Transplant‐related survival benefit should influence prioritization for liver transplantation especially in patients with hepatocellular carcinoma

Transplant‐related survival benefit is calculated as the difference between life expectancy with transplantation and life expectancy without transplantation. Determining eligibility and prioritization for liver transplantation based on the highest survival benefit is a superior strategy to prioritization based on the highest urgency (ie, the highest wait‐list mortality) or the highest utility (ie, the highest posttransplant survival) because prioritization based on the highest survival benefit maximizes the overall life expectancy of all patients in need of liver transplantation. Although the Model for End‐Stage Liver Disease (MELD)–based prioritization system was designed as an urgency‐based system, in practice it functions to a large extent as a survival benefit–based system, when the natural MELD score is used without exceptions. Survival benefit considerations should be used to determine the consequences of deviating from prioritization based on the natural MELD score, such as when exception points are awarded to patients with hepatocellular carcinoma (HCC) that are independent of MELD score or tumor burden, or the appropriateness of expanding eligibility for transplantation. The most promising application of survival benefit–based prioritization would be to replace the current system of prioritization of patients with HCC by one that uses their natural MELD score and tumor characteristics such as HCC tumor burden, serum alpha fetoprotein level, and response to locoregional therapies to predict the impact on survival benefit caused by the presence of HCC and adjust the natural MELD score for prioritization accordingly. Liver Transplantation 23 652–662, 2017 AASLD.

This review focuses primarily on the application of survival benefit concepts to determine prioritization in patients with HCC, although the concepts can be applied more broadly. This is a timely topic due to widespread concerns that the current system of prioritization in patients with HCC assigns inappropriately high priority, (6,7) which have led to recent changes in HCC prioritization incorporated by the Organ Procurement and Transplantation Network (OPTN)/ United Network for Organ Sharing (UNOS) in October 2015 (the "HCC Delay" and "Cap HCC" policies), (8) and even more changes in HCC prioritization currently being proposed by the OPTN/UNOS Liver and Intestinal Organ Transplantation Committee. (9) Principles Guiding Prioritization for Liver Transplantation: Urgency, Utility, and Survival Benefit How can we maximize the benefit achieved by the precious limited resource of liver transplantation and what principle should be employed to determine eligibility for liver transplantation or prioritization once a patient is deemed eligible? Three potential principles have been proposed (10) -urgency, utility, and survival benefit- (Table 1), and are further described below.

Urgency: The Current Model for End-Stage Liver Disease-Based Prioritization System
The Model for End-Stage Liver Disease (MELD)based prioritization system is based on urgency, which prioritizes the sickest patient with the greatest need for liver transplantation and the highest wait-list mortality. Since its introduction in 2002, the MELD-based system has been successful in reducing wait-list mortality. (11) Furthermore, it achieves this goal by using only simple, objective, readily available, and inexpensive laboratory tests, although other factors such as liver disease etiology can also potentially affect wait-list mortality. (12) The downside of all urgency-based systems, including the MELD-based system, is that they do not consider posttransplantation survival. Thus, 2 patients with identical MELD scores who happen to have widely disparate expected posttransplant survival will have identical prioritization for transplantation, even though transplantation will result in much greater life expectancy in one versus the other.

Utility: A System Prioritizing Posttransplant Survival
"Utility-based" systems prioritize patients according to their expected posttransplantation survival such Overall survival of all patients in need of transplantation that the patient with the highest expected posttransplantation survival receives the highest priority. Although such a system is not currently employed, multiple scoring systems have been developed to predict posttransplantation survival and could, in theory, be used in utility-based prioritization. (13,14) The limitation of utility-based prioritization systems is that they do not consider the urgency of transplantation, that is, the expected survival without transplantation.

