AN EXAMINATION OF CONTEMPORARY CHALLENGES IN DECEASED DONOR KIDNEY ALLOCATION

Transcription

1 AN EXAMINATION OF CONTEMPORARY CHALLENGES IN DECEASED DONOR KIDNEY ALLOCATION by Caren Lee Rose MSc, Dalhousie University, 2003 BSc, University of British Columbia, 2000 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Health Care and Epidemiology) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) October 2014 Caren Lee Rose, 2014

2

3 Abstract Transplantation is the preferred treatment for patients with kidney failure, but the need for transplantation exceeds the organ supply. Strategies to address the organ shortage include: preventing end-organ failure, increasing the number of deceased donor kidneys (DDK) for transplantation, and ensuring appropriate allocation to avoid organ waste. This thesis: develops an improved metric for deceased donation activity; describes an agematching allocation strategy to reduce organ waste; and examines the impact of selected wait-list policies on disparities in access to DDK transplantation. Using national administrative databases, we: Estimated and validated the number of potential deceased organ donors among in-hospital deaths using diagnostic codes; Calculated differences in DDK and transplant candidate survival by age using Cox regression to determine the area between survival curves (ABSC), and combined these measures with information on DDK and candidate ages to define age cut-points for DDK allocation; Described the use and outcomes of older DDKs ( 65 years) in countries with different allocation systems using Cox regression and ABSC, and identified patients that achieved a lifetime of transplant function from older DDKs; and Examined longitudinal use and outcomes of wait-listing candidates at multiple transplant centres using logistic and Cox regression. Three percent of Canadians who die in-hospital were identified as potential organ donors, suggesting significant potential to increase deceased donation. We determined DDK and ii

4 candidate age cut-points for Canadian allocation, and estimated that implementation of these cut-points could have eliminated 500 years of wasted donor kidney function, and prevented 800 years of post-transplant dialysis compared to the current allocation strategy. We found that older DDKs provided a lifetime of kidney function for patients aged >60 years, suggesting targeted use of these organs could safely increase transplantation. Finally, we determined that multiple wait-listing helped minimize geographic disparities in accessing transplantation and may be an important policy consideration in countries that do not currently allow multiple listing. As transplantation wait-lists grow at unprecedented rates, the potential to increase deceased donation, implement allocation policies to decrease organ wastage, safely expand the use of older deceased donors and promulgate wait-list policies to increase access to transplantation will become more important. iii

5 Preface This statement is to certify that the work in this thesis was conceived, conducted and written by Caren Rose. Ethics approval was granted from the University of British Columbia Research Ethics Board (UBC REB # H A002). Caren Rose was entirely responsible for Chapters 1,3,4, and 9. A version of Chapter 2 has been submitted for publication and is awaiting revisions. The manuscript is entitled: Estimating the number of potential deceased donors in Canada ; Caren Rose, Peter Nickerson and John Gill. Caren Rose was responsible for the study design, data access, data analysis, writing and revising the manuscript. Dr. Gill assisted in the study design and revisions of the manuscript. Dr Nickerson approved the study design, assisted in data access and edited the manuscript. The findings of Chapter 2 were presented at the Canadian Society of Nephrology annual general meeting in Vancouver April The findings of Chapters 5 and 6 have been written up and presented to the Canadian Kidney Working Group as policy recommendations for the allocation of deceased donor kidneys by age matching. A version of Chapter 7 has been submitted for publication. The manuscript is entitled: Lifetime of allograft function- a new metric to inform the use of expanded criteria donor iv

6 transplantation ; Caren Rose, Elke Schaeffer, Jagbir Gill, Ulrich Frei and John Gill. Caren Rose was responsible for the study design, data analysis, writing and revising the manuscript. Dr John Gill assisted in the study design, writing and revisions of the manuscript. Dr Schaeffer and Dr Frei provided access to Eurotransplant data and edited the manuscript. Dr Jagbir Gill edited the manuscript. A version of Chapter 8 was presented at the World Transplant Congress in San Francisco in July The project is entitled: Multiple wait-listing ; Caren Rose, John Gill. Caren Rose was responsible for the study design, data analysis, writing and revising the manuscript. Dr Gill participated in designing the study, and revising the manuscript. Dr Joel Singer and Dr Jacek Kopec provided epidemiologic input and editing of all chapters. v

7 Table of contents Abstract... ii Preface... iv Table of contents... vi List of tables... ix List of figures... xi Abbreviations... xiii Acknowledgements... xiv Dedication... xv 1 Introduction The epidemiology of end-stage renal disease Renal replacement therapies Supply of organs Allocation of kidneys for transplantation Ethical considerations for allocation: utility and justice Outstanding questions and study justification Challenges in allocation Recipient and deceased donor kidney life expectancies The problem with age matching- introduction of inequities in access to transplantation Attitudes about deceased donor kidney allocation Regional disparities in access to transplantation Summary Research hypotheses and study questions Outline of thesis Estimating the number of potential deceased organ donors Introduction Methods Identification of potential donors Accuracy of the study method estimates Conversion of potential donors to actual donors Results Accuracy of the study method estimates Conversion of potential donors to actual donors Interpretation The ethics of deceased donor kidney allocation Principles Justice Utility Examining utility through different lenses...43 vi

8 3.3 Transplant candidates Likelihood of accessing transplantation Life expectancy after transplantation Quality of life Summary The allocation of deceased donor kidneys for transplantation Overview Eurotransplant Senior Program Changing a deceased donor kidney allocation system- An American narrative Allocation prior to Need for change Kidney allocation score Proposed national allocation policy American allocation summary Summary The type and quantity of inefficiency in donor-candidate age matching: measuring the area between the curves Introduction Methods Study population and data source Analytical methods Results Summative utility Interpretation Defining equitable-based utility cut-points for donor-candidate age matching Introduction Deceased donor kidney and candidate ages Considerations for allocation by donor and candidate age matching The Canadian perspective Definitions Nominal definitions of young and old Operational definitions of young and old Quantifying ages of donors and candidates for age matching Results Donor-candidate age matching cut-points Recommendations and hypothetical redistribution of deceased donor kidneys to recipients Interpretation vii

9 7 Lifetime of allograft function a new metric to inform the optimal use of expanded criteria donor kidney transplantation Introduction Methods Comparison of the use and outcomes of ECD kidney transplantation in the Eurotransplant Senior Program and the United States Analysis of recipient outcomes after ECD kidney transplantation in the United States Results Comparison of the use and outcomes of ECD kidney transplantation in the Eurotransplant Senior Program and the United States Analysis of recipient outcomes after ECD kidney transplantation in the United States Interpretation Disclosure Multiple wait-listing Introduction Methods Data source and study population Statistical analyses Results Factors associated with multiple wait-listing Association of multiple wait-listing with access to deceased donor transplantation Discussion Conclusion Disclosure Conclusion Summary Key findings and contributions Estimating the number of potential deceased donors Defining equitable utility-based cut-points for age matching Implications of not age-matching Multiple wait-listing Strengths and limitations Study implications and future work Conclusion References Appendices viii

10 List of tables Table 2.1. Regional metrics for death and deceased organ donation...36 Table 2.2. The identification of potential donors by age group using the study method...37 Table 2.3. Number of potential donors identified by chart audit and study method in Winnipeg area hospitals compared to actual number of donors...37 Table 2.4. Number of potential donors identified by the chart audit and study method in Winnipeg area hospitals by age at death Table 2.5. The conversion of potential donors to actual donors by age group...38 Table 5.1 Recipient and donor characteristics...86 Table 5.2. Study sample size by recipient and donor kidney age categories...87 Table 5.3. Area under and between recipient survival and graft survival curves...88 Table 5.4. The summative utility of two allocation systems...88 Table 6.1. Area under and between recipient survival and graft survival curves Table 6.2. The designation of transplant candidates as young or old relative to the age of the deceased donor kidney Table 7.1. Characteristics of patients who received kidneys from donor aged 65 years Table 7.2. Characteristics of patients aged 65 years who received kidneys from donor aged 65 years Table 7.3. Relative hazard of transplant failure in ESP compared to U.S. transplant recipients Table 7.4. Unadjusted mean difference in patient survival versus graft survival in elderly ESP versus U.S. recipients after five years of transplantation Table 7.5. Patient survival ten years after transplantation with a donor kidney <65 years stratified by kidney donor profile index and after transplantation from a donor kidney 65 years ix

11 Table 8.1. Characteristics of multiply versus singly listed wait-listed candidates (N=310,475) Table 8.2. The multivariate odds ratio for a candidate being multiply (two or more centres) versus singly wait-listed for deceased donor transplantation Table 8.3. Among candidates who are multiply wait-listed for deceased donor kidney transplantation, the multivariate odds ratio of being listed singly at more than two centres versus two centres Table 8.4. Median time to deceased donor kidney transplantation among multiply versus singly wait-listed candidates. Table shows median (q1, q3) years x

12 List of figures Figure 2.1. A flow chart describing the study method used to estimate the number of potential deceased organ donors...35 Figure 2.2. Potential donors, actual donors and the age-standardized donor conversion Ratio by province of transplantation...35 Figure 3.1. A description of the intersection of life expectancy, access to transplantation, and quality life from the transplant candidates perspective...51 Figure 5.1 Panel A shows possible differences in predicted patient and deceased donor kidney survival between donor kidneys and their recipients under current allocation rules. Panel B shows possible differences in predicted patient and deceased donor kidney survival for the same aged donor kidneys and recipients using allocation by life expectancy matching Figure 5.2 A description of the inefficiencies that occur when the same recipient is transplanted with deceased donor kidney grafts with differential expected survival.the dark grey bars represent the two types of inefficiency that occur when the recipient and the donor kidney have unequal survival...82 Figure 5.3 Left panel: Area between the recipient and graft survival curves represents the time a patient would be required to return to dialysis and await repeat transplantation. Right panel: area between the recipient and graft survival curves represents the loss of potential kidney function Figure 5.4. Histograms showing the distribution of recipient (top panel) and deceased donor kidney (bottom panel) ages...84 Figure 5.5. Left panel: Recipient survival by recipient age categories. Right panel: Graft survival by donor age categories Figure 6.1 The distribution of Canadian recipient age and deceased donor kidney age by year of transplantation. Recipient mean age (std): 44 (14.7) in 1995, 53 (15.1) in 2010; Donor kidney mean age (std): 36 (16.9) in 1995, 43 (17.0) in P-value for trends in recipient age and deceased donor kidney age over time p< xi

