ORIGINAL ARTICLE LIVER TRANSPLANTATION 21: , Additional supporting information may be found in the online version of this article.

Transcription

1 LIVER TRANSPLANTATION 21: , 2015 ORIGINAL ARTICLE Utility-Based Criteria for Selecting Patients With Hepatocellular Carcinoma for Liver Transplantation: A Multicenter Cohort Study Using the Alpha-Fetoprotein Model as a Survival Predictor Alessandro Vitale, 1 Fabio Farinati, 2 Patrizia Burra, 2 Franco Trevisani, 3 Edoardo G. Giannini, 4 Francesca Ciccarese, 5 Fabio Piscaglia, 6 Gian Lodovico Rapaccini, 7 Mariella Di Marco, 8 Eugenio Caturelli, 9 Marco Zoli, 6 Franco Borzio, 10 Giuseppe Cabibbo, 11 Martina Felder, 12 Rodolfo Sacco, 13 Filomena Morisco, 14 Gabriele Missale, 15 Francesco Giuseppe Foschi, 16 Antonio Gasbarrini, 17 Gianluca Svegliati Baroni, 18 Roberto Virdone, 19 Maria Chiaramonte, 20 Gaya Spolverato, 1 and Umberto Cillo, 1 for the Italian Liver Cancer group 1 Dipartimento di Chirurgia Generale e Trapianto d Organo, Unit a di Chirurgia Epatobiliare e Trapianti Epatici, Universit a di Padova, Padova, Italy; 2 Divisione di Gastroenterologia, Azienda Universit a di Padova, Padova, Italy; 3 Unit a di Semeiotica Medica, Dipartimento di Scienze Mediche e Chirurgiche, Alma Mater Studiorum, Universit a di Bologna, Bologna, Italy; 4 Dipartimento di Medicina Interna, Unit a di Gastroenterologia, Universit a di Genova, Genova, Italy; 5 Divisione di Chirurgia, Policlinico San Marco, Zingonia, Italy; 6 Unit a di Medicina, Dipartimento di Scienze Mediche e Chirurgiche, Alma Mater Studiorum, Universit a di Bologna, Bologna, Italy; 7 Unit a di Medicina Interna e Gastroenterologia, Complesso Integrato Columbus, Universit a Cattolica di Roma, Roma, Italy; 8 Divisione di Medicina, Azienda Ospedaliera Bolognini, Seriate, Italy; 9 Unit a Operativa di Gastroenterologia, Ospedale Belcolle, Viterbo, Italy; 10 Dipartimento di Medicina, Unit a di Radiologia, Ospedale Fatebenefratelli, Milano, Italy; 11 Dipartimento Biomedico di Medicina Interna e Specialistica, Unit a di Gastroenterologia, Universit a di Palermo, Palermo, Italy; 12 Ospedale Regionale di Bolzano, Unit a di Gastroenterologia, Bolzano, Italy; 13 Unit a Operativa Gastroenterologia e Malattie del Ricambio, Azienda Ospedaliero-Universitaria Pisana, Pisa, Italy; Additional supporting information may be found in the online version of this article. Abbreviations: AFP, alpha-fetoprotein; AIC, Akaike information criterion; BCLC, Barcelona Clinic Liver Cancer; BSC, best supportive care; CTP, Child-Turcotte-Pugh; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HR, hepatic resection; ICER, incremental cost-effectiveness ratio; IQR, interquartile range; ITA.LI.CA, Italian Liver Cancer group; LRT, locoregional therapy; LT, liver transplantation; NHB, net health benefit; NHS, National Health System; QALY, quality-adjusted life year; QoL, quality of life; SE, standard error; TACE, transarterial chemoembolization; WTP, willingness to pay. Alessandro Vitale had full access to all the data in this study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Alessandro Vitale, Franco Trevisani, Fabio Farinati, Edoardo G. Giannini, Francesca Ciccarese, Fabio Piscaglia, Gian Lodovico Rapaccini, Mariella Di Marco, Eugenio Caturelli, Marco Zoli, Franco Borzio, Giuseppe Cabibbo, Martina Felder, Antonio Gasbarrini, Rodolfo Sacco, Francesco Giuseppe Foschi, Gabriele Missale, Filomena Morisco, Gianluca Svegliati Baroni, Roberto Virdone, and Umberto Cillo: data collection. Alessandro Vitale, Patrizia Burra, and Umberto Cillo: study design. Alessandro Vitale, Patrizia Burra, Gaya Spolverato, and Umberto Cillo: data analysis. Alessandro Vitale, Patrizia Burra, Gaya Spolverato, and Umberto Cillo: article writing. Grants and funding support: Nothing to report. Potential conflict of interest: Nothing to report. Address reprint requests to Patrizia Burra, M.D., Divisione di Gastroenterologia, Azienda Universita di Padova, via Giustiniani 2, Padova, Italy. Telephone: ; FAX: / ; burra@unipd.it DOI /lt View this article online at wileyonlinelibrary.com. LIVER TRANSPLANTATION.DOI /lt. Published on behalf of the American Association for the Study of Liver Diseases VC 2015 American Association for the Study of Liver Diseases.

