ABSTRACT
-
Background/Aims
Various expanded criteria (EC) for liver transplantation (LT) in patients with hepatocellular carcinoma (HCC) have been proposed to avoid the narrow nature of the Milan criteria (MC). To investigate which EC predicts more favorable outcomes in terms of overall survival (OS) and recurrence-free survival (RFS), we conducted a network meta-analysis (NMA).
-
Methods
A database search was conducted on PubMed, Embase, and the Cochrane Library, to identify studies comparing OS and RFS between patients within the MC and those exceeding the MC but within the EC. Hazard ratios (HRs) were pooled using a random-effects NMA and validated in an in-house cohort of 1,008 LT recipients.
-
Results
Among 22,466 articles identified, 35 studies with 45 pairwise comparisons were included in the NMA along with 8 different EC. The University of California San Francisco (HR, 1.43; 95% CI, 1.19–1.71), Up-to-Seven (HR, 1.50; 95% CI, 1.15–1.97), and Hangzhou criteria (HR, 1.69; 95% CI, 1.11–2.57) showed inferior OS to the MC. The MC ranked highest for both OS and RFS, followed by Metroticket 2.0 for OS and the Asan criteria for RFS. In the validation cohort, both Metroticket 2.0 and AFP model yielded more favorable HCC-specific mortality than other EC.
-
Conclusions
Several EC, of which those of Metroticket 2.0 were the best, yielded comparable outcomes to the MC. AFP-based EC such as Metroticket 2.0 and AFP model appeared to be useful in both the NMA and the validation cohort, suggesting a potential role in identifying selected low-risk patients beyond the MC.
-
Keywords: Carcinoma, hepatocellular; Liver transplantation; Prognosis; Meta-analysis
Study Highlights
• We performed a network meta-analysis comparing the Milan criteria with eight expanded criteria for LT in HCC and validated the findings in an independent cohort of 1,008 LT recipients at Asan Medical Center.
• In the network meta-analysis, Metroticket 2.0 showed the second highest rank probability for OS, following the Milan criteria. In the validation cohort, the AFP model and Metroticket 2.0 showed relatively lower HCC-specific mortality than the other expanded criteria.
• This study suggests that AFP-based expanded criteria, such as Metroticket 2.0 and AFP model, may help identify selected low-risk patients beyond the Milan criteria.
Graphical Abstract
INTRODUCTION
Liver transplantation (LT) is an effective treatment option for unresectable, early-stage hepatocellular carcinoma (HCC) patients that can simultaneously resolve the tumor itself and the underlying cirrhotic liver. Overall survival (OS) after LT for patients with HCC within the Milan criteria (MC) reaches 80% at 5 years, justifying cadaveric organ transplantation for patients with HCC [
1]. To avoid excluding patients from transplant opportunities because of the strictness of the MC, several expanded criteria (EC) have been proposed for patients exceeding the MC. These EC yield outcomes comparable to those achieved under the MC [
2]. Living-donor liver transplantation (LDLT), widely performed in Asia due to the scarcity of deceased donor grafts, has demonstrated safety and outcomes comparable to deceased donor liver transplantation (DDLT) [
3]. This availability has enabled the development of diverse EC across centers.
There are EC that only expand the size limit and number of tumors compared to the MC [
4-
7]. Other EC incorporate components reflecting tumor biology such as alpha-fetoprotein (AFP), protein-induced by vitamin K absence or antagonist-II, and data on differentiation in biopsy [
8-
11]. Several of EC, University of California San Francisco (UCSF) criteria, Up-to-Seven criteria, and Metroticket 2.0 model, have been validated in large series studies, and a number of other EC are being applied in practice [
4,
6,
9]. Few studies have directly compared these EC. Lozanovski et al. [
12] reported that the Metroticket 2.0 model achieved the highest recurrence-free survival (RFS) among EC; however, they compared patients within the MC to those within the EC, which resulted in considerable overlap between the two groups, potentially reducing statistical power.
It is important to clearly define the extent to which the MC can be expanded. To evaluate whether specific EC offer more favorable outcomes in terms of OS and RFS, we conducted a network meta-analysis (NMA) comparing the performance of various EC among patients within the MC and those exceeding the MC but within each EC. We further validated the NMA findings using a large in-house cohort of LT recipients.
MATERIALS AND METHODS
This study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension statement for network meta-analysis (PRISMA-NMA) [
13], and it was registered in the International prospective register of systematic reviews (PROSPERO) as registration number CRD42024569587. All research was conducted in accordance with both the Declaration of Helsinki and Istanbul. The Institutional Review Board of Asan Medical Center, Seoul, Republic of Korea, approved this study-level NMA, and waived the informed consent for individual patients (IRB No. 2020-1501).
Studies that compared the performance of EC with the MC were eligible for this NMA. PICO (population, intervention, comparison, and outcome) were as follows. Participants were patients diagnosed with HCC who underwent LT. The intervention groups were patients who exceeded the MC but fulfilled one or other EC, and the comparison groups were patients within the MC. Outcomes were OS, RFS, and time to recurrence (TTR). Studies were excluded for the following reasons: (1) no reported Kaplan-Meier curve; (2) publications not in English; (3) conference abstracts, review articles, and retracted articles; (4) observation periods of less than 5 years, to ensure adequate follow-up for capturing both early and late post-transplant events and maintaining consistency in survival endpoint definitions across studies [
14]; and (5) studies employing EC described in only one article and therefore not validated, to minimize potential small-study effects and ensure the validity of our network estimates.
The databases of PubMed, Embase, and the Cochrane Library were searched (last search date: December 4, 2024). The key search terms for the various EC were obtained by a scoping search. The three keywords “hepatocellular carcinoma,” “liver transplantation,” and “LT criteria” were combined using the AND operator. To maximize the sensitivity of the search, we combined MeSH term and text word using the OR operator. The detailed search strategy is presented in
Supplementary Table 1. Two independent reviewers (DY and JHS) checked all titles and abstracts identified by the database searches and then carried out a full-text review of potentially relevant studies. At the full-text review step, the reasons for excluding studies were recorded.
The following data were extracted from each selected study: author, publication year, country, applied EC, number of patients, method of staging (radiologic or pathologic), type of LT donor (deceased or living), age (years), sex (%), proportion of patients with hepatitis B virus (HBV) and hepatitis C virus (HCV) infection, and alcoholic liver disease (%), proportion of patients who received pre-LT treatment (%), follow-up duration (months), and 1-, 3-, and 5-year OS, RFS, and TTR. OS was defined as the time between LT and death; otherwise, patients were censored at the time of last known contact. RFS was defined as the time from LT to either recurrence or death. TTR was defined as the time from LT to HCC recurrence.
Cases where two different OS or RFS outcomes were reported depending on the method of staging were considered to involve different cohorts. All percentage survivals were extracted from published Kaplan-Meier curves using GetData Graph Digitizer (version 2.24) software.
Risk of bias assessment
The Risk of Bias in Non-randomized Studies of Interventions Version 2 (ROBINS-I V2) tool was used to assess the risk of bias [
15]. Bias was assessed across seven domains: confounding, participant selection, intervention classification, deviations from intended interventions, missing data, outcome measurement, and selective reporting. Each bias domain included signaling questions guiding the domain-level risk assessment and was evaluated independently by two reviewers (DY and YSH). The overall bias was categorized as low, moderate, or serious.
To validate the NMA findings, we utilized a registry of LT recipients aged ≥18 years who were prospectively enrolled at Asan Medical Center between January 2010 and December 2017 [
16]. Inclusion criteria were: (1) treatment-naïve patients with Liver Imaging Reporting and Data System (LIRADS) category 5 (LR-5, definitely HCC) or biopsy-confirmed HCC, and (2) patients who underwent downstaging or bridging therapy prior to transplantation for HCC, following multidisciplinary team decision-making [
9,
14]. For all included patients, hepatic lesions were evaluated using dynamic contrast-enhanced computed tomography (CT) within one month prior to LT. A total of 134 patients were excluded due to either inapplicability of LI-RADS or challenges in correlating radiologic and pathologic findings, as illustrated in
Supplementary Figure 1 [
17,
18]. The final cohort included 1,008 patients with HCC risk factors as defined by LI-RADS, including chronic hepatitis B, liver cirrhosis of any etiology, or a prior history of HCC. Post-transplantation surveillance for HCC recurrence consisted of routine imaging and serological assessments. Dynamic CT or magnetic resonance imaging (MRI) was typically performed every 2–3 months during the first year, with the interval extended to every 3–6 months thereafter, depending on the risk of recurrence. Further details regarding the radiologic and pathologic assessment of the validation cohort are provided in the
Supplementary Materials.