13 Figure 6.2 The distribution of deceased donor kidneys to recipients by age. Top panel shows how deceased donor kidneys from were allocated in practice to recipients. Bottom panel shows the redistribution of the same aged deceased donor kidneys to the same aged recipients following the recommendations for donor-candidate age matching in this chapter Figure 7.1. Kaplan-Meier plots comparing transplant outcomes among the N = 1520 ESP and n=446 U.S. recipients who were 65 years of age at the time of kidney transplantation from a deceased donor 65 years. (P< for each comparison) Figure 7.2. Distribution of age of recipients of kidney transplantation from ECD kidneys from Figure 7.3. Ten year mean patient and death censored graft survival and difference between curves Figure 8.1. The odds ratio of multiple listing by year of candidate wait-listing compared to candidates wait-listed between 1995 and 1998 (error bars represent 95% confidence intervals). The model included the following covariates: candidate demographics (i.e. age at wait-listing, sex, race, cause of end-stage renal disease, body mass index), biologic factors (i.e. peak panel reactive antibody titre, ABO blood group), socioeconomic factors (i.e. level of education, employment status, health insurance provider), and quintile of median waiting time for deceased donor transplantation in the candidate s primary centre of wait-listing Figure 8.2. The proportion of candidates multiply listed by organ procurement organization of wait-listing Figure 8.3. The relative hazard of transplantation over time for multiply versus singly wait-listed candidates (reference=1.00 singly wait-listed) in different years of transplantation. The model was adjusted for: candidate demographics (i.e. age at wait-listing, sex, race, cause of endstage renal disease, body mass index), biologic factors (i.e. peak panel reactive antibody titre, ABO blood group), socioeconomic factors (i.e. level of education, employment status, health insurance provider), and quintile of median waiting time for deceased donor transplantation in the candidate s primary centre of wait-listing Figure 8.4. The proportion of patients who were transplanted at five years after wait-listing date xii

14 Abbreviations ABSC BMI CCDT CCI CI CKD CORR CVA DAD DCD DDK DRPM DT ECD ESP ESRD GED HLA HR HRSA ICD-9-CM ICD-10- CA KAS KDPI KDRI KPD LYFT OPO OPTN OPTNKC OR PRA Q1 Q3 SCD SRTR STD USRDS Area between survival curves Body mass index Canadian Council for Donation and Transplantation Canadian classification of health intervention Confidence interval Chronic kidney disease Canadian Organ Replacement Register Cerebrovascular disease Discharge Abstract Database Donation after cardio-circulatory death Deceased donor kidney Donor rate per million population Dialysis time Expanded criteria donor Eurotransplant Senior s Program End-stage renal disease General educational development Human leukocyte antigen Hazard ratio Health Resources and Services Administration International Classification of Diseases, Ninth Revision, Clinical Modification International Classification of Diseases, 10 th Revision, Canadian enhanced version Kidney allocation score Kidney donor profile index Kidney donor risk index Kidney paired exchange Life years from transplant Organ procurement organization Organ procurement and transplantation network Organ procurement and transplantation network kidney committee Odds ratio Panel reactive antibody First quartile Third quartile Standard criteria donor Scientific Registry of Transplant Recipients Standard deviation United States Renal Data System xiii

15 Acknowledgements The research presented in this thesis would not have been possible without the limitless support of my thesis committee (Drs John Gill, Joel Singer and Jacek Kopec), who provided their time and expertise to share feedback on the work presented in this thesis and ultimately helped improve the research. In particular, I will be forever grateful for the mentorship I have received over the past decade from Dr Gill. Dr Gill has taught me how to think critically, challenged me to solve problems by thinking outside every box, and helped refine my writing skills. I cannot thank Dr Gill enough for his enduring support throughout the PhD process and in the development of my career. I would also like to thank my colleagues and funding programs for giving me opportunities to discuss my work, and vet my solutions and ideas. I have learned an immense amount through conversation with my peers and mentors in my scholarship programs. Most importantly, I would like to thank my family for their love and patience during the writing of this dissertation. They provided me with constant encouragement and time, both essential to the completion of this degree. Thank you for always believing in me, and nurturing the belief in myself that I was capable of anything I could imagine. You bring love and happiness into my life, and your humour keeps me grounded. xiv

16 A Dedication To my Family xv

17 1 Introduction 1.1 The epidemiology of end-stage renal disease Chronic kidney disease (CKD) is the progressive and irreversible loss of kidney function over time. The leading causes of CKD are diabetes, cardiovascular disease, hypertension and obesity. CKD is the 10 th leading cause of death in Canada and the 12 th leading cause of death worldwide, 1, 2 contributing to 3,803 and 927,592 deaths in 2009 in Canada and the world respectively. 1, 3 The burden of this disease may be underestimated as there are a large number of disability adjusted life years lost with CKD, and many patients with CKD die of other causes (e.g. cardiovascular disease). Patients who survive to develop end stage renal disease (ESRD) will die if they do not receive renal replacement therapy. Estimating the prevalence of ESRD worldwide is difficult because many developing countries do not have the means to provide renal replacement therapy. Among countries that do provide renal replacement therapy, many collect data on patients with ESRD in national registries. These national registries provide data on the incidence and prevalence of ESRD, as well as the characteristics of the population and outcomes data. Unfortunately, the data across registries are collected differentially and are of variable quality. Nonetheless, these registries provide the best available data on ESRD, and allow for international comparisons 4. Canadian data is collected in the Canadian Organ Replacement Register (CORR). CORR is a national information system for renal dialysis and solid organ transplantation housed 1

18 at the Canadian Institute for Health Information. The register collects, processes, analyzes and reports the level of activity and outcomes of solid organ transplantation and renal dialysis activities. 5 CORR obtains patient- level data through voluntary data submission from hospital dialysis programs, regional transplant programs, organ procurement organizations and independent providers of dialysis. Further details on CORR, including coverage, reliability and validity, can be found in Appendix A. In Canada in 2009, there were incident cases of ESRD, an increase of 137% since In the same period, the prevalence of ESRD also increased more than 200% from to As the leading causes of CKD continue to increase, the impact and repercussions of ESRD on the Canadian health care system will be significant Renal replacement therapies The primary functions of the kidney are: to maintain the body s internal equilibrium of water and minerals, to excrete end-products of metabolism, and to function as part of the endocrine system. 7 For patients with ESRD, whose kidneys no longer provide these functions, there are two types of renal replacement therapy: dialysis and kidney transplantation. Dialysis uses a machine to mimic the major roles of the kidney: diffusion (eliminating waste) and ultrafiltration (eliminating fluids). Dialysis is an imperfect replacement to natural kidney function because it cannot replicate the endocrine functions of the kidney. 2

19 Kidney transplantation is the engraftment of a human kidney from a donor to a recipient with the goal of restoring kidney function. Transplant recipients require continuous immunosuppressive drug therapy to prevent graft (i.e. transplanted kidney) rejection. Compared to dialysis, transplantation is life prolonging, quality of life enhancing, costsaving and can be life saving for patients who are unable to dialyze. 8, 9 Therefore, despite the requirement for on-going immunosuppression, transplantation is the preferred form of renal replacement therapy for most patients with ESRD without a contraindication to transplantation, such as infection, cancer or a short life expectancy due to advanced age or multi-system disease. Both deceased and living people can donate kidneys for transplantation. Deceased donor kidneys come from patients who die in hospital with a diagnosis of brain death or cardiocirculatory death, who have no absolute contraindications to donation (e.g. metastatic cancer or seropositivity for human immunodeficiency virus). 10 In Canada, these patients or their families must provide consent for donation. Approximately 2.5% of Canadian patients who die in hospital are eligible for donation (roughly per year) (see Chapter 2). Unfortunately, primarily due to difficulty with identification, only 15% of these potential donors become actual donors. Living donor kidney transplantation provides timely access to transplantation; and living donor kidney transplant outcomes are superior to those achieved with transplantation from a deceased donor. Living kidney donors are most often related to their recipient through a blood or emotional bond. Potential living donors must be ABO blood group 3

20 and Human Leukocyte Antigen (HLA) compatible with their recipient, and must undergo a thorough medical and psychological evaluation before being accepted for living organ donation. In British Columbia approximately 50% of all kidney transplants are from a living donors (19% of living donors are spouses, 59% blood related, and 22% emotionally related (e.g. friend, in-law)) Supply of organs The need for transplantation exceeds the availability of transplantable organs. In Canada, there are more than people waiting for kidney transplantation, 12 but only 1000 of these patients receive transplantation yearly (34% living donor, 66% deceased donor) 6 ; The waiting list provides a limited view of the need for transplantation, as many patients are never wait-listed, and there are more than patients whose life is sustained with dialysis in Canada. 12 Further, the fact that only 2% of patients die on the waiting list underestimates the magnitude of the organ shortage problem as many other patients are removed from the waiting list prior to death for declining health concerns. 12 The median waiting time for kidney transplantation varies three-fold across Canadian provinces. 6 This variation occurs because deceased donor kidneys are allocated within seven regions that follow provincial boundaries (i.e. they are not shared nationally). Ongoing strategies to address the shortage of kidneys for transplantation include: 1) decreasing the incidence of ESRD, 2) increasing the number of deceased and living donor organs for kidney transplantation, and 3) decreasing the need for repeat kidney 4

21 transplantation by improving kidney transplant survival. Improvements in immunosuppression and general medical and surgical management have greatly increased kidney survival, and patient survival in the early period posttransplant. However, despite these successes little further gain has been made in increasing patient life with kidney function in the long term. Unfortunately, diabetes, hypertension and obesity are leading causes of CKD/ESRD, and despite the best efforts of the medical community the prevalence of kidney disease continues to increase worldwide Expanding deceased kidney donation Because of the organ shortage, medical professionals endeavour to use every viable deceased donor kidney for transplantation. As such, the criteria for accepting deceased donors has been expanded to include organs from older aged donors or donors with medical conditions, such as hypertension, that would not have routinely been used for transplantation in the past. These donors are termed expanded criteria donors (ECDs). Although ECD kidney transplant outcomes are inferior to outcomes from non-ecd donors, these organs are safe to transplant and provide a survival advantage relative to treatment with dialysis for elderly transplant candidates, patients residing in regions with extremely long waiting times, and patients who tolerate dialysis poorly. 16 Deceased organ donors usually have sustained severe brain injuries and meet strict 5

