Liver Transplantation for Hepatocellular Carcinoma: Beyond the Milan Criteria


  • F. Y. Yao

    Corresponding author
    1. Division of Gastroenterology, University of California, San Francisco, CA 94143-0538
      * Corresponding author: Francis Y. Yao,
    Search for more papers by this author

* Corresponding author: Francis Y. Yao,


Liver transplantation represents a cornerstone in the management of early-stage hepatocellular carcinoma (HCC). Expansion beyond the Milan criteria for liver transplantation (1 lesion ≤ 5 cm, or 2 to 3 lesions each ≤ 3 cm) remains controversial. This review covers several key areas: (1) Recent developments and published data on expanded criteria for deceased donor and live-donor liver transplantation, with emphasis on criteria that have been applied to preoperative imaging. (2) Independent testing of expanded criteria, where published data are largely limited to the proposed University of California, San Francisco criteria (1 lesion ≤ 6.5 cm, 2–3 lesions each ≤ 4.5 cm with total tumor diameter ≤ 8 cm). (3) Response to loco-regional therapy and tumor downstaging. (4) The fundamental questions and answers in resolving the controversy over expanded criteria. The key issue pertains to whether acceptable outcome can be achieved on a broader scale beyond single center experience, which appears to support modest expansion beyond the Milan criteria. The foundation of the debate over expanded criteria may rest upon what the transplant community would consider to be the acceptable threshold for patient survival using expanded criteria, without causing significant harm to other transplant candidates without HCC.


The early experience with liver transplantation for hepatocellular carcinoma (HCC) with extensive tumor burden was associated with dismal outcome due to tumor recurrence (1). It was only after the adoption of the restrictive Milan criteria (1 lesion ≤ 5 cm, or 2 to 3 lesions ≤ 3 cm) that liver transplantation became widely accepted as the best curative treatment for early stage HCC (2). The growing experience and success of liver transplantation for HCC in recent years also fueled controversies regarding whether the Milan criteria are too restrictive, thus excluding many candidates from liver transplantation who would have otherwise done well with an acceptably low risk of posttransplant tumor recurrence (3–11). Expanding the limits of tumor size and/or number for deceased donor liver transplantation (DDLT) or live-donor liver transplantation (LDLT), and the idea of tumor ‘downstaging’ prior to liver transplantation now take center stage in a widespread debate. This article reviews recent developments and published data on this highly controversial topic.

Beyond the Milan Criteria—the ‘Metro Ticket’

Although many studies have demonstrated that patients with HCC meeting the Milan criteria could achieve excellent 5-year survival of 70% or better after liver transplantation (1), others have suggested that a ‘controlled’ or modest expansion of tumor size beyond the Milan criteria could achieve posttransplant survival comparable to that with the Milan criteria (3,4). Several groups have proposed expansion beyond the Milan criteria for DDLT (3–9) (Table 1). Some are just beyond the borders of the Milan criteria, while others are much more liberal. The idea of expanded criteria has been compared to the European Metro system, with the paradigm of ‘the further the distance, the greater the price’ (12) (Figure 1). The projected survival figures, however, are not entirely consistent with published data. The 75–80% 5-year expected survivals for the Milan criteria represent the best reported outcomes, whereas other studies have reported 5-year survivals in the 50–70% range using the Milan criteria (7,13,14), similar to the projected survival figures for modest expansion of tumor criteria (Figure 1).

Table 1.  Proposed expanded criteria based on tumor size and number in deceased donor liver transplantation
Author (year), institution (reference)Proposed expanded criteriaRadiology/pathology5-year patient survival (number of patients)
  1. UCSF = University of California, San Francisco; EC = expanded criteria; NA = not available.

  2. The number of patients in the expanded criteria (EC) group represents only patients beyond the Milan criteria but within expanded criteria.

  3. 1Recurrence-free survival.