A Survival Benefit-Based System Maximizes the Overall Survival of All Patients in Need of Transplantation
Assuming that we could accurately determine urgency, utility, and survival benefit, which one is the preferable system? To answer that, we need to consider what each system achieves. An urgency-based system maximizes wait-list survival because patients with the highest wait-list mortality are transplanted first (Table 1). A utility-based system maximizes posttransplant survival because the patient with the highest posttransplant survival receives the next available organ. Although it is not intuitively obvious, a survival benefit-based system maximizes the overall survival (ie, before and after transplantation) of all patients on the waiting list and by extension of all patients in need of a liver transplantation. This can be illustrated by a simple example adapted from Schaubel et al. (15) (Table 2) and B (6 years), for a total of 17.5 years. Thus, the overall life expectancy of the whole group is highest when the patient with the highest survival benefit undergoes transplantation, rather than the patient with highest urgency or utility. Table 2 shows that the life expectancy of the whole group is also equal to the sum of the wait-list life expectancies of each patient in the group plus the survival benefit of the patient who is transplanted. In other words, each time a person undergoes transplantation, the life expectancy of the entire population increases by the survival benefit achieved by that transplantation. It follows that in order to maximize the life expectancy of the entire population, each organ has to be allocated to the person with the highest expected survival benefit at that point in time.
Although the example shown in Table 2 may appear to be an oversimplification of a very complicated  process involving thousands of patients listed for, or in need of, transplantation around the country, it actually applies directly to real life. In real life, just as in the example shown in Table 2, only 1 organ is allocated at a time and the benefit derived from each transplantation is added to the sum of all the baseline life expectancies without transplantation of the entire population. Thus, the same principle applies regardless of how many patients are in need of transplantation, how many of them are listed for transplantation, or how many eventually undergo transplantation.

The MELD-Based System Functions in Practice as a Survival Benefit-Based System
Although the MELD-based prioritization system was explicitly designed as an urgency-based system, in practice it functions to a large extent as a survival benefit-based system, when the natural MELD score is used without exceptions. This is because the eligibility criteria that all centers have in place for listing patients for transplantation generally serve to exclude patients with low expected posttransplant survival (ie, low utility), such as patients with serious medical comorbidities and adverse psychosocial characteristics. The fact that center-specific posttransplant outcomes are publicly available (16) and scrutinized by oversight bodies, thirdparty payers, and patients further ensures stringent eligibility criteria aimed at selecting patients with high posttransplant survival. At the same time, prioritization based on the MELD score (urgency) ensures that patients who are transplanted have the lowest life expectancy without transplantation because they have the highest MELD score. Therefore, this yields a very high survival benefit in almost all patients transplanted with high MELD scores under the MELD prioritization system, calculated as the difference between a high life expectancy with transplantation (ensured by listing criteria) minus a very low life expectancy without transplantation (ensured by high MELD score). In fact we (17) and others (18) have shown that transplantrelated survival benefit increases dramatically with MELD score. It is important to emphasize that the success of the MELD-based system depends as much on the fact that it does not predict posttransplant survival (19) as it does on the fact that it does predict wait-list survival. If increasing MELD scores turned out to be associated with decreasing posttransplantation survival, then the MELD-based allocation system would have failed because survival benefit would have been diminished, especially with the ever-increasing wait-list MELD scores that are an inevitable result of a MELD-based allocation system. This was not necessarily a forgone conclusion when the MELD-based system was introduced, and it might have turned out differently. However, thanks to the ever-increasing expertise of transplant surgeons and transplant intensive care units, patients with very high MELD scores undergo liver transplantation with relatively little excess mortality despite being extremely debilitated at the time of transplantation.

