Survival benefit of liver transplantation: One size fits all or fits none?

Authors

  • Sumeet K. Asrani M.D.,

    1. Division of Gastroenterology and Hepatology, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author
  • W. Ray Kim M.D.,

    Corresponding author
    1. Division of Gastroenterology and Hepatology, Mayo Clinic College of Medicine, Rochester, MN
    2. William J. von Liebig Transplant Center, Mayo Clinic College of Medicine, Rochester, MN
    • 200 First Street SW, Rochester, MN 55905
    Search for more papers by this author
    • fax: 507–538–3974.

  • Julie K. Heimbach M.D.

    1. William J. von Liebig Transplant Center, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author

  • See Article on Page 400

  • Potential conflict of interest: Nothing to report.

In the past few years, an organ allocation process based on survival benefit has been advocated by the Scientific Registry of Transplant Recipients (SRTR).1 For liver transplantation (LTx), this method of organ allocation proposes to replace the current system that places an emphasis on equity (performing transplantation on patients with the greatest medical urgency) with a system based on utility (performing transplantation on patients expected to derive the greatest benefit from LTx). The latter system holds promise to ensure the most efficient use of a scarce resource (donated organ for transplantation), providing society with the most “bang for the buck”.

Abbreviations

ALD, alcoholic liver disease; HCV, hepatitis C virus; LTx, liver transplantation; MELD, Model for End-Stage Liver Disease; SRTR, Scientific Registry of Transplant Recipients.

In simplistic terms, a survival benefit represents the balance between waiting list mortality and post-LTx mortality. To implement such a system, however, complex statistical modeling is required, including a model to predict waitlist mortality on the basis of various candidate characteristics and another model to predict survival following transplantation according to an array of recipient and donor variables. With these models, the difference between survival predicted with and without LTx can be calculated as the predicted benefit of transplantation.

In this issue of HEPATOLOGY, Lucey et al.2 considered the effect of etiology of liver disease, namely alcoholic liver disease (ALD) and hepatitis C virus (HCV) infection, on the survival benefit associated with liver transplantation. They examined 38,899 patients placed on the transplant waiting list for deceased donor LTx between September 2001 and December 2006 using data from the SRTR. Waiting list survival was calculated as the time from initial registration to either transplantation, death, loss to follow-up, meeting criteria for exclusion, or end of follow-up and posttransplant survival as the time from transplantation to either death or retransplantation (median posttransplant follow-up = 21 months). The estimated transplant survival benefit was the ratio of the hazard of death after transplant relative to death on the waiting list.

The results of the analysis showed that both waiting list mortality (hazard ratio = 1.19) and posttransplant mortality (hazard ratio = 1.26) was increased in patients with HCV as compared to those without HCV. However, ALD had no impact on death for patients on the waiting list or who had undergone LTx. Having both HCV and ALD increased the risk of waitlist mortality even more than having either diagnosis separately. Finally, survival benefit of LTx for patients with HCV compared to those without HCV was decreased at Model for End-Stage Liver Disease (MELD) scores of 9–29 (15%–33% increased hazard of death) and then increased for MELD scores > 30 with a 41% survival benefit. A diagnosis of ALD did not change the survival benefit compared to patients without ALD. The authors conclude that HCV status has a negative impact on survival after transplantation over a large range of MELD scores.

Despite the complexity of the statistics involved in the analysis, the article clearly shows that the etiology and severity of underlying liver disease affect waitlist mortality, post-LTx mortality, and survival benefit of LTx. Transplant candidates with ALD, who represent a highly selected subgroup among patients with ALD, are expected to behave like those with “bland” cirrhosis on the waitlist, whereas recidivism to alcohol may threaten long-term outcome in some patients. Many patients with cirrhosis who have HCV infection, on the other hand, continue to have active liver disease, leading to progressive deterioration to hepatic insufficiency.3 The phenomenon of recurrent hepatitis C following LTx remains one of the biggest challenges faced by liver transplant physicians today, because it decreases patient and graft survival. These disease-specific considerations modify the transplant benefit in patients with different diagnoses.

These results raise several questions whether the proposed system based on the benefit calculation can be implemented equitably and practically. First of all, allocating organs based on the projected survival benefit represents a true paradigm shift. By and large, the current principle in organ allocation, namely transplanting the sickest first, has served well because it makes the allocation process transparent and because the benefit of LTx is mostly dominated by the waitlist mortality. Following the initial learning curve about the MELD score, the liver transplant community has embraced the scoring system because of its simplicity and objectivity. MELD has become a part of standard language in hepatology such that most transplant physicians and surgeons can draw a mental picture of a patient with a given MELD score. It is easy for patients and their families to understand that organs will be given patients who need LTx the most.