2 LIVER TRANSPLANTATION, Vol. 21, No. 10, 2015 VITALE ET AL Dipartimento di Medicina Clinica e Chirurgia, Unit a di Gastroenterologia, Universit a di Napoli Federico II, Napoli, Italy; 15 Unit a di Malattie Infettive ed Epatologia, Azienda Ospedaliero, Universitaria di Parma, Parma, Italy; 16 Dipartimento di Medicina Interna, Ospedale per gli Infermi di Faenza, Faenza, Italy; 17 Unit a di Medicina Interna e Gastroenterologia, Policlinico Gemelli, Universit a Cattolica di Roma, Roma, Italy; 18 Clinica di Gastroenterologia, Universit a Politecnica delle Marche, Ancona, Italy; 19 Dipartimento Biomedico di Medicina Interna e Specialistica, Unit a di Medicina Interna 2, Azienda Ospedaliera Ospedali Riuniti Villa Sofia-Cervello, Palermo, Italy; and 20 Unit a di Gastroenterologia, Ospedale Sacro Cuore Don Calabria, Negrar, Italy The lifetime utility of liver transplantation (LT) in patients with hepatocellular carcinoma (HCC) is still controversial. The aim of this study was to ascertain when LT is cost-effective for HCC patients, with a view to proposing new transplant selection criteria. The study involved a real cohort of potentially transplantable Italian HCC patients (n selected from the Italian Liver Cancer group database) who received nontransplant therapies. A non-lt survival analysis was conducted, the direct costs of therapies were calculated, and a Markov model was used to compute the cost utility of LT over non-lt therapies in Italian and US cost scenarios. Post-LT survival was calculated using the alpha-fetoprotein (AFP) model on the basis of AFP values and radiological size and number of nodules. The primary endpoint was the net health benefit (NHB), defined as LT survival benefit in quality-adjusted life years minus incremental costs (US $)/willingness to pay. The calculated median cost of non-lt therapies per patient was US $53,042 in Italy and US $62,827 in the United States. On Monte Carlo simulation, the NHB of LT was always positive for AFP model values 3 and always negative for values > 7 in both countries. A multivariate model showed that nontumor variables (patient s age, Child-Turcotte-Pugh [CTP] class, and alternative therapies) had the potential to shift the AFP model threshold of LT cost-ineffectiveness from 3 to 7. LT proved always costeffective for HCC patients with AFP model values 3, whereas the cost-ineffectiveness threshold ranged between 3 and 7 using nontumor variables. Liver Transpl 21: , VC 2015 AASLD. Received February 5, 2015; accepted June 3, Liver transplantation (LT) is theoretically the best treatment for hepatocellular carcinoma (HCC) in patients with cirrhosis. 1 The posttransplant outcome of patients with HCC beyond the Milan criteria undergoing LT is still controversial, however, because most of the published data are based on the pathological examination of the explanted liver. 2 A novel model has been recently proposed for predicting outcomes in patients undergoing LT for HCC the alpha-fetoprotein (AFP) model 3 that has 2 main advantages: first, it was obtained using a robust methodology (ie, a large multicenter cohort of patients, with a training and a validation set); and second, it is based on preoperative factors such as radiological size and number of nodules and AFP levels (Table 1). The developers of the model proposed a cutoff of 2 for rejecting patients for LT. The main limitation of the AFP model TABLE 1. The AFP Model 3 Variables Points Largest diameter, cm >6 4 Number of nodules >4 2 AFP level, ng/ml > lies in that it only predicts the outcome of patients undergoing LT; it does not consider the survival prospects of patients unable to receive a transplant. Resource allocation for alternative health interventions should rely on measuring their relative efficacy in the production of health. 4 The term utility is often used in transplant ethics when discussing efforts to maximize the benefit for the largest number of patients, considering both the effectiveness and the economic burden of such surgery. In this context, the true utility of LT is not simply a matter of posttransplant outcome but should be assessed in terms of years of life gained over the best alternative therapeutic option. 4 This is very important because, although traditional LT inclusion/exclusion criteria are based exclusively on tumor characteristics, the utility of LT may also be influenced by nontumor variables such as a patient s age, liver function, and the availability of alternative therapies. 5 Some survival-benefit studies on LT for HCC patients 1,5-8 performed in recent years have suggested that LT is of limited utility in HCC patients with compensated cirrhosis who are amenable to alternative radical therapies such as hepatic resection (HR). The concept of transplant utility may be expanded to lifetime and may include quality of life (QoL) and costs, thus becoming a cost-effectiveness analysis. Lim et al. 9 showed in a recent article that LT is not cost-effective in patients with compensated cirrhosis for whom HR is an option when their HCC is within the Milan criteria. The lifetime utility of LT, considering all the HCCrelated aspects (within and beyond Milan criteria) and the nontumor variables (patient s age, liver function, and alternative therapies), nonetheless remains an