22 medical criteria for neurological brain death. Donation after circulatory death (DCD) describes the retrieval of organs for transplantation using circulatory, not neurologic, criteria. The use of DCD differs across provinces. Kidneys from DCD donors currently comprise more than 20% of all deceased donors in Ontario and Quebec, but are not used in other provinces. 17 Despite consensus internationally and in Canada, 18 some physicians maintain ethical objections to DCD and believe more research is necessary. Widespread efforts to increase deceased donation and the limited introduction of DCD in Canada, have produced a modest 25% increase in the number of Canadian deceased donors over the last decade, however, this increase remains inadequate to meet the growing need for kidney transplantation Measuring deceased donation Efforts to increase deceased donation are critically dependent on timely, accurate and insightful information about deceased donor activity and the success rate of converting potential donors to actual donors. The currently reported metric for deceased donation activity is the donor rate per million living population (DRPM). Wide variation exists in the DRPM internationally with Canada performing in the bottom half of all reported countries (DRPM=15). 3 In addition, there is a three-fold variation in DRPM across Canadian provinces. 3 Although the DRPM may be used as a comparative metric in the same population over time, it is limited for comparisons between populations. 19 Importantly, it does not account for baseline differences in population characteristics among patients who die and are eligible for donation. 20 A measure of deceased donation 6

23 that more accurately reflects the population of potential donors is needed. Such a metric could be used to compare donation activity across regions, and inform strategies to increase donation. A new method to estimate deceased donation activity will be introduced in Chapter Expanding living kidney donation Several strategies have been used to increase living donation. Living donation was expanded from blood-related donors to include emotionally related, but biologically unrelated donors. Over the last decade the use of unrelated donors in Canada has increased by 64%. 3 In addition, older age living donors are now more frequently transplanted (i.e. the number of Canadian living donors aged 55 years has tripled since 2003). Another expansion of living kidney donation is the inclusion of non-directed anonymous donors. These donors have no relationship with a specific transplant recipient, and do not decide which patient will receive their kidney. Kidney donations from non-directed anonymous donors were first accepted in Vancouver in 2003, and are now accepted in most transplant programs in Canada. Kidneys from non-directed anonymous donors were originally allocated to the first matched person on the deceased donor waiting list, but are now most commonly allocated as part of another program intended to increase living kidney donation: a Canadian national living kidney donor paired exchange program (KPD), launched by Canadian Blood Services as a pilot project in More than 20% of individuals that come forward to donate their kidney to a blood or 7

24 emotionally related person in need of transplantation are unable to do so because of a biologic incompatibility. 11 KPD facilitates donation from these willing but incompatible donor-recipient pairs by matching these pairs with other similarly incompatible pairs. 22 Entry into the program is voluntary, and a computer algorithm is run with all incompatible pairs every three months that determines feasible chains of pairs for donation. Although the use of non-directed anonymous donors and KPD have helped to promote living donation, it is unclear whether the impact of these programs has led to an increase in the number of living kidney donors or simply changed the type of living donor kidney transplantations. The number of living donors in Canada increased from 435 to 539 from , 6 and 240 kidney transplantations were facilitated by KPD between Although the new initiatives to expand both living and deceased kidney donation have been marginally successful in increasing the absolute number of transplanted kidneys, their successes have been overwhelmed by the increasing number of patients with renal disease. As a result, addressing the widening gap between supply and demand for organs requires alternative and additional strategies. 1.2 Allocation of kidneys for transplantation Living donors are selected based on their psychological and medical suitability, as well as their biologic compatibility (i.e. blood group and tissue compatibility) with their intended recipient. Aside from the less common non-directed anonymous donors, a living donor has an emotional relationship (either blood related or unrelated) with their recipient. 8

25 Therefore, transplantation from living donors is directed, i.e. the recipient is specified a priori. In contrast, there is no relationship between deceased donors and patients on deceased organ transplant waiting lists and decisions must be made about which transplant candidates will receive the limited number of available deceased donor organs. In Canada, deceased donor kidneys are allocated to biologically compatible transplant candidates, almost exclusively within geographic regions. Allocation within these groups then follows national consensus recommendations, assigning overriding priority for kidney transplantation to patients with the greatest medical need. 24 This comprises patients who are unable to receive dialysis treatment, pediatric patients whose physical and psychological development can be severely stunted on dialysis, and highly sensitized patients (patients who are estimated to be tissue incompatible with a high percentage of deceased organ donors). Within prioritized groups, the need for kidney transplantation is assumed to be equal and ties between transplant candidates are decided by regional point systems that include various patient factors. Allocation rules are updated intermittently to address the changing needs of the transplant candidate population Ethical considerations for allocation: utility and justice Two key ethical principles guide the ranking systems for allocation of deceased donor organs to transplant candidates: utility and justice 25. In the ethics literature on resource allocation, utility commonly refers to making optimal use of the resources so that the greatest total benefit is obtained and justice to the equal treatment of people so as not to advantage or disadvantage any group. 26 9

26 For non-renal organ failure (e.g. heart, lung, liver), transplantation is the only long-term treatment and the sickest patients are prioritized. Utility is the precedent focus in allocation of these organs, and is measured by the prevention of imminent death. In contrast, ESRD patients can be treated chronically with dialysis and can live years waiting for kidney transplantation. In this setting, allocation rules seek to determine an acceptable balance between utility and justice. In kidney transplantation, utility most commonly refers to allocating a kidney to the patient in whom it will survive the longest and justice to ensuring that each patient who could benefit from transplantation would have equal opportunity to receive one. 14 The balance of these two allocation principles has changed historically, and is driven by the successes of kidney transplantation and the availability of deceased donor kidneys. The definitions of utility and justice will be explored more thoroughly in Chapter 3. In the 1970s and early 80s, clinical transplantation was in its infancy and graft and patient survival outcomes following deceased donor kidney transplantation were mediocre. Use of transplantation as a therapy was therefore reserved for young, otherwise healthy patients who were expected to have superior survival 27 At this time, few patients were waiting for transplantation and deceased donor kidneys were allocated to obtain optimal outcomes. For example, graft rejection was a significant issue and allocation was driven by biologic considerations (i.e. matching transplant candidates to the most immunologically compatible donors-defined by compatibility of Human Leukocyte Antigens (HLA)) to decrease the risk of rejection. As a result of medical and surgical advancements in the mid-80s and 90s, graft rejection became less common and short- 10

27 term graft and patient survival increased. Transplantation has become the gold standard treatment for ESRD and is no longer reserved for young and healthy dialysis patients. There is no absolute age restriction for transplantation, and many of the sickest patients (e.g. diabetics) derive the greatest relative survival benefit from kidney transplantation, because their dialysis survival is limited. 28 The liberalization of transplant eligibility combined with an increased prevalence of ESRD has led to an incessant growth in the demand for kidney transplantation that outstrips the supply of transplantable organs. In consequence, many countries have all but eliminated HLA-matching in their allocation algorithms in favour of emphasizing time on the kidney transplant waiting list- a shift from utility to justice based allocation. There is no or limited attempt to optimize postkidney transplantation outcomes by matching the anticipated longevity of the deceased donor kidney with that of the potential recipient. 1.3 Outstanding questions and study justification Challenges in allocation The current allocation system, which prioritizes fairness in access to kidney transplantation, allows the possibility of discordance in recipient survival and survival of the deceased donor kidney. For example, transplantation of a 20-year old deceased donor kidney to a patient aged > 60 years would result in loss of potential kidney transplant function because the anticipated life expectancy of the transplanted organ is greater than that of the transplant recipient. 29 Loss of potential kidney function occurs when a patient dies with a transplant that could have continued to function had the patient not died. Transplanted kidneys cannot be re-transplanted after the death of the recipient. 11

28 Alternatively, if a 60-year old deceased donor kidney is transplanted into a 20-year old recipient, the recipient s life expectancy will exceed that of the deceased donor kidney and the recipient will require another transplant. Repeat transplantation is a significant burden on the Canadian organ supply, accounting for 11% of kidney transplantation 6. In addition, failed transplant recipients may develop immunological barriers to transplantation that lessen their likelihood of repeat transplantation, leading to prolonged requirement for dialysis after transplant failure. Therefore, matching the anticipated life expectancies of transplant recipients and deceased donor kidneys has the potential to avoid organ wastage and minimize the need for renal replacement therapy after deceased donor kidney failure Recipient and deceased donor kidney life expectancies Accurately predicting post-kidney transplant recipient and deceased donor kidney life expectancies prior to transplantation is imperfect, in large part because life expectancies are dependent on post-transplant factors, which are unpredictable. Recipient and donor ages are the most important predictors of post-kidney transplant recipient and deceased donor kidney survival 15, 30 and have the advantage of being objectively measurable and transparent factors for both patients and medical personnel. The inclusion of additional variables in models estimating recipient and deceased donor kidney survival, has been proposed but rejected. The basis for rejection of this approach is that it is too complex to be understood by patients, subjective, and only marginally superior to models including donor and recipient ages alone. 31, 32 For these reasons, donor and recipient ages will be 12

29 used as the primary determinants of recipient and deceased donor kidney life expectancies in this thesis. Although the independent effects of deceased donor kidney and recipient ages on posttransplant outcomes are well described, it is not clear if deceased donor kidney age modifies the relationship between recipient age and post-transplant outcomes. 33, 34 For example, although older age for both donors and recipients is associated with worse deceased donor kidney and recipient survival, it is not clear if the effect of increasing deceased donor kidney age on deceased donor kidney survival is as marked in older recipients with shorter expected survival, as it would be in younger recipients with longer 33, 34 expected survival The problem with age matching- introduction of inequities in access to transplantation In this thesis, age matching refers to the preferential allocation of deceased donor kidneys of a given age to ABO blood group and human leukocyte antigen (HLA) compatible transplant candidates in given age categories. For example, an age matching policy might allocate deceased donor kidneys aged 40 years to transplant candidates aged 55 years and deceased donor kidneys aged < 40 years to transplant candidates < 55 years. Within organ allocation algorithms age matching would be superseded by established criteria that prioritize certain wait-listed candidates for transplantation such as pediatric candidates and highly sensitized candidates. 13

