Total tumor volume predicts risk of recurrence following liver transplantation in patients with hepatocellular carcinoma

Authors


  • See Editorial on Page 1071

Abstract

Criteria for the selection of candidates for liver transplantation in the presence of hepatocellular carcinoma (HCC) should accurately predict posttransplant recurrence while not excluding excessive numbers of patients from candidacy. Existing criteria are challenged by the limited accuracy of radiological assessment. The total tumor volume (TTV) was calculated by the addition of the volume of each individual tumor. A preliminary analysis was carried out on HCC patient data from the Alberta Liver Transplant Program (52 patients) and then validated on the populations of the Universities of Toronto and Colorado programs (154 and 82 patients). A TTV cutoff of 115 cm3 was chosen on the basis of the risk of recurrence with use of a receiver operating characteristic curve. Radiology correlated more closely to pathology with TTV than with Milan and University of California at San Francisco (UCSF) criteria (91% versus 69% and 75% of patients, P < 0.0001). Although more patients met qualifying criteria for transplant with TTV (28%-53% more than Milan and 16%-26% more than UCSF), no deterioration of outcome was demonstrated in an analysis of patients within TTV ≤ 115 cm3 in comparison with those meeting Milan or UCSF classifications at all institutions. Patients with TTV > 115 cm3 experienced more recurrences and lower patient survival in the Alberta and Colorado series (P < 0.05). When TTV with a cutoff of 115 cm3 is used for candidate selection, the accuracy of pretransplant radiological assessment is enhanced, with posttransplant outcomes not different from those achieved with Milan and UCSF classifications despite a more inclusive patient population. Liver Transpl 14:1107–1115, 2008. © 2008 AASLD.

Liver transplantation is a rational therapeutic option in selected patients with hepatocellular carcinoma (HCC) and underlying cirrhosis. Various candidacy evaluation systems are currently available to assess patients with HCC. The Milan criteria were first described in 1996 and include patients with a single tumor up to 5 cm in diameter or up to 3 tumors with none larger than 3 cm.1 This classification has been widely used but has been felt by many, including Mazzafero in a recent report,2 to be too restrictive, denying access to liver transplantation to a substantial number of patients who may expect benefit. A more extended set of inclusion criteria was subsequently proposed by the University of California at San Francisco (UCSF) group.3 The latter includes patients with a single tumor up to 6.5 cm diameter or ≤3 tumors with none larger than 4.5 cm and with a cumulative diameter up to 8 cm. Although these selection criteria were originally designed with single-center patient populations, they have now been applied in several subsequent reports.4–6 Unfortunately, the clinical applicability of both of these systems is challenged by the low accuracy of the pretransplant radiological assessment, on which transplant listing decisions are based. Using these classifications, radiology matches the pathology in only 40% to 60% of patients.6–8 There is thus a need for a pretransplant selection classification achieving transplant outcomes at least equivalent to the Milan and UCSF criteria but with improved pretransplant radiological accuracy and with candidacy that does not exclude patients with expectation of good outcome. Such a classification could improve today's practice and could in the future be combined with other tests currently under development, including assessment of tumor cell ploidy or gene expression from tumor biopsy or blood.9

This 3-center study investigated the total tumor volume (TTV) as a pretransplant selection tool, defined a cutoff value, and compared it to other staging systems to predict transplant outcome. It also assessed the ability of these classifications, when applied to radiological studies carried out before transplantation, to predict both the tumor staging on explant pathology and ultimate tumor-free survival.

Abbreviations

CI, confidence interval; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HR, hazard ratio; NASH, nonalcoholic steatohepatitis; NPV, negative predictive value; NS, not significant; PPV, positive predictive value; r, maximum radius of hepatocellular carcinoma; ROC, receiver operating characteristic; TACE, transarterial chemoembolization; TTV, total tumor volume; UCSF, University of California at San Francisco.

PATIENTS AND METHODS

Patient Inclusion

Only liver recipients documented to have HCC before transplantation were included in the study. Patients with incidental HCC discovered on the explanted liver were excluded. They were transplanted at the University of Alberta, Edmonton, Canada (Alberta, from December 1996 to February 2006), at the University of Toronto, Toronto, Canada (Toronto, from January 1996 to January 2006), and at the University of Colorado Health Sciences Center, Denver, CO (Colorado, January 2002 to January 2007). This study was approved by all institutional ethical review board committees.