3 1252 VITALE ET AL. LIVER TRANSPLANTATION, October 2015 open question. The aim of this study was to ascertain when LT is cost-effective (net health benefit [NHB] > 0) or cost-ineffective (NHB 0) for HCC patients with a view to revising the criteria for selecting patients for LT. To do so, we calculated the lifetime NHB of LT for HCC, considering both tumor-related (AFP model components) and nontumor variables (patient s age, CTP classes, and alternative treatment options). Two cost scenarios were considered in Italy and the United States to improve the generalizability of our findings. PATIENTS AND METHODS The study involved a real cohort of Italian HCC patients receiving nontransplant therapies. A survival analysis was performed, and the direct cost of their therapies was calculated. Then a Markov model was used to compute the survival benefit and cost-utility of LT over non-lt therapies in cases of HCC. The Institutional Review Board of the participating institutions approved the study. Study Population and Nontransplant Survival Model Prospectively collected data on 5134 consecutive Italian patients with HCC (enrolled in the Italian Liver Cancer group [ITA.LI.CA] database from 1988 to 2012) were retrospectively assessed. Among them, 2419 patients receiving nontransplant therapies who had a complete set of follow-up data were judged potentially eligible for LT according to the following inclusion criteria: 1. Age 70 years and no relevant extrahepatic comorbidities. 2. HCC with a noninfiltrating pattern or involving less than 50% of the liver, no metastases or macroscopic vascular invasion at preoperative imaging. 3. A deteriorated performance status (>0) due to cancer symptoms because such patients are considered at an advanced or terminal stage of liver disease. 10 The following variables were recorded for each patient: age, sex, etiology of underlying liver disease, main serological parameters (total bilirubin, creatinine, prothrombin time and/or international normalized ratio, and AFP levels), CTP class, radiological features of the tumor (number and size of lesions), Barcelona Clinic Liver Cancer (BCLC) stage, and main treatment strategy (HR, percutaneous ablation, transarterial chemoembolization [TACE], or other options). As in the Duvoux et al. 3 study, the number of nodules (1-3 versus >3 nodules), diameter of the largest nodule (3, 3-6, and >6 cm), and AFP levels (100, , and >1000 ng/ml) were modeled as ordinal variables. The alternative therapy was also considered as an ordinal variable. We analyzed 4 main treatment subgroups: 1. Patients undergoing HR (resection group, n 5 357) were followed up from the time of resection onward. These patients remained in the resection group even if they subsequently received other nonsurgical HCC therapies. 2. Patients undergoing at least 1 percutaneous ablation procedure, such as radiofrequency ablation (n 5 798), were followed up from the time of first ablation onward, regardless of whether they received other nonsurgical treatments afterward. 3. Patients undergoing at least 1 TACE (n 5 791). 4. Patients receiving none of the above treatments but only receiving best supportive care (BSC; n 5 473). All patients in the BSC group had decompensated cirrhosis (133 had Child C and 340 had Child B), which contraindicated any locoregional HCC therapy. A multivariate survival model was developed for these non-lt patients, including AFP model variables (AFP, nodule diameter, and nodule number), CTP classification, and alternative treatments, to calculate the beta values (and standard errors [SEs]) of the covariates. Details on the statistical analysis are given in the Supporting Tables and Figure. Markov Model A Markov prediction model was developed to estimate the lifetime utility of LT in our cohort of patients receiving nontransplant therapies. The model was constructed by simulating the ideal clinical scenario of a randomized trial in which 2 similar but independent populations were considered: one immediately undergoing LT, the other receiving the best nontransplant standard of care. Our study group of 2419 HCC patients was analyzed in terms of tumor characteristics (AFP, nodule size, and nodule number), liver function, and alternative therapy to estimate the survival outcome with and without LT. Survival calculations in the 2 arms were performed: in the non-lt scenario, using beta values and SEs derived from the multivariate non-lt survival model; in the post-lt scenario, using the AFP model. 3 The primary endpoint of this study was NHB, defined as the LT survival benefit in quality-adjusted life years (QALYs) minus the incremental costs (US $)/willingness to pay. Details on the costs and utilities used in this study for the Italian and US scenarios are given in the Supporting Tables and Figure. A probabilistic sensitivity analysis was then performed, simulating a comparison between a cohort of 10,000 HCC patients undergoing LT and a cohort of 10,000 patients with the same baseline characteristics receiving non-lt therapies. RESULTS Characteristics of the Study Group and Survival Analysis Most of the 2419 patients treated for HCC at institutions contributing to the ITA.LI.CA database had

