In 2008, the United Network for Organ Sharing issued a request for information regarding a proposed revision to kidney allocation policy. This plan described combining dialysis time, donor characteristics and the estimated life years from transplant (LYFT) each candidate would gain in an allocation score that would rank waiting candidates. Though there were some advantages of this plan, the inclusion of LYFT raised many questions. Foremost, there was no clear agreement that LYFT should be the main criterion by which patients should be ranked. Moreover, to rank waiting candidates with this metric, long-term survival models were required in which there was no incorporation of patient preference or discounting for long survival times and for which the predictive accuracy did not achieve accepted standards. The American Society of Transplant Surgeons was pleased to participate in the evaluation of the proposal. Ultimately, the membership did not favor this proposal, because we felt that it was too complicated and that the projected slight increase in overall utility was not justified by the compromise in individual justice that was required. We offer alternative policy options to address some of the unmet needs and issues that were brought to light during this interesting process.
Five years ago, the Organ Procurement and Transplantation Network (OPTN) charged its Kidney Transplant Committee to perform a complete review of kidney allocation policy. This was carried out by a subcommittee titled the Kidney Allocation Review Subcommittee (KARS). The KARS process identified several areas where kidney allocation might be improved (Table 1). The Committee chose to focus primarily on addressing the mismatch for some kidney transplants where the expected posttransplant patient survival is less than the expected graft survival time (i.e. the current system allocates some organs with expected long-term survival to recipients with a relatively short life expectancy). The Committee noted that, with the aging of the population and the increased burden of hypertension and diabetes, the number of patients with end-stage renal disease (ESRD) has ballooned to more than 300 000 and is increasing at a rate of 12% per year. The change in the ESRD demographic has resulted in a shift in the age distribution of patients registered on kidney transplant list with increasing numbers of older patients enlisted and little change in the rate of addition of younger patients (Figure 1). Under the current policy, this has resulted in a shift of kidneys to the more numerous older candidates. In addition, the current system does not proscribe the use of kidneys from better donors for older patients where the expected survival time of the kidney may not be as long as if it was used for a younger recipient.
Table 1. Areas identified by the KARS where kidney allocation could be improved
(1) The general inefficiency of placing kidneys from ECD, leading to a high discard rate.
(2) Lack of predictability of kidney allocation, making maintenance of the list difficult.
(3) Great variability in access to transplantation by blood group and geographic location.
(4) Inefficiency of current methods of identifying and allocating kidneys to sensitized candidates.
(5) Mismatch between expected graft and patient survival.
(6) Variation in access to the donor pool by geographic location.
In an effort to alleviate some of this mismatch between donor grafts and older candidates, the KARS initially determined that the central goal for kidney allocation should be to maximize the LYFT, calculated as the difference between the estimated number years of life gained with a transplant compared with staying on dialysis without a transplant. The Committee recognized that kidneys with longer projected function time would have to be directed to younger patients with longer projected survival time to maximize LYFT. After further deliberation, the KARS refined their initial proposal and the OPTN issued a request for information (RFI) (1) outlining the concepts that were being considered for inclusion in a revised kidney allocation policy in August of 2008. The RFI described a revised plan that included a new Kidney Allocation Score (KAS), which the Committee offered as a compromise between the purely utilitarian goal of maximizing LYFT and maintaining some degree individual justice. Thus, the original central goal was replaced with a much less clearly defined stated goal of ‘providing equitable access for kidney transplant candidates to deceased donor kidneys for transplantation while improving the outcomes of recipients of such kidneys’.
The proposed KAS score was to be calculated by combining LYFT with a donor profile index (DPI) that defined the risk of graft failure based on donor characteristics and also the amount of time a candidate had spent on dialysis (dialysis time or DT). The DPI included the following donor variables: age, gender, race, height, weight, creatinine, history of smoking, donation after cardiac death, current extended criteria donor (ECD) definition, hepatitis C virus, history of hypertension, history of diabetes and cause of death (i.e. anoxia, stroke, central nervous system tumor, other) that would all be included to calculate a continuous measure of donor quality. Including the DPI in kidney allocation would replace the established dichotomous definition of ECD and more realistically represent the continuum of donor quality (2). Inclusion of DT was felt to represent a more specific patient-based variable and was proposed so that patients referred well after initiation of dialysis would not be disadvantaged.
The proposal combined DPI, DT and LYFT along with the degree of sensitization to calculate a final KAS for each candidate at the time of an organ offer. As mentioned above, DT and LYFT were given different weight in allocation for different donor kidneys in an effort to maintain some degree of justice. As the DPI increased for a given donor, more weight was given to DT and less is given to LYFT in calculating the overall KAS. As the DPI decreased, the increased weight given to LYFT would provide kidneys from the more favorable donors (e.g. those with lower DPI) to patients with higher LYFT scores (e.g. primarily the younger patients). Time on the list and donor characteristics are already part of current kidney allocation policy, however, the proposed policy markedly decreased the emphasis on time on the list (especially for the younger candidates) and broadened the use of donor criteria.