Yao (2001), UCSF (3)1 nodule ≤ 6.5 cm, or 2–3 nodules ≤ 4.5 cm and total tumor diameter ≤ 8 cmPathologyMilan (n = 46), 72%
  EC (n = 14), 73%
Yao (2007), UCSF (8)Same as aboveRadiologyMilan (n = 130), 80%
  EC (n = 38), 82%
Herrero (2001), Pamplona, Spain (4)1 nodule ≤ 6 cm, or 2–3 nodules each ≤ 5 cmRadiologyMilan (n = 49), NA
  EC (n = 12), NA
  Entire group, 79%
Roayaie (2002), Mt. Sinai, NY (5)1 or more nodules 5 to 7 cmRadiologyMilan, NA
  EC (n = 31), 55%1
Kneteman (2004), Edmonton, Canada (6)1 nodule < 7.5 cm any number < 5 cmRadiologyMilan (n = 19), 87% (4-year)
  EC (n = 21), 83% (4-year)
Onaca (2007), International Tumor Registry (7)1 nodule ≤ 6 cm 2–4 nodules each ≤ 5 cmPathologyMilan (n = 1038), 62%1
  EC (n = 114), 54.3%1
Figure 1.

Beyond the Milan criteria—the HCC ‘Metro Ticket’. Modified and updated from reference (12).

Fundamental Questions and Answers in Resolving the Controversy over Expanded Criteria

1. What is the most clinically relevant end-point for justifying expansion of the Milan criteria?

Rather than debating whether the expanded criteria are better or worse than the Milan criteria, a more relevant clinical question is whether expanded criteria can result in survival after DDLT that is deemed acceptable by the transplant community to justify utilization of a scarce resource. Bruix and colleagues have proposed a 5-year posttransplant patient survival of 50% to be the minimal acceptable cutoff (15). Recently, a Markov model comparing the survival benefit of DDLT for a patient with HCC beyond the Milan criteria versus harm to other patients on the waiting list has demonstrated that a higher threshold of 61% 5-year survival is needed to justify expansion of the Milan criteria in the United States (16).

2. How should the expanded criteria be validated?

Many of the proposed expanded criteria were based on explant tumor pathology. The limitations of imaging studies, exemplified by tumor stage underestimation in 20–30% of patients undergoing liver transplantation (1,2,8), have been repeatedly cited as a major concern for ‘pushing the envelope’ even to a modest degree (12,15). Imaging may also overestimate tumor stage in 10–20% of cases (1,8). Proponents of maintaining the status quo might argue that the Milan criteria have so far been able to maintain acceptable outcome after liver transplantation, and offer a wider ‘safety margin’ for tumor understaging by preoperative imaging. Proposed expanded criteria should therefore be validated independently based on application to preoperative radiologic staging. Focus should be on the subgroup meeting the proposed expanded criteria but exceeding the Milan criteria to avoid a ‘dilutional’ effect (including a majority of HCC meeting the Milan criteria).

3. Should expanded criteria be judged by the intention-to-treat outcome?

It has been suggested that expansion would result in a higher rate of dropout from the waiting list for DDLT and therefore worse intention-to-treat survival (15). Currently, patients with HCC beyond the Milan criteria do not have equal access to DDLT as those with tumors within the Milan criteria under the model for the end-stage liver disease (MELD) system of organ allocation, and thus the results of an intention-to-treat analysis would be heavily biased against those with HCC outside the Milan criteria. Even for tumors within the Milan criteria, those with a single lesion > 3 cm or 2–3 lesions had a much greater risk of dropout than those with a single tumor ≤ 3 cm according to a study from the pre-MELD era (17). Consequently, in the subgroup with tumors beyond the Milan criteria who could achieve acceptable posttransplant outcome, the risk of dropout from the waiting list would almost certainly be greater than those within the Milan criteria. Rather than interpreting that these are worse candidates for DDLT by virtue of their greater risk for wait-list dropout, future efforts should be directed toward improving their access to timely DDLT.

Proposed Expanded Criteria Based on Tumor Size and Number in DDLT

Among the proposed expanded criteria (3–8) (Table 1), the ones that have been applied to preoperative imaging are selectively reviewed here in greater detail.