30 The age distribution of deceased donors is a moving target and somewhat unpredictable, but is generally increasing over time, while the age distribution of transplant candidates has been increasing more rapidly over time. An age allocation system that is based on matching deceased donor kidney and recipient ages therefore has the potential to introduce inequities in access to transplantation between candidates of different age groups. The most recently proposed deceased donor kidney allocation rules in the United States emphasize utility by including components that match estimated donor kidney and transplant candidate life expectancies (detail provided in Chapter 4). Although the results from preliminary simulations showed improvement in lifespan benefit per kidney transplantation compared to current allocation, the results also showed that there was a reduction in the proportion of older recipients receiving kidney transplantation (10% and 35, 36 20% reduction in the number of recipients aged and 65 respectively). Even if age matching could increase the utility of deceased donor kidney allocation, the magnitude of the change in utility and its effect on equity in Canada is unknown. A sample of data from all deceased donor kidney transplantation in the United States in 2006, showed that recipients aged 18 to 35 received fewer than 5% of kidneys from deceased donors aged 55 years, and only 1% from deceased donors aged 60 years 37. In contrast, recipients aged 65 years received more than 40% of their kidneys from deceased donors aged < 35 years (this represented 24% of all deceased donor kidneys aged < 35 years). Similar statistics showing the donor-recipient pairings at age extremes are not currently available for Canada. The potential increase in utility arising from an 14

31 age matching allocation strategy needs to be estimated, and considered in the context of its potential impact on access to transplantation among patients in different age groups, including the number of patients transplanted and waiting times for transplantation Attitudes about deceased donor kidney allocation Attitudes among stakeholders vary on how to appropriately balance utility and justice in the allocation of deceased donor kidneys. Specifically in the United States, this has made it difficult to implement change to current allocation policies. Some believe utility to be of major importance, advocating for a greater emphasis on donation to younger and healthier patients, while others are steadfast in their belief that equal access to this scarce health resource is a necessity. A survey in the United States in 1996 reported that the public believe small improvements in post-transplant outcomes are not justified at the expense of longer waiting times in some patient subgroups. 38 In another small study (N=33) in the United States, transplant candidates had varying views about whether longer waiting times were justified in order to ensure a better tissue matched kidney for transplantation. 39 However, this study may have been biased by including a cohort of patients who were waiting for repeat kidney transplantation (i.e. they may have been familiar with the agony of graft failure, and the difficulty in accessing another donor kidney), and therefore, the study s external validity is uncertain. A larger survey of dialysis patients and transplant recipients (N=232) in the United Kingdom, found that participants favoured equitable allocation (waiting time) 2:1, compared to efficient allocation (tissue matching), but 66% favoured allocation by longer waiting times when 15

32 this factor was the only difference between candidates. 40 The same survey asked patients to decide who should receive a deceased donor kidney transplant between two similar candidates, one aged 20 years and one aged 60 years. Fifty percent of respondents thought age should not be a determinant in allocation, while 35% allocated in favour of the younger candidate and only 7% to the older candidate (8% did not know). Interestingly, study patients aged > 70 years were the most likely to allocate the deceased donor kidney to the younger patient (56%), suggesting that older patients may not be averse to some inequity in access to transplantation by age. However, only one of nine questions in this study was related to kidney allocation by age, and the study did not explore the extra waiting time or increased probability of death that an older candidate may experience in this scenario. To date, no studies have been published in Canada examining the attitudes of transplant candidates about deceased donor kidney allocation. The only record of Canadian patient perceptions is from focus groups in Manitoba. The focus group study results suggested that candidates wanted allocation policies to be transparent and free of unjust discrimination. 41 No comment was made about utility and justice. Although rules for allocating deceased donor kidneys to transplant candidates are predetermined, decisions about which ESRD patients to wait-list for transplantation, and which kidneys to accept for transplantation are at the discretion of the transplant nephrologist or surgeon. In addition, policy makers rely on expert opinion of these physicians to inform changes in allocation policy. Therefore, transplant physicians are major stakeholders in the allocation of deceased donor kidneys. A recent qualitative 16

33 survey of Australian nephrologists (N=25) ascertained their opinions about wait-listing and deceased donor kidney allocation. 42 Overall, the nephrologists wanted to limit discrimination among sub-populations access to transplantation as well as increase the survival benefit gained from each kidney. Age matching was mentioned as a strategy to increase utility, suggesting preference in allocation for the young; although criticism arose at the potential disadvantage of transplanting older kidneys into older candidates (i.e. reduced patient survival). In contrast, matching older deceased donor kidneys to older candidates was also suggested as a strategy to make better use of older kidneys and reduce discard of these organs. Most importantly, the nephrologists did not want to be responsible for decision making regarding the appropriate balance of utility and equity in allocation, but believed that this was a necessary role for policy makers. ESRD patients and their physicians have varying views about weighting the conflicting principles of utility and justice. Although age matching is a recurring theme in surveys about the allocation of deceased donor kidneys, there is no comprehensive data describing the considerations of health practitioners and patients regarding specific age matching policies, nor the degree of equity loss (reduced access to, or longer waiting times for, deceased donor kidney transplantation) that patients and physicians would be willing to accept for a given improvement in post-transplant survival outcomes Regional disparities in access to transplantation With the exception of highly sensitized candidates, deceased donor kidneys are shared regionally, but not nationally in Canada. Therefore, the likelihood of accessing deceased 17

34 donor kidney transplantation depends on regional factors (e.g. wait-listing practices, the number of candidates on regional waiting lists, the number of living and deceased donors available for transplantation) as well as biologic factors (e.g. blood type, and genetic tissue profile of available donors and candidates). Despite similar allocation rules in practice across Canada, there is four-fold variation in the likelihood of accessing deceased donor transplantation across Canadian regions, 43 and the median time to deceased donor kidney transplantation varies three-fold (median time in Nova Scotia 2.1 years; median time in British Columbia 5.9 years). 3 Given the small population of some transplant regions in Canada, it is not known if the regional incorporation of new allocation policies (e.g. donor-candidate age matching) would reduce or exaggerate geographic disparities in access to transplantation. An example of a policy in the United States that may reduce geographic inequity in access to deceased donor kidney transplantation is the practice of multiple wait-listing (i.e. the ability to appear on a deceased donor waiting list in more than one region). American candidates and health care practitioners have used the practice of multiple wait-listing, as a mechanism to improve individual patient access to transplantation. 44 For example, the median waiting time for deceased donor transplantation was significantly longer in the candidates first centre of wait-listing suggesting that patients are exploiting the policy of multiple wait-listing to achieve quicker access to transplantation and reduce geographic disparities in access to transplantation. 45 Multiple wait-listing was associated with an 88% increase in access to deceased donor transplantation in the United States (data up to June 2000). 45 It is not known if the use and impact of multiple wait-listing on access to transplantation has increased in the current era where there has been rapid growth of the 18

35 deceased donor kidney waiting lists in the United States. Understanding the current impact of this policy in the U.S. would help inform the need for such a policy in Canada Summary The number of kidneys available for transplantation is not sufficient to transplant all patients who would benefit. The donor rate per million population is an imperfect measure of deceased donation and more accurate measures of the number of potential donors who die in Canadian hospitals are needed to more closely reflect our maximum possible supply of deceased donors. Accurate measurement of donation activity will help to inform strategies to increase deceased donation. In the absence of an adequate donor supply, decisions must be made about how best to use (i.e. allocate) the organs that are available. There is consensus in the transplant community that avoiding large discrepancies between expected organ and patient survival is desirable. Although, individual programs and clinicians may actively try to restrict this practice, mismatches in life expectancy still occur. Age matching, in some format, is consistently proposed as a possible solution. There is a discord in opinion on how to balance utility and justice in deceased donor kidney allocation, providing challenges for policy makers. Gill 46 proposes that there is a need to revisit the strategies by which increased utility will be incorporated into allocation policies, while maintaining justice. The solution that may be the most currently acceptable is a strategy that would not change the population characteristics of future recipients from current recipients, but 19

36 would instead allocate deceased donor kidneys differentially among the same pool of transplant candidates. 1.4 Research hypotheses and study questions The overarching goals of this thesis are: 1) to develop a new method to estimate deceased donation activity, 2) to examine the potential benefits and harms of allocating deceased donor kidneys by age matching, and inform the safe use of older donor kidneys for transplantation, and 3) to describe the impacts of wait-listing policies on regional disparities. The primary determinant of patient survival after kidney transplantation is recipient age, and the primary determinant of deceased donor kidney survival after kidney transplantation is the age of the donor kidney. To date, age has not been widely accepted as a criterion for the allocation of deceased donor kidneys in North America. As a result, older recipients are likely to die with a functioning deceased donor kidney, leading to varying amounts of unrecognized kidney function. Similarly, young recipients are likely to outlive their donor kidneys and return to dialysis for varying times or require another transplant. A strategy that allocates deceased donor kidneys by age matching would minimize both kidney function waste and time back on dialysis after kidney transplant failure. The specific research questions addressed in this thesis are: 1) To develop and validate a new method to estimate the number of deceased donor kidneys available for transplantation in Canada, including stratification by age. 20

37 2) To develop a new metric for the utility of deceased donor kidney transplantation, and use this to define equitable utility-based age cut-points for the allocation of deceased donor kidneys. 3) To compare the utility of allocating older deceased donor kidneys in a system that incorporates age matching versus one that does not; to provide evidence to inform the safe use of older donor kidneys for transplantation, and decrease organ discard; and to determine recipient outcomes after transplantation when older donor kidneys are not allocated using age-matching. 4) To describe longitudinal changes in the use of the multiple wait-listing for deceased donor transplantation, and assess whether the impact of multiple waitlisting on access to deceased donor transplantation has changed over time. 1.5 Outline of thesis The research chapters in this thesis are intended for publication elsewhere. As such, they have been formatted to resemble stand-alone papers, with reduced repetition across chapters (except as needed for clarity), and with transition pieces. This chapter serves as an overview to native kidney disease and kidney transplantation. In addition, outstanding questions in the field of deceased donor kidney allocation were presented, and the thesis objectives and an outline of the thesis are presented. Chapter 2 focuses on the measurement of deceased donation activity. Specifically, this chapter develops a new method to estimate the number of potential deceased donors in Canada, and assesses the accuracy of this method in one Canadian region. In addition, the 21