HCC-MELD Exceptions Result in Transplanting HCC Patients With Low Natural MELD Score and Low Survival Benefit
Problems arise when we deviate from the MELDbased allocation system, the most notable example being "MELD exceptions" for HCC, that is, artificially high MELD scores assigned to patients with early stage HCC. Patients transplanted with HCC MELD exceptions often have a low natural MELD score, "early stage" HCC (stage II, or Milan criteria) because this is a requirement for transplantation eligibility and in most cases have already received locoregional treatments for HCC. These patients actually have a substantial life expectancy without transplantation resulting in a very low transplantrelated survival benefit. (17,20) In addition, multiple studies have shown that patients with HCC also have a reduced posttransplantation life expectancy compared with patients without HCC (21) due to the small but finite probability of tumor recurrence after transplantation, which further reduces their survival benefit. We have recently demonstrated that patients transplanted in the United States with HCC-MELD exceptions gained a much lower survival benefit than patients transplanted without HCC, due to their lower natural MELD score. (17) Impact of the "Wait But Not Ablate" Strategy for Small HCCs on Survival Benefit According to the "wait but not ablate" strategy, (22) patients with tumors < 2 cm, who are not eligible for expedited priority by MELD exception points, do not receive HCC treatment and instead are observed allowing the tumors to grow to a size >2 cm, thus allowing the patient to get exception points. This strategy enabled many patients to undergo transplantation, and they experienced excellent 3-year posttransplant survival of 75.5%. (22) However, in addition to the problem of potentially leaving some patients untreated who turn out to have rapidly growing tumors that grow beyond Milan criteria (9% at 1 year), (22) this approach offered transplantation to patients with some of the lowest transplant-related survival benefit, that is patients with very small tumors that could have been successfully ablated resulting in substantial survival even without transplantation.
In general, any patient group that has significant life expectancy without transplantation will have reduced survival benefit, even if the patients have adequate posttransplant life expectancy, making them inappropriate liver transplant candidates.

Impact of "HCC Delay" and "Cap HCC" Policies on Survival Benefit
Two policy changes in HCC prioritization were introduced by OPTN/UNOS in October 2015. (8) First, the "HCC Delay" policy mandated a 6-month delay after listing for transplantation before assigning an exception MELD score of 28, followed by the scheduled 3-monthly progression of exception points on the "MELD elevator" (28!30!32!34). Second, the "Cap HCC" policy capped the scheduled progression on the "MELD elevator" to a maximum assigned score of 34, so that these patients would not be candidates for default regional sharing under the Share 35 policy, introduced in June 2013. (23) These changes may slightly reduce the disparities between HCC and non-HCC patients in adverse wait-list outcomes. However, these policies do very little to ensure that patients are prioritized according to their survival benefit or that the "exception score" assigned to patients with HCC is in any way directly related to their expected survival benefit.

Impact of Currently Proposed Changes in HCC Prioritization and Eligibility on Survival Benefit
Three additional policy changes that have been proposed by the OPTN/UNOS Liver and Intestinal Organ Transplantation Committee were up for public comment from August 15, 2016 to October 15, 2016, and were due for presentation to the OPTN board meeting on December 1, 2016. The first proposes that patients with a single, small HCC 2-3 cm in size should be required to undergo locoregional therapy and should only be eligible for automatic standardized MELD exception if the HCC persists or recurs. This proposal is consistent with survival benefit principles because such small HCCs that are completely ablated have little or no impact on survival benefit and therefore should not receive any additional prioritization. The second proposes that certain patients who exceed the current T2 HCC criteria (specifically, 1 lesion 5-8 cm; 2-3 lesions each < 5 cm and total diameter 8 cm; or 4-5 lesions each < 3 cm and total diameter 8 cm) and who are successfully downstaged to T2 by locoregional treatments should be eligible for automatic standardized MELD exception. This proposal is probably the most controversial and is not directly consistent with survival benefit principles because the presence of a downstaged HCC in these patients does not necessarily have a positive impact on their survival benefit over and above their natural MELD score (see Table 4). This would only occur if these patients with downstaged tumors truly have a posttransplant survival similar to those meeting T2 criteria without downstaging, as has been suggested in highly selected patients from centers with special expertise. (24) The third proposal recommends that patients with HCC and alpha fetoprotein (AFP) > 1000 ng/mL should not be eligible for standardized MELD exception, unless the AFP is downstaged to a value 500 ng/mL by locoregional therapies. This proposal is reasonable because the presence of a tumor with AFP > 1000 ng/mL most likely has a negative impact on survival benefit.
Although these proposed changes are reasonable, the prioritization process for patients with HCC would still be entirely based on "standardized MELD exceptions," which assign HCC patients an arbitrary priority score regardless of the natural MELD score or tumor characteristics or the actual impact that the presence of HCC is expected to have on survival benefit.