To simultaneously assess the comparative effects of the MC and EC on LT, a frequentist NMA with a random effects model was conducted based on weighted least squares regression. Hazard ratios (HRs) for survival outcomes between patients within the MC and those beyond the MC but within the EC were estimated using the method described by Tierney et al. [
19] The survival data up to 5 years were used to estimate the HR. Statistical heterogeneity was assessed using I
2 statistics. The pooled HRs for survival outcomes were depicted by forest plots. The probabilities of the criteria for each possible ranking were visualized using a composite line chart. League tables showing all pairwise comparisons were also generated. To detect moderator effects, the association between demographic and outcome variables was examined using a meta-regression analysis of the EC with data available from at least two studies per covariate [
20]. The demographic variables used were as follows: age, sex, publication year, method of assessing the criteria, study region, HBV (%), HCV (%), DDLT (%), and pre-LT treatment (%). Stratified analyses were performed to assess whether study-level characteristics influenced the comparative effectiveness of the LT selection criteria. Studies were grouped according to key predefined characteristics, including study region (Asia or non-Asia), predominant etiology (HBV-or HCV-predominant cohorts, defined as ≥60% of cases), LT type (LDLT or DDLT, defined as ≥60% of cases), use of pre-transplant anti-HCC treatment (≥60% of patients), method of assessing the criteria (imaging-based or pathology-based), and transplantation period, categorized into pre- and post-tyrosine kinase inhibitor (TKI) eras (cutoff year 2008) and pre- and post-pangenotypic direct-acting antivirals (DAA) eras (cutoff year 2014). Publication bias was assessed using comparison-adjusted funnel plots and Egger’s tests to explore potential publication bias or small-study effects across comparisons [
21,
22].
Since the boundary of the AFP model in a subset of included studies does not fully encompass the MC, four subgroups could arise when applying the MC and AFP model simultaneously (i.e., within the MC+AFP score≤2, beyond the MC+AFP score≤2, within the MC+AFP score>2, and beyond the MC+AFP score>2). To examine post-LT outcomes in patients who met both the MC and the AFP model versus those who exceeded the MC but met the AFP model, we conducted a pairwise meta-analysis of these studies using OS and TTR as outcome variables.
Competing risk analyses, accounting for non–HCC-related mortality, were performed to evaluate HCC-specific deaths within the validation cohort for both radiology- and pathology-based LT eligibility criteria. Deaths occurring after documented post-transplant HCC recurrence were classified as “HCC-related,” whereas all other causes were categorized as “non–HCC-related.” Cumulative incidence functions and subdistribution hazard ratios (sHRs) for HCC-specific mortality were estimated and compared across the MC and each expanded criterion using Fine–Gray’s method. 23 In addition, subgroup analyses were performed within the clinically predominant strata of the validation cohort, including patients with HBV infection, those undergoing LDLT, and those who received pre-transplant anti-HCC treatments.
The ‘netmeta’ package was used to perform the frequentist NMA; the ‘rankinma’ package was used to estimate the ranking of LT eligibility criteria; the ‘meta’ package was used to perform the pairwise meta-analysis and meta-regression analysis; and the ‘cmprsk’ package was used to perform competing risk analysis. A P-value of <0.05 was considered statistically significant. All data analyses were performed using R software (version 4.2.2).
RESULTS
Study selection and characteristics
Figure 1 presents the flowchart of the study selection process. A total of 22,466 potentially relevant articles were identified through the database search. After removing duplicates, 17,836 records remained for title and abstract screening, leaving 179 articles for full-text review. Of these, 140 full-text articles were excluded for the reasons detailed in
Figure 1. Ultimately, 35 studies comprising 45 pairwise comparisons were included in the NMA. Together with the MC, eight EC were included: the AFP model, Asan, Hangzhou, Metroticket 2.0, R4 T3, Shanghai, UCSF, and Up-to-Seven. The definitions of these criteria are summarized in
Supplementary Table 2. In addition, four studies providing five pairwise comparisons were included in a separate pairwise meta-analysis comparing patients who met both the MC and the AFP model with those who exceeded the MC but met the AFP model [
8,
24-
26]. The full list of the studies included in this meta-analysis is provided in the
Supplementary Material. The studies excluded based on the predefined criteria are summarized in
Supplementary Table 3 and
Supplementary Table 4.
The characteristics of the studies included in the NMA are summarized in
Table 1. Of the studies (n=35) analyzed, 14 (40%) were conducted in Europe, nine (26%) in North America, eight (23%) in Asia, and one (3%) in South America. The remaining three studies were published in Iran, Australia, and Egypt. The median follow-up duration ranged from 23 to 79.2 months. The proportion of patients receiving pre-LT treatment ranged from 20.2% to 100% and was more than 50% in 18 studies. Twenty studies mainly included patients with HCV, whereas nine studies focused on HBV infection as the primary cause. The data collection periods ranged from 1984 to 2019. Most studies reported OS (31/35) and nearly half reported RFS (18/35). Thirteen studies applied radiologic staging, 16 applied pathologic staging, and four applied both radiologic and pathologic staging, reporting their outcomes separately. DDLT and LDLT were performed in 14 and 3 studies, respectively, with 8 studies incorporating both. The network geometry of the studies is presented in
Figure 2 and the MC served as the basis for comparison in all the included studies. Nineteen studies compared the MC with UCSF, eight with Up-to-Seven, four with Hangzhou, and three with Asan criteria. The AFP model, Metroticket 2.0, R4 T3, and Shanghai criteria were compared with the MC in two studies, respectively.
Thirty-nine pairwise comparisons were included in the NMA for OS (
Fig. 3A). The UCSF (HR: 1.43; 95% confidence interval [CI]: 1.19–1.71), Up-to-Seven (HR: 1.50; 95% CI: 1.15–1.97), and Hangzhou criteria (HR: 1.69; 95% CI: 1.11–2.57) resulted in significantly poorer OS than the MC. The MC had the highest rank probability of being the most effective followed by the Metroticket 2.0 model in terms of OS (
Fig. 3C). The following criteria had significantly poorer RFS than the MC by 22 pairwise comparisons: UCSF (HR: 1.41, 95% CI: 1.20–1.66), Up-to-Seven (HR: 2.18, 95% CI: 1.54–3.08), Hangzhou criteria (HR: 1.74, 95% CI: 1.46–2.08), and AFP model (HR: 2.86, 95% CI: 1.06–7.72) (
Fig. 3B). The MC had the highest rank probability of being the best followed by the Asan criteria in terms of RFS (
Fig. 3D). In pairwise comparisons as a league table (
Table 2), there was no statistically significant difference among the EC except for direct comparisons of the MC with UCSF, Up-to-Seven, and Hangzhou criteria in terms of OS. In terms of RFS, R4 T3 (HR: 0.61, 95% CI: 0.39–0.96), Shanghai (HR: 0.62, 95% CI: 0.39–0.99), and UCSF criteria (HR: 0.65, 95% CI: 0.44–0.95) were superior to Up-to-Seven criteria.
Meta-regression analysis was further conducted to assess whether specific study characteristics influenced the overall effect size when comparing the EC with the MC (
Supplementary Table 5). The covariates considered included sex, age, year of publication, study region, method of assessing the criteria, proportion of DDLT, etiology of liver disease, and proportion of patients receiving pre-transplant treatment. Among these, only HCV infection showed a significant association with RFS in studies evaluating the UCSF criteria (coefficient=–0.021; 95% CI, –0.039 to –0.002;
P=0.026). No significant associations were observed for OS or RFS in studies assessing the other EC.
To assess the consistency of the findings across different patient groups, stratified analyses were performed according to the study region, predominant etiology, type of LT, method of assessing the criteria, and transplantation period. Regarding the study region, both the Metroticket 2.0 and AFP models showed no significant differences in OS compared with the MC in studies conducted outside Asia, consistent with the main analysis (
Supplementary Fig. 2). However, no studies from Asia were available for these two models. When stratified by etiology, in studies where HCV infection accounted for ≥60% of cases, the Metroticket 2.0 demonstrated comparable RFS to the MC (
Supplementary Fig. 3). In contrast, in studies including ≥60% of HBV-infected patients, the Asan, UCSF, Hangzhou, Shanghai, and Up-to-Seven criteria were eligible for OS analysis; among these, only the Asan criteria showed no significant difference from the MC.
With respect to LT type, among studies with ≥60% DDLT, the UCSF criteria were associated with significantly worse OS and RFS outcomes compared with the MC (
Supplementary Fig. 4). Conversely, among studies with ≥60% LDLT, the UCSF criteria showed no significant difference from the MC. Data on the LT type were insufficient for the stratified analyses of the AFP model and Metroticket 2.0. When restricting the analysis to studies in which ≥60% of patients received pre-transplant anti-HCC treatment, both the UCSF and Up-to-Seven criteria were associated with inferior OS compared with the MC, whereas RFS did not significantly differ among these criteria (
Supplementary Fig. 5). Furthermore, when stratifying studies according to the method used to assess the LT selection criteria, we found that in studies evaluated using pre-transplant radiologic imaging, both the AFP model and the Metroticket 2.0 model demonstrated survival outcomes comparable to those of the MC (
Supplementary Fig. 6).
Regarding the transplantation period, the analyses were stratified by the pre- and post-TKI eras, using 2008 as the cutoff, and the pre- and post-DAA eras, using 2014 as the cutoff. In the pre-TKI era, the UCSF criteria were associated with significantly worse OS and RFS compared with the MC, whereas this difference was no longer observed in the post-TKI era (
Supplementary Fig. 7). Before the introduction of DAAs, both the AFP and Metroticket 2.0 models showed comparable OS to that of the MC (
Supplementary Fig. 8). However, only Hangzhou-based studies were available in the post-DAA era, limiting meaningful cross-criteria comparisons.