38 number of deceased donors estimated using the study method are compared to the number of actual deceased donors in Canada to calculate the conversion rate of potential donors to actual donors. Chapters 3-7 focus on deceased donor kidney allocation. Chapter 3 defines the concepts of utility and justice as they are commonly used in the transplantation ethics literature, and provides thesis specific definitions of these concepts. In addition, this chapter explores the perspectives of transplant candidates and policy makers in regards to the allocation of deceased donor kidneys. Chapter 4 provides an overview of the policy of deceased donor kidney allocation, including an extensive look at current and proposed allocation rules in the United States. Chapter 5 and 6 use Canadian data to describe a new metric of utility, and the use of this utility measure along with information about deceased donor kidney and recipient ages to define and develop equitable utility based cut-points for the allocation of deceased donor kidneys. In Chapter 7, the allocation of older donor kidneys (i.e. which candidates are receiving transplantation from these kidneys), as well as the recipient survival and donor kidney survival are compared between two allocation systems: one which allocates older donor kidneys by age matching, and one which does not allocate older donor kidneys by age matching. In addition, post-transplant outcomes for recipients of older deceased donor kidneys in the system that does not require age matching are described. In the United States, patients can be wait-listed for deceased donor transplantation at more than one transplant centre. Chapter 8 describes the advantage of being multiply 22

39 wait-listed for deceased donor kidney transplantation, and the factors associated with multiple wait-listing. The implications of this policy of geographic disparities in access to transplantation will be discussed. The final chapter, Chapter 9, summarizes the findings of the thesis, and integrates the discussion. In addition, Chapter 9 reviews the strengths and limitations of the research, synthesizes the implications of the research, and presents suggestions for future work. 23

40 2 Estimating the number of potential deceased organ donors 2.1 Introduction Despite numerous initiatives, there has been no substantial increase in the number of deceased organ donors in Canada during the past decade. 47, 48 To what extent this stagnation is related to: 1) fewer potential deceased donors (individuals who have sustained severe brain injuries without medical contraindications to organ donation) due to aging of the Canadian population and public safety initiatives that have reduced the number of traumatic deaths, or 2) failure to identify and obtain consent for donation from potential deceased donors remains uncertain. The understanding of these issues is impeded by the lack of national data regarding the number of potential donors who die in Canadian hospitals. Information regarding the potential pool of deceased donors in Canada is essential to determine system performance, and devise strategies to increase deceased donation. The most commonly reported metric of deceased donation, the donor rate per million inhabitants living in a region (DRPM), 48 does not account for regional or secular differences in mortality, or for the cause of death among hospitalized patients, 19, 49 and therefore may lead to regions being misclassified as underperformers. 50 There is more than three-fold variation in the DRPM across Canadian provinces (Table 2.1). 51 The gold standard method of obtaining information on potential donors is by prospective audit of all in hospital deaths; however, with more than 150,000 deaths annually in Canadian hospitals, 52 it is extremely challenging to obtain this information on an ongoing basis. 24

41 These audits are performed periodically in different Canadian regions, 53 but are intermittent, costly, and not standardized between provinces. For example, a different estimate of potential donors would be generated if only individuals referred to an organ procurement organization were considered for chart review, versus all individuals who died in hospital. In order for a potential donor to become an actual donor, the potential donor needs to be identified, and then undergo formal medical assessment to confirm eligibility for organ donation. Finally, consent for donation from the potential donor has to be established. Various metrics have been developed to measure health system performance in deceased donation based on the above process. The objectives of this chapter are: 1) to develop a practical and timely method to estimate the number of potential deceased donors using information already collected for patients who die in Canadian hospitals; 2) to estimate the accuracy of this study method to identify potential deceased donors; and 3) to compare the number of potential deceased donors identified by the study method with the number of actual deceased donors to determine the conversion ratio of potential donors to actual donors. 2.2 Methods This is a retrospective analysis of all in-hospital deaths captured in the Discharge Abstract Database (DAD) between April 1, 2005 and March 31 st, The DAD contains demographic, administrative and clinical data for all acute care hospital 25

42 separations, excluding emergency room admissions, and still births for all provinces and territories, excluding Quebec. 54 Diagnostic and procedural information in the DAD is recorded in a standardized format using International Statistical Classification of Diseases, 10 th Revision, Canadian enhanced version (ICD-10-CA) and Canadian Classification of Health Intervention (CCI) codes Identification of potential donors The primary outcome measure, the number of potential deceased organ donors, was determined using a unique study method based on previous work from Australia by Holt et al, 55 that defined a limited number of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes that were prevalent among organ donors. These ICD-9-CM codes are also frequently recorded in the hospital separation records of Canadian organ donors 56. For the purposes of this analysis, the Canadian Institute for Health Information supplied a mapping of ICD-9-CM codes specified by Holt et al to ICD-10-CA codes recorded in the DAD. The list of ICD-10-CA codes used to identify included causes of death are listed in Appendix B. Figure 2.1 provides an overview of the study method used to identify potential donors. ICD codes, equivalent to those published by Holt et al, 55 were used to identify a subgroup of patients who died in hospital with diagnoses compatible with organ donation. We further excluded individuals with absolute and relative contraindications to donation (e.g. metastatic cancer or seropositivity for human immunodeficiency virus), based on a 26

43 list of ICD codes described by the Canadian Standards Association (Appendix C). 10 With the understanding that donation is highly unlikely unless patients have access to critical care services, we further restricted our identification of potential donors to individuals for whom there was a CCI code for mechanical ventilation. As only a small proportion of inhospital deaths in individuals aged >70 years are eligible for donation, the primary analyses were restricted to hospital deaths in patients aged 70 years. 57 The number of potential donors was calculated overall, by age group, by gender and by province Accuracy of the study method estimates To determine the accuracy of our potential donor estimates, we compared the number of potential donors identified in a chart audit of in-hospital deaths in six greater Winnipeg area hospitals during the study time frame with the estimated number of potential donors in the same hospitals identified in the DAD using the study method. These estimates were compared overall and by age group. The identified potential donors from each method were not linkable at the patient level, and thus it is not possible to determine if the cases did or did not represent the same patients in both cohorts Conversion of potential donors to actual donors The donor conversion ratio represents the number of potential donors who become actual donors (actual donors potential donors). We compared our estimates of the number of potential organ donors in the DAD with the number of actual organ donors recorded in 27

44 the Canadian Organ Replacement Register (CORR) in the study period, to calculate the donor conversion ratio. CORR receives information annually for all deceased organ donors in Canada, directly from provincial or regional organ procurement organizations 57. CORR defines an actual deceased donor as an individual from whom 1 organ is transplanted. Actual donors from CORR were not directly linkable to potential donors identified in the DAD. Donor conversion ratios were calculated by age group, by gender and by province. Provincial donor conversion ratios were standardized for age distribution using direct standardization. All analyses were performed using SAS 9.4 (Carey, NC). 2.3 Results There were in-hospital deaths captured in the DAD during the study period. After excluding patients aged > 70 years, patients without diagnostic codes consistent with organ donation, patients with absolute or relative contraindications to organ donation defined by the Canadian Standards Association standards, and patients without an intervention code for mechanical ventilation, we identified potential donors. The potential donors represented 2.5% of all deaths among hospitalized patients during the study period (Figure 2.1). 28

45 Table 2.2 shows the number of in-hospital deaths in different age groups, and summarizes the number of patients remaining after each step in our algorithm identifying potential donors. The proportion of potential donors among all individuals who died in hospital ranged from 5.8 % among patients years to 19.5%, among patients aged years. Potential donors were disproportionately male; among all individuals who died in hospital, 2.9% of males were identified as potential donors, compared to only 2.0% of females Accuracy of the study method estimates The chart review of patients who died in greater Winnipeg area hospitals from , identified 126 potential donors (Table 2.3), among whom 52 individuals were actual organ donors (Table 2.3). Application of the study method identified 266 potential donors from the same hospitals in the same time period, indicating the study method overestimates the number of potential organ donors by approximately two-fold. The overestimate of potential donors using the study method was similar across years (Table 2.3), but was exaggerated with the increasing age of individuals who died in hospital (Table 2.4) Conversion of potential donors to actual donors There were actual deceased organ donors aged 70 identified in CORR during the study period - producing an estimated overall donor conversion ratio of 15% (

46 actual potential donors). The donor conversion ratio varied by age, ranging from 5% in patients aged years to 43% among patients < 18 years of age (Table 2.5). Females had a higher donor conversion ratio (19%) compared to males (14%). Information on actual donors was not available for the intersections of age group and gender so these ratios were not age-standardized. Figure two shows that the provincial age-standardized donor conversion ratios ranged from a low of 11% in British Columbia, to a high of 21% in Saskatchewan. 2.4 Interpretation This study provides a method to estimate the number of potential deceased organ donors and the conversion of potential donors to actual donors using pre-existing administrative data. Overall, 2.5% of individuals who died in hospital (7.7% of all patients aged 70 years) were identified as potential organ donors. A comparison of the study method to the gold standard method of identifying potential donors by chart audit showed that the study method overestimated the number of potential donors by a factor of approximately two. However, even after accounting for this overestimation, the results suggest there is significant potential to increase deceased organ donation in Canada. The conversion of potential donors to actual donors was highest in younger age groups, and females; there was a two-fold variation in the age-standardized conversion ratio of potential donors to actual donors between provinces. Compared to the deceased donor rate per million population, the estimate of potential deceased donors and the donor conversion ratios in 30

47 this study provide significantly more information that may prove useful in improving the delivery of deceased donor services in Canada. Our findings extend recent observations showing low organ procurement rates among potential donors in the province of Ontario, 58 but differ from recent findings in a study from Calgary suggesting few critically ill patients with severe brain injuries qualify as potential donors. 59 However, the Calgary study relied on electronic charting of a neurological diagnosis of brain death to identify potential donors, and it was unclear how often patients were in fact evaluated for neurologic brain death, or how complete the capture of brain death diagnoses were in the electronic charts included in the study. 59 Our study method, which does not require a neurologic diagnosis of brain death to identify potential donors, provides a more comprehensive estimate of potential deceased donors in Canadian hospitals, including individuals who would meet criteria for neurological brain death as well as those who may qualify for donation after circulatory death. Donation after circulatory death comprises fifteen percent of deceased organ donors in the United States 60 and forty-three percent in the United Kingdom. 61 Although organ donation after circulatory death varies between provinces, the most recent CORR data indicates that such donors comprised more than twenty percent of all donors in Ontario and Quebec in indicating that methods to estimate potential donors should ideally not be restricted to neurologically brain dead donors. Importantly, the study method of estimating potential donors does not obviate the need for detailed chart audits in Canadian hospitals. The study method is based on ICD codes, 31