All institutions used extended inclusion criteria. Selection criteria for transplantation at the University of Alberta were a single HCC up to 7.5 cm or multiple tumors (without number restriction) up to 5 cm in diameter. When an HCC was larger than 5 cm in diameter, a biopsy was obtained to rule out poorly differentiated tumor; having a tumor that was both larger than 5 cm and poorly differentiated would disqualify the patient from transplant. At the University of Toronto, Milan selection criteria were followed till September 2004. After this date, no size and number restrictions were used, but patients were excluded if they had large (>5 cm) tumors that were also poorly differentiated on liver biopsy. At the University of Colorado, Milan criteria were followed, except for living-related donor transplants, who were included patient by patient, without size and number restrictions. In all 3 centers, patients with extrahepatic disease or major vascular invasion on imaging were excluded from candidacy throughout the study time period.

Immunosuppression

In Alberta, immunosuppression was sirolimus-based (target trough range of 12-15 μg/L) with low-dose calcineurin inhibitors. Early patients also received corticosteroids for 3 months; induction therapy with daclizumab replaced steroids for the last 35 patients.

In Toronto, immunosuppression included calcineurin inhibitors and steroids. Hepatitis C virus patients were put on cyclosporin, and non–hepatitis C virus patients were put on tacrolimus. Living-related donor recipients received induction therapy with thymoglobulin or Simulect.

In Colorado, immunosuppression was sirolimus-based with calcineurin inhibitors and 3-day corticosteroid taper till 2003. It subsequently included tacrolimus, mycophenolate mofetil, and short-term (usually less than 1 week) corticosteroids.

Outcome Assessment and Statistical Analysis

Data were collected prospectively in an electronic database (OTTR, Hickman-Kenyon Systems, Omaha, NE) and were analyzed retrospectively. Median follow-up was 25 (1-112) months.

Radiological staging was assessed from reports of contrast-enhanced computed tomography and/or magnetic resonance imaging obtained closest to the time of transplant or prior to pretransplant ablative therapy. A lesion was considered an HCC when it demonstrated an arterial enhancement with washout on delayed images or when a significant increase in size was documented. A stable, nonenhancing small lesion was not considered an HCC. In cases of pretransplant tumor treatment, the pathological estimate of tumor volume included the size of the remaining HCC together with the size of the necrotic tumor. The TTV was calculated as the sum of the volumes of all tumors [(4/3)πr3, where r is the maximum radius of each HCC]. Patient survival was defined as survival with or without HCC recurrence. Patients were classified according to Milan and UCSF criteria as previously described.1, 3

Analysis was initially performed on the Alberta Liver Transplant Program patient population only (52 patients) in order to assess the TTV and to select a cutoff value. The cutoff was determined according to the risk of HCC recurrence on a receiver operating characteristic (ROC) curve. It was selected to achieve the best sensitivity/specificity combination.

Pretransplant radiological staging was next compared to the pathological staging. The radiological accuracy was determined as the rate of patients correctly classified by radiology according to Milan, UCSF, and TTV scores. Patients with radiology upstaging pathology were patients beyond the specific score on radiology but within it on pathology. Patients with radiology downstaging pathology were patients within the specific score on radiology but beyond it on pathology. Survival was analyzed by the Kaplan-Meier method, and differences between groups were further tested by univariate analysis using the log-rank test. Multivariate analysis using a Cox proportional hazards model was used between the prognostic factors reaching at least 0.1 on univariate analysis. These analyses were conducted on the Alberta patient population. As a validation process, the model was used to predict recurrence and survival on 2 independent populations of patients from Toronto and Colorado. Analyses were also performed with both pathological and radiological assessments.

Further tests were performed with the chi-square or Fisher test for categorical variables and the t test for continuous variables. P values less than 0.05 were considered significant. Calculations used Stata (College Station, TX) and Statistica (Statsoft, Berikon, Switzerland) software.