4 LIVER TRANSPLANTATION, Vol. 21, No. 10, 2015 VITALE ET AL TABLE 2. Characteristics of Patients and Tumors Variables (n ) Sex, female 519 (21) Age, years 63 (57-67) Alcohol use 709 (29) HBV positivity 497 (21) Anti-HCV positivity 1395 (57) CTP class A 1553 (64) B 733 (30) C 133 (6) Diameter of largest lesion, cm (60) (32) >6 190 (8) Multinodularity (>3 nodules) 300 (12) AFP, ng/ml (81) (14) > (5) AFP model score (79) (17) (5) BCLC (9) A 1034 (43) B 1030 (42) D 133 (6) Main therapy HR 357 (15) LRT 1589 (65) BSC 473 (20) hepatitis C virus (HCV) infection (57%) and Child A liver cirrhosis (64%). Concerning their tumors, the largest nodule exceeded 3 cm in diameter in 40% of patients; 12% of patients had more than 3 nodules; and 19% had AFP levels higher than 100 ng/ml (Table 2). The enrolled patients median survival after therapy was 26 months (13-48 months). Table 3 shows the prognostic impact of the variables related to liver function, tumor, and treatment on non-lt survival in the study group on the basis of a Log-logistic parametric multivariate survival model. All variables included in the model had a strong prognostic effect on non-lt survival (P < 0.001). The Loglogistic model was chosen for this survival analysis because it has a greater capacity for discrimination (lower Akaike information criterion [AIC] value) than the conventional Cox s model or other parametric (exponential, Weibull, Log-normal) models. NHB of LT Over Alternative Therapies Study Group NOTE: Data are given as n (%) and median (IQR). The real-life median cost of non-lt therapies per patient calculated in the study group was US $15,700 TABLE 3. Parametric Log-Logistic Survival Model Predicting Non-LT Survival Beta Estimate SE P Value Diameter of largest <0.001 lesion, cm 3-6 versus <0.001 >6 versus AFP, ng/ml < versus <0.001 >1000 versus Multinodularity <0.001 Alternative therapy <0.001 Ablation versus HR TACE versus ablation BSC versus TACE <0.001 CTP class <0.001 B versus A <0.001 C versus B <0.001 AIC Cox s model 20,238 AIC exponential model 14,993 AIC Weibull s model 14,885 AIC Log-normal model 14,740 AIC Log-logistic model 14,682 for Italy and US $53,042 for the United States (Supporting Table 1). The Monte Carlo simulation produced 10,000 NHB values of LT over alternative therapies for HCC patients for each cost scenario (Italy and United States). We first explored the overall impact of the AFP model on the NHB of LT in the 2 countries. The NHB of LT was always positive (ie, the whole range of NHB values was above 0) when the AFP model values were 3, whereas the former was always negative when the latter values were >7, in both Italy and the United States (Fig. 1A,B). Table 4 shows the results of Monte Carlo simulation in the 2 cost scenarios in detail. The variation in the incremental cost-effectiveness ratio (ICER) and the NHB due to the values of the AFP model were almost entirely attributable to variations in the survival benefit rather than to variations in the incremental costs of LT versus alternative therapies. Although the survival benefit was significantly lower as the AFP model value became higher (especially for values higher than 3), the incremental costs remained much the same for each AFP model value and in both of the considered countries. The impact of clinically relevant variables on the distribution of the NHB for the 10,000 outcomes obtained with the Monte Carlo simulation was examined using the multivariate standard least square regression method. The results are represented graphically in Fig. 2 and described in detail in Table 5. In both the Italian and the US scenarios, tumor features showed a strong negative impact on the NHB of LT. Conversely, patient s age 60 years, liver dysfunction (Child B and C classes), and the lack of

5 1254 VITALE ET AL. LIVER TRANSPLANTATION, October 2015 Figure 1. NHB of LT in HCC patients. NHB of LT based on the AFP model in patients with HCC in the (A) Italian and (B) United States cost scenarios. Error bars and box plots represent the overall values and IQRs, respectively, of the NHB for each class of AFP model values. The NHB of LT was always positive for AFP model values 3 and always negative for values > 7 in both countries. TABLE 4. Results of Monte Carlo Simulation (With 10,000 Samples) in the Italian and US Scenarios Survival Incremental AFP Model n Benefit, QALYs Cost, US $ ICER, US $ NHB (range) Italy ,704 19, (0.4, 5.3) ,619 19, (0.5, 5.2) ,566 20, (0.5, 4.7) ,404 20, (0.3, 4.4) ,150 23, (20.3, 4.0) ,060 25, (20.4, 2.8) ,827 31, (21.2, 2.1) ,783 38, (21.5, 0.7) ,047 59, (22.4, 20.8) ,026 77, (22.7, 