The proposal to include LYFT in ranking patients on the waiting list represented a major shift in allocation policy. Current kidney allocation policy does not place much emphasis one prioritizing candidates or grafts to improve posttransplant survival, except in circumstances such as allocation for pediatric patients, 0 mismatch kidneys and providing points for DR matching. In contrast, LYFT was entirely based on estimating survival probabilities for kidney transplant candidates. Consequently, since long-term survival is dependent on age, the major factor influencing the LYFT score was candidate age. As was the intent, using LYFT in an allocation policy would have shifted priority for lower risk kidney grafts preferentially to younger patients as these patients would have had the highest LYFT scores.
The OPTN asked for public comment on these concepts for proposed changes to kidney allocation policy in December of 2008 and subsequently held a public forum in January 26, 2009 in St. Louis. The written input and discussion at the public forum pointed up numerous concerns with the proposal. On January 27, the Kidney Transplant Committee voted to table further consideration of a LYFT-based allocation plan and to evaluate other options. Nevertheless, this process has illuminated some important principles that should be considered in any new kidney allocation policy and has highlighted some new areas where further research is required.
Concerns With a LYFT-Based System
There are several reasons that LYFT may not be a viable basis for kidney allocation. The debate around the LYFT proposal highlighted the fact that underlying demand for kidney transplantation, and therefore the foundation for kidney allocation, is driven by different motivations than exist for liver or lung transplants where allocation policy is heavily weighted toward mortality risk. In contrast, candidates for kidney transplantation do not have immediate short-term mortality risks because of the availability of dialysis. Their motivation, as has been documented in recent patient preference studies (3,4), for seeking a kidney transplant is much more driven by the more immediate gains in quality of life that a kidney transplant can offer. Therefore, allocation policy based on LYFT, where long-term survival estimates are required to accumulate enough LYFT to distinguish between high and low priority patients on the list, does not have much relevance and does not offer adequate consideration for the more looming quality of life decisions patients with ESRD face.
Another confounding aspect to the LYFT score is that dialysis takes a toll on patients’ life years. The toll is higher in older patients rather than younger patients and in diabetics as compared to nondiabetics. It has been shown that kidney transplantation decreases the mortality of patients compared with remaining on dialysis (5). In the shorter term, this effect is relatively greater in the older patient because older patients have more comorbidities that increase their risk of dying overall compared with younger patients who have relatively better survival times on dialysis even though they have excellent long-term success with kidney transplantation also. Given the above, the KARS committee found that very long time horizons had to be employed so that younger patients would accumulate enough LYFT to overcome the short-term gains in LYFT experienced by older patients. This phenomenon is illustrated by the differences in overall years gained when different time horizons were used to calculate LYFT (Figure 2). When a shorter time frame is used, older patients actually get more LYFT while a progressive lengthening of the time horizon allows the younger patients to accumulate more LYFT.
These age-based differences further highlight the difficulty with any LYFT allocation system where the age of candidates necessarily must carry significant weight. Older patients will always have less years ‘at risk’ to gain benefit compared with younger patients and the longer the time frame for calculation, the more the older patients will be disadvantaged. Generally in utility-based health care research, a discounting method or truncation of follow-up is used (6) to address these concerns. This was not done with the LYFT proposal. Age is not a unique biological criterion with which to select patients; it is just a fact of life that older living things will always have less life to live than younger organisms. Age is not a measure of health status, but is just a covariate in the assessment of health status and, as such, does not a priori mean it should be included in allocation policy. In fact, age was intentionally left out of liver allocation policy, even though it is clearly associated with liver waiting list and transplant outcome, to avoid the very value judgments that many found problematic with LYFT-based kidney allocation policy. Since death from severe liver disease is a short-term phenomena, it was possible to base liver allocation on patient-specific variables that do not include age. But because dialysis removes short-term mortality risks for waiting kidney candidates, survival benefit can only be used to segregate patients if longer time frames are used. Using longer timeframes, by definition, means older patients will get less priority since they will always have less life time left. Thus, long-term survival-based allocation policy will only be acceptable if it is agreed that it is acceptable to limit access based on age. It may be reasonable to use age among other variables to measure the results of allocation policy as a measure of equity, but doing so does not necessarily justify using age for the allocation policy itself.