By design, there is an exponential increase in waiting list mortality with increasing MELD scores. In contrast, pre-LTx MELD score has much smaller influence on mortality after LTx. For example, as compared to mortality risk with a MELD score of 6-11, the mortality risk at a MELD score of 40 is 300-fold higher on the waiting list and 1.5-fold higher after transplantation.4 Thus, allocating organs to the patient with the highest MELD score would reduce the waitlist mortality, while affecting the transplant outcome modestly at best.

In contrast, according to the transplant benefit model, data presented in this article suggest that compared to patients with ALD who have comparable MELD score, patients with HCV should be given a lower priority when their MELD is intermediate (scores of 9–29), whereas patients with HCV who have higher MELD should be given an even higher priority than candidates at the same MELD score with another diagnosis. If a benefit-based transplant policy is to be adopted, the transplant community must be willing to accept this departure from the traditional thinking: because some patients with hepatitis C will experience poor outcome, they will be placed at lower priority than patients without HCV who are faced with the same (or even lower) risk of death while waiting.

In such a context, the heterogeneity of outcome even within the same diagnostic categories is worth careful attention. This is especially true with HCV; there is a wide range in the outcome of LTx recipients with HCV. Whereas the majority of LTx recipients with HCV do well, some patients have extremely poor outcome, including those who develop fibrosing cholestatic hepatitis C. More importantly, it is not possible to predict post-LTx outcome in an individual patient based on pre-LTx variables. This raises a question whether it is fair to assign a lower priority to all patients with HCV because some with the same diagnosis have severe recurrence, driving the average post-LTx survival lower for the group. Schaubel et al. recently described the current state of the benefit model proposal on behalf of SRTR.1 The posttransplant model (in patients with all diagnosis) has a concordance statistic of 0.63, implying that 4 of 10 times, the model would incorrectly predict the first patient to die out of a randomly drawn pair of patients. We anticipate the concordance would be even lower if patients with HCV were analyzed separately.

These and other analytical concerns about the difficulty of implementing the benefit model as an allocation standard for LTx have been discussed previously.5 In addition, in the day-to-day practice of liver transplantation, there is a substantial degree of selection that goes into accepting an organ for a given recipient. One potential explanation for the variance in LTx benefit at different MELD scores among patients with HCV may be a conscious donor selection by the transplant surgeon that is embedded in the LTx data. A higher survival benefit may be derived by this active matching that takes into account a myriad of factors, including patient motivation and surgeon's prior experiences, which are unquantifiable by any degree of statistical modeling. Alternatively, as recent studies suggest, if more marginal (“extended criteria”) donors were given in patients with HCV and lower MELD, a part of the decrement in LTx benefit in patients with HCV with lower MELD may be attributable to donor factors.6 If one is going to take the purist point of view of maximizing the benefit of individual organs, all of the factors that influence that criterion must be taken into account. These include “sensitive” factors such as age, race, and socioeconomic and insurance status of the recipient, as well as characteristics of the individual transplant center. Although some of these are considered “off-limits”, their impact on transplant benefit may be as large as the difference between patients with HCV and patients with ALD. For example, the impact of recipient age would vary based on the time of post-LTx benefit under consideration. On the one hand, if 1-year survival benefit is measured, older recipients would be favored given their relatively higher mortality on the waiting list as compared to younger recipients. On the other hand, if one considers a survival benefit measured over 5–10 years, we would preferentially end up performing transplantation on younger recipients. Finally, a large proportion of patients on the waiting list may simply trade in their risk of early post-LTx mortality for an opportunity to improve their functional status and quality of life after transplant. The impact of such factors on the derivation of a benefit model remains to be seen.

The work by Lucey et al. is an important step in the continued debate on (1) which variables matter in predicting survival benefit of LTx and (2) whether an allocation system based on such a prediction can be equitably implemented. The ultimate question of whether the benefit model should be adopted will depend on the magnitude of the benefit: how many lives will eventually be saved by directing some of the organs from one group of patients, namely those at the highest risk of waitlist mortality, to another, i.e., those who are expected to have better post-LTx outcome. It is essential that the benefits of the “benefit model” are large enough to be embraced by practicing physicians and surgeons, policy makers, and more importantly, by the patients and their families.

Ancillary