Thirty studies (76.9%) were assessed as having a moderate overall risk of bias, primarily due to baseline confounding or deviations from the intended intervention. Five studies (12.8%) were rated as having a low overall risk of bias, whereas four studies (10.3%) were judged to have a serious overall risk of bias, mainly attributable to baseline confounding. Details of the quality assessment using the ROBINS-I V2 tool are presented in
Supplementary Figure 9 and
Supplementary Table 6. Sensitivity analyses were conducted after stratifying the studies according to the overall risk of bias. Restricting the analysis to studies with low-to-moderate risk of bias yielded results consistent with the main findings (
Supplementary Fig. 10). The symmetrical pattern of the comparison-adjusted funnel plots for OS and RFS, together with the non-significant Egger’s test results (
P=0.262 and
P=0.391, respectively), indicated no evidence of small-study effects or publication bias (
Supplementary Fig. 11).
The AFP model does not fully encompass the MC; some patients may exceed the AFP model while meeting the MC, whereas others may satisfy the AFP model but exceed the MC. Four studies have compared post-LT outcomes between patients meeting the MC as well as having an AFP score≤2 (meeting the AFP model) and those exceeding the MC but with AFP score≤2 and characteristics are summarized in
Table 3 [
8,
24-
26]. Two studies were conducted in Europe, one was in South America, and another one was in China. Duvoux et al.’s [
8] study population was composed of a training cohort and a validation cohort and we handled them as different cohorts in the meta-analysis. No significant differences in OS (HR: 1.19, 95% CI: 0.59–2.42,
Fig. 4A) or TTR (HR: 1.33, 95% CI: 0.72–2.48,
Fig. 4B) were observed, suggesting that meeting the AFP model indicates a comparable prognosis, even for patients exceeding the MC.
The validation cohort consisted of 1,008 patients: 83.7% of whom were male and the median age at LT was 55 years (interquartile range, 51–60 years). HBV infection was the predominant cause of chronic liver disease, affecting 84.5% of the patients, and LDLT was performed in 95.5% of the cases (
Supplementary Table 7). Prior to LT, 855 patients (84.8%) received anti-HCC treatments. Of these patients, 915 patients (90.8%) met the MC by LI-RADS, while 801 (79.5%) met the MC based on pathology. Both pathology- and radiology-based assessments were used to compare HCC-specific mortality between patients within the MC and those exceeding the MC but within the EC using competing risk analysis. Across nearly all the groups exceeding the MC but within the EC (except for the AFP model based on radiologic basis), HCC-specific mortality was significantly higher than in the patients within the MC (Fine and Gray test;
P<0.05). Of the EC, the AFP model (pathology-based sHR: 2.46, 95% CI: 1.49–4.05,
P<0.001; radiology-based sHR: 1.79, 95% CI: 0.92–3.50,
P=0.087) and the Metroticket 2.0 (pathology-based sHR: 2.21, 95% CI: 1.25–3.93,
P=0.006; radiology-based sHR: 2.68, 95% CI: 1.38–5.21,
P=0.004) yielded relatively lower sHRs than the other EC in both the radiologic and pathologic assessments (
Fig. 5 and
Supplementary Figs. S12,
S13). Furthermore, subgroup analyses restricted to patients with HBV infection (n=852), those who underwent LDLT (n=963), and those who received pre-transplant HCC treatment (n=855) showed that the AFP model and Metroticket 2.0 continued to yield the lowest sHRs among the EC (
Supplementary Fig. 14).
DISCUSSION
Various criteria have been proposed in order to expand the eligibility for LT of patients with HCC exceeding the MC, although global practice guidelines do not yet recommend any specific EC [
2,
14,
27,
28]. In the present NMA, unlike previous studies that primarily compared patients within the MC and within the EC groups or relied on effect measures such as odds ratios that do not adequately capture time-to-event outcomes [
12,
29-
31], we aimed to provide an objective evaluation of the prognostic performance of each expanded criterion. This was achieved by comparing post-transplant outcomes between patients within the MC and those beyond the MC but fulfilling each EC, using HRs to appropriately model the time-to-event data. Notably, among the EC, the Metroticket 2.0 demonstrated the most favorable predictive performance for both OS and RFS. The pairwise meta-analysis demonstrated no significant differences in OS or TTR between patients who met both the MC and the AFP model and those who exceeded the MC but met the AFP model, indicating comparable prognostic performance. Consistently, in the validation cohort, patients beyond the MC but fulfilling the Metroticket 2.0 or AFP model achieved the most favorable post-transplant survival among EC. Taken together, these findings support the potential utility of biologically integrated models for refined candidate selection beyond conventional morphologic criteria.
In our analysis and a previous analysis [
12], the Metroticket 2.0 model yielded more favorable outcomes in terms of survival, compared with other EC. This finding implies that considering both morphology and biology is a more effective approach for selecting patients for LT in the context of HCC. This model raises the transplantation cutoff to 3 (within the up-to-7 criteria, if AF
P<200 ng/mL; within the up-to-5 criteria, if AFP 200–400 ng/mL; within the up-to-4 criteria, if AFP 400−1,000 ng/mL) and offers personalized survival predictions at various time points both in treatment-naïve patients and patients who had undergone downstaging or bridging therapies. Consistent with its performance in the overall analysis, Metroticket 2.0 remained robust in the stratified analyses, demonstrating favorable outcomes particularly in non-Asian cohorts and HCV-predominant populations. Moreover, its predictive accuracy was further supported by our independent validation in an Asian, HBV-predominant cohort, in which the model also performed well. These results reinforce the generalizability and clinical applicability of the Metroticket 2.0 across geographically and etiologically diverse LT populations.
The AFP model, alongside Metroticket 2.0 using the same biomarker, has been shown to outperform the MC in identifying candidates with low risk of HCC recurrence or who will survive for 5 years after LT [
32]. Interestingly, 5-year recurrence probabilities varied depending on whether patients within or beyond the MC met the AFP model [
24]. In another study, use of the AFP model showed that patients with HCC who fulfilled the MC and received LT could be distinguished as low-risk or high-risk for tumor recurrence at 2 years [
33]. Although the AFP model did not demonstrate superior RFS compared with the MC in our NMA, this result warrants cautious interpretation. Patients beyond the MC but within the AFP model exhibited an unusually high prevalence of microvascular invasion, a pivotal predictor of post-LT recurrence [
6], approaching 50% in the study by Grąt et al. [
34], which was the sole study contributing AFP model data to the NMA for RFS, compared with 30.3% reported in another study of similar patient populations and 17.3% observed in our validation cohort (
Supplementary Table 8) [
24]. Nevertheless, both our pairwise meta-analysis and the findings from our validation cohorts were consistent with prior observations, showing comparable post-transplant outcomes between patients who met both the MC and the AFP model and those who exceeded the MC but fulfilled the AFP model. These observations suggest that the AFP model may help identify biologically favorable, low-risk patients beyond the MC. In particular, patients exceeding the MC but within the AFP model may still be considered suitable candidates for LT, especially in LDLT-dominant regions such as Asia, where broader yet biologically rational selection criteria may be clinically applicable.
The significance of serum AFP levels in selecting appropriate candidates for LT among patients with HCC has been emphasized in liver allocation policies across various countries globally. For instance, an AFP level greater than 1,000 ng/mL is considered an absolute contraindication for LT in both the United States [
35] and the United Kingdom [
36]. In France, the AFP model is officially implemented for candidate selection [
37], while in Japan, eligibility criteria for insurance coverage of LT have been expanded to include patients who meet either the MC or the 5-5-500 criteria (≤5 cm, ≤5 tumors, AFP≤500 ng/mL) [
38]. The predictive value of AFP for tumor recurrence following LT has also been demonstrated in several other models, such as the Warsaw criteria in Poland [
39], Munich criteria in Germany [
40], the TTV/AFP criteria in Canada [
41], and Wan et al.’s [
42] criteria in China which were not included in our final NMA due to no validation reports on their performance. Additional NMA including these four models supported the comparable outcomes of all the criteria incorporating AFP (
Supplementary Fig. 15 and
Supplementary Table 9). Notably, AFP is generally not expressed in well-differentiated HCC but is expressed in moderately to poorly differentiated HCC [
43]; thus, excluding patients with elevated AFP levels may help prevent transplantation in cases with unfavorable tumor biology. Moreover, AFP has been shown to be an independent predictor for tumor recurrence after LT [
44]. Therefore, it is prudent to simultaneously use tumor burden and AFP levels when selecting candidates for LT for HCC to address the limitations of the MC, which could lead to the transplantation of patients at high risk of recurrence. In accordance with global organ allocation policies, our NMA findings support the notion that complacency regarding MC could jeopardize post-LT outcomes, and that efforts should be made to adopt AFP-based models that surpass the performance of the MC.