21.5) United States ,111 32, (0.2, 4.5) ,369 32, (0.4, 4.4) ,290 34, (0.1, 4.0) ,265 34, (0.0, 3.8) ,477 38, (20.4, 3.1) ,748 41, (20.6, 2.4) ,005 49, (21.3, 1.6) ,428 60, (21.6, 0.4) ,220 91, (22.4, 20.9) , , (22.7, 21.6) NOTE: Two different WTP values were used to calculate the NHB: US $35,926 for Italy and US $53,042 for the United States. effective alternative therapies (TACE or BSC) had a significant positive impact on the NHB of LT. The constant term in the final model, representing the estimated NHB of LT at the baseline (patients > 60 years old, Child A, undergoing HR) and an AFP model value of 0, was quite similar for Italy (2.2 QALYs) and the United States (1.7 QALYs). The effect of any given covariate was expressed as the variance from the constant term NHB. For example, among patients who represented the baseline scenario, except that they were 60 years old, the NHB of LT relative to alternative therapies was 1.1 QALYs higher in Italy or 0.9 QALYs higher in the United States. Similarly, if a patient presented with a nodule >6 cm in diameter, the NHB of LT relative to alternative therapy was 0.2 QALYs in Italy and 0.5 in the United States (2.4 and 2.2 QALYs lower than the constant term, respectively). Using these NHB calculators (Table 5) thus enabled us to see when LT was cost-effective or costineffective in the gray area of AFP model values

6 LIVER TRANSPLANTATION, Vol. 21, No. 10, 2015 VITALE ET AL Figure 2. Monte Carlo simulation. Multivariate analysis of the results of the Monte Carlo simulation in the (A) Italian and (B) US cost scenarios. The estimated NHB describes the overall contribution of each covariate to a multivariate standard least square regression model. between 3 and 7. To illustrate this relevant point graphically, we depicted the variation in the predicted NHB (on isometric profiler graphs) by AFP model value (x axis), and the predicted NHB when the AFP model value is 0 (y axis). The y axis thus represents the effect of nontumor covariates (Table 5) on the predicted NHB. This ranged from 2.2 QALYs (Child A patient > 60 years old undergoing HR) to 4.9 QALYs (Child C patient 60 years old undergoing BSC) in Italy and from 1.7 to 4.2 QALYs in the United States (Fig. 3A,B). On the basis of the characteristics of the nontumor variables (y axis), the AFP model threshold for LT cost-ineffectiveness (gray area) ranged from 3 to 7 in both of the considered countries. DISCUSSION Current criteria for selecting HCC patients for LT 2 are based on an arbitrary 5-year post-lt survival endpoint (ie, 5-year survival after LT >50%-70%), which was acceptable in a pioneering phase of LT but is clearly anachronistic nowadays when costeffectiveness studies are fundamental in helping health systems to allocate limited medical resources efficiently. 4 On the basis of the results of this study, we classified HCC patients in terms of transplant utility criteria: LT was always cost-effective for HCC patients with AFP model values 3 and was always cost-ineffective for values > 7, whereas a gray area emerged for patients with AFP model values between 3 and 7. This finding clearly demonstrates that tumor characteristics alone may not give a precise indication of the cost-effectiveness of LT in HCC patients. Hence, we proposed a NHB calculator (Table 5) as a means for precisely ascertaining the cost utility of LT on a patient-by-patient basis. We clearly show that, although conventional transplant selection criteria 2 are influenced by tumor characteristics alone, a comprehensive cost-utility analysis of LT for HCC patients should also include nontumor variables, such as patient s age, liver function, and possible alternative therapies, because these aspects are prognostically relevant to patient outcome with or without LT. 11,12 A flexible AFP model threshold of LT cost-ineffectiveness ranging between 3 and 7 was identified using our model covariates (Table 5), showing that LT was costineffective in Child A HCC patients >60 years old undergoing HR (patients with a limited need for LT) if their AFP model values were >3 and that LT was cost ineffective in Child C HCC patients 60 years old undergoing BSC (patients in severe need of LT) if their AFP model values were >7 (Fig. 3); this applied to the cost scenarios in both Italy and the United States. We found that the futility 11 limit, from a costeffectiveness standpoint, was largely less restrictive than the one suggested by an approach based on post-lt disease-free survival alone. In fact, Duvoux et al. 3 suggested an AFP model threshold of 2 for including patients on the waiting list for LT, based on the arbitrary assumption that patients meeting this criterion would have the same post-lt outcome as patients meeting the Milan criteria. 2 Our findings extend this limit, however, to a range of 3-7 points in the AFP model, adopting a true utility-based criterion. This extended limit assigns a considerable prognostic value to nontumor variables such as patient s age, liver function, and available alternative therapies, all of which are factors that measure the real patient s need to receive a transplant. 12 Unlike Lim et al., 9 we found that LT was always cost-effective for patients with a relatively early HCC (AFP model 3), even when these patients underwent HR. Lim et al. 9 calculated that patients within the Milan criteria had a lifetime, QoL-adjusted survival of 5.3 QALYs after LT, whereas we found a much higher figure of approximately 9 QALYs after LT for such patients whose AFP model values were 3 (data not shown). This marked difference may be due to how