The University of California, San Francisco (UCSF) criteria (1 lesion ≤ 6.5 cm in diameter, or 2–3 lesions each ≤ 4.5 cm with total tumor diameter ≤ 8 cm) were initially derived from explant tumor characteristics (3). These criteria were recently validated in a separate cohort at UCSF as selection criteria based on imaging (8). For the entire group of 168 patients, the overall 5-year recurrence-free probabilities and patient survival were 91% and 81%, respectively. The 5-year recurrence-free probabilities were 90% for the 130 patients with preoperative tumor stage within the Milan criteria, and 94% for the other 38 patients meeting the UCSF criteria but exceeding the Milan criteria (p = 0.58). The risk of tumor understaging was not significantly different between the two subgroups (20% versus 29%, respectively, p = 0.26). When explant tumors exceeded the UCSF criteria (15%), the 5-year recurrence-free probabilities were 60% versus 97% for those within the UCSF criteria (p < 0.0001). High alpha-fetoprotein > 1000 ng/mL was a strong independent predictor of tumor recurrence (8).

Herrero et al. from Pamplona, Spain (4), reported the posttransplant outcome of 47 patients with HCC meeting the following criteria: a single lesion ≤ 6 cm or 2–3 lesions each ≤ 5 cm. The 5-year recurrence-free survival rate and actuarial patient survival rate were 70% and 79%, respectively, which were not significantly different from the actuarial survival of 107 patients without HCC who underwent DDLT for other indications. This study did not separately analyze the outcome of the subgroup with HCC meeting the expanded criteria but exceeding the Milan criteria.

Roayaie et al. from Mount Sinai Medical Center, New York (5), applied a multimodality neoadjuvant protocol of systemic chemotherapy with doxorubicin and chemoembolization in 80 patients with one or more tumors > 5 cm. Thirty-seven patients (46%) were excluded from DDLT mainly due to tumor progression, and 43 patients underwent DDLT. The 5-year recurrence-free survival among the subgroup of 32 patients with the largest tumor diameter between 5 and 7 was significantly better than the 12 patients with tumors > 7 cm in diameter (55% versus 34%, respectively, p = 0.024).

Kneteman et al. from Edmonton, Canada, reported a consecutive series of 19 patients with HCC meeting the Milan criteria and 21 patients beyond the Milan criteria with a single lesion < 7.5 cm or multiple tumors < 5 cm transplanted under a sirolimus-based immunosuppressive protocol. The 4-year patient survival rates were 87.4% and 82.9% in the Milan group and the extended criteria group, respectively. The 4-year recurrence-free survival rates were 81.1% and 76.8% in the Milan and extended criteria group, respectively. Further studies are needed to determine if sirolimus actually reduces the risk of tumor recurrence after liver transplantation.

Independent Testing of Expanded Criteria

Among the proposed expanded criteria, only the UCSF criteria have been independently tested based on either explant tumor pathology (9,13,14,18) or preoperative radiologic staging (9,13,14). Since the UCSF criteria represent only a modest expansion of the Milan criteria, only a few studies have adequate sample size for the subgroup meeting the UCSF but exceeding the Milan criteria (9,14,18). According to a report from 14 French transplant centers (14), 39 of the 461 patients (8.5%) had explant tumors beyond the Milan but within UCSF criteria. Their 5-year survival rate was similar to the 184 patients meeting the Milan criteria only (64% versus 70%, p = 0.33), but significantly better than the 34% 5-year survival rate among 238 patients exceeding both criteria (p < 0.0001). However, based on pretransplant imaging data on 468 patients, the 44 patients (9.4%) meeting the UCSF but exceeding the Milan criteria had a 5-year intention-to-treat survival rate of 46%, compared to 60% for the 279 patients within the Milan criteria (p = 0.1) and 35% for the 121 patients with pretransplant tumor stage beyond both criteria (p < 0.0001) (Figure 2). Although the authors suggested that the UCSF criteria should not be used for patient selection because of the 5-year intention-to-treat survival of < 50% by radiologic staging, the difference in survival between the UCSF and Milan criteria did not achieve statistical significance (46% versus 60%, p = 0.1). The 5-year cumulative risk of HCC recurrence was also not significantly different between the two groups (27% versus 20%, p = 0.15).