48 the recording of which may vary in accuracy and completeness. We propose that chart audits, such as that from Winnipeg in this study, be used to estimate the accuracy of the study method in different regions. In fact, we are currently working with BC Transplant and the Canadian Institute for Health Information to further validate this work. The proposed validation in BC will be able to directly link patients in the provincial chart audit, with patients in the DAD, and allow for determination of sensitivity and specificity of the study method in this region. Determination of the accuracy in different regions could be used to modify the study method to improve the comparability of estimates between different regions, while periodic validation exercises could also be used to longitudinally monitor the accuracy of the study method. Hospital chart audits are also necessary to understand the reasons why potential donors are not converted to actual donors, which may include failure to identify potential donors or failure to obtain consent for donation. Importantly the DAD excludes deaths in outpatient hospital areas as well as in emergency rooms, and thus potential donors in these settings are not captured by the study method. The major strength of the study is that it provides a feasible strategy to estimate the number of potentials donors in all Canadian provinces included in the DAD. In contrast to the donor rate per million population, the study method provides information that is insensitive to regional and secular variations in hospitalized mortality, and provides an estimate of the potential to increase deceased organ donation. Although hospital chart audits are acknowledged as the best method to obtain information regarding potential organ donors, such data are self-reported by hospitals, collected by multiple abstractors, 32

49 and not standardized between regions. These limitations together with the cost and workload required to perform chart audits, make it unlikely that this source of information will be longitudinally available in most provinces on an ongoing basis. Therefore, the study method represents a pragmatic solution to the need for comprehensive and timely information regarding deceased organ donation activity in Canada. We are currently working with the Canadian Institute for Health Information and the CORR Board to annually report the number of potential donors estimated by the study method along with donor rate per million. Importantly, we were not able to link individual actual donors in CORR with potential donors identified in the DAD. Therefore, it is not clear if all actual donors are identified as potential donors using the study method, or if misclassification exists. In addition, we were also not able to link individuals in the Winnipeg chart audit with individuals in the DAD. The ability to do so may provide greater information about which information in the DAD could be used to refine our estimates of potential donors. A future step would be to determine the donor conversion ratios within provinces for different subgroups, such as age and gender, in order to determine whether the provincial differences are consistent across subgroups. This could target education strategies to increase deceased donation. In conclusion, the study identified 2.5% of patients dying in Canadian hospitals as potential organ donors. The number of actual donors was only 15% of the estimated number of potential donors. Although the study method likely overestimates the number 33

50 of potential organ donors by more than two-fold, there is still significant potential to increase deceased organ donation in Canadian hospitals. For example, if the true conversion ratio was closer to 30% (i.e. two times 15%), 70% of potential donors would still not proceed to donation. Estimating the number of potential deceased organ donors from data already collected from hospital separations represents a feasible strategy to obtain information needed to inform improvements to the deceased organ donation system in Canada. 34

51 Figure 2.1. A flow chart describing the study method used to estimate the number of potential deceased organ donors Figure 2.2. Potential donors, actual donors and the age-standardized donor conversion Ratio by province of transplantation Potential Donors Actual Donors Donor Conversion Rate Number of Donors % 5 In-Hospital Deaths 0 0 BC AB SK MB ON NS N=55,610 N=37,385 N=16,918 N=19,666 N=161,384 N=44,227 Province 35

52 Table 2.1. Regional metrics for death and deceased organ donation Donor rate per million Crude death rate per Study method donor population (2012) a thousand population b conversion ratio c Canada e British Columbia Alberta Saskatchewan Manitoba Ontario Quebec N/A Nova Scotia 15.5 d d New Brunswick 9.0 Newfoundland 9.3 Prince Edward Island 8.6 a Data from Canadian Institute of Health Information/ Canadian Organ Replacement Register 2014 b Data from Statistics Canada 2010 c Donor conversion ration obtained using actual donor from Canadian Organ Replacement Register and potential donors identified in the Discharge Abstract Database using the study method d Includes New Brunswick, Newfoundland, Prince Edward Island e Excludes Quebec 36

53 Table 2.2. The identification of potential donors by age group using the study method All in-hospital Deaths (N) Without contraindication to transplantation** Medical diagnoses compatible with organ donation*** Potential organ donors (i.e. mechanically ventilated patients) Age Group (years) < (6.7%)* (19.5%) (10.7 %) (7.5%) (5.8%) * Percentage of in-hospital patient deaths identified as potential donors by the study method. **Appendix B *** Appendix A Table 2.3. Number of potential donors identified by chart audit and study method in Winnipeg area hospitals compared to actual number of donors Fiscal Year Chart Audit of Winnipeg Hospitals Potential Donors Study Method Ratio (Study method: chart audit) Actual Donors Canadian Organ Replacement Register Total

54 Table 2.4. Number of potential donors identified by the chart audit and study method in Winnipeg area hospitals by age at death. Age Chart Audit Study Method Ratio (Study Method: Chart Audit) <18 years years years years years Total 123* *3 individuals missing age data Table 2.5. The conversion of potential donors to actual donors by age group Age Group N < Potential donors Actual Donors Donor Conversion 43% 26% 20% 13% 5% Ratio 38

55 3 The ethics of deceased donor kidney allocation 3.1 Principles The early allocation of medical resources, such as in war times, was based on triage whereby patients with the greatest expected benefit would be treated first 63. Now, in an era of evidence informed practice, there is an eagerness to use objective measures to allocate resources fairly with the greatest benefit. 64 Although competing theories exist, in democratic nations, such as Canada and the United States, the allocation of deceased donor organs has been founded on the principles of autonomy (patient choice), fidelity (keeping commitments), veracity, transparency, utility and justice. 25 These former five principles should underlie or constrain any allocation policy. Thereafter, the latter two principles (utility and justice) can be incorporated, and are the primary factors guiding allocation policies for deceased donor organs. Utility and justice are constantly being pitted against one another in an effort to achieve a balance in allocation priorities that is acceptable to all stakeholders (transplant candidates, donors and their families, and the public). For non-renal organ failure (e.g. heart, lung, liver), deceased donor transplantation is the only (or significant) long-term treatment and the sickest patients are prioritized. In contrast, ESRD patients can be treated chronically with dialysis or living donor transplantation and can wait years for deceased donor kidney transplantation. In this context, the competing and complementary roles of utility and justice in the allocation of deceased donor kidneys is different from other solid organs and the following chapters 39

56 will be restricted to the exploration of these principles as related to deceased donor kidneys Justice In the transplantation ethics literature justice refers to equal treatment of people so as not to advantage or disadvantage any group. 26 Equal treatment here is taken to be equity of access to transplantation, where access is proportional to the medical need of the patient. 65 Some authors believe that organ allocation policies should diverge from equal access ONLY where there are substantial differences in patient benefit, 66 however, substantial has not been quantified. Examples of unequal (i.e. especially high) medical need are: 1) a patient who cannot access another form of treatment, and will therefore die imminently without transplantation, or 2) a pediatric patient who will suffer developmentally without transplantation. Inequities in access to transplantation exist in deceased donor kidney allocation. For example, certain blood types and rarer genetic profiles are common amongst racialspecific candidate groups, but are less common among deceased donors, and therefore some racial groups have inevitable inequity in access to transplantation. In addition, candidates who are sensitized (i.e. will mount an immunologic response to a large proportion of the donor pool), are less likely to find an acceptable donor match and thus have inequitable access to transplantation. Allocation policies strive to achieve justice for these patient groups, including registries for highly sensitized patients, but are not sufficient to overcome these biologic barriers. 40

57 3.1.2 Utility In the transplantation ethics literature utility refers to medical utility. Medical utility has been defined solely with respect to the transplant candidate as maximizing the welfare of patients suffering from end-stage organ failure, 66 and with respect to the deceased donor organ and transplant candidate in tandem as making optimal use of the resources so that the greatest total benefit is obtained. 26 These definitions differentiate medical utility from social utility, which would rank eligible candidates for transplantation based on their potential to benefit society or the greater good. A comparable task was undertaken at the University of Washington in the 1960s when the allocation of scarce hemodialysis machines was determined by an anonymous public representative committee who ranked patients with kidney failure based on social and health determinants. 67 For example, the committee decided that a young, male, businessman was considered to have greater social worth than a housewife; and an aircraft worker with six kids was voted to have greater social worth compared to a professional with two kids. As a result the businessman and aircraft worker were offered the elusive dialysis treatment, while many others were denied. Deciding the potential worth of an individual is controversial, and may lead to greater disutility if included as a factor in allocation. 25 In addition, the consideration of social utility in allocation would undermine any sense of justice. Therefore, in this thesis the definition of utility is restricted to medical utility and defined by: the optimal use of deceased donor kidneys, with maximal recipient survival with a functioning deceased donor kidney allograft. 41

58 Measures of utility The medical utility of deceased donor kidney transplantation can be measured by the following: 1) patient survival after transplantation, 2) deceased donor kidney survival after transplantation, 3) patient death on the kidney transplant waiting list, 4) patient quality of life, and 5) wasted organ function (i.e. loss of potential kidney function that occurs when the transplant candidate dies with a functioning kidney). 68 These outcomes can be measured in isolation or in combination using different metrics. The first three outcomes can be measured by estimating group life expectancies, or death/failure rates. This can be done in a number of ways. For example, one could use median expected survival, mean survival (area under the survival curve) or mortality rate per 100 patient years. 31 The fourth metric of utility, quality of life, is commonly measured using quality adjusted life years. Quality of life is difficult to measure objectively because its significance varies between individuals, and it is also difficult to quantify. Importantly, it may be equally weighted with the extension of patient life as a reason for pursuit of transplantation. A final metric that combines all four outcomes is life years (gained) from transplantation (LYFT)- the additional years of life a patient is expected to live with a transplant compared to remaining on dialysis (see LYFT, described in Chapter The final utility measure of organ waste is more difficult to measure. A method to measure organ waste will be described in Chapter 5. Integrating rules into allocation algorithms with the intent of increasing these measures of utility without consideration of justice may lead to greater inequity in access to transplantation amongst groups with less favourable expected utility outcomes. In 42

