Reply

Abstract

The following reply is in response to the letter by Firl et al, published on page 1856 in the May issue (Volume 154, Issue 6). The following reply is in response to the letter by Firl et al, published on page 1856 in the May issue (Volume 154, Issue 6). We thank Firl et al for their thoughtful comments on our study proposing the Metroticket 2.0 Model1Mazzaferro V. et al.Gastroenterology. 2018; 154: 128-139Abstract Full Text Full Text PDF PubMed Scopus (285) Google Scholar: a prognostic tool for patients undergoing liver transplantation (LT) for hepatocellular carcinoma (HCC). Some of their observations are worth a reply. First, we would like to reassure the readers of Gastroenterology that the research strategy guiding our effort for more than 20 years has not changed. LT is a relevant option for cure in patients with HCC if transplant criteria adhere to restrictions based of tumor characteristics rather than liver function.2Mazzaferro V. et al.N Engl J Med. 1996; 334: 693-699Crossref PubMed Scopus (5670) Google Scholar After all, the Milan criteria improved post-LT patient survival worldwide through drastic reduction of the post-transplant HCC recurrence rates. Introducing “HCC-specific survival” as primary end point in the Metroticket study aimed at maintaining survival as primary end point while dissecting—through competitive risk analysis—the tumor-related causes of death after transplant. Because death after LT may precede the event of cancer recurrence, the competing risk analysis weighted appropriately the oncologic determinants of post-LT death and decreased the need of variables recalibration among transplant series of different origin. For what regards allocation, this was not the aim of our study. Through an algorithm based on a large sample of transplanted HCC patients, the Metroticket tool predicts post-LT survival (usefulness) and—as any study lacking intention-to-treat analysis of the entire listed population, including the Hazard Associated with Liver Transplantation for HCC (HALT-HCC)3Sasaki K. et al.Lancet Gastroenterol Hepatol. 2017; 2: 595-603Abstract Full Text Full Text PDF PubMed Scopus (43) Google Scholar—cannot infer on how to prioritize patients, unless allocation is based only on post-transplant usefulness. As we have stated elsewhere,4Cillo U. et al.Am J Transplant. 2015; 15: 2552-2561Crossref PubMed Scopus (128) Google Scholar, 5Mazzaferro V. Hepatology. 2016; 63: 1707-1717Crossref PubMed Scopus (86) Google Scholar the allocation in HCC patients should target a balance of urgency (risk of death/dropout before LT) and survival benefit, after a cap of post-transplant usefulness (ie, survival or tumor-specific survival) is agreed locally, regionally, or nationally according to the different policies. We convene on the fact that “loco-regional therapies (LRT) can induce decompensation, coagulopathy etc,” but these concepts have low or null impact when predicting post-transplant usefulness. Second, prospectively collected data available for analysis in the training cohort were clearly stated in the Methods section of the paper. The readers may observe that data on neutrophil-lymphocyte ratio and Model for End-Stage Liver Disease score with sodium level were not available, thus impeding comparison of the accuracy of the Metroticket 2.0 model with that of the Model of Recurrence After Liver transplantation (MORAL)6Halazun K.J. et al.Ann Surg. 2017; 265: 557-564Crossref PubMed Scopus (168) Google Scholar and HALT-HCC3Sasaki K. et al.Lancet Gastroenterol Hepatol. 2017; 2: 595-603Abstract Full Text Full Text PDF PubMed Scopus (43) Google Scholar scores. Paucity of variables was the drawback of collecting a very large sample size at 3 different centers, in spite of using a single-center series with a broader spectrum of available variables. The robustness of the Metroticket 2.0 model, however, is demonstrated by c-statistics of >0.7 in both training and external validation sets that showed significantly different baseline characteristics. The formulas used to calculate HCC-specific and overall survival in the Metroticket 2.0 model are clearly stated in the article. Through them, the Metroticket 2.0 model accuracy could be compared with that of HALT-HCC or MORAL scores with the limitation that our model did not include living-related LT. With respect to the University of California at Los Angeles recurrence score,7Bodzin A.S. et al.Ann Surg. 2017; 266: 118-125Crossref PubMed Scopus (93) Google Scholar we find it inappropriate to compare that model (which predicts survival in patients presenting with HCC recurrence after LT) with the Metroticket, which predicts survival in HCC candidates before LT. Finally, we cannot agree with Firl et al when stating that the Metroticket 2.0 was not validated at multiple time points during the pre-LT period. The model was developed on a consecutive series of HCC patients who underwent LT at different time points of the HCC history, according to different organ availability and Centers’ policies. In addition, tumor characteristics and alpha-fetoprotein (AFP) levels were intentionally captured at the last staging before LT (median, 2.3 months) when large heterogeneity of pre-LT treatments (and responses) was detectable. Thus, the generated model predictions refer to the mean of different timepoints in the longitudinal history of HCC, implying that the model can be used longitudinally with an implicit sacrifice in accuracy. For example, a patient with 3 nodules, a maximum size of 5 cm, and an AFP of 800 ng/mL would have a prediction of HCC-specific survival at 5 years of 40% (95% confidence interval, 36%-44%), whereas the same patient presenting after LRT 2 vital nodules, the largest of being 5 cm, and an AFP of 30 ng/mL, would have a HCC-specific survival prediction of 76.1% (95% confidence interval, 73.9%-78.3%). Notably, the latter prediction would be valid also in case of a patient presenting at first diagnosis with 2 vital nodules of HCC, the largest being 5 cm and an AFP of 30 ng/mL. Whether our model, which generates an individualized prediction of post-LT outcomes, performs worse or better than other models that incorporate radiologic response and AFP slopes but generate binary predictions8Lai Q. et al.Liver Transpl. 2013; 19: 1108-1118Crossref PubMed Scopus (12) Google Scholar can be surely a matter of debate. We strongly believe that response to LRTs and correspondent AFP slopes should guide prioritization of HCC patients awaiting LT,5Mazzaferro V. Hepatology. 2016; 63: 1707-1717Crossref PubMed Scopus (86) Google Scholar but prediction of dropout and mortality on the waiting list were not the scope of our study. The road toward a precise prediction of post-LT outcomes in patients with HCC remains long. However, the Metroticket 2.0 model is an accessible and solid tool made on pre-LT variables that, integrated with the local allocation policies, may help with decision making in LT for HCC. Predicting Outcomes of Patients Undergoing Liver Transplantation for Hepatocellular CarcinomaGastroenterologyVol. 154Issue 6PreviewWe read with great interest the manuscript by Mazzaferro et al1 entitled “Metroticket 2.0 Model for Analysis of Competing Risks of Death Following Liver Transplantation for Hepatocellular Carcinoma.” The authors established a preoperatively assessed metric specific to “hepatocellular carcinoma (HCC)-related mortality” after liver transplantation (LT) for HCC using 1018 patients from three Italian centers followed by validation in 341 patients from Shanghai, China. The model was described as an improvement over past work and the authors concluded that their model, “could update current recommendations for cirrhotic patients with HCC.” Although the authors should be congratulated on a substantial effort and nuanced approach, we feel several points warrant emphasis to place their work in proper context. Full-Text PDF