Figure 2.

UCSF expanded criteria—5-year patient survival based on preoperative imaging in the three largest studies. The dotted line represents the minimum 5-year patient survival threshold of 50% proposed by Bruix et al. (15).

By far the largest single institutional study by Duffy et al. from the University of California, Los Angeles (9), has independently validated the UCSF criteria. According to pretransplant imaging, 185 patients met the UCSF but exceeding the Milan criteria, 173 met only the Milan criteria and 109 exceeded both criteria. The 5-year patient survival rates were 64% in the group meeting the UCSF but exceeding the Milan criteria, compared to 79% for the group meeting the Milan criteria only (p = 0.061), and 41% for those exceeding both UCSF and Milan criteria (p < 0.01) (Figure 2). Based on explant tumor pathology, the 208 patients meeting the UCSF but exceeding the Milan criteria achieved 5-year survival of 81% versus 86% for the 126 patients meeting only the Milan criteria (p = 0.057), and 32% for the remaining 133 patients with HCC exceeding both criteria (p < 0.01).

Expanded Criteria—Beyond Tumor Size and Number

Other tumor characteristics, including poorly differentiated grade and microvascular invasion, have been associated with poor outcome after liver transplantation, independent of the tumor stage (1). In a study by Cilo et al. (10), poorly differentiated tumor grade was used as an exclusion criterion for liver transplantation, and 100 patients (including 40 patients exceeding the Milan criteria) selected by this approach underwent a multi-modal adjuvant treatment protocol prior to liver transplantation. They reported a low dropout rate and favorable 3-year intention-to-treat survival irrespective of whether these patients met the Milan criteria (69% within Milan and 85% exceeding Milan). None of these patients had HCC recurrence after a median follow-up of 16 months, but longer follow-up data are needed to provide further support of this approach.

Marsh et al. from the University of Pittsburgh reported favorable outcome using a modified TNM staging classification (11). It requires histologic information related to microvascular invasion, which limits its application as preoperative selection criteria. The Pittsburgh group also applied microdissection techniques to assess for loss of heterozygosity for targeted microsatellites (19). This could serve as a prognostic marker to identify a subgroup of patients with advanced HCC beyond the Milan criteria with an acceptably low risk of posttransplant tumor recurrence. Further studies are needed before these results can be extrapolated into the pretransplant clinical arena to support using these markers as selection criteria for liver transplantation.

Response to Loco-Regional Therapy and Downstaging of HCC beyond the Milan Criteria

Majno and colleagues from Hospital Paul Brousse, France, were the first to apply the concept of downstaging of tumors to facilitate liver transplantation (20). Recently data have emerged suggesting that response to loco-regional therapy for HCC exceeding the Milan criteria could serve as a prognostic marker for improved posttransplant outcome and for the selection of candidates for liver transplantation (21,22). These findings also support the concept and feasibility of tumor downstaging.

Millonig et al. from Insbruck, Austria (21), applied exclusively chemoembolization to 68 patients with HCC meeting the Milan criteria, 33 patients meeting the UCSF but exceeding the Milan criteria and 15 patients exceeding both criteria. Treatment response was based on the degree of tumor necrosis, classified as complete response or partial response. Among patients meeting the UCSF criteria, those with complete or partial response had survival similar to the partial responders in the Milan group, and better than those who did not respond or had tumor progression although the difference did not achieve statistical significance.