Survival Benefit Considerations in Other Patient Subgroups: Older Patients
Thinking in terms of survival benefit can lead to "counterintuitive" conclusions about the appropriateness of liver transplantation in certain patient subgroups that are contrary to the conclusions we would reach if we consider only pretransplant survival (urgency) or only posttransplant survival (utility; Table 3). For example, when considering transplantation in "elderly" patients, we might conclude that elderly patients should be given additional priority because they have a higher probability of wait-list dropout or death. (25) Conversely, if we consider only posttransplant survival, we might reach the opposite conclusion because elderly patients clearly have reduced posttransplantation survival. However, we have recently shown that older age diminishes both posttransplantation life expectancy and pretransplantation life expectancy approximately equally and consequently does not affect survival benefit. (25) Therefore, older age in isolation should neither be a contraindication for transplantation nor a reason for assigning increased priority for transplantation over and above that afforded by the patient's natural MELD score. Survival Benefit Combines "Eligibility" and "Prioritization" Into a Single Metric We currently think of "eligibility" for transplantation and "prioritization" as 2 distinct processes. Eligibility is based on patient characteristics that are meant to predict good posttransplantation survival, although these are unfortunately not combined in predictive models in clinical practice. Once a patient is listed for transplantation, the MELD score is used for prioritization, ie, determining who gets the next available organ. However, survival benefit combines "eligibility" and "prioritization" in 1 metric and can be applied to entire populations (listed or not) both to determine whether to list a patient or not and to determine subsequent prioritization. For example, patients with low survival benefit (whether due to high life expectancy without transplantation or low posttransplant life expectancy or both) should not get listed for transplantation. For patients listed, the same metric-survival benefitwould determine their prioritization.

Future Directions: Implementing a Survival Benefit-Based Prioritization System for Patients With HCC
In theory, the entire MELD-based prioritization system could potentially be replaced by a system based on survival benefit resulting in an estimated 2000 lifeyears saved per year. (15) However, this is unlikely to happen in practice because we can predict wait-list survival in patients with high MELD scores with much greater accuracy than posttransplant survival, such that we would be reluctant to bypass a patient with very high MELD score who has a very high likelihood of dying for a "stable" patient with lower MELD score   *The impact of HCC on survival benefit is calculated as the difference between the survival benefit with HCC minus the survival benefit without HCC. If there is a positive impact by the HCC on survival benefit, then the natural MELD score should be increased accordingly. If there is zero impact, the natural MELD score should be unchanged. If there is a negative impact, the natural MELD score should be reduced.
who we predict has greater posttransplant survival, resulting in greater survival benefit. Therefore, it is probably unwise to use relatively small differences in survival benefit to override the MELD-based prioritization, which as described above functions largely as a survival benefit system in practice in the absence of MELD exceptions. However, I believe that a modified survival benefitbased system can and should replace the current system of "standardized HCC-MELD-exceptions" used to prioritize patients with HCC. The fundamental concept is to predict the impact that the presence of a particular HCC has on the survival benefit of a patient and use it to adjust the natural MELD score accordingly (Table 4). This requires determining the impact of a particular HCC on posttransplant survival and wait-list survival. It is critical to point out that although it is difficult to predict accurately the absolute survival of a patient as shown by Schaubel et al., we can predict much more accurately the excess wait-list mortality and excess posttransplant mortality caused by a given HCC. Some hypothetical examples of this are shown in Table 4. The critical concept of this proposal is to determine the excess survival benefit caused by the presence of HCC in a patient, all other factors being equal, and then use it to modify their natural MELD score. For example, in a hypothetical patient with a MELD score of 20, the presence of a 4-cm HCC might be expected to result in an increase in survival benefit of 0.5 years, relative to an otherwise identical patient who does not have HCC. Therefore, the patient's natural MELD score of 20 needs to be adjusted by adding points corresponding to an additional 0.5-year survival benefit. This is determined by finding the MELD score that corresponds to a survival benefit greater by 0.5 years than the survival benefit of a MELD score of 20 in patients without HCC. If, for example, this corresponds to a MELD score of 23, then 3 points would be added.
Models can be developed that predict the impact that the presence of HCC has on transplant-related survival benefit based on the following characteristics: In the future, these models could include even better markers of tumor biology, progression, or microvascular invasion as such markers are discovered and developed. Such a system would have the flexibility of dealing with tumors that exceed or do not yet reach Milan criteria, as well as tumors within Milan criteria, by appropriately calculating the impact of such tumors on survival benefit. Such a system would be far superior to the current system of assigning a fixed HCC-MELD-exception score regardless of natural MELD score and tumor characteristics that critically affect survival benefit (as well as wait-list mortality). Also, this system would eliminate the current practice of the "MELD elevator," whereby the "HCC-MELDexception" score increases arbitrarily as time accrues on the waiting list, even though this does not predict wait-list mortality or dropout. Gaming of this system could potentially occur by intentionally avoiding locoregional treatments, if successful treatments lead to fewer or no additional MELD points, but such gaming can be avoided by requiring locoregional treatments for certain HCCs or awarding additional MELD points for receipt of locoregional treatment, regardless of the response to it.
In fact, models using natural MELD score, tumor burden, and serum AFP level have already been developed to predict wait-list outcomes in patients with HCC and yield a score that is equivalent to the natural MELD score of patients without HCC in terms of the risk of wait-list death or dropout. (26)(27)(28)(29) These models need to be modified to predict survival benefit rather than wait-list mortality, to avoid the trap of prioritizing patients with very high wait-list mortality who also have unacceptably low posttransplant survival (eg, very large tumors, very high AFP). At least 1 scoring system has been proposed that predicts survival benefit in patients with stage II HCC as a function of natural MELD score and serum AFP and equates it to the survival benefit of patients without HCC based on their MELD score. (18) This represents the most promising approach for replacing the current system and should continue to be improved, refined, and validated before it can be used in clinical practice.