In our analysis, the application of the UCSF and Up-to-Seven criteria, together with the Hangzhou criteria, was associated with significantly poorer OS and RFS compared with the MC. However, these findings should be interpreted with caution. First, the number of studies comparing the UCSF (20 direct comparisons) and Up-to-Seven criteria (seven direct comparisons) with the MC was substantially larger than for other EC (1–3 direct comparisons) (
Fig. 2), raising the possibility of a small-study effect that may have underestimated the performance of the UCSF and Up-to-Seven criteria while overestimating that of others [
45]. Second, in our meta-regression analysis, HCV prevalence was found to influence the RFS outcomes of studies evaluating the UCSF criteria. Stratified analysis by pre- and post-DAA eras could not be performed because no UCSF-based studies were conducted in the post-DAA period. Considering previous reports demonstrating reduced recurrence and mortality after HCV cure in the LT setting [
46,
47], it is reasonable to assume that the impact of HCV infection on post-transplantation outcomes would be attenuated in the current DAA era. In addition, our stratified analyses suggested that UCSF outcomes tended to improve in studies with a high proportion of patients receiving pre-transplant anti-HCC treatment and in those conducted during the post-TKI era, further supporting the possibility that therapeutic advances may mitigate the inferior outcomes associated with the UCSF criteria. Despite these potential limitations and influences, in our large independent validation cohort, which was predominantly HBV-related, the Metroticket 2.0 and AFP models consistently outperformed the UCSF and Up-to-Seven criteria, in line with the meta-analysis findings. This concordance strengthens the credibility of our results and supports the broader applicability of these prognostic models.
In the validation cohort, competing risk analysis was applied to assess HCC-specific mortality, allowing oncologic outcomes to be distinguished from overall post-LT survival. This approach provides a clearer understanding of the prognostic significance of the selection criteria and was also the basis for the Metroticket 2.0 model, which demonstrated consistent prognostic value [
9,
48]. Furthermore, standardized pre-LT radiologic evaluation using LI-RADS, with explant pathologic correlation, strengthened prognostic assessment and addressed the limitations of most NMA studies that did not specify radiologic methodology [
2,
14,
49]. In addition, because the majority of donors were living donors, potential bias related to the more frequent allocation of marginal cadaveric organs to patients beyond the MC was minimized. Collectively, these strengths support the robustness of our validation analysis and suggest that survival differences primarily reflect tumor biology and selection criteria, rather than organ allocation variability.
There are several limitations to this study. First, many other EC were not included in our NMA because survival data regarding patients within the MC and those exceeding the MC but meeting other EC were not reported in the original studies, and such data are essential for conducting an NMA (e.g., Toronto criteria [
10], Kyoto criteria [
11], and Kyushu criteria [
50]). Secondly, the imaging modalities used for diagnosing HCC may have varied across studies. Unfortunately, some studies did not provide specific comments regarding diagnostic criteria and modalities. Additionally, the methods for evaluating treatment responses in patients who underwent loco-regional therapy prior to LT, may have differed across studies (e.g., Response Evaluation Criteria in Solid Tumors [RECIST v1.1] or modified RECIST). Third, some degree of error is inevitable when extracting data from Kaplan-Meier curves. To minimize this error, we used digitization software, which is known to provide accurate estimates of survival probabilities, medians and HRs [
51]. We also included numbers of patients at risk on each arm at regular time intervals, if available, to improve the accuracy of the approximated time-to-event data [
52]. Fourth, because the network plot was star-shaped with a single common comparator, i.e., MC, statistical assessment of inconsistency was not feasible. Finally, although potential effect modifiers (e.g., pre-LT anti-HCC therapy, LT type, underlying liver disease etiology, and geographic region) were examined through meta-regression, subgroup analyses, and validation in an independent cohort, their effects could not be fully delineated, highlighting the need for future individual patient data meta-analyses.
In conclusion, our comprehensive NMA showed that the use of several sets of the EC led to comparable outcomes to the MC, thus supporting the use of such EC in the relevant centers. In particular, AFP-based EC such as the Metroticket 2.0 and AFP model appeared to be useful in both the meta-analyses and the validation cohort. Further analyses such as meta-analysis of individual patient data may offer an effective approach to addressing significant confounding factors, including differences in the etiology of liver disease.
FOOTNOTES
-
Authors’ contributions
Study concept and design, data acquisition, analysis, and interpretation: Dongman Yu, Yeongseok Hwang, JinSung Ju, Subin Heo, Sang Hyun Choi, Ju Hyun Shim, and Jihyun An. Manuscript draft: Dongman Yu, Yeongseok Hwang, Jin-Sung Ju, Seon-Ok Kim, Subin Heo, Sang Hyun Choi, Gi-Won Song, Jihyun An, and Ju Hyun Shim. Critical revision of the manuscript for important intellectual content: Dongman Yu, Yeongseok Hwang, Jin-Sung Ju, Seon-Ok Kim, Subin Heo, Sang Hyun Choi, Gi-Won Song, Jihyun An, and Ju Hyun Shim. Verification of the underlying data: Dongman Yu, Yeongseok Hwang, Jin-Sung Ju, Subin Heo, Sang Hyun Choi, Gi-Won Song, Jihyun An, and Ju Hyun Shim. Statistical analysis: Dongman Yu, Yeongseok Hwang, and Seon-Ok Kim.
-
Acknowledgements
This study was supported by grants from the National Research Foundation of Korea funded by the Ministry of Science and ICT (NRF-2022R1A2C3008956 and NRF-2021R1A6A1A03040260), Asan Institute for Life Sciences (grant number: 2022IP0046) from the Asan Cancer Institute of Asan Medical Center, and Scientific Research Fund of the Korean Liver Cancer Association (2025). The grant sources were not involved in the design of the study, the collection, analysis, and interpretation of data, the writing of the report, or the decision to submit the paper for publication.
-
Conflicts of Interest
The authors have no conflicts to disclose.
SUPPLEMENTARY MATERIAL
Supplementary material is available at Clinical and Molecular Hepatology website (
http://www.e-cmh.org).
Supplementary Table 5.
Meta-regression analysis according to overall survival and recurrence-free survival comparing use of the expanded criteria with the Milan criteria
Supplementary Figure 1.
Patient flow diagram in the validation cohort. HCC, hepatocellular carcinoma; LT, liver transplantation; LI-RADS, Liver Imaging Reporting and Data System.
Supplementary Figure 2.
Forest plots of the network meta-analysis stratified by study region for overall survival and recurrence-free survival. (A) Studies conducted in Asia. (B) Studies conducted outside Asia. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Supplementary Figure 3.
Forest plots of the network meta-analysis stratified by predominant etiology for overall survival and recurrence-free survival. (A) Studies including ≥60% hepatitis B virus-infected patients. (B) Studies including ≥60% hepatitis C virus-infected patients. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Supplementary Figure 4.
Forest plots of the network meta-analysis stratified by type of liver transplantation for overall survival and recurrence-free survival. (A) Studies including ≥60% deceased-donor liver transplantation cases. (B) Studies including ≥60% living-donor liver transplantation cases. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco.
Supplementary Figure 5.
Forest plots of the network meta-analysis restricted to studies in which ≥60% of patients received pre-transplant anti-HCC treatment for overall survival and recurrence-free survival. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco.
Supplementary Figure 6.
Forest plots of the network meta-analysis stratified by the method of criteria assessment for overall survival and recurrence-free survival. (A) Radiologic assessment. (B) Pathologic assessment. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Supplementary Figure 7.
Forest plots of the network meta-analysis stratified by tyrosine kinase inhibitor (TKI) era, based on the liver transplantation period of enrolled patients, for overall survival and recurrence-free survival. (A) Pre-TKI era (patients transplanted before 2008). (B) Post-TKI era (patients transplanted from 2008 onward). HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco.
Supplementary Figure 8.
Forest plots of the network meta-analysis stratified by pangenotypic direct acting antivirals (DAA) era, based on the liver transplantation period of enrolled patients, for overall survival and recurrence-free survival. (A) Pre-DAA era (patients transplanted before 2014). (B) Post-DAA era (patients transplanted from 2014 onward). HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Supplementary Figure 9.
Quality assessments of the included studies using the ROBINS-I V2 tool. ROBINS-I V2, Risk of Bias in Non-randomized Studies of Interventions Version 2.
Supplementary Figure 10.
Sensitivity analyses restricted to studies at low-to-moderate risk of bias. (A) Overall survival. (B) Recurrence-free survival. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Supplementary Figure 11.
Comparison-adjusted funnel plots for assessing outcomes. (A) Overall survival. (B) Recurrence-free survival. AFP, alpha-fetoprotein; UCSF, University of California San Francisco.
Supplementary Figure 12.
HCC-specific mortality after LT according to expanded criteria in the validation cohort on a radiologic basis. Comparisons of HCC-specific mortality after LT were made between patients within the Milan criteria (MC-in) and those who met the following expanded criteria but were beyond the MC: (A) within the Shanghai criteria (Shanghai-in); and (B) within the Hangzhou criteria (Hangzhou-in). The subdistribution hazard ratio (sHR) with Gray’s P-values is also presented. HCC, hepatocellular carcinoma; LT, liver transplantation; CI, confidence interval.
Supplementary Figure 13.
HCC-specific mortality after liver transplantation according to expanded criteria in the validation cohort on a pathology basis. Comparisons between patients within the Milan criteria (MC-in) and those who met the following expanded criteria but were beyond the MC: (A) within the University of California San Francisco criteria (UCSF-in); (B) within the up-to-seven criteria (UTS-in); (C) within the Asan criteria; (D) within the AFP model; (E) within the Metroticket 2.0 model; (F) within the R4T3 model; (G) within the Shanghai criteria; and (H) within the Hangzhou criteria. The subdistribution hazard ratio (sHR) with Gray’s P-values is also presented. HCC, hepatocellular carcinoma; LT, liver transplantation; CI, confidence interval.