59 addition, it is most often the positive components of utility that are examined (e.g. allocating kidneys of higher quality to candidates with longer expected survival after transplantation), thus ignoring the disutility that may occur when less healthy candidates receive transplantation from poorer quality kidneys. For example, lower quality kidneys reduce recipient survival with a functioning graft 69 and also shorten recipient survival compared to transplantation with higher quality kidney Examining utility through different lenses Optimizing the use of deceased donor kidneys while maximizing the benefit and minimizing the harm to the transplant candidate population is the common overarching utility goal of deceased donor kidney allocation. However, the operational definitions of these concepts may conflict for different stakeholders, and should be integrated when formulating decisions about which quality deceased donor kidneys should be allocated to different transplant candidates. This section describes my assessment of the considerations for offering and accepting deceased donor kidneys for various players. 3.3 Transplant candidates Eligible transplant candidates make an informed decision to be added to the deceased donor kidney waiting list. As such we can assume that transplantation is their preferred treatment. However, various considerations apply from the patient perspective when determining what quality of kidney a patient would be willing to accept. Here, quality 43

60 refers to the expected survival of the deceased donor kidney 1. Some transplant candidates may wish to wait longer to receive a deceased donor kidney that will maximize their total expected survival, while others may be willing to trade future life years for immediate quality-improved years. 71 Parfit s claim about compensation 72 argues that one s burdens are not compensated for by benefits provided for someone else, but candidates may be willing to accept a lesser quality kidney that meets their needs, perhaps at the expense of increased waiting time, in order to allow another patient to benefit from a higher quality deceased donor kidney. 39 A patient s choice in accepting or declining an offer of transplantation from a given quality deceased donor kidney can be described by the candidate s: 1) Likelihood of accessing transplantation (surviving the waiting time) 2) Life expectancy after transplantation, and 3) Quality of life (with and without transplantation) Likelihood of accessing transplantation Likelihood of accessing transplantation is a continuous concept composed of three components: 1) the candidate s risk of death on the transplant waiting list, 2) the candidate s probability of finding a compatible deceased donor kidney, and 3) the average waiting time in the candidate s region of wait-listing. An individual who has a high risk of death on the waiting list, or one who has a low probability of finding a donor match is unlikely to access transplantation (poor access), while a healthier individual with 1 The quality of a deceased donor kidney may also refer to the risk of unknown comorbidities such as HIV or hepatitis. Complete knowledge about donor health is not always available for donors that die quickly upon arrival in hospital. Laboratory test results to rule out more serious comorbidities may take time, and 44

61 a high probability of finding a donor match is likely to access transplantation (good access). Access to transplantation is not directly related to time spent on dialysis. For example, a patient with poor access due to a low deceased donor kidney match probability will most likely spend more time on dialysis compared to a patient with poor access due to a high likelihood of death on the waiting list. Therefore, patient utility factors related to dialysis (i.e. quality of life) are independent from access to transplantation, and need to be considered separately. Geography is another factor impacting the likelihood of accessing deceased donor transplantation. The rates of deceased donation vary regionally across Canada, 51 as well as in the United States. 50 Importantly, in the United States transplant candidates can be wait-listed for kidney transplantation at multiple centres simultaneously, whereas in Canada candidates can only be wait-listed in one province at a time. The implications of multiple wait-listing as a strategy to increase likelihood of access transplantation will be explored in the United States in Chapter Life expectancy after transplantation Life expectancy after transplantation is a continuous concept representing the predicted years of survival a given candidate will realize if they are successfully transplanted. This concept is correlated with the requirement for repeat transplantation. For example, a candidate with a long life expectancy is likely to outlive their graft, return to dialysis and require repeat transplantation. Thus, all candidates will need to consider not only their risk of graft failure with deceased donor kidneys of different qualities, but also their 45

62 future likelihood of repeat transplantation. Repeat transplantation is not without surgical risk, and candidates will be more highly sensitized after transplantation, decreasing their likelihood of another match and thus reducing their access to future transplantation. Therefore, it may be in the best interest of a candidate with longer life expectancy to refuse a poorer quality kidney and wait on dialysis for a deceased donor kidney with a longer life expectancy Quality of life Quality of life is defined as an individual s perception of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns. 73 The definition includes consideration of psychological and physical health, independence, and social and spiritual environment. 73 Therefore, quality of life on dialysis relative to living with a transplant varies significantly between individuals. Several factors may lead to an individual s determination of their quality of life; only a couple of possibilities are described here. First, this could depend on the level of independence and freedom an individual felt prior to requiring dialysis. For example an able-bodied, active or jet-setting patient who suddenly requires long, restrictive in-hospital dialysis sessions three times per week may feel more burdened by the change in lifestyle, compared to a less active individual or a lonely individual seeking socialization. Another factor may be the level of comfort a patient feels with the medical aspects of the treatment (e.g. needles, hospitals). Patients will need to consider quality of life on dialysis prior to transplantation, as well as quality 9, 74 of life after transplantation, including time back on dialysis after graft failure. 46

63 It is not possible to predict the importance of each of these concepts in the evaluation of utility for each candidate. Nevertheless, an exploration of the important determinants is discussed below (Figure 3.1). Note, this thought experiment does not consider: 1) medically urgent candidates who are unable to be treated with dialysis and thus require immediate transplantation, 2) pediatric candidates in whom transplantation may impact physiologic or anatomic development, nor, 3) differences in surgical risk for transplantation (assumed to be low for all candidates). Figure 3.1 panel A gives a simplified description of characteristics of transplant candidates that might occur in the intersection created by combining and dichotomizing the continuous concepts of access to transplantation and life expectancy. (Quadrant I: age 25, blood type A; Quadrant II: age 25, blood type B; Quadrant III: age 60, blood type B; Quadrant IV: Age 60, blood type A) Figure 3.1 Panel B shows the intersection of access to transplantation and life expectancy, stratified for two extremes of quality of life on dialysis. Figure 3.1 Panel B shows that for the highest quality of life on dialysis patients in quadrant I would be tempted to wait for a pristine deceased donor kidney because they will likely receive an offer for transplantation quickly, and want to maximize their graft survival because they have a long life-expectancy. In contrast, the candidate in quadrant 3 would be tempted to take the first offered deceased donor kidney because they will likely not receive another offer, either because of their low likelihood of finding a match, or because of their poor life expectancy. Candidates in quadrants II and IV can take a slightly more selective approach, perhaps only reneging offers of the poorest quality 47

64 deceased donor kidneys. As quality of life on dialysis deteriorates (Figure 3.1 Panel B), all candidates may persuade themselves to accept a wider range of quality deceased donor kidneys to offset increased time waiting on dialysis. In this example, the healthiest candidates with good access, may no longer wait for an ideal deceased donor kidney, but might only refuse the lowest quality deceased donor kidneys; and candidates with either poor access or poor life expectancy would be more apt to accept a wider range of deceased donor kidneys quality. The candidate s decision making for accepting a deceased donor kidney of different quality will depend on where they are positioned in the 3-dimensional continuum of these factors The policy maker Similar to the transplant candidate, utility from the policy maker perspective depends on the overarching goal. If the primary goal is saving lives, then deceased donor kidneys would be prioritized to patients with medical urgency (unable to undergo dialysis) or pediatric patients for whom transplantation drastically improves neurologic and physical development. These patients are currently prioritized in North American allocation systems and are not considered in the discussion below. Therefore, this section discusses other outcome goals such as: short term patient survival (e.g. avoidance of acute rejection), total life years gained from transplantation, total years of graft life, total number of patients transplanted or total quality adjusted life years. It may be argued that the first priority from the policy maker perspective is to get the most use (i.e. graft survival) out of each deceased donor kidney donated. That is, the focus should be on reducing the waste of kidneys that occurs when these kidneys are transplanted into 48

65 candidates with a life expectancy that is shorter than that of the kidney they received. In theory, transplanting all deceased donor kidneys into transplant candidates with the greatest life expectancies would minimize kidney waste. It has been suggested that there may be an interaction between recipient and deceased donor kidney life expectancies, such that the youngest recipients may use up the graft life of their deceased donor kidneys more quickly (i.e. due to physiologic stress placed on the kidney by the active immune system of the younger recipient), thus reducing the total graft survival of older kidneys. 75 Although this hypothesis was not supported in a large American cohort study, 33 if an interaction does exist, deceased donor kidneys should be transplanted into candidates who will realize their maximum benefit. Efficient allocation (better use of deceased donor kidneys) will also result in improved patient outcomes. For example, in the European transplant program preferentially allocating lower quality deceased donor kidneys (age 65) to candidates with shorter life expectancies (age 65) has been proven to decrease the rate of discard of these older kidneys by facilitating their acceptance. 76 This has resulted in more candidates receiving kidney transplantation. In addition, this old to old prioritization has led to reduced time between organ retrieval (i.e. organ surgically removed from the donor) and transplantation as well as delayed graft function (often defined by a requirement of dialysis in the first week after transplantation). These factors are related to the health of the graft, and their reduction is associated with improved recipient survival. 76 After constraining allocation policies such that the potential waste of deceased donor kidneys is avoided, the policy maker will want to ensure that other utility measures are 49

66 maximized. Measuring the total utility of different strategies will depend on perceived benefit. A solution that optimizes short-term patient survival (e.g. matching genetic tissue) for the candidate population, may compromise total population survival or quality of life. With continued advancements in short-term transplant survival, the focus on maximizing longer-term outcomes will probably be more heavily weighted. The determination of which long-term measure to highlight at this level is most likely subjective. In contrast to transplant candidates, whose focus is primarily on individual benefit, policy makers focus on the benefit to the system or the entire transplant candidate population. 3.4 Summary Two ethical principles jockey for their place in the allocation of deceased donor kidneys. Justice, defined in this thesis as equitable access to transplantation, refers primarily to the selection of candidate subgroups for transplantation (i.e. justice determines the likelihood with which different candidate subgroups will be offered transplantation). Utility, defined in this thesis as reducing organ waste while maximizing patient outcomes, refers more so to the optimal distribution of deceased donor kidneys to these candidates. The definitions of justice and utility discussed in the remainder of this thesis are from the perspective of the policy maker, with consideration of the value of different utility outcomes for individual candidates. The roles of utility and justice in deceased donor kidney allocation have been dynamic over time. The following chapters discuss the roles of utility and justice in historical and 50

67 current deceased donor kidney allocation. Figure 3.1. A description of the intersection of life expectancy, access to transplantation, and quality life from the transplant candidates perspective. 51