RESULTS

Patient Characteristics

All studied patient populations demonstrated similar characteristics, with more males and median ages of 54, 57, and 54 years (Table 1). Primary underlying liver diseases were most often related to viral hepatitis (Table 1). Roughly half of the patients in each program had pretransplant treatment to control tumor growth. There were differences in practice between the 3 institutions, with use of alcohol injection more frequent in Alberta, use of radio frequency ablation more frequent in Toronto, and use of transarterial chemoembolization more frequent in Colorado (Table 1).

Table 1. Patient and Tumor Characteristics
 Alberta (%)Toronto (%)Colorado (%)P
  • Abbreviations: HBV, hepatitis B virus infection; HCV, hepatitis C virus infection; NASH, nonalcoholic steatohepatitis; TACE, transarterial chemoembolization.

  • *

    Denver versus Alberta or Toronto: P ≤ 0.05.

  • Denver versus Alberta: P ≤ 0.05.

  • Alberta versus Toronto or Denver: P < 0.01.

  • §

    Toronto versus Alberta or Denver: P < 0.01.

Patients (number)5215482 
Median age (years)54 (range 37–67)57 (range 28–70)54 (range 33–65)NS
Gender7/45 female/male21/133 female/male11/71 female/maleNS
Cause of liver disease (%)    
 HCV (±alcohol, ±HBV)23 (44)83 (54)58 (71)*
 HBV16 (31)29 (19)7 (9)*
 Alcohol4 (7.5)25 (16)7 (9)NS
 Cryptogenic3 (5.5)6 (3.5)2 (2)NS
 NASH05 (3)2 (2)NS
 Alpha 1-antitrypsin deficiency03 (2)0NS
 Primary biliary cirrhosis3 (6)1 (0.5)0
 Hemochromatosis2 (4)1 (0.5)1 (1)NS
 Primary sclerosing cholangitis1 (2)02 (2)NS
 Various01 (0.5)3 (4)NS
Pretransplant treatment (%)    
 Ethanol injection13 (25)14 (9)0
 TACE5 (9.5)6 (4)41 (51)*
 Surgical resection3 (6)7 (5)0NS
 Radio frequency ablation3 (6)42 (27)1 (1)§
 None28 (53.5)85 (55)39 (48)NS

Tumor Recurrence and Survival

Seven patients experienced a tumor recurrence in the Alberta series (13%), 15 in Toronto (10%), and 10 in Colorado [12%, P = not significant (NS)]. The median time between transplant and diagnosis of recurrence was 11 months (1.3-51.7). Twelve, 28, and 13 patients died during the posttransplant observation period in Alberta, Toronto, and Colorado (P = NS); among them, 6, 11, and 10 had experienced a recurrence. Patient survivals were similar at all 3 institutions. The 1- and 5-year survivals were 87% and 75% in Alberta, 90% and 72% in Toronto, and 90% and 80% in Colorado (P = NS).

Selection of a TTV Cutoff Value

The selection of a TTV cutoff was performed on the Alberta study population data only. According to pathology, patients with higher calculated TTV were more at risk of recurrence (Fig. 1). ROC analysis identified an optimal cutoff value of 115 cm3 to differentiate HCC-free patients from those with recurrence (Fig. 2A). This could be achieved with good diagnostic accuracy (area under the curve: 0.8). With this cutoff, sensitivity was 71%, specificity was 84%, the positive predictive value (PPV) was 42%, and the negative predictive value (NPV) was 95%. For comparison, the application of Milan and UCSF criteria to predict recurrence in the same data set gave sensitivities of 86% and 57%, specificities of 47% and 60%, PPVs of 20% and 18%, and NPVs of 95% and 90%.

Figure 1.

Distribution of patients with and without hepatocellular carcinoma recurrence in Alberta according to the total tumor volume based on pathological staging.

Figure 2.

Receiver operating characteristic curve assessments of (A) the pathological and (B) radiological total tumor volume scores to predict posttransplant hepatocellular carcinoma recurrence. The areas under the curves were 0.8 and 0.6, respectively. A cutoff of 115 cm3 was selected (*). It provided a sensitivity and specificity of (A) 71% and 84% with pathology and (B) 29% and 98% with radiology. Diagonal segments were produced by ties.

A TTV ROC assessment was also performed on the basis of radiology (Fig. 2B). The area under the curve was 0.6. The same cutoff of 115 cm3 was found to be appropriate and achieved a sensitivity of 29%, a specificity of 98%, a PPV of 67%, and an NPV of 90%.