Supplementary Figure 14.
HCC-specific mortality after liver transplantation according to expanded criteria in subgroups of the validation cohort. (A) Patients with hepatitis B virus infection (n=852). (B) Patients receiving living-donor liver transplantation (n=963). (C) Patients receiving pre-transplant HCC therapy (n=855). Competing-risk analysis was performed to compare outcomes between patients within the Milan criteria (MC-in) and those who fulfilled each expanded criterion but were beyond the MC. HCC, hepatocellular carcinoma; CI, confidence interval; EC, expanded criteria; SHR, subdistribution hazard ratio; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Supplementary Figure 15.
Forest plot for the network meta-analysis among expanded criteria including additional LT eligibility criteria incorporating AFP (Munich, TTV/AFP, Warsaw and Wan et al.’s criteria) for overall survival. The boxes and horizontal lines represent HRs with 95% CIs. The higher the P-score value, the higher the likelihood that a criterion is in the top rank. AFP, alpha-fetoprotein; CI, confidence interval; HR, hazard ratio; LT, liver transplantation; TTV, total tumor volume; UCSF, University of California San Francisco.
Figure 1.Flow diagram for study selection process. NMA, network meta-analysis; AFP, alpha-fetoprotein.
Figure 2.Network geometry of included studies. Each node represents a certain LT eligibility criterion, and the node size is proportional to the number of included patients divided by the number of studies. The numbers in parentheses indicate the number of patients in each LT eligibility criterion. The line represents direct comparisons and line thickness is proportional to the number of direct comparisons. (A) Overall survival. (B) Recurrence-free survival. LT, liver transplantation; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Figure 3.Forest plot and ranking probability line chart for the network meta-analysis among expanded criteria. Forest plot for (A) overall survival and (B) recurrence-free survival. The boxes and horizontal lines represent HRs with 95% confidence intervals. The higher the P-score value, the higher the likelihood that a criterion is in the top rank. Ranking probability plot for the network meta-analysis in terms of (C) overall survival and (D) recurrence-free survival. The X-axis shows the rank, and the Y-axis shows the probability (%) that a certain liver transplantation criterion holds that rank. HR, hazard ratio; CI, confidence interval; UCSF, University of California San Francisco; AFP, alpha-fetoprotein.
Figure 4.Forest plot of the pairwise meta-analysis comparing patients who met both the Milan criteria and the AFP model (AFP score≤2) with those who exceeded the Milan criteria but met the AFP model (AFP score≤2). Forest plot for (A) overall survival and (B) time to recurrence. The boxes and horizontal lines represent hazard ratios with 95% CIs. The diamond and vertical dotted lines represent the pooled hazard ratios with 95% CIs. AFP, alpha-fetoprotein; CI, confidence interval; HR, hazard ratio; SE, standard error.
Figure 5.HCC-specific mortality after LT according to expanded criteria in the validation cohort on a radiologic basis. Comparisons of HCC-specific mortality after LT were made between patients within the Milan criteria (MC-in) and those who met the following expanded criteria but were beyond the MC: (A) within the University of California San Francisco criteria (UCSF-in); (B) within the Up-to-Seven criteria (UTS-in); (C) within the Asan criteria (Asan-in); (D) within the AFP model (AFP model-in); (E) within the Metroticket 2.0 model (Metroticket 2.0-in); and (F) within the R4 T3 criteria (R4 T3-in). HCC, hepatocellular carcinoma; LT, liver transplantation; AFP, alpha-fetoprotein.
Table 1.Baseline characteristics of the studies included in the network meta-analysis
Table 1.
|
First author (year) |
Country |
No. of patients |
Method of staging |
Time of staging |
Donor type |
Prior therapy (%) |
Median follow-up (months) |
1-, 3-, and 5-year OS (%) |
1-, 3-, and 5-year RFS (%) |
Etiology of liver disease* (%) |
LT period |
|
Decaens (2006) |
France |
Milan: 279 |
Radiology |
NA |
DDLT |
75 |
48.8 |
80.1/67.4/60.4 |
NA |
20/38/32/10 |
1985–1998 |
|
UCSF: 44 |
66 |
82.1/65.6/47.4 |
|
Decaens (2006) |
France |
Milan: 184 |
Pathology |
NA |
DDLT |
75 |
48.8 |
86.8/77.7/70.1 |
NA |
20/38/32/10 |
1985–1998 |
|
UCSF: 39 |
71 |
79.3/69.3/63.2 |
|
Duffy (2007) |
USA |
Milan: 173 |
Radiology |
NA |
DDLT |
49 |
79.2 |
91.7/79.2/68.2 |
91.2/80.3/70.5 |
17/55/13/15 |
1984–2006 |
|
UCSF: 185 |
88.8/71.0/58.9 |
89.1/70.7/55.1 |
|
Duffy (2007) |
USA |
Milan: 126 |
Pathology |
NA |
DDLT |
49 |
79.2 |
96.6/89.7/80.8 |
90.5/74.9/62.7 |
17/55/13/15 |
1984–2006 |
|
UCSF: 208 |
93.7/83.1/69.5 |
89.2/70.6/55.3 |
|
Parfitt (2007) |
Canada |
Milan: 50 |
Pathology |
NA |
NA |
NA |
96.0†
|
85.1/83.3/83.1 |
NA |
17.3/42.7/21.3/29.3 |
1985–2003 |
|
UCSF: 9 |
85.1/44.4/14.9 |
|
Yao (2007) |
USA |
Milan: 130 |
Radiology |
NA |
DDLT: 157 |
73 |
26.1 |
NA |
94.7/89.2/89.2 |
23.1/64.6/3.1/9.2 |
2001–2006 |
|
UCSF: 38 |
LDLT: 11 |
79 |
96.5/94.7/94.7 |
23.7/57.9/10.5/7.9 |
|
Lee (2008) |
Republic of Korea |
Milan: 152 |
Pathology |
NA |
LDLT |
73.8 |
43.0 |
93.3/85.6/81.8 |
NA |
93.2/6.8/0/0 |
1997–2004 |
|
Asan: 22 |
100/88.9/80.0 |
|
Lee (2008) |
Republic of Korea |
Milan: 152 |
Pathology |
NA |
LDLT |
73.8 |
43.0 |
93.3/85.6/81.8 |
NA |
93.2/6.8/0/0 |
1997–2004 |
|
UCSF: 10 |
100/90.1/78.7 |
|
Fan (2009) |
China |
Milan: 394 |
Pathology |
NA |
DDLT |
23.1 |
35.3†
|
85.6/76.0/65.1 |
79.0/67.6/55.9 |
84.9/2.8/NA/NA |
2001–2007 |
|
Shanghai: 176 |
82.9/64.8/51.2 |
76.1/54.3/46.0 |
|
Mazzaferro (2009) |
Italy |
Milan: 444 |
Pathology |
NA |
DDLT: 1,404 |
NA |
53.0 |
86.5/78.8/73.3 |
NA |
NA |
1984–2006 |
|
UTS: 283 |
LDLT: 121 |
85.7/77.7/71.3 |
|
NA: 31 |
|
Cescon (2010) |
Italy |
Milan: 224 |
Radiology |
At listing |
DDLT |
78 |
42 |
NA |
86.2/78.4/75.1 |
25/54/NA/NA |
1997–2009 |
|
UTS: 43 |
88.1/76.2/71.4 |
|
Cescon (2010) |
Italy |
Milan: 208 |