7 1256 VITALE ET AL. LIVER TRANSPLANTATION, October 2015 TABLE 5. Multivariate Analysis Including Variables With a Significant Impact on NHB of LT Versus Alternative Therapy Terms Estimate SE T Ratio P Value Italy Constant <0.001 Age 60 years <0.001 Child B <0.001 Child C <0.001 Ablation <0.001 TACE <0.001 BSC <0.001 Size 3-6 cm <0.001 Size >6cm <0.001 AFP , <0.001 ng/ml AFP > 1000, <0.001 ng/ml >3 nodules <0.001 United States <0.001 Constant <0.001 Age 60 years <0.001 Child B <0.001 Child C <0.001 Ablation <0.001 TACE <0.001 BSC <0.001 Size 3-6 cm <0.001 Size > 6cm <0.001 AFP , <0.001 ng/ml AFP > 1000, <0.001 ng/ml >3 nodules <0.001 NOTE: The constant term in the final model represents the estimated NHB of LT at the baseline(patient>60 years old, Child A, undergoing HR, single HCC < 3 cm with AFP < 100 ng/ml). The covariates effects for the estimated NHB are assumed to be additional to the constant term. the model was constructed. Most HCCs recurring after LT do so within the first 2 years, and only 5-year post-lt survival or disease-free survival data are available in the literature 2,3 ; so, for our study, we assumed a low mortality rate beyond 5 years after LT of 2% per year (see Supporting Tables and Figure), although there is no evidence of such an assumption being made in the Lim et al. 9 study. A positive aspect of the present study lies in the fact that cost scenarios in 2 different countries were explored, and it is noteworthy that the results of our cost-utility analysis were very similar for the 2 scenarios: Italy and the United States (Figs. 2 and 3). This gives us confidence concerning the generalizability of our results. However, our study also has several potential limitations. First, though it contributes to defining utilitybased criteria for selecting HCC patients for LT, the allocation of liver grafts to patients with end-stage liver disease is also driven by another pivotal ethical principle, and that is equity. 13 The equity criterion relies on an urgency principle (the sickest first ), prioritizing patients with more severe disease. 13,14 This point is particularly important when we consider that a relevant proportion of patients waiting for a LT suffer from nonmalignant liver disease, and these patients are usually selected and prioritized for LT according to this sickest first principle. On the other hand, lifetime cost-utility analyses are mainly influenced by posttransplant outcome, and a pure utilitybased allocation would suffer from the biases peculiar to cost-effectiveness studies: focusing only on maximizing life years alone would discriminate against the very old and sick patients, who generally have worse outcomes. 4 The final decision should therefore be based on a balanced approach to graft allocation: on the one hand, utility criteria help clinicians to identify patients who should be excluded from a transplant program; on the other, equity-weighted criteria are useful for deciding the priority of patients with and without HCC on the waiting list. Our findings may consequently be useful at the time of patient listing, but other models are needed to decide their priority on the waiting list. Although a large proportion of HCC patients could potentially be placed on the waiting list according to our criteria (AFP model values up to 7), the final decision should be based not only on case-by-case considerations, but also and more importantly on a population basis, bearing in mind the length of local waiting lists, the number of non- HCC patients with high MELD scores, and donor resources. When the resources available for HCC patients are limited, only HCC patients with no radical treatment alternatives and those with poor liver function should be considered for LT. 1 As in the Duvoux et al. 3 article, we did not consider the risk of dropout before LT. We preferred to disregard this aspect because the risk of dropout or death before LT is also related to local waiting-list issues (the balance between patients with and without HCC), local rules for prioritizing patients for LT, and local donor resources. Another limitation of our study lies in that we used a relatively recently introduced model, the AFP model, to predict patient outcome after LT in a non-lt population. This was done on the grounds of a methodological approach already used in several LT survival benefit models published in recent years, 1,5-7 in which a post- LT survival model (the Metroticket calculator) was applied to non-lt populations to simulate their potential post-lt outcome. In the present study, we likewise applied the AFP model to a large Italian non-lt population that was potentially transplantable in terms of patient and tumor characteristics. The advantage of the AFP model over the Metroticket model is that it is based on preoperative variables. The AFP model was developed with a robust statistical methodology (with training and internal validation cohorts) in France, and a value >2 is currently adopted in France 19 and the