Pretransplant Radiological Selection Accuracy

The accuracy of pretransplant imaging was similar at all 3 institutions. Overall, radiological staging better matched explant pathology with application of the TTV criteria in comparison with Milan and UCSF classifications (91% versus 69% and 75% of patients, P ≤ 0.0001; Table 2). Errors of classification were most often due to understaging on radiology in comparison with pathology. In keeping with previous reports, the absolute number of tumors was frequently assessed incorrectly with radiology; this was the case in 60%, 47%, and 59% of patients in Alberta (31/52), Toronto (71/154), and Colorado (48/82), respectively (this rate included any patient with an HCC number mistake, whether this impacted on Milan-UCSF staging or not).

Table 2. Performance of Pretransplant Radiological Assessment Compared to Pathology
 Milan (%)UCSF (%)TTV (%)P: TTV versus MilanP: TTV versus UCSF
  1. Abbreviations: NS, not significant; TTV, total tumor volume; UCSF, University of California at San Francisco.

Alberta, n: 52     
 Radiological accuracy32 (62)37 (71)45 (87)≤0.005NS
 Radiological upstaging of pathology5 (10)3 (6)0≤0.05NS
 Radiological downstaging of pathology15 (28)12 (23)7 (13)NSNS
Toronto, n: 154     
 Radiological accuracy110 (71)123 (80)146 (95)≤0.0001≤0.0001
 Radiological upstaging of pathology15 (10)5 (3)1 (0.5)≤0.001NS
 Radiological downstaging of pathology29 (19)26 (17)7 (4.5)≤0.0001≤0.001
Colorado, n: 82     
 Radiological accuracy57 (70)57 (70)70 (85)≤0.01≤0.01
 Radiological upstaging of pathology7 (9)8 (10)1 (1)≤0.05≤0.05
 Radiological downstaging of pathology18 (22)17 (21)11 (13)NSNS
Combined results, n: 288     
 Radiological accuracy199 (69)217 (75)261 (91)≤0.0001≤0.0001
 Radiological upstaging of pathology27 (9)16 (6)2 (1)≤0.0001≤0.001
 Radiological downstaging of pathology62 (22)55 (19)25 (9)≤0.0001≤0.001

Analyses of Variables Predicting Survival

With univariate analyses, several pathology-based variables predicted survival, and they are summarized in Table 3. On multivariate Cox analysis, TTV was the only predicting factor in both Alberta (P = 0.01; Table 4) and Colorado (P ≤ 0.0001). In Toronto, the alpha-fetoprotein level (P ≤ 0.05) and macrovascular invasion (P ≤ 0.0001) were significant.

Table 3. Univariate Analysis of Factors Impacting on Patient Survival from Pathological Staging
 AlbertaTorontoColoradoAll Three Centers Combined
 Patients (%)Five-Year Patient Survival (%)PPatients (%)Five-Year Patient Survival (%)PPatients (%)Five-Year Patient Survival (%)PPatients (%)Five-Year Patient Survival (%)P
  • NOTE: Milan criteria were as previously defined by Mazzaferro et al.1 UCSF criteria were as previously defined by Yao et al.3

  • Abbreviations: NS, not significant; UCSF, University of California at San Francisco.

  • *

    Grade I: well differentiated; grade II: moderately differentiated; grade III: poorly differentiated.