Lung Allocation Score: An Example of an Existing Survival Benefit-Based Transplant Prioritization System
Lung transplant allocation in the United States has been based on survival benefit calculations since 2005. The UNOS lung allocation score (LAS) estimates 1year posttransplant life expectancy using 8 patient characteristics and 1-year wait-list life expectancy using 13 patient characteristics and calculates a normalized difference between the 2, ie, a normalized survival benefit with a score ranging from 0 to 100. (30) In fact, the variables that are used in the LAS calculations are too numerous and complicated to determine the MELD score and HCC tumor characteristics (tumor burden, serum AFP, response to locoregional therapies), which are the variables that we think could be used to develop a robust survival benefit model in patients with HCC. On the other hand, the 1-year posttransplant horizon used in LAS would be too short for liver transplantation because excess mortality related to HCC recurrence would more likely occur beyond the first posttransplant year.

Conclusions
Demonstration of high posttransplant survival (ie, high utility) alone should not be a sufficient criterion for eligibility for liver transplantation, although this is the argument most frequently made for "expanding" access to transplantation to a new group. Taken to extremes, this strategy would offer transplantation to patients with little need for it simply because they have good posttransplant outcomes. At the same time, high waitlist mortality or dropout rate, over and above what is predicted by the MELD score, should not be used alone as an argument for awarding excess priority over and above what is afforded by the patient's natural MELD score. Taken to extremes this strategy might expedite the transplantation of patients with very poor posttransplantation survival. Instead, posttransplantation life expectancy and wait-list life expectancy can be combined into a single metric, the transplant-related survival benefit, which when used to guide our decisions about eligibility and prioritization can maximize the survival of all patients in need of liver transplantation. In practice, the most promising application of survival benefit-based prioritization will be in improving our current system of prioritizing patients with HCC. This would be best achieved by determining the impact of a given HCC on survival benefit and using that to adjust the natural MELD score accordingly.