Pathology |
NA |
DDLT |
78 |
42 |
NA |
88.4/83/78.4 |
25/54/NA/NA |
1997–2009 |
|
UTS: 41 |
87.3/69.8/69.8 |
|
Unek (2011) |
Turkey |
Milan: 34 |
Pathology |
NA |
DDLT: 18 |
26 |
39.5 |
91.0/87.6/87.6 |
91.3/87.7/87.7 |
NA |
1998–2009 |
|
UCSF: 7 |
LDLT: 23 |
85.7/53.4/53.4 |
71.4/53.7/53.7 |
|
Lei (2012) |
China |
Milan: 21 |
Radiology |
NA |
LDLT |
100 |
72 |
90.5/76.3/71.5 |
NA |
85.7/4.8/NA/NA |
2003–2009 |
|
UCSF: 23 |
91.4/74.0/69.6 |
91.3/8.7/0/0 |
|
Piardi (2012) |
France |
Milan: 106 |
Pathology |
NA |
NA |
73.2 |
74 |
100/87.1/81.4 |
NA |
15.1/38.0/37.0/NA |
1997–2007 |
|
UCSF: 28 |
100/86.4/77.2 |
|
Gao (2013) |
China |
Milan: 88 |
Radiology |
NA |
DDLT: 119 |
NA |
44 |
90.9/86.3/82.6 |
NA |
100/NA/NA/NA |
2005–2009 |
|
Shanghai: 24 |
LDLT: 32 |
91.6/70.8/70.8 |
|
Gao (2013) |
China |
Milan: 88 |
Radiology |
NA |
DDLT: 119 |
NA |
44 |
90.9/86.3/82.6 |
NA |
100/NA/NA/NA |
2005–2009 |
|
Hangzhou: 39 |
LDLT: 32 |
87.2/79.2/76.5 |
|
Gugenheim (2013) |
France |
Milan: 299 |
Pathology |
NA |
NA |
69.1 |
52.3 |
90.6/82.6/75.9 |
NA |
18.4/29.9/29.7/22.0 |
1985–2001 |
|
UTS: 84 |
85.8/77.8/68.7 |
|
Kim (2013) |
USA |
Milan: 176 |
Radiology |
After LRT |
DDLT |
66 |
59 |
93.2/83.9/79.1 |
NA |
9.7/64.2/11.9/10.8 |
2002–2008 |
|
R4 T3: 49 |
92 |
87.7/73.3/69.2 |
8.2/71.4/8.2/6.1 |
|
Bittermann (2014) |
USA |
Milan: 537 |
Radiology |
Worst stage |
DDLT |
NA |
NA |
87.8/71.8/68.7 |
NA |
11.2/61.1/7.4/20.3 |
2005–2011 |
|
UCSF: 348 |
91.2/82.0/75.5 |
7.9/52.1/8.8/31.2 |
|
Zhang (2014) |
China |
Milan: 114 |
Pathology |
NA |
NA |
21.9 |
67 |
89.6/82.5/77.2 |
92.8/86.1/85.1 |
100/NA/NA/NA |
2002–2006 |
|
UTS: 89 |
20.2 |
73.0/60.7/57.3 |
70.2/60.2/58.9 |
|
Bonadio (2015) |
Belgium |
Milan: 39 |
Pathology |
NA |
DDLT: 48 |
71 |
70.5 |
88.6/83.1/74.2 |
94.7/88.4/85.6 |
11.8/55.3/25.0/NA |
2000–2007 |
|
Asan: 19 |
LDLT: 28 |
86.6/73.2/66.7 |
93.9/80.6/80.6 |
|
Marques (2015) |
Portugal |
Milan: 177 |
NA |
NA |
DLT: 114 |
53.1 |
34 |
82.1/72.3/67.4 |
82.0/69.8/66.6 |
14.6/45.0/NA/NA |
2001–2014 |
|
UCSF: 26 |
DDLT: 146 |
84.0/74.3/68.1 |
84.3/74.4/68.5 |
|
León Díaz (2016) |
Spain |
Milan: 74 |
Pathology |
NA |
NA |
NA |
NA |
85.2/71.7/58.1 |
83.7/70.3/58.1 |
14.9/37.8/16.2/4.1 |
2002–2010 |
|
UTS: 12 |
91.8/66.7/58.3 |
91.7/58.4/50.1 |
8.3/50.0/16.7/16.7 |
|
Xu (2016) |
China |
Milan: 2,626 |
Radiology |
Closest to LT |
Mostly DDLT |
34.7 |
31.9 |
90.7/81.3/77.0 |
87.2/77.0/72.9 |
91.2/NA/NA/NA |
2000–2012 |
|
UCSF: 429 |
89.4/73.5/68.3 |
83.6/70.8/64.2 |
|
Xu (2016) |
China |
Milan: 2,626 |
Pathology |
Closest to LT |
Mostly DDLT |
34.7 |
31.9 |
90.7/81.3/77.0 |
87.2/77.0/72.9 |
91.2/NA/NA/NA |
2000–2012 |
|
Hangzhou: 1,352 |
89.6/70.8/62.4 |
81.7/64.2/56.3 |
|
Xia (2017) |
China |
Milan: 144 |
Pathology |
NA |
DDLT: 302 |
49.6‡
|
57.7†
|
NA |
84.9/73.2/72.1 |
93.3/NA/NA/NA |
2003–2013 |
|
Hangzhou: 49 |
LDLT: 41 |
76.1/62.5/62.5 |
|
Abdelfattah (2018) |
Egypt |
Milan: 60 |
Pathology |
NA |
DDLT: 38 |
31.7 |
45†
|
84.9/75.1/75.1 |
98.0/90.2/90.2 |
25.0/58.3/NA/NA |
2003–2013 |
|
UCSF: 16 |
LDLT: 50 |
56.3 |
100/75.0/75.0 |
100/85.7/85.7 |
43.8/43.8/NA/NA |
|
Commander (2018) |
USA |
Milan: 1,888 |
Radiology |
NA |
DDLT |
79 |
NA |
92.9/80.8/72.8 |
90.5/77.6/69.1 |
NA |
2002–2013 |
|
R4 T3: 180 |
80 |
89.2/76.2/68.1 |
85.5/71.1/62.5 |
|
Daoud (2018) |
USA |
Milan: 11,555 |
Radiology |
NA |
DDLT |
59 |
NA |
91.7/79.2/71.6 |
NA |
8.6/60.7/NA/NA |
2005–2011 |
|
UCSF: 291 |
93.0/75.2/63.8 |
|
Piñero (2018) |
Argentina |
Milan: 354 |
Radiology |
At listing |
DDLT |
43.9 |
37 |
87.8/76.8/71.4 |
NA |
24.3/33.0/19.2/22.7 |
1990–2016 |
|
UCSF: 40 |
23.1 |
97.0/83.6/83.6 |
|
Pommergaard (2018) |
Denmark |
Milan: 5,662 |
Pathology |
NA |
NA |
NA |
23 |
91.6/83.6/76.8 |
NA |
NA |
1990–2016 |
|
UTS: 1,319 |
89.1/78.8/69.1 |
|
Sternby Eilard (2018) |
Sweden |
Milan: 205 |
Radiology |
Closest to LT |
NA |
40.1 |
63.6†
|
90.8/77.3/69.2 |
NA |
11/63/32/11 |
1996–2014 |
|
UCSF: 51 |
76.5/61.5/52.3 |
|
Degroote (2020) |
Belgium |
Milan: 436 |
Radiology |
NA |
NA |
53 |
56.1†
|
88.0/78.7/71.6 |
NA |
9.5/31.7/43.0/6.3 |
1999–2016 |
|
Asan: 49 |
92.1/81.0/67.2 |
|
Degroote (2020) |
Belgium |
Milan: 436 |
Radiology |
NA |
NA |
53 |
56.1†
|
88.0/78.7/71.6 |
NA |
9.5/31.7/43.0/6.3 |
1999–2016 |
|
UTS: 46 |
91.7/82.8/71.9 |
|
Degroote (2020) |
Belgium |
Milan: 436 |
Radiology |
NA |
NA |
53 |
56.1†
|
88.0/78.7/71.6 |
NA |
9.5/31.7/43.0/6.3 |
1999–2016 |
|
AFP model: 60 |
88.6/73.0/61.3 |
|
Degroote (2020) |
Belgium |
Milan: 436 |
Radiology |
NA |
NA |
53 |
56.1†
|
88.0/78.7/71.6 |
NA |
9.5/31.7/43.0/6.3 |
1999–2016 |
|
MT2.0: 44 |
91.4/82.3/75.1 |
|
Grąt (2020) |
Poland |
Milan: 170 |
NA |
Closest to LT |
NA |
48.6 |
50.4 |
NA |
97.1/93.1/87.1 |
43.3/69.5/NA/NA |
2001–2017 |
|
MT2.0: 43 |
97.1/79.7/75.2 |
|
Grąt (2020) |
Poland |
Milan: 170 |
NA |
Closest to LT |
NA |
48.6 |
50.4 |
NA |
97.1/93.1/87.1 |
43.3/69.5/NA/NA |
2001–2017 |
|
AFP model: 48 |
92.8/82.0/74.0 |
|
Morgul (2020) |
Germany |
Milan: 74 |
Pathology |
NA |
DDLT |
62.5 |
53 |
83.0/77.6/71.1 |
NA |
5.8/11.7/62.5/20.0 |
1994–2013 |
|
UTS: 12 |
75.5/66.7/66.7 |
|
Victor (2020) |
USA |
Milan: 138 |
Pathology |
NA |
DDLT |
93.2 |
NA |
94.0/84.1/78.6 |
100/95.1/91.9 |
5.0/69.5/8.2/20.0 |
2008–2017 |
|
UCSF: 23 |
100/81.1/81.1 |
95.4/88.7/88.7 |
|
Dastyar (2021) |
Iran |
Milan: 82 |
Pathology |
NA |
DDLT |
60.5 |
32 |
88.8/81.0/69.2 |
88.8/78.3/68.1 |
55.6/12.1/NA/NA |
2004–2019 |
|
UCSF: 22 |
86.4/67.6/67.6 |
86.4/53.2/53.2 |
|
Barreto (2022) |
Australia |
Milan: 313 |
Radiology |
NA |
DDLT |
NA |
NA |
94.8/88.3/83.2 |
NA |
19.9/57.2/8.5/12.8 |
1997–2015 |
|
UCSF: 124 |
87.8/69.9/63.7 |
|
Lyu (2022) |
China |
Milan: 44 |
Radiology |
NA |
DDLT |
75.5 |
38.9 |
100/87.4/78.7 |
92.9/80.9/73.2 |
85.5/3.4/NA/NA |
2013–2017 |
|
Hangzhou: 22 |
97.3/85.2/56.8 |
88.0/71.0/55.7 |
|
Ivanics (2024) |
Canada |
Milan: 275 |
Radiology |
NA |
LDLT |
72 |
NA |
90.8/82.4/78.5 |
88.6/80.1/73.2 |
8.1/65.0/11.7/17.5 |
1999–2019 |
|
UCSF: 35 |
88.6/80.0/73.5 |
82.9/74.4/68.5 |
Table 2.League table for overall survival and recurrence-free survival
*
Table 2.
|
Recurrence-free survival |
|
Overall survival |
AFP model |
2.03 (0.37–11.16) |
1.64 (0.60–4.51) |
1.05 (0.25–4.47) |
2.86 (1.06–7.72)†
|
2.16 (0.77–6.08) |
2.12 (0.75–5.99) |
2.02 (0.74–5.54) |
1.31 (0.46–3.77) |
|
1.29 (0.53–3.11) |
Asan |
0.81 (0.20–3.25) |
0.52 (0.09–2.94) |
1.40 (0.35–5.60) |
1.06 (0.26–4.36) |
1.04 (0.25–4.29) |
0.99 (0.25–4.00) |
0.65 (0.16–2.69) |
|
1.05 (0.47–2.34) |
0.81 (0.41–1.62) |
Hangzhou |
0.64 (0.22–1.86) |
1.74 (1.46–2.08)†
|
1.32 (0.94–1.85) |
1.29 (0.91–1.84) |
1.23 (0.97–1.57) |
0.80 (0.54–1.18) |
|
1.27 (0.44–3.66) |
0.99 (0.37–2.61) |
1.21 (0.49–3.00) |
Metroticket 2.0 |
2.72 (0.95–7.78) |
2.05 (0.69–6.12) |
2.01 (0.67–6.02) |
1.92 (0.66–5.58) |
1.25 (0.41–3.78) |
|
1.77 (0.89–3.52) |
1.37 (0.79–2.38) |
1.69 (1.11–2.57)†
|
1.39 (0.62–3.10) |
Milan |
0.76 (0.57–1.01) |
0.74 (0.55–1.00) |
0.71 (0.60–0.83)†
|
0.46 (0.32–0.65)†
|
|
1.30 (0.57–2.98) |
1.01 (0.49–2.08) |
1.24 (0.66–2.33) |
1.02 (0.40–2.59) |
0.74 (0.46–1.17) |
R4 T3 |
0.98 (0.64–1.49) |
0.94 (0.67–1.30) |
0.61 (0.39–0.96)†
|
|
1.08 (0.45–2.58) |
0.84 (0.39–1.80) |
1.03 (0.52–2.04) |
0.85 (0.32–2.23) |
0.61 (0.36–1.04) |
0.83 (0.41–1.69) |
Shanghai |
0.96 (0.68–1.35) |
0.62 (0.39–0.99)†
|
|
1.24 (0.61–2.52) |
0.96 (0.54–1.71) |
1.18 (0.75–1.87) |
0.97 (0.43–2.22) |
0.70 (0.58–0.84)†
|
0.95 (0.58–1.57) |
1.15 (0.65–2.02) |
UCSF |
0.65 (0.44–0.95)†
|
|
1.18 (0.56–2.46) |
0.91 (0.49–1.68) |
1.12 (0.68–1.85) |
0.93 (0.40–2.16) |
0.67 (0.51–0.87)†
|
0.90 (0.53–1.55) |
1.09 (0.60–1.99) |
0.95 (0.69–1.32) |
Up-to-Seven |
Table 3.Baseline characteristics of the studies included in the pairwise meta-analysis of comparing the AFP model and Milan criteria
Table 3.
|
First author (year) |
Country |
No. of patients |
Method of staging |
Time of staging |
Donor type |
Prior therapy (%) |
Median follow-up (mo) |
1-, 3-, and 5-year OS (%) |
1-, 3-, and 5-year TTR (%) |
Etiology of liver disease* (%) |
LT period |
|
Duvoux (2012)†
|
France |
MC-in/AFP-in: 340 |
Radiology |
At listing |
DDLT |
66 |
57.9 |
NA |
6.7/11.4/13.7 |
Alcohol 30 |
1988–2001 |
|
MC-out/AFP-in: 44 |
2.6/7.6/7.6 |
Posthepatitic 61 |
|
Duvoux (2012)†
|
France |
MC-in/AFP-in: 325 |
Radiology |
At listing |
DDLT |
59 |
44.3 |
NA |
1.2/6.1/7.6 |
Alcohol 45 |
2003–2004 |
|
MC-out/AFP-in: 65 |
3.6/11.1/14.6 |
Posthepatitic 44 |
|
Piñero (2016) |
Argentina |
MC-in/AFP-in: 238 |
Radiology |
At listing |
DDLT |
47.4 |
45 |
76.4/64.1/61.2 |
6.2/11.6/16.1 |
28.7/27.2/17.7/26.3 |
2005–2011 |
|
MC-out/AFP-in: 19 |
89.5/82.5/82.5 |
6/6/6 |
|
Notarpaolo (2017) |
Italy |
MC-in/AFP-in: 415 |
Radiology |
At listing |
DDLT |
84.7 |
40.9 |
NA |
3.4/8.9/12.8 |
24.0/58.7/11.7/5.6 |
2002–2010 |
|
MC-out/AFP-in: 97 |
4.7/15.7/15.7 |
|
Ren (2020) |
China |
MC-in/AFP-in: 69 |
Radiology |
Closest to LT |
NA |
54 |
43 |
88.6/75.3/72.2 |
NA |
100/NA/NA/NA |
2010–2015 |
|
MC-out/AFP-in: 19 |
94.9/63.3/56.3 |
Abbreviations
deceased donor liver transplantation
living-donor liver transplantation
Liver Imaging Reporting and Data System
magnetic resonance imaging
Risk of Bias in Non-randomized Studies of Interventions Version 2
subdistribution hazard ratio
tyrosine kinase inhibitor
University of California San Francisco
REFERENCES
- 1. Mehta N. Liver transplantation criteria for hepatocellular carcinoma, including posttransplant management. Clin Liver Dis (Hoboken) 2021;17:332-336.
- 2. Singal AG, Llovet JM, Yarchoan M, Mehta N, Heimbach JK, Dawson LA, et al. AASLD practice guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma. Hepatology 2023;78:1922-1965.
- 3. Liang W, Wu L, Ling X, Schroder PM, Ju W, Wang D, et al. Living donor liver transplantation versus deceased donor liver transplantation for hepatocellular carcinoma: a meta-analysis. Liver Transpl 2012;18:1226-1236.
- 4. Yao FY, Xiao L, Bass NM, Kerlan R, Ascher NL, Roberts JP. Liver transplantation for hepatocellular carcinoma: validation of the UCSF-expanded criteria based on preoperative imaging. Am J Transplant 2007;7:2587-2596.
- 5. Lee SG, Hwang S, Moon DB, Ahn CS, Kim KH, Sung KB, et al. Expanded indication criteria of living donor liver transplantation for hepatocellular carcinoma at one large-volume center. Liver Transpl 2008;14:935-945.
- 6. Mazzaferro V, Llovet JM, Miceli R, Bhoori S, Schiavo M, Mariani L, et al. Predicting survival after liver transplantation in patients with hepatocellular carcinoma beyond the Milan criteria: a retrospective, exploratory analysis. Lancet Oncol 2009;10:35-43.
- 7. Guiteau JJ, Cotton RT, Washburn WK, Harper A, O’Mahony CA, Sebastian A, et al. An early regional experience with expansion of Milan criteria for liver transplant recipients. Am J Transplant 2010;10:2092-2098.
- 8. Duvoux C, Roudot-Thoraval F, Decaens T, Pessione F, Badran H, Piardi T, et al. Liver transplantation for hepatocellular carcinoma: a model including α-fetoprotein improves the performance of Milan criteria. Gastroenterology 2012;143:986-994.e3 quiz e14-e15.
- 9. Mazzaferro V, Sposito C, Zhou J, Pinna AD, De Carlis L, Fan J, et al. Metroticket 2.0 model for analysis of competing risks of death after liver transplantation for hepatocellular carcinoma. Gastroenterology 2018;154:128-139.
- 10. Sapisochin G, Goldaracena N, Laurence JM, Dib M, Barbas A, Ghanekar A, et al. The extended Toronto criteria for liver transplantation in patients with hepatocellular carcinoma: a prospective validation study. Hepatology 2016;64:2077-2088.
- 11. Kaido T, Ogawa K, Mori A, Fujimoto Y, Ito T, Tomiyama K, et al. Usefulness of the Kyoto criteria as expanded selection criteria for liver transplantation for hepatocellular carcinoma. Surgery 2013;154:1053-1060.
- 12. Lozanovski VJ, Ramouz A, Aminizadeh E, Al-Saegh SA, Khajeh E, Probst H, et al. Prognostic role of selection criteria for liver transplantation in patients with hepatocellular carcinoma: a network meta-analysis. BJS Open 2022;6:zrab130.
- 13. Hutton B, Salanti G, Caldwell DM, Chaimani A, Schmid CH, Cameron C, et al. The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: checklist and explanations. Ann Intern Med 2015;162:777-784.