Another major concern with calculating LYFT is that the statistical models used for estimating pre- and posttransplant survival did not meet the usual standards for mathematical model predictive accuracy as measured by concordance (the area under the receiver operating curve) (7). While there are potentially many reasons for this, including the fact that the LYFT and DPI calculations have not been published or peer reviewed, the discussion around this proposal highlighted the deficiencies of the OPTN data. The current OPTN kidney candidate and recipient data collection was not designed for constructing the types of survival models required and all of these data are potentially biased since they are observer-reported. It has become increasingly clear that these data do not discriminate well among various important covariates such as cardiovascular disease, severity of diabetes and some would argue, race. Furthermore, the above concerns with LYFT-based allocation would suggest that pure survival-driven calculations, even if they could be accurately estimated, may not be appropriate for kidney allocation policy since they do not account for patient preference (They are very useful for informed consent and patient education purposes, however!). But if there is ever a consensus that survival-based kidney allocation policy is acceptable, a clear understanding has evolved that precedents for subjecting models to peer review and outside validation that were established with development of liver allocation policy should be followed with future model development. Any survival model used for organ allocation should meet accepted standards for accuracy to ensure that such a policy maintains fairness and credibility.
Another drawback of the KAS was that the type of donor, as defined by the DPI, would determine which patients would get the highest priority regardless of any patient preference. In order to construct a priority list using the proposed KAS, the DPI of the donor at hand would determine the proportions that DT and LYFT would contribute to the final KAS thereby determining the rank of each candidate for that organ offer. This means that, while the candidate being offered a kidney would always have the option for refusing that kidney, the system would determine the quality of kidney being offered to individual patients. If the candidate refused a given offer, it would be unlikely that a better offer would ever come along because LYFT decreases as the candidate would be older at the time of the next offer. Many saw this as a reduction in patient autonomy relative to the current system. Moreover, since there would be no way to predict what type of donor would be offered, there would be limited ability to predict which waiting candidates were most likely to be offered the next organ. Because of this, transplant centers would have much more difficulty managing their lists and would not be able to offer any realistic estimate of waiting time for their patients. This would not improve the predictability of the system.
The development and inclusion of DPI in kidney allocation represents an improvement in concept, however, at the time this proposal was issued, the DPI had not been peer reviewed and published nor had it been validated. As mentioned above, validation of a model's accuracy is critical to its applicability in clinical decision-making and even more important if such a model is to be included in an allocation policy. Moreover, the KAS developers included observer reported donor data in the DPI calculation (Donor Race, History of Smoking, Diabetes and Hypertension), whereas the same observer defined candidate variables were deemed too unreliable for inclusion in LYFT models. This contradiction in defining data validity in the OPTN database remains unresolved.
Additional concern was raised regarding DT. On the one hand, in the current system, patients referred for transplantation months or years after starting dialysis may be disadvantaged since they are eligible for listing when their calculated creatinine clearance falls below 20 mL/min (8) but they do not receive the lost time. Using time on dialysis would solve this problem. However, there were many cases where patients initiating dialysis treatment were not suitable for transplantation at the time of dialysis initiation for many reasons and, only over time, became acceptable. Allowing these patients increased priority even though they were not deemed candidates for transplant at when dialysis started may not be fair. Moreover, giving very high priority for the best kidney grafts to young candidates may reduce the incentive to pursue living donation as has been seen with pediatric candidates since initiating higher priority policy for these candidates (9).
Although not the major focus of our discussion, there were concerns related to the other parts of the proposal put forward by the OPTN. Equating candidates for simultaneous pancreas-kidney (SPK) transplant to candidates waiting for life saving organs has not been widely accepted. Allowing SPK candidates to have a relatively higher priority for receiving a kidney allograft with pancreas allograft requires analyses aimed at defining more rigorous minimal listing criteria for SPK transplants to eliminate the potential ‘gaming’ of the system. Another unresolved issue was incorporating calculated panel reactive antibody systems for prioritizing sensitized patients because this change had not been tested widely. Doing this, in combination with all of the other proposed changes represents significant alterations in the current system and opens many avenues for unintended consequences to emerge. And finally, there was no elaboration of a transition plan to move the current system to this new proposed allocation policy. This transition would be critical since it would result in an immediate and significant redistribution of priority among waiting candidates creating confusion and potentially evoke a perception of unfair treatment.
Ultimately, the LYFT-based KAS proposal was also too complicated and too inaccurate to be implemented for kidney allocation policy. Nonetheless, the process did yield some productive and useful information with which future policy proposals can be developed. There numerous potential changes that could be modeled with the SRTR techniques to assess their impact.