Age 
 ≤5526 (50)780.566 (43)790.543 (52)860.3135 (47)81NS
 >5526 (50)69 88 (57)70 39 (48)75 153 (53)71 
Number of tumors 
 ≤338 (73)74NS119 (77)71NS69 (84)84NS226 (78)76NS
 >314 (27)76 35 (23)87 13 (16)61 62 (22)74 
Tumor size 
 ≤540 (77)79NS140 (91)71NS67 (82)90≤0.0001247 (86)780.002
 >512 (23)62 14 (9)100 15 (18)36 41 (14)61 
Milan criteria 
 Fulfilled22 (42)83NS85 (55)75NS50 (61)91NS157 (55)820.02
 Beyond30 (58)69 69 (46)68 32 (39)59 131 (45)66 
UCSF 
 Fulfilled30 (58)79NS106 (69)72NS57 (70)92≤0.0001193 (67)800.02
 Beyond22 (42)70 48 (31)72 25 (30)50 95 (33)65 
Total tumor volume, cm3 
 ≤11542 (81)87≤0.005141 (91)71NS68 (83)90≤0.0001251 (87)80≤0.0001
 >11510 (19)39 13 (9)100 14 (17)32 37 (13)46 
Tumor grade* 
 I12 (24.5)72NS7 (6)100NS17 (36)940.0236 (17)880.002
 II23 (47)88 91 (76)86 26 (55)61 140 (65)82 
 III14 (28.5)64 21 (18)78 4 (9)50 39 (18)63 
Microvascular invasion 
 No25 (50)78NS113 (73)82≤0.0531 (70)85≤0.01170 (69)810.001
 Yes25 (50)77 41 (27)44 13 (30)25 78 (31)56 
Macrovascular invasion 
 No51 (98)76NS147 (95)79≤0.000181 (99)81≤0.0001279 (97)79≤0.0001
 Yes1 (2)0 7 (5)0 1 (1)0 9 (3)0 
Alpha fetoprotein level, ng/mL 
 ≤1018 (35)78NS55 (40)87≤0.00129 (41)77NS102 (39)820.002
 >10, ≤10017 (32.5)78 53 (39)78 17 (24)91 87 (34)83 
 >10017 (32.5)73 28 (21)43 25 (35)66 70 (27)61 
Table 4. Step-by-Step Multivariate Cox Analysis Using Pathology Staging (Significant Variables)
 HR95% CIP
  1. Abbreviations: CI, confidence interval; HR, hazard ratio.

Alberta   
 Total tumor volume, cm3 (≤115 versus >115)4.91.4–16.30.01
Toronto   
 Macrovascular invasion (no versus yes)2.81.6–4.9≤0.0001
 Alpha fetoprotein level, ng/mL (≤10 versus >10)3.31.1–10.2≤0.05
Colorado   
 Total tumor volume (≤115 versus >115)13.94.3–44.9≤0.0001

Using radiology-based classifications, we further performed the same analysis, looking at variables that might predict survival. The median time between images and transplantation was 2 (0.03-13) months. TTV tended to predict patient survival in Alberta (P ≤ 0.09) and in Colorado (P ≤ 0.05). Besides tumor size (P ≤ 0.01 in Colorado), none of the other variables reached statistical significance.

When combining all centers and using both pathology-based and radiology-based staging, we found that patients beyond Milan but within TTV ≤ 115 cm3 had survivals similar to those of patients within Milan (Figs. 3A and 4A). On the contrary, patients with TTV > 115 cm3 demonstrated lower survival than those within TTV ≤ 115 cm3 when pathology (5-year: 47% versus 79%, P ≤ 0.001) and radiology staging (5-year: 53% versus 76%, P = 0.1) was used.

Figure 3.

Pathology-based staging: (A) cumulative survival (log rank: P = 0.2) and (B) cumulative risk of HCC recurrence (P = 0.1) for all patients grouped as within Milan or beyond Milan–within TTV (≤115 cm3). Numbers of patients at risk are reported under each specific time point. Abbreviations: HCC, hepatocellular carcinoma; TTV, total tumor volume.

Figure 4.

Radiology-based staging: (A) cumulative survival (log rank: P = 0.3) and (B) cumulative risk of HCC recurrence (P = 0.3) for all patients grouped as within Milan or beyond Milan–within TTV (≤115 cm3). Numbers of patients at risk are reported under each specific time point. Abbreviations: HCC, hepatocellular carcinoma; TTV, total tumor volume.

Analyses of Variables Predicting Tumor Recurrence

With both pathological and radiological staging, the overall risk of HCC recurrence remained stable at all institutions for patients within TTV classification but increased for patients with a TTV exceeding 115 cm3:

When pathological staging was used, no significant difference could be observed (P = NS between all conditions) between patient groups divided into within Milan (1/22, 5% in Alberta; 6/85, 7% in Toronto; and 2/50, 4% in Colorado), beyond Milan–within UCSF (2/8, 25%; 2/21, 9.5%; and 0/7, 0%), and beyond UCSF–within TTV (0/12, 0%; 5/35, 14%; and 1/11, 9%) classifications. However, the risk increased when patients had a TTV exceeding 115 cm3, reaching 50% in Alberta, 15% in Toronto, and 50% in Colorado. This was significantly higher than recurrence rates for patients who were within Milan, UCSF, and TTV (≤115 cm3) criteria in Alberta (P ≤ 0.01, P ≤ 0.05, and P ≤ 0.001) and in Colorado (P ≤ 0.001, P ≤ 0.005, and P ≤ 0.00001).