- 14. European Association for the Study of the Liver. EASL clinical practice guidelines on the management of hepatocellular carcinoma. J Hepatol 2025;82:315-374.
- 15. Sterne J, Higgins J. The Risk of Bias in Non-randomized Studies – of Interventions (ROBINS-I) V2 [Internet]. riskofbias.info; 2024;[cited 2025 Aug 16]. Available from: https://www.riskofbias.info/welcome/robins-i-v2.
- 16. An J, Shim JH, Kim SO, Lee D, Kim KM, Lim YS, et al. Prevalence and prediction of coronary artery disease in patients with liver cirrhosis: a registry-based matched case-control study. Circulation 2014;130:1353-1362.
- 17. Chernyak V, Fowler KJ, Kamaya A, Kielar AZ, Elsayes KM, Bashir MR, et al. Liver Imaging Reporting and Data System (LI-RADS) version 2018: imaging of hepatocellular carcinoma in at-risk patients. Radiology 2018;289:816-830.
- 18. American College of Radiology. LI-RADS® CT/MRI Nonradiation TRA v2024 Core. 2024. American College of Radiology; published online [cited 2025 Aug 16]. Available from: https://ref.mcbradiology.com/Body/Liver/HCC%20&%20LI-RADS/LIRADS%202024%20CT%20&%20MRI%20Nonradiation%20TRA.pdf.
- 19. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 2007;8:16.
- 20. Huizenga HM, Visser I, Dolan CV. Testing overall and moderator effects in random effects meta-regression. Br J Math Stat Psychol 2011;64:1-19.
- 21. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629-634.
- 22. Chaimani A, Salanti G. Using network meta-analysis to evaluate the existence of small-study effects in a network of interventions. Res Synth Methods 2012;3:161-176.
- 23. Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988;16:1141-1154.
- 24. Notarpaolo A, Layese R, Magistri P, Gambato M, Colledan M, Magini G, et al. Validation of the AFP model as a predictor of HCC recurrence in patients with viral hepatitis-related cirrhosis who had received a liver transplant for HCC. J Hepatol 2017;66:552-559.
- 25. Ren A, Li Z, Zhou X, Zhang X, Huang X, Deng R, et al. Evaluation of the alpha-fetoprotein model for predicting recurrence and survival in patients with hepatitis B virus (HBV)-related cirrhosis who received liver transplantation for hepatocellular carcinoma. Front Surg 2020;7:52.
- 26. Piñero F, Tisi Baña M, de Ataide EC, Hoyos Duque S, Marciano S, Varón A, et al. Liver transplantation for hepatocellular carcinoma: evaluation of the alpha-fetoprotein model in a multicenter cohort from Latin America. Liver Int 2016;36:1657-1667.
- 27. Omata M, Cheng AL, Kokudo N, Kudo M, Lee JM, Jia J, et al. Asia-Pacific clinical practice guidelines on the management of hepatocellular carcinoma: a 2017 update. Hepatol Int 2017;11:317-370.
- 28. Korean Liver Cancer Association (KLCA) and National Cancer Center (NCC) Korea. 2022 KLCA-NCC Korea practice guidelines for the management of hepatocellular carcinoma. Korean J Radiol 2022;23:1126-1240.
- 29. Bento de Sousa JH, Calil IL, Tustumi F, da Cunha Khalil D, Felga GEG, de Arruda Pecora RA, et al. Comparison between Milan and UCSF criteria for liver transplantation in patients with hepatocellular carcinoma: a systematic review and meta-analysis. Transl Gastroenterol Hepatol 2021;6:11.
- 30. Lerut J, Foguenne M, Lai Q. Hepatocellular cancer selection systems and liver transplantation: from the tower of Babel to an ideal comprehensive score. Updates Surg 2021;73:1599-1614.
- 31. Frazão LP, Quaresma MC, Pereira-Leal JB, Duvoux C, Cardoso J. Hepatocellular carcinoma transplant criteria show poor negative predictive value: a systematic review and meta-analysis. J Pers Med 2025;15:449.
- 32. Brusset B, Dumortier J, Cherqui D, Pageaux GP, Boleslawski E, Chapron L, et al. Liver transplantation for hepatocellular carcinoma: a real-life comparison of Milan criteria and AFP model. Cancers (Basel) 2021;13:2480.
- 33. Al-Ameri AAM, Wei X, Liu P, Lin L, Shao Z, Xie H, et al. Prediction of early recurrence of hepatocellular carcinoma in patients with cirrhosis who had received deceased donor liver transplantation: a multicenter study. Ann Transplant 2019;24:489-498.
- 34. Grąt M, Stypułkowski J, Morawski M, Wronka KM, Wasilewicz M, Lewandowski Z, et al. Shadows behind using simple risk models in selection of hepatocellular carcinoma patients for liver transplantation. Ann Surg 2020;271:1124-1131.
- 35. Health Resources and Services Administration (HRSA). OPTN Policies & Bylaws [Internet]. U.S. Department of Health & Human Services; 2025;[cited 2025 Aug 16]. Available from: https://www.hrsa.gov/optn/policies-bylaws/policies.
- 36. Millson C, Considine A, Cramp ME, Holt A, Hubscher S, Hutchinson J, et al. Adult liver transplantation: a UK clinical guideline - part 1: pre-operation. Frontline Gastroenterol 2020;11:375-384.
- 37. Durand F, Antoine C, Soubrane O. Liver transplantation in France. Liver Transpl 2019;25:763-770.
- 38. Kudo M, Kawamura Y, Hasegawa K, Tateishi R, Kariyama K, Shiina S, et al. Management of hepatocellular carcinoma in Japan: JSH consensus statements and recommendations 2021 update. Liver Cancer 2021;10:181-223.
- 39. Grąt M, Wronka KM, Stypułkowski J, Bik E, Krasnodębski M, Masior Ł, et al. The Warsaw proposal for the use of extended selection criteria in liver transplantation for hepatocellular cancer. Ann Surg Oncol 2017;24:526-534.
- 40. Schoenberg MB, Anger HJW, Bucher JN, Denk G, De Toni EN, Seidensticker M, et al. Liver transplantation for extended criteria hepatocellular carcinoma using stable response to locoregional therapy and alpha-fetoprotein as selection criteria. Visc Med 2020;36:506-515.
- 41. Toso C, Asthana S, Bigam DL, Shapiro AM, Kneteman NM. Reassessing selection criteria prior to liver transplantation for hepatocellular carcinoma utilizing the Scientific Registry of Transplant Recipients database. Hepatology 2009;49:832-838.
- 42. Wan P, Xia Q, Zhang JJ, Li QG, Xu N, Zhang M, et al. Liver transplantation for hepatocellular carcinoma exceeding the Milan criteria: a single-center experience. J Cancer Res Clin Oncol 2014;140:341-348.
- 43. Fujioka M, Nakashima Y, Nakashima O, Kojiro M. Immunohistologic study on the expressions of alpha-fetoprotein and protein induced by vitamin K absence or antagonist II in surgically resected small hepatocellular carcinoma. Hepatology 2001;34:1128-1134.
- 44. Ruch B, Wagler J, Kumm K, Zhang C, Katariya NN, Garcia-Saenz-de-Sicilia M, et al. Hepatocellular carcinoma, alpha fetoprotein, and liver allocation for transplantation: past, present and future. Curr Oncol 2022;29:7537-7551.
- 45. Lu L, Phua QS, Bacchi S, Goh R, Gupta AK, Kovoor JG, et al. Small study effects in diagnostic imaging accuracy: a meta-analysis. JAMA Netw Open 2022;5:e2228776.
- 46. Tabrizian P, Saberi B, Holzner ML, Rocha C, Kyung Jung Y, Myers B, et al. Outcomes of transplantation for HBV-vs. HCVrelated HCC: impact of DAA HCV therapy in a national analysis of >20,000 patients. HPB (Oxford) 2022;24:1082-1090.
- 47. Piñero F, Boin I, Chagas A, Quiñonez E, Marciano S, Vilatobá M, et al. Direct-acting antivirals and hepatocellular carcinoma: no evidence of higher wait-list progression or posttransplant recurrence. Liver Transpl 2020;26:640-650.
- 48. Sposito C, Cucchetti A, Mazzaferro V. Assessing competing risks for death following liver transplantation for hepatocellular carcinoma. Dig Dis Sci 2019;64:1001-1007.
- 49. Kim DH, Kim B, Choi JI, Oh SN, Rha SE. Li-RADS treatment response versus modified RECIST for diagnosing viable hepatocellular carcinoma after locoregional therapy: a systematic review and meta-analysis of comparative studies. Taehan Yongsang Uihakhoe Chi 2022;83:331-343.
- 50. Soejima Y, Taketomi A, Yoshizumi T, Uchiyama H, Aishima S, Terashi T, et al. Extended indication for living donor liver transplantation in patients with hepatocellular carcinoma. Transplantation 2007;83:893-899.
- 51. Guyot P, Ades AE, Ouwens MJ, Welton NJ. Enhanced secondary analysis of survival data: reconstructing the data from published Kaplan-Meier survival curves. BMC Med Res Methodol 2012;12:9.
- 52. Wei Y, Royston P. Reconstructing time-to-event data from published Kaplan-Meier curves. Stata J 2017;17:786-802.
, Jihyun An6