In the study by Otto et al. from Mainz, Germany (22), 62 of 96 patients (65%) treated with chemoembolization had HCC exceeding the Milan criteria. Treatment response was based on the Response Evaluation Criteria in Solid Tumors (RECIST) criteria, characterized by a 30% reduction in the sum of the largest tumor diameter. Exclusion from liver transplantation was based upon a 20% increase in the sum of the largest tumor diameter or the development of new lesions. Treatment response or the absence of tumor progression was associated with a significantly better 5-year recurrence-free survival compared to those with minimal progression but not meeting the RECIST criteria for exclusion (94.5% vs. 35%). A very high dropout rate of 55% was observed in the group with HCC exceeding the Milan criteria, possibly due to the absence of an upper limit in the tumor size and number in determining eligibility for their treatment protocol.

In 2005, the UCSF group published the first prospective study on downstaging in 30 patients as a test of concept (23). The general guidelines in this downstaging protocol are summarized in Table 2. Longer follow-up data were recently presented in a larger cohort of 61 patients (24). Successful downstaging was observed in 70%, and the remaining 30% were classified as treatment failures, largely due to tumor progression. Among the 35 patients transplanted, 2 patients had died unrelated to HCC recurrence, while the other 33 patients were alive without tumor recurrence after a median posttransplant follow-up of 25 months. The 4-year intention-to-treat survival of the entire cohort was 69%, and the 4-year posttransplant patient survival was 92%. These results suggest that excellent outcome can be achieved with tumor downstaging in a highly selective group of patients.

Table 2.  UCSF downstaging protocol for hepatocellular carcinoma
  1. 1In a patient with 4 nodules, for example, successful downstaging requires obliteration of at least one nodule so that there would be no more than 3 nodules with viable tumor, none > 3 cm, to meet UNOS T2 criteria.

Inclusion criteria for downstaging
 (a) 1 lesion > 5 cm and ≤ 8 cm
 (b) 2 or 3 lesions, at least one > 3 cm but ≤ 5 cm + total tumor diameter ≤ 8 cm
 (c) 4 or 5 lesions, all ≤ 3 cm + total tumor diameter ≤ 8 cm
 (d) No vascular invasion by imaging
Criteria for successful downstaging by imaging studies1
 (1) Tumor size and number meeting UNOS T2 criteria
 (2) Complete tumor necrosis without contrast enhancement to suggest residual tumor, equivalent to obliteration of the tumor irrespective of the tumor size
Additional guidelines
 (1) A minimum observation period of 3 months after downstaging is required before deceased donor liver transplantation, and if imaging studies meet the above defined criteria for successful downstaging
 (2) Patients can undergo live donor liver transplantation at 3 months after downstaging, and if imaging studies meet proposed the UCSF criteria (3)
 (3) Those with acute hepatic decompensation after downstaging procedures are not eligible for liver transplantation unless they meet above criteria

Live-Donor Liver Transplantation and Expanded Criteria

The majority of the largest series on LDLT for HCC have come from the Far East (25–27), where the rate of deceased organ donation is negligible. Whether LDLT for HCC would achieve the same outcome as DDLT is still not entirely clear. According to a multicenter study from Korea involving 237 LDLT and 75 DDLT recipients, the 3-year survival rate was 88% for DDLT and 91% for LDLT when the UCSF criteria were met (26). In the largest series from 49 centers in Japan including a total of 316 patients, the 3-year patient survival rates were 79% for patients meeting the Milan criteria and 61% for those who did not (25). Lo et al. from Hong Kong, China (27), however, found a trend for worse 5-year patient survival among 43 recipients of LDLT versus 17 patients who underwent DDLT (58% vs. 94%, p = 0.19) using the UCSF criteria for patient selection. In a retrospective analysis by the A2ALL study group from the United States (28), the 58 patients undergoing LDLT for HCC had a significantly higher risk for tumor recurrence compared to the 34 DDLT recipients with HCC (29% versus 0%). About 50% of LDLT recipients had HCC beyond the UCSF criteria in the liver explant, which could be an important contributing factor for the inferior outcome.

The issue of expanded criteria in LDLT is even more complicated. The primary objective is to justify the risk to the donor based upon the likelihood of favorable outcome for the recipient. This decision has no significant effect on the deceased donor pool. It is therefore more difficult to define the minimal acceptable outcome. As shown in Table 3, the proposed expanded criteria in LDLT are far more liberal than the UCSF criteria, and have shifted more toward multifocal tumors than a larger solitary lesion (29,30). The number of patients meeting the proposed expanded criteria but exceeding the Milan criteria is generally very small, and these results therefore require independent validation.

Table 3.  Live-donor liver transplantation for HCC within and beyond the Milan criteria
Author (year), institution (reference)5-year patient survivalProposed criteria
Within Milan> Milan
  1. Studies with < 3 years of follow-up or < 20 patients are not included. NA = not applicable.

  2. 1Only recurrence-free probabilities were available.

  3. 2The number of patients meeting expanded criteria but beyond the Milan criteria.

  4. 3From the 3rd Japan–Korea Liver Symposium (Digestive Diseases 2007; 5: 96–319).

Todo (2004), Japanese Study Group (49 centers) (25)(n = 137), 79% (3-year)(n = 172), 60% (3-year)NA
Hwang (2005), Seoul, Korea (26)(n = 151), 91% (3-year)(n = 62), 63% (3-year)NA
(Within UCSF)(>UCSF) 
(n = 167), 91% (3-year)(n = 46), 59% (3-yr) 
Soejima (2007), Fukuoka, Japan (29)(n = 20), 100%1(n = 40), 74%1Any number each ≤ 5 cm
Jonas (2007), Berlin, Germany (30)(n = 8), 75% (3-year)(n = 13), 62% (3-year)Any number each ≤ 6 cm & total tumor diameter ≤ 15 cm (n = 11)2
Kwon (2007), Seoul, Korea3(n = 99), 80%(n = 40), 80%Any number each ≤ 5 cm & ≤ 5 nodules each ≤ 5 cm (n = 4)2
Sugawara (2007), Toyko, Japan3(n = 68), 94% (3-year)1(n = 10), 70% (3-yr)1 
Takada (2007), Kyoto, Japan3(n = 74), 73%(n = 62), 67%≤ 10 nodules each ≤ 5 cm (n = 9)2

Beyond Milan: Summary

For over a decade, the Milan criteria for HCC have been widely used for the selection of candidates for liver transplantation, and have limited the risk of tumor recurrence to an acceptable level (1,2). More recently, large single center studies on modest expansion beyond the Milan criteria based on preoperative imaging have also shown acceptable 5-year patient survival of 60–80% after DDLT (8,9), supporting a modified view of the ‘HCC-Metro ticket’(12)—‘We can afford the price’ that other transplant candidates without HCC might have to pay, ‘if we don't go too far’ in expanding the HCC criteria for DDLT (Figure 3).

Figure 3.

Modified perspective on the HCC ‘Metro Ticket’.

At this time, whether acceptable outcome using expanded criteria for patient selection can be achieved on a broader scale beyond single center experience remains highly controversial (14,16). The foundation of this debate may rest upon what the transplant community would consider to be the acceptable threshold for patient survival using expanded criteria, and whether a benchmark higher than the proposed 50% 5-year survival (15) should be set across the nation to avoid causing significant harm to other transplant candidates without HCC (16).

Finally, loco-regional therapy has been routinely used as a bridge to liver transplantation even though the benefits of these treatments have not been confirmed (1). The potential role of loco-regional therapy in preserving favorable outcome using expanded criteria should be further investigated and factored into future considerations for expanding the existing criteria for liver transplantation. Persistently high alpha-fetoprotein > 1000 ng/mL despite loco-regional therapy predicts a high risk for tumor recurrence (8), and should be considered an exclusion criterion for liver transplantation. Downstaging of HCC is an attractive alternative to simply expanding the tumor size limits in lieu of loco-regional therapy, since response to downstaging may serve as a prognostic marker and may help select a subgroup with more favorable tumor biology for liver transplantation (22–24).