. The extended rule of 7 by Mazzaferro et al.3 (based on explant data) does not help. When the original Milan criteria are used with pretransplant imaging and without biopsy for the assessment of microvascular invasion (which is the norm), they are the maximum inclusion criteria for the worst hypothesis (ie, microvascular invasion), and an average survival rate of only 50% is achieved at 5 years.
New studies should conform to standards for reporting methods and for the type and timing of imaging and the number of nodules (with the maximum diameters based on pretransplant imaging and explant specimens). The methods of surveillance for recurrence and the sites of recurrence should be described. Deaths due to recurrence and other causes should be recorded.