8 LIVER TRANSPLANTATION, Vol. 21, No. 10, 2015 VITALE ET AL Figure 3. Impact of tumor-related (AFP model) and nontumor variables on the NHB of LT. Variation in NHB (isometric profiler graphs) by AFP model value versus nontumor variables (described by the NHB predicted when the AFP model value was 0) in (A) Italy and (B) the United States. The red area in the graphs represents LT cost-ineffectiveness; the white area represents LT cost-effectiveness. The y axis represents the range of predicted NHB when the AFP model value was 0, derived from Table 5: it ranged from 2.2 to 4.9 QALYs for Italy and from 1.7 to 4.2 QALYs for the United States. United Kingdom 20 as a criterion for including HCC patients on the waiting list for LT; it recently underwent external validation in Spain 21 and Italy too. 22 A recent US study also indirectly confirmed some important characteristics of the AFP model, particularly showing the strong prognostic power of the AFP 1000 ng/ml threshold in HCC patients undergoing LT. 23 It is important to emphasize, however, that a potential limitation of using the AFP model relates to the non-lt population. It is unclear whether this model would work equally well in patients who do not typically receive a transplant, such as patients with advanced intrahepatic tumor burden (multifocal HCC but <50% involvement) who are usually treated with locoregional therapies alone, and/or those without portal hypertension who would undergo HR. Another potential weakness of our study concerns the concept of transplantable HCC that we used to select our study group. For instance, we did not define any specific number and size of nodules. On the one hand, we included T1 HCC patients not awarded exception points in the United States, but such patients are not rejected for LT in Italy and France. 24,25 It is also well known that many patients with very early or early HCC are treated with TACE if they are considered unresectable and when percutaneous ablation is contraindicated (due to the tumor s location, for example). 1,5 On the other hand, we also included patients beyond currently accepted criteria. 2 We see this not as a limitation but probably as the most important original aspect of our study. TACE is considered the standard of care for patients with intermediate-stage HCC (beyond the Milan criteria), but many patients in this stage may benefit greatly from LT. 1,26,27 As for BSC, many patients with decompensated cirrhosis may not be given non-lt therapy for their HCC. Paradoxically, they may be offered only 2 options: LT or BSC. To give an example, a patient with a single HCC nodule of 3 cm in diameter but with Child C cirrhosis is a case of BCLC class D and can only choose between LT or BSC; but, such patients stand to benefit the most from LT, in terms of survival 1 when they are within the transplant criteria in terms of tumor characteristics. Other factors, such as financial and psychosocial barriers to LT, were also not considered in our definition of transplantable HCC because of the retrospective nature of our study. We wish to emphasize, however, that there are no financial barriers to LT in Italy (where the National Health System [NHS] covers the costs of LT for all Italian citizens) and that all enrolled patients had a complete follow-up obtained at tertiary referral hepatology units, thus limiting any potential bias due to psychosocial barriers. In conclusion, this study proposes new criteria for selecting patients with HCC as candidates for LT. Our criteria are based on fair and objective utility thresholds derived from a cost-effectiveness analysis of a real-life patient population. REFERENCES 1. Vitale A, Morales RR, Zanus G, Farinati F, Burra P, Angeli P, et al.; for Italian Liver Cancer group. Barcelona Clinic Liver Cancer staging and transplant survival benefit for patients with hepatocellular carcinoma: a multicentre, cohort study. Lancet Oncology 2011;12:

9 1258 VITALE ET AL. LIVER TRANSPLANTATION, October Mazzaferro V, Llovet JM, Miceli R, Bhoori S, Schiavo M, Mariani L, et al.; for Metroticket Investigator Study Group. Predicting survival after liver transplantation in patients with hepatocellular carcinoma beyond the Milan criteria: a retrospective, exploratory analysis. Lancet Oncol 2009;10: Duvoux C, Roudot-Thoraval F, Decaens T, Pessione F, Badran H, Piardi T, et al.; for Liver Transplantation French Study Group. Liver transplantation for hepatocellular carcinoma: a model including a-fetoprotein improves the performance of Milan criteria. Gastroenterology 2012;143: Keller EJ, Kwo PY, Helft PR. Ethical considerations surrounding survival benefit-based liver allocation. Liver Transpl 2014;20: Cillo U, Vitale A, Volk ML, Frigo AC, Grigoletto F, Brolese A, et al. The survival benefit of liver transplantation in hepatocellular carcinoma patients. Dig Liver Dis 2010; 42: Vitale A, Cucchetti A, Qiao GL, Cescon M, Li J, Ramirez Morales R, et al. Is resectable hepatocellular carcinoma a contraindication to liver transplantation? A novel decision model based on number of patients needed to transplant as measure of transplant benefit. J Hepatol 2014;60: Vitale A, Huo TL, Cucchetti A, Lee YH, Volk M, Frigo AC, et al. Survival benefit of liver transplantation versus resection for hepatocellular carcinoma: impact of MELD score. Ann Surg Oncol 2015;22: Berry K, Ioannou GN. Are patients with Child s A cirrhosis and hepatocellular carcinoma appropriate candidates for liver transplantation? Am J Transplan 2012;12: Lim KC, Wang VW, Siddiqui FJ, Shi L, Chan ES, Oh HC, et al. Cost-effectiveness analysis of liver resection versus transplantation for early hepatocellular carcinoma within the Milan criteria. Hepatology 2015;61: Bruix J, Sherman M; for Practice Guidelines Committee, American Association for the Study of Liver Diseases. Management of hepatocellular carcinoma. Hepatology 2005;42: Merion RM. When is a patient too well and when is a patient too sick for a liver transplant? Liver Transpl 2004;10(suppl 2):S69-S Vitale A, Burra P, Frigo AC, Trevisani F, Farinati F, Spolverato G; for Italian Liver Cancer (ITA.LI.CA) group. Survival benefit of liver resection for patients with hepatocellular carcinoma across different Barcelona Clinic Liver Cancer stages: a multicenter study. J Hepatol 2015;62: Persad G, Wertheimer A, Emanuel EJ. Principles for allocation of scarce medical interventions. Lancet 2009;373: James C, Carrin G, Savedoff W, Hanvoravongchai P. Clarifying efficiency-equity tradeoffs through explicit criteria, with a focus on developing countries. Health Care Anal 2005;13: Toso C, Mazzaferro V, Bruix J, Freeman R, Mentha G, Majno P. Toward a better liver graft allocation that accounts for candidates with and without hepatocellular carcinoma. Am J Transplant 2014;14: Toso C, Dupuis-Lozeron E, Majno P, Berney T, Kneteman NM, Perneger T, et al. A model for dropout assessment of candidates with or without hepatocellular carcinoma on a common liver transplant waiting list. Hepatology 2012; 56: Toso C, Majno P, Berney T, Morel P, Mentha G, Combescure C. Validation of a dropout assessment model of candidates with/without hepatocellular carcinoma on a common liver transplant waiting list. Transpl Int 2014;27: Vitale A, Volk ML, De Feo TM, Burra P, Frigo AC, Ramirez Morales R, et al. A method for establishing allocation equity among patients with and without hepatocellular carcinoma on a common liver transplant waiting list. J Hepatol 2014;60: Agence de la Biomedecine. Le rapport medical et scientifique de l Agence de la Biomedecine Greffe Hepatique. Accessed August NHS Blood and Transplant Liver Advisory Group. Liver transplantation for hepatocellular carcinoma in the UK report from a national consensus meeting (Birmingham January 2014). National Health Service. recommendations_ir_ts_b_nas_work_in_progress.pdf. Accessed August Varona MA, Soriano A, Aguirre-Jaime A, Garrido S, Oton E, Diaz D, et al. Risk factors of hepatocellular carcinoma recurrence after liver transplantation: accuracy of the alpha-fetoprotein model in a single-center experience. Transplant Proc 2015;47: Notarpaolo A, Bizouard G, Gambato M, Montalti R, Magini G, Miglioresi L, et al. Prediction of recurrence after liver transplantation for HCC: Validation of the AFP model in an Italian cohort. Liver Transpl 2014;20(suppl 1):S Hameed B, Mehta N, Sapisochin G, Roberts JP, Yao FY. Alpha-fetoprotein level > 1000 ng/ml as an exclusion criterion for liver transplantation in patients with hepatocellular carcinoma meeting the Milan criteria. Liver Transpl 2014;20: Angelico M, Cillo U, Fagiuoli S, Gasbarrini A, Gavrila C, Marianelli T, et al.; for Liver Match Investigators. Liver Match, a prospective observational cohort study on liver transplantation in Italy: study design and current practice of donor-recipient matching. Dig Liver Dis 2011;43: Francoz C, Belghiti J, Castaing D, Chazouillères O, Duclos-Vallee JC, Duvoux C, et al. Model for end-stage liver disease exceptions in the context of the French Model for End-Stage Liver Disease score based liver allocation system. Liver Transpl 2011;17: Yao FY, Mehta N, Flemming J, Dodge J, Hameed B, Fix O, et al. Downstaging of hepatocellular cancer before liver transplant: long-term outcome compared to tumors within Milan criteria. Hepatology 2015;61: Toso C, Meeberg G, Hernandez-Alejandro R, Dufour JF, Marotta P, Majno P, Kneteman NM. Total tumor volume and alpha-fetoprotein for selection of transplant candidates with hepatocellular carcinoma: A prospective validation. Hepatology 2015;62: