About the Milan Criteria for Liver Transplantation in Hepatocellular Carcinoma
One lesion no larger than 5 cm, or up to three lesions none larger than 3 cm, with no macrovascular invasion and no extrahepatic spread — that is the whole of the Milan criteria. They are a yes-or-no eligibility rule for liver transplantation in hepatocellular carcinoma rather than a score, and they exist because the 1996 study that defined them showed patients meeting them achieved 4-year overall survival of 85% and recurrence-free survival of 92%, against 50% and 59% for those who did not. Being beyond them is not the end of the discussion: expanded criteria exist, and downstaging with locoregional therapy back into Milan is an established route to transplantation.
Formula
Within Milan = (single lesion ≤ 5 cm OR ≤ 3 lesions each ≤ 3 cm) AND no macrovascular invasion AND no extrahepatic spread- OR
- The two burden patterns are alternatives. A single 4 cm lesion qualifies; three lesions of 3 cm each qualify; two lesions of 4 cm do not, because neither pattern allows a lesion over 3 cm once there is more than one.
- AND
- All three clauses are required. A single 2 cm lesion with portal vein tumour thrombus is beyond criteria, and no amount of favourable size compensates.
- The size limits are not interchangeable. Two lesions of 3.5 cm each fall outside Milan even though both are under 5 cm, because the multi-lesion arm caps each lesion at 3 cm.
- Total tumour diameter is not part of the Milan criteria. That is a UCSF concept, and importing it produces a different rule.
- Microvascular invasion cannot be assessed before explant and is not part of the criteria, though it is the strongest predictor of recurrence in every series — which is a large part of why imaging-based criteria have an irreducible error rate.
- Assessment is on imaging at the time of listing. Progression while waiting is a recognised failure mode, and it is why bridging therapy is offered rather than optional in a long wait.
Interpreting the result
Within Milan means standard transplant candidacy applies, and the patient should go to a hepatobiliary multidisciplinary meeting for a decision between transplantation, resection and ablation — being eligible is not the same as transplantation being the best option, and a patient with preserved hepatic function and a single peripheral lesion may do better with resection. Beyond Milan means standard allocation under these criteria does not apply, but the reason matters enormously. Tumour burden beyond limits with no vascular invasion is the situation in which downstaging with transarterial chemoembolisation or ablation is an accepted pathway, with transplantation offered if the tumour responds and stays responsive through a period of observation. Macrovascular invasion or extrahepatic spread is a different matter and generally excludes transplantation outright. Expanded criteria — UCSF, up-to-seven, models incorporating alpha-fetoprotein — are used by some centres and jurisdictions, and which set applies is a programme-level and regulatory decision rather than a clinical judgement made at the bedside.
| Score | Band | What it means | Action |
|---|---|---|---|
| Single lesion ≤ 5 cm, no invasion, no spread | Within Milan criteria | 4-year overall survival 85% and recurrence-free survival 92% in the derivation series, comparable to transplantation for non-malignant disease | Refer to a transplant centre and discuss at a hepatobiliary MDT; plan bridging therapy if a wait is expected |
| ≤ 3 lesions each ≤ 3 cm, no invasion, no spread | Within Milan criteria | The multi-lesion arm of the criteria, carrying the same eligibility as the single-lesion arm | As above. Note that a fourth lesion, or any lesion over 3 cm, moves the patient beyond criteria |
| Burden beyond limits, no vascular invasion or spread | Beyond Milan — potentially downstageable | Standard allocation does not apply, but response to locoregional therapy can restore eligibility | MDT discussion for downstaging with TACE or ablation, then reassessment and a period of observed stability |
| Macrovascular invasion or extrahepatic spread present | Beyond Milan — absolute contraindication | Implies disease beyond what transplantation can clear; excluded under these criteria irrespective of lesion size | Assess for systemic therapy by BCLC stage; transplantation is generally not offered |
What the Milan Criteria needs (3 inputs)
- Tumour burden on contrast-enhanced imaging
- Three options: a single lesion 5 cm or smaller; two or three lesions each 3 cm or smaller; or anything beyond both. Measured on the largest diameter on multiphase CT or MRI, and lesions must meet radiological criteria for hepatocellular carcinoma — an indeterminate nodule is not counted as a tumour, but it is not ignored either.
- Macrovascular invasion
- Tumour thrombus in the portal or hepatic vein. An absolute contraindication under these criteria regardless of tumour size, because it implies dissemination that imaging cannot map.
- Extrahepatic spread
- Including regional nodal disease. Also an absolute contraindication. Nodes are the difficult judgement in practice, since reactive nodal enlargement is common in cirrhosis and needs interpreting rather than measuring.
What it returns
- Within or beyond Milan criteria
- A binary result. There is no partial credit and no score — all three conditions must be satisfied simultaneously.
- Which criterion failed
- Named explicitly when the result is beyond, because the three failures have entirely different implications: excess burden is potentially downstageable, macrovascular invasion generally is not.
How it is calculated
The criteria were derived empirically rather than modelled. Forty-eight patients with cirrhosis and small hepatocellular carcinoma underwent transplantation, and survival was analysed against tumour characteristics assessed on the explanted liver. The size and number thresholds that emerged — a solitary tumour of 5 cm or less, or up to three nodules each 3 cm or less — separated a group whose outcomes matched transplantation for benign disease from a group whose did not. Nothing is weighted or summed because nothing needed to be: the effect of burden was large and largely categorical. The absence of vascular invasion and extrahepatic spread was treated as a precondition rather than a variable, since both indicate disease that has already left the field a transplant can clear.
Facts & figures
| Group | n | 4-year overall survival | 4-year recurrence-free survival |
|---|---|---|---|
| Meeting criteria on pathology | 35 (73%) | 85% | 92% |
| Exceeding criteria on pathology | 13 (27%) | 50% | 59% |
P = 0.01 for overall survival and P = 0.002 for recurrence-free survival. Note that the grouping was by assessment of the explanted liver, not by preoperative imaging — the criteria were derived from pathology and are applied prospectively to imaging, which is the source of their unavoidable staging error.
| Criteria | Definition | Reported survival |
|---|---|---|
| Milan (1996) | Single ≤ 5 cm, or ≤ 3 lesions each ≤ 3 cm; no macrovascular invasion or extrahepatic spread | 4-year overall survival 85% |
| UCSF (Yao 2001) | Solitary ≤ 6.5 cm, or ≤ 3 nodules with largest ≤ 4.5 cm and total diameter ≤ 8 cm | 90% at 1 year, 75.2% at 5 years in 70 patients |
| Up-to-seven (Mazzaferro 2009) | Sum of the size of the largest tumour in cm and the number of tumours ≤ 7 | 5-year overall survival 71.2% (95% CI 64.3–77.0) in 283 patients without microvascular invasion |
The up-to-seven figure applies specifically to patients without microvascular invasion — a variable knowable only after explant. That caveat is why the metroticket concept refined the understanding of risk without replacing Milan as the allocation rule.
Evidence
Derivation — Milan, single-centre transplant series
1996 · n = 48Patients with cirrhosis and small hepatocellular carcinoma undergoing liver transplantation at a single Italian centre, analysed against tumour size and number assessed on the explanted liver.
Of 48 patients, 35 (73%) met the criteria on pathological review and had 4-year overall survival of 85% and recurrence-free survival of 92%; the 13 (27%) who exceeded them had 50% and 59% respectively (P = 0.01 and P = 0.002).
Expanded criteria — UCSF, Yao 2001
2001 · n = 70Seventy consecutive patients with cirrhosis and hepatocellular carcinoma transplanted over a 12-year period at a single US centre, used to test whether the size limits could be widened without harming outcomes.
Patients meeting the expanded criteria — solitary tumour ≤ 6.5 cm, or ≤ 3 nodules with the largest ≤ 4.5 cm and total tumour diameter ≤ 8 cm — had survival of 90% at 1 year and 75.2% at 5 years; those exceeding them had 50% at 1 year.
Metroticket and up-to-seven — Mazzaferro 2009
2009 · n = 1,556Multicentre retrospective study of 1,556 transplanted patients, of whom 1,112 exceeded the Milan criteria and 444 met them on pathology review, used to model survival as a continuous function of size and number rather than a threshold.
Among 283 patients without microvascular invasion who met the up-to-seven criterion, 5-year overall survival was 71.2% (95% CI 64.3–77.0), supporting a graded rather than binary relationship between burden and outcome.
Adding a biomarker — French AFP model, Duvoux 2012
2012French multicentre study incorporating alpha-fetoprotein alongside size and number, in patients transplanted for hepatocellular carcinoma.
A model including alpha-fetoprotein outperformed the Milan criteria for predicting recurrence and survival, and the resulting AFP score has been adopted for allocation in France.
How it compares
Milan Criteria vs UCSF criteria
UCSF widens the size limits and reported 5-year survival of 75.2%, but Milan remains the reference standard because the expansion rests on a much smaller and single-centre evidence base.
UCSF permits a solitary tumour up to 6.5 cm, or up to three nodules with the largest up to 4.5 cm and a total tumour diameter up to 8 cm. In the original 70-patient series, patients meeting those limits had survival of 90% at one year and 75.2% at five years, while those exceeding them had 50% at one year. The case for the expansion is that Milan excludes patients who would do well; the case against is that a 70-patient single-centre series is thin evidence on which to redistribute scarce organs, and that survival at five years was numerically lower than Milan's four-year figure in a differently selected population. Which set applies is a programme and jurisdiction decision, not a bedside one.
Milan Criteria vs Up-to-seven criteria (metroticket)
Up-to-seven shows that the relationship between tumour burden and survival is continuous rather than a threshold — but its headline figure applies only to patients without microvascular invasion, which cannot be known before transplantation.
The up-to-seven criterion sets the sum of the size of the largest tumour in centimetres and the number of tumours at seven or less. In a 1,556-patient multicentre series, the 283 patients meeting it without microvascular invasion achieved 5-year overall survival of 71.2% (95% CI 64.3–77.0). The conceptual contribution is real: it demonstrates there is no biological cliff at the Milan limits, and it produced the metroticket calculators that estimate survival across a continuum. The practical limitation is equally real, because the qualifying condition is a pathological finding available only after the organ has been used. It refines how risk is understood rather than providing a directly applicable allocation rule.
Milan Criteria vs ALBI grade
Two halves of one assessment — Milan describes the tumour and says nothing about the liver, ALBI describes the liver and says nothing about the tumour.
The Milan criteria are purely radiological, and a patient can sit comfortably within them with a liver that will not tolerate a bridging embolisation or a resection while waiting. ALBI supplies that axis from albumin and bilirubin alone. In practice the pair determines the route: within Milan with good hepatic reserve leaves resection and bridging therapy on the table, while within Milan with ALBI grade 3 argues for transplantation without an aggressive bridge, because there is no functional margin to spend. Neither tool ranks waiting-list priority — that remains MELD 3.0 with hepatocellular carcinoma exception points.
Milan Criteria vs French AFP model
Adding alpha-fetoprotein to size and number outperformed Milan for predicting recurrence, and France allocates on that basis — evidence that a biomarker captures biology imaging misses.
The French multicentre study showed that a model incorporating alpha-fetoprotein alongside tumour size and number predicted recurrence and survival better than the Milan criteria, and the resulting AFP score was adopted for national allocation. This matters conceptually because it addresses Milan's core weakness directly: size and number are proxies for aggressiveness, and a rapidly rising alpha-fetoprotein identifies unfavourable biology in a tumour that looks acceptable on imaging. The counter-argument is that alpha-fetoprotein is normal in a substantial minority of hepatocellular carcinomas, so it adds discrimination without providing it universally.
Pearls & pitfalls
- Two lesions of 4 cm are beyond Milan. The 5 cm limit applies only to a solitary lesion; once there is more than one, every lesion must be 3 cm or smaller.
- Total tumour diameter is a UCSF concept, not a Milan one. Adding diameters together and comparing with 8 cm applies a different rule than the one you think you are applying.
- Macrovascular invasion overrides everything. A 2 cm tumour with portal vein thrombus is beyond criteria, and favourable size does not mitigate it.
- The criteria were derived from explant pathology and are applied to imaging. Understaging on preoperative imaging is common and is the principal reason real-world outcomes fall short of the 1996 figures.
- Microvascular invasion is the strongest recurrence predictor and is invisible before explant. Any imaging-based rule therefore has an error floor that better scanners will not remove.
- Nodal enlargement in cirrhosis is often reactive. Treating every visible node as extrahepatic spread wrongly excludes patients; treating none as suspicious wrongly includes them.
- Beyond criteria is not a discharge. Downstaging is an established pathway with published outcomes, and a patient told they are ineligible without a discussion of it has been under-served.
- Within Milan does not mean transplantation is the right treatment. Resection or ablation may serve a patient with well-preserved liver function better, which is what the multidisciplinary meeting is for.
- Progression while waiting happens. Reassess at intervals and bridge with locoregional therapy rather than assuming eligibility established at listing persists.
Critical actions
- Stage on multiphase contrast-enhanced CT or MRI, and confirm each counted lesion meets radiological criteria for hepatocellular carcinoma rather than being an indeterminate nodule.
- Record which criterion failed when the result is beyond, because burden and macrovascular invasion lead to completely different pathways.
- Take every case to a hepatobiliary multidisciplinary meeting, whether within criteria or beyond — the decision between transplantation, resection, ablation and downstaging is not a single-clinician call.
- Calculate MELD 3.0 for allocation priority, remembering that hepatocellular carcinoma exception points are handled separately from the laboratory score.
- Assess hepatic reserve alongside tumour burden — ALBI grade or Child-Pugh — because the criteria say nothing about whether the liver will tolerate a bridge or a resection.
- Offer bridging locoregional therapy where a wait is expected, to reduce the risk of progression beyond criteria before an organ becomes available.
- Where a patient is downstaged, document a period of observed stability before transplantation rather than proceeding on the first favourable scan.
- Check which criteria your jurisdiction and programme actually use — UCSF, up-to-seven and AFP-based models govern allocation in some systems, and Milan is not universal.
Why this score exists
The authors were addressing a field that had largely given up. Transplantation for hepatocellular carcinoma had been attempted, had produced high recurrence rates, and was being abandoned on the reasonable grounds that a scarce organ should not go to a patient who would relapse within two years. What the Milan group argued was that the failures reflected patient selection rather than the operation: if tumour burden was restricted enough, the biology was favourable enough that transplantation was curative. The restrictiveness was the point, and the authors were explicit that they had chosen limits conservative enough to make the comparison with non-malignant indications defensible. Every subsequent attempt to widen the criteria has had to argue against that original logic, and the reason Milan has survived is not that it is optimal — the metroticket work from the same group shows the relationship is continuous, not a cliff — but that any widening trades a defensible equity position for additional transplants of uncertain benefit, and that is a policy judgement rather than a clinical one.
About the creator
First author, Milan criteria and the metroticket studies
Led both the 1996 study that defined the criteria and the 2009 multicentre analysis that reframed tumour burden as a continuous predictor.
Co-author, 1996 derivation series
Co-authored the original Milan transplant series in patients with small hepatocellular carcinoma.
First author, UCSF expanded criteria
Proposed and validated the UCSF expansion, and much of the subsequent work establishing downstaging as a route to transplantation.
Limitations
- Derived from a 48-patient single-centre series in the mid-1990s, with grouping based on explant pathology rather than the preoperative imaging to which the criteria are now applied.
- Imaging understages tumour burden in a meaningful proportion of patients, so real-world recurrence exceeds what the derivation figures suggest.
- Size and number are proxies for tumour biology, not measures of it. Microvascular invasion and differentiation drive recurrence and neither is assessable before explant.
- No biomarker is included, despite evidence that alpha-fetoprotein adds independent prognostic information — the French AFP model exists precisely because of this gap.
- Deliberately restrictive, so patients who would benefit are excluded. The metroticket analysis makes clear there is no biological discontinuity at the thresholds.
- Says nothing about hepatic function, which determines whether a patient can tolerate the wait, the bridge, or an alternative to transplantation.
- Derived in a population where viral hepatitis predominated; the rising proportion of hepatocellular carcinoma arising in metabolic liver disease differs in tumour behaviour and in competing mortality.
- Downstaging protocols, the observation period required after response, and which expanded criteria are permitted all vary between programmes, so 'beyond Milan' means different things in different systems.
If you are the patient
The Milan criteria are the rules doctors use to decide whether a liver transplant is a suitable treatment for liver cancer. Scans are used to check three things: that there is either one tumour no bigger than 5 cm across, or no more than three tumours with none bigger than 3 cm; that the cancer has not grown into the large blood vessels of the liver; and that it has not spread outside the liver. If all three are true, you are described as 'within criteria', and transplantation is an option worth considering — in the study that established these rules, 85 out of every 100 patients within the criteria were alive four years later. If you are outside the criteria, that is not the end of the conversation. Where the issue is that the tumours are a bit too large or too many, treatments given directly to the tumour — such as blocking its blood supply — can sometimes shrink it back within the limits, after which a transplant may become possible again. Where the cancer has grown into a major vein or spread beyond the liver, a transplant is generally not offered, because the operation cannot remove disease that has already travelled. Every case goes to a specialist team meeting rather than being decided by one doctor, and being within the criteria does not automatically mean transplantation is the best choice — surgery to remove just the affected part of the liver is sometimes better.
Frequently asked questions
What are the Milan criteria?#
A single hepatocellular carcinoma lesion 5 cm or smaller, or up to three lesions each 3 cm or smaller, with no macrovascular invasion and no extrahepatic spread. All three conditions must be met simultaneously. They define eligibility for liver transplantation in hepatocellular carcinoma and govern standard allocation in most jurisdictions.
What survival do the Milan criteria predict?#
In the 1996 derivation series of 48 transplanted patients, the 35 who met the criteria on pathological review had 4-year overall survival of 85% and recurrence-free survival of 92%, against 50% and 59% for the 13 who exceeded them. Those differences were statistically significant, and the finding that outcomes matched transplantation for non-malignant disease is what made hepatocellular carcinoma a transplant indication.
Are two 4 cm lesions within the Milan criteria?#
No. The 5 cm limit applies only to a solitary lesion. Once there is more than one lesion, each must be 3 cm or smaller, so two lesions of 4 cm fall outside the criteria despite both being under 5 cm. This is the most frequent misapplication of the rule.
What happens if a patient is beyond the Milan criteria?#
It depends which criterion failed. If the problem is tumour burden alone, downstaging with transarterial chemoembolisation or ablation is an accepted pathway — if the tumour responds and remains stable, transplantation can be reconsidered. If there is macrovascular invasion or extrahepatic spread, transplantation is generally not offered, and management follows BCLC stage toward systemic therapy. Some centres and jurisdictions also apply expanded criteria such as UCSF, up-to-seven or an alpha-fetoprotein-based model.
How do the UCSF criteria differ from Milan?#
UCSF permits a solitary tumour up to 6.5 cm, or up to three nodules with the largest up to 4.5 cm and a total tumour diameter up to 8 cm. It also introduces total tumour diameter, which Milan does not use at all. Reported survival in the original 70-patient series was 90% at one year and 75.2% at five years for patients meeting the expanded limits.
What is the up-to-seven criterion?#
The sum of the size of the largest tumour in centimetres and the number of tumours, at seven or less — so a single 6 cm lesion with one nodule, or three lesions with a 4 cm largest. In a 1,556-patient multicentre series, the 283 patients meeting it without microvascular invasion had 5-year overall survival of 71.2%. The caveat is significant: microvascular invasion is only knowable after the liver is explanted, so the figure cannot be applied prospectively.
Do the Milan criteria account for alpha-fetoprotein?#
No, and that is a recognised gap. The French AFP model showed that adding alpha-fetoprotein to size and number improved prediction of recurrence and survival over Milan, and France now allocates on that basis. Alpha-fetoprotein is normal in a substantial minority of hepatocellular carcinomas, so it adds discrimination for many patients rather than all.
Does meeting the Milan criteria mean a transplant is the best treatment?#
No — it means transplantation is an option. A patient with well-preserved liver function and a single peripheral lesion may achieve equivalent outcomes with resection or ablation while keeping a transplant in reserve. The criteria establish eligibility; the choice between transplantation, resection and ablation depends on hepatic reserve, tumour location and portal hypertension, and belongs in a hepatobiliary multidisciplinary meeting.
Are the Milan criteria assessed on imaging or on the explanted liver?#
They are applied prospectively to contrast-enhanced CT or MRI, but they were derived from assessment of the explanted liver. That mismatch is the source of their main practical weakness — imaging understages a meaningful proportion of patients, which is why observed recurrence rates exceed those in the derivation series.
References
Expanded criteria and validation
- Yao FY, Ferrell L, Bass NM, Watson JJ, Bacchetti P, Venook A, Ascher NL, Roberts JP. Liver transplantation for hepatocellular carcinoma: expansion of the tumor size limits does not adversely impact survival. Hepatology. 2001;33(6):1394-1403 (UCSF criteria).
- Mazzaferro V, Llovet JM, Miceli R, Bhoori S, Schiavo M, Mariani L, Camerini T, Roayaie S, Schwartz ME, Grazi GL, Adam R, Neuhaus P, Salizzoni M, Bruix J, Forner A, De Carlis L, Cillo U, Burroughs AK, Troisi R, Rossi M, Gerunda GE, Lerut J, Belghiti J, Boin I, Gugenheim J, Rochling F, Van Hoek B, Majno P. Predicting survival after liver transplantation in patients with hepatocellular carcinoma beyond the Milan criteria: a retrospective, exploratory analysis. Lancet Oncol. 2009;10(1):35-43 (metroticket / up-to-seven).
- Duvoux C, Roudot-Thoraval F, Decaens T, et al. Liver transplantation for hepatocellular carcinoma: a model including α-fetoprotein improves the performance of Milan criteria. Gastroenterology. 2012;143(4):986-994.