When radiological staging was used, except for an isolated increase in Colorado patients beyond Milan and within UCSF, the risk of HCC recurrence also remained stable when radiology-based Milan (4/32, 12.5% in Alberta; 8/99, 8% in Toronto; and 6/61, 10% in Colorado), beyond Milan–within UCSF (1/7, 14%; 5/29, 17%; and 2/5, 40%), and beyond UCSF–within TTV (0/10, 0%; 2/21, 9.5%; and 0/12, 0%) staging systems were used. In Alberta, 2 of 3 patients with TTV higher than 115 cm3 on radiology experienced a recurrence. This rate was significantly higher than the rate of 10.2% in patients with TTV smaller than 115 cm3 (P ≤ 0.05). Similarly, in Colorado, 2 of 4 patients with TTV higher than 115 cm3 on radiology experienced a recurrence (P ≤ 0.05 versus patients with TTV smaller than 115 cm3).

When assessing all groups together and using both pathology- and radiology-based staging, we found that the cumulative risk of HCC recurrence (assessed by Kaplan-Meier) was similar in patients beyond Milan but within TTV ≤ 115 cm3 compared to those within Milan (Figs. 3B and 4B). On the contrary, patients with TTV > 115 cm3 demonstrated a higher cumulative risk of HCC recurrence than those within TTV ≤ 115 cm3 when pathology (5-year: 58% versus 13%, P ≤ 0.001) and radiology staging (5-year: 39% versus 17%, P ≤ 0.01) was used.

Number of Patients Included for Transplantation

More patients would have been selected for transplantation with TTV compared to Milan and UCSF. With pretransplant radiology, 49, 147, and 78 patients would have qualified for transplantation with TTV classification, 32, 99, and 61 would have qualified with Milan classification, and 39, 127, and 66 would have qualified with UCSF classification in Alberta, Colorado, and Toronto. The number of transplant candidates increased between 28% and 53% in comparison with Milan and between 16% and 26% in comparison with UCSF.

DISCUSSION

This study suggests that TTV may be a useful tool for assessing patients with HCC for consideration of liver transplant. With this staging system, with a cutoff of 115 cm3, the accuracy of pretransplant radiology can be enhanced, with posttransplant outcomes not different from those achieved with Milan and UCSF classifications, despite the inclusion of more patients for transplant.

The present study, like others,6, 10, 11 determined that macrovascular tumor invasion is strongly associated with posttransplant tumor recurrence. The discussed TTV classification should therefore be considered only in patients without major vascular involvement by a tumor on pretransplant imaging. In addition, both lymph node and other extrahepatic metastases remain contraindications for transplantation.

The use of current selection criteria for HCC patients has been challenged by the low accuracy of staging by pretransplant radiology in comparison with explant pathology, usually reaching only 40% to 60%.6–8 In the current report, accuracy was higher than in previous studies, reaching 60% to 70% for Milan and 70% to 80% for UCSF classifications. However, these outcomes were significantly poorer than those achieved with the TTV system, with which 85% to 95% accuracy of pretransplant radiological diagnostic staging was achieved.

The improved accuracy is substantially related to the fact that TTV gives more power to the largest tumors, as the calculated volume increases much faster than the related diameter, the mathematical formula for volume being (4/3)πr3. As a result, the likelihood of correct radiological staging is increased because larger tumors can be more accurately assessed than smaller ones, and larger tumors impact much more strongly on the calculation of TTV. Sotiropoulos et al.7 reported sensitivities of radiology to detect HCC seen on pathology of 0% for tumors smaller than 1 cm, 21% for tumors between 1 and 2 cm, 77.5% for tumors between 2 and 5 cm, and 100% for tumors larger than 5 cm. False positive tumors also tend to involve smaller lesions.12, 13