Many observers noted that the OPTN proposal only addressed one of the identified deficiencies of the current allocation system outlined in Table 1. During the discussions, a general agreement has been reached that some method for better allocating kidneys with long-projected function time to recipients with long-projected lifespan is a reasonable goal. One straightforward option that might achieve most of the purported advantages of KAS could be to require that deceased donor kidneys from donors less than 35 years old be preferentially offered to candidates within a more defined age range, for example, candidates less than 35 years old. This suggestion is attractive because the number of donors in this age group approximates the number of candidates less than 35 years of age added to the waitlist. Some will argue, correctly, that this is an arbitrary solution that could disadvantage certain groups (e.g. 36-year-olds). However, this simple modification is consistent with the existing, also arbitrary, donor age of 35 policy that mandates that kidneys from these donors be first allocated to pediatric candidates who are equally arbitrarily defined as less than 18 years of age at the time of listing (10). Such a modification requires no discounting of longer term life gained and does not require that there be justification for giving a 40-year-old more priority than a 45-year-old. There is no need for developing accurate survival models to implement such a proposal and it would still allow for the possibility for a range of donor quality to be offered to every candidate. Geographic differences, another problem with the current system that has not be addressed with the current proposals (Table 1), might possibly be alleviated somewhat by allowing regional sharing for kidneys from donors less than 35 that are allocated to candidates within the defined age range. Importantly, this would be a very simple, easy-to-understand, system.
A potential concern with this approach is that it might reduce younger candidates’ incentive for pursuing living donor transplantation since they could chose to wait for an ideal deceased donor transplant as has been observed with the pediatric renal transplant waiting list since the current pediatric policy was activated (9). To address this, modifications could be made so that some period of waiting was required for the younger candidate before they achieve the priority for the less than 35-year-old donor kidney analogous to the pediatric waiting time policy (10) in effect now. This would provide some incentive to pursue living donor transplantation.
Another, not necessarily mutually exclusive, change could be to require that all candidates indicate what range of donor risk (DPI) he or she is willing to accept. Some might suggest that this would be very similar to the LYFT-based KAS proposal, but there is a fundamental difference. By allowing patients to designate the DPI they are willing to accept before a donor organ is offered, the system becomes a patient-based system, in contrast to the LYFT-based proposal, where the candidate's age mostly determines the type of kidney he or she will be offered, regardless of the patient's preference. Some critics have voiced concern that under such a system, candidates might all designate broad ranges of donor risk that they are willing to accept. But this is precisely why kidney allocation is different. As the patient preference data cited above indicates, many candidates are more concerned about their quality of life and being given the opportunity for any kidney with reasonable risk characteristics rather than getting the one with the best match to their own projected lifetimes. Moreover, such a system could be dynamic whereby candidates are allowed to change their donor risk preferences over time as their own personal quality of life and survival benefit calculations may change. If they are well informed in making such a decision (another advantage of this proposal is that it would, out of necessity, require that candidates are well informed about the spectrum of donor quality that could be offered to them), and the younger candidates are willing to accept a higher risk graft, there is no reason that the system should prevent this. To address concerns that some candidates with very limited survival times might choose to designate broad ranges of donor types and thereby open the possibility for receiving a long functioning kidney, we could just require that candidates greater than a certain age cannot designate accepting kidneys with the lowest donor risk ranges. This type of patient driven system may not achieve an overall increase in LYFT, but it will preserve patient autonomy and individual justice, and would require much less reorganization and reeducation.
Where Do We Go From Here?
Change always engenders fear. Whatever changes in kidney allocation policy are put forward, open and frequent communication, presentation and publication in peer-reviewed venues and careful planning for transition can go a long way to allay these fears. In addition, smaller stepwise implementation of changes may provide time for observation and stabilization of the system without a complete disruption of patients and their hopes for a kidney transplant. We, the ASTS, are pleased to have had the opportunity to present our thoughts on the proposed allocation system. An open, transparent process, allowing all the stakeholders the opportunity to discuss and refine these thoughts, has allowed the voices of the transplant community to be heard. The American Society of Transplant Surgeons’ members interacts with kidney transplant candidates face-to-face everyday and are direct messengers to patients for explaining allocation policy for every type of deceased donor transplant. We are pleased to be active, ongoing participants in the future development of these policies.
The KARS process has highlighted many important issues for kidney allocation policy as we move forward. It is clear that a policy based purely on survival calculations is not likely to be acceptable to most of the adult candidates since they have short-term quality of life concerns that often carry equal importance to the long-term quantity of life gains transplantation can offer. Moreover, accurately estimating long-term survival for patients with ESRD and after kidney transplant has proven to be more difficult than originally thought. Nonetheless, there is wide agreement that while less than accurate LYFT calculation can be useful in counseling patients, a higher degree of accuracy is required if these predictive models are to be incorporated into policy. Patient preference and health utility research may provide variables other than the age-based long-term survival parameters to better serve the waiting kidney transplant candidate.