As a result of the difficulty in assessment of small HCC, the number of tumors had been correctly determined in only 40% to 53% of patients. The TTV staging system does not exclude patients solely on the basis of the number of lesions. For example, a patient with 1 tumor of 4- or 5-cm diameter and 3 tumors of 0.5-cm diameter would be ruled ineligible for liver transplant under both the Milan and UCSF criteria. Such a patient would be eligible under the TTV criteria as the 3 small tumors have little impact on overall tumor volume (TTV = 33.5 and 65.4 cm3, respectively). Our data also suggest that such a patient could be expected to have a very good long-term tumor-free survival and as such would be poorly served by current systems of candidacy evaluations. Our findings that the number of tumors per se does not accurately predict tumor recurrence and survival after liver transplant is consistent with other studies.1, 5, 6, 11, 14

Of note, about half of the patients included in our study received a pretransplant HCC treatment, which, by its nature, has the potential to impact on the size of the lesion. As such, some discrepancy between radiology and pathology may be related to treatment. Importantly, however, this would have impacted similarly on the assessment of the 3 scores (Milan, UCSF, and TTV).

TTV was selected for evaluation as a potentially useful predictor of posttransplant outcome on the basis of the observation that the occurrence of microsatellite HCC metastasis increases exponentially with tumor size, matching tumor volume.15 Several groups further reported that patients with large tumors have diminished tumor-free survivals.3, 10, 14, 16 Survival after liver resection of colorectal metastasis can also be predicted by the tumor volume.17

We first determined a cutoff value from the patient population of the Alberta Liver Transplant Program according to the risk of recurrence and with an ROC curve. Both pathological and radiological assessments provided a total volume cutoff of 115 cm3. By the choice of a higher cutoff, more patients would have been selected for transplant, but the risk of recurrence would have increased significantly, decreasing the specificity of the test. In contrast, had a lower cutoff value been used, several patients with a high likelihood of good long-term outcome would have been excluded from transplant, and this would have decreased the sensitivity of the test. A cutoff of 115 cm3 corresponds to a single tumor of approximately 6 cm in maximum diameter or 3 lesions of approximately 4.2 cm.

As a validation process, assessments were performed separately on the 3 patient populations. Of note, all 3 teams used extended selection criteria, including patients beyond Milan, UCSF, and TTV classifications. Analyses were done with both pathological and radiological staging.

TTV accurately predicted recurrence and patient survival in both the Alberta and Colorado series. This was not the case in the Toronto series. This may be linked to the low number of patients included with TTV > 115 cm3 in the Toronto series (only patients within Milan criteria were accepted in Toronto before 2004) and possibly also to the fact that patients with poorly differentiated large tumors were excluded from transplant candidacy (they would be predicted to have a higher risk of recurrence). As such, the Toronto validation was probably underpowered.

Along the same line and although large, poorly differentiated tumors were also excluded in Alberta, the reader must be aware of this specificity of the study, and further validation in a non-preselected group appears of interest. Of note, one would expect higher rates of recurrence, especially in the groups including large tumors (patients with TTV exceeding 115 cm3).

Importantly, when respecting a TTV cutoff of 115 cm3, we observed overall similar rates of recurrence and patient survival with TTV, Milan, and UCSF classifications at all 3 institutions despite a more inclusive patient population with the TTV system. Recurrence rates in all 3 centers remained around or lower than 10% with both radiological and pathological staging and are in accordance with previously published results.1, 4, 5, 11, 16, 18–20

A point of caution has to be raised as the TTV cutoff was determined on the basis of the Alberta population, which includes only 7 recurrences. An error in 1 or 2 cases could potentially have modified the assessment of the cutoff. However, the selected cutoff in the Alberta population ultimately proved to be the ideal cutoff in the Colorado population as well. Interestingly, TTV is a continuous scale, and the cutoff could be modified by the addition of further variables as well as centers' local policies and results, with the application of higher or lower levels of expected recurrence.

Although prospective and more consistent trials are required, our results suggest that TTV, with a cutoff of 115 cm3, may be a useful tool for selection of patients with HCC for transplant candidacy. TTV would allow the transplantation of more patients than Milan and UCSF classifications and would improve the accuracy of radiological staging while preserving equivalent and acceptable rates of tumor recurrence and posttransplant patient survival.

Acknowledgements

The authors thank Abdul Salam, M.Sc., for the statistical support.

Ancillary