68 4 The allocation of deceased donor kidneys for transplantation 4.1 Overview In the 1970s and early 80s, clinical transplantation was in its infancy and graft and patient survival outcomes following deceased donor kidney transplantation were mediocre (i.e. five year combined outcome of graft or patient survival <50%). 77 Use of kidney transplantation as a therapy was therefore reserved for young, otherwise healthy patients who were expected to have superior survival 27. At that time, few patients were waiting for transplantation and deceased donor kidneys were allocated to obtain optimal outcomes by matching transplant candidates to donors with the most genetically compatible tissues. 78 As a result of advances in immunosuppression therapy in the mid- 80s and 90s, rates of acute and chronic graft rejection decreased, resulting in increases in short-term graft and patient survival (American graft and patient survival in 2010: one year 95% and 96%; five year 84% and 86%). 15 As a result of these advances, kidney transplantation is no longer reserved for the healthiest patients, but is considered a viable treatment for most ESRD patients, including those of advanced age or with comorbid diseases. In fact, the relative benefit of transplantation is greater in patients with disease conditions such as diabetes who do very poorly on dialysis. In particular, there is no absolute age restriction for transplantation, and many of the sickest patients (e.g. diabetics) derive the greatest relative survival benefit from kidney transplantation, because their dialysis survival is limited

69 The liberalization of transplant eligibility combined with an increased prevalence of ESRD has created the current situation where the number of ESRD patients who could benefit from transplantation exceeds the number of kidneys available for transplantationresulting in a waiting list. In Canada, there are close to incident ESRD patients annually (a 20% increase since 2000); 3, 79 and more than prevalent ESRD patients in 2012 were on a Canadian waiting list for kidney transplantation. 80 Comparatively in the United States, there are close to incident ESRD patients annually and more than prevalent patients awaiting kidney transplantation. 81 As a result of this inequality between supply and demand, many transplant programs in different countries have shifted their focus in allocation away from utility to increased equity. The expanded eligibility criteria for deceased donor kidney transplant candidates, has created the need to utilize every organ that can be safely transplanted. This has led to a liberalization of donor acceptance (i.e. older donors or donors with greater comorbid burden). Deceased donors that previously may not have been considered for transplantation are now routinely used. 76 In 2011, more than 40% of Canadian deceased donor kidney transplant recipients were aged 60 years (compared to 27% of recipients aged 60 years in 2003), and 25% of deceased donors were aged 60 years (compared to 20% of deceased donors aged 60 years in 2003). 3 Despite this increase in expanded criteria donors, there remain only about 750 deceased donor kidneys available for transplantation per year; 3 only enough to transplant roughly 20% of current wait-listed patients. 53

70 Patient (recipient) life expectancy varies by transplant candidate case-mix, and many older transplant recipients die with a functioning kidney. The current justice based allocation system allows the transplantation of candidates with deceased donor kidneys that either do not function long enough to treat the patient for the remainder of their lifetime, or would be expected to function longer than the patient s life expectancy. Some countries have explored matching candidates and deceased donor kidneys with similar life expectancies, in an effort to increase the utility of the available donor pool and these approaches are described below. 4.2 Eurotransplant Senior Program Eurotransplant is the organization responsible for facilitating allocation and cross-border exchange of deceased donor organs in eight counties: Austria, Belgium, Croatia, Germany, Hungary, Luxembourg, the Netherlands and Slovenia). 82 Eurotransplant is responsible for a population of close to 120 million, of those have ESRD, and close to patients are wait-listed for deceased donor transplantation. In January 1999, in response to an increasing number of elderly transplant candidates on deceased donor kidney waiting lists, Eurotransplant established the Eurotransplant Senior Program (ESP) as a means to increase transplantation for elderly candidates. 83 ESP is intended to facilitate the matching of the shorter life expectancies of older deceased donor kidneys (age 65 years) and older transplant candidates (age 65 years). Enrolment in ESP for the allocation of deceased donor kidneys aged 65 years was voluntary by center in the first two years, but became mandatory in Meanwhile, 54

71 transplant candidates aged 65 years have the option of enrolling in either ESP for prioritized access to donor kidneys aged 65 years, or in the Eurotransplant kidney allocation scheme where they will compete for deceased donor kidneys with all other candidates. 82, 84 In ESP, allocation is restricted within close regions by blood type and waiting time alone with the goals of: 1) increasing the utilization of older donor kidneys, 2) decreasing time to deceased donor kidney transplantation for older candidates, and 3) ensuring graft and patient outcomes after transplantation are not negatively impacted. 76 The outcomes of ESP after five years were positive: the number of deceased donors aged 65 years doubled, waiting times for deceased donor kidney transplantation among ESP recipients decreased, and recipient (age 65 years) survival was similar with deceased donor kidneys aged 65 years and < 65 years Changing a deceased donor kidney allocation system- An American narrative Allocation prior to 2014 In the United States prior to 2002, the acceptability of different quality deceased donor kidney for transplantation was decided by physicians and individual organ procurement organizations. In October 2002 the expanded donor criteria program was established with the mandate of classifying and allocating poorer quality kidneys. 85 The objectives of the expanded criteria donor program were: 1) to decrease the number of marginal quality kidneys that were retrieved (removed from the donor for transplantation), but then discarded prior to transplantation (i.e. to reduce the number of wasted kidneys), and 2) to increase system efficiency by facilitating the allocation of these kidneys such that the 55

72 time from retrieval to transplantation would be minimized, and graft survival would be increased. 85 With this program deceased donors in the United States were classified into two quality categories for the purpose of kidney allocation: standard criteria donors (SCD) and expanded criteria donors (ECD). ECD kidneys by definition portend a 70% greater risk of graft failure compared to SCD kidneys, 69 and are classified by a donor age 60 years, or donor age years with at least two of the following comorbidities: a high donor serum creatinine, donor death due to cereberovascular accident or a donor history of hypertension. In practice, all non-ecd donors are considered SCD. 86 A separate kidney transplant candidate waiting list exists for each type of deceased donor kidney. Transplant candidates can be on both waiting lists at the same time, but must give explicit consent to be a member of the ECD list. The intended recipients of ECD kidneys were candidates with poorer survival on dialysis, or candidates in regions with long waiting times on dialysis who would benefit from receiving such kidneys (i.e. kidneys at increased risk of graft failure), if they were transplanted faster and had less exposure to dialysis. 16 The reduced waiting time on dialysis was expected to result from fewer candidates vying for transplantation from these ECD kidneys. SCD kidneys are allocated first to candidates with perfect genetic tissue matches anywhere in the country. Allocation is then based on a system that awards prioritization points for a) pediatric candidates, b) time on the SCD kidney transplant wait-list, c) degree of genetic tissue mismatch, d) high sensitization (i.e. expectation of tissue incompatibility with a high percentage of deceased organ donors) and e) previous living donation. 87 Of note, SCD kidneys aged < 35 are prioritized to pediatric recipients unless 56

73 they are promised to another candidate with perfect genetic tissue match. SCD kidneys are first offered locally, then regionally and nationally, except for perfect genetically matched kidneys as noted above. In contrast, ECD kidneys are allocated within blood groups by waiting time alone Need for change Over the last decade the difference between the number of wait-listed transplant candidates and the number of deceased donor kidney transplantations has widened. 88 The current American kidney allocation system has neither minimized death on the kidney transplant waiting-list, nor maximized survival after transplantation. 31 In 2004, the American organ procurement and transplantation network (OPTN) board of directors charged the OPTN kidney committee (OPTNKC) with reviewing the allocation policies for deceased donor kidneys. Particular concern was voiced about increasing utility of the current system, particularly by avoiding extreme age mismatches (or the allocation of deceased donor kidneys with long life expectancies into candidates with much shorter life expectancies). 36 Members of the Scientific Register of Transplant Recipients collaborated with the OPTNKC to develop new tools for allocation that would increase the utility gained from transplanted deceased donor kidneys. During the multi-year review, the OPTNKC considered various proposals in conjunction with the members of the transplant community and other stakeholders. The review culminated with a proposal presentation of a new allocation policy (the Kidneys Allocation Score) at a public forum in St Louis, MO in

74 During the development of the Kidney Allocation Score, OPTNKC was guided by the OPTN Final Rule- a regulatory framework implemented by the U.S. Department of Health and Human Services. 64 The Final Rule states that the OPTN should develop policies for the equitable allocation of deceased donor organs that must 1) be based on sound medical judgement, 2) seek to achieve the best use of donated organs, 3) be designed to avoid wasting of organs, 4) avoid futile transplants, 5) promote patient access to transplantation and 6) promote the efficient management of organ placement Kidney allocation score The primary proposal entertained and put forth by the OPTNKC was the Kidney Allocation Score (KAS). The goal of KAS was to provide equitable access to deceased donor kidneys for all transplant candidates, while improving the outcomes of recipients of such kidneys. 90 Therefore, the KAS ranking system for priority for deceased donor kidneys contains components of both utility and justice. Three components (described below) are included in the calculation of KAS: patient life years (gained) from transplant, the donor profile index (which is used to estimate the organ quality and survival), and dialysis time. KAS is calculated so that deceased donor kidneys with the greatest expected survival (i.e. top 20%) are allocated to candidates with the greatest expected survival (i.e. top 20%), while deceased donor kidneys with lower expected lifetimes are allocated by the justice principle to candidates with the longest times on dialysis. 58

75 Life years from transplant- measure of deceased donor kidney transplant utility Life years from transplant (LYFT) is the number of extra years of life a transplant candidate would be expected to live with a deceased donor kidney transplant compared to never receiving a transplant. 8, 31, 91 Each candidate has a unique LYFT score based on his/her patient characteristics, for specified deceased donor kidney characteristics (e.g. the same patient would have a different LYFT score with two donor kidneys that differed by age alone). LYFT is intended to be calculated for matched candidates at the top of the deceased donor kidney waiting lists after a deceased donor kidney has been identified for transplantation. The concept of LYFT can be adjusted to account for differences in expected quality of life with and without a functioning transplant, using a fractional constant Q. LYFT is calculated in the literature with Q=0.8 as: LYFT = 1.0 x (median expected lifetime of candidate with a functioning transplant) x (median expected lifetime of candidate after transplant failure) x (median expected lifetime of candidate without a transplant) LYFT methodology Estimates of kidney transplantation survival benefit prior to LYFT were biased, using referent populations of all dialysis patients. 31 Transplant candidates are a highly selected healthy dialysis population, and as such LYFT calculations use the more appropriate population of patients wait-listed for deceased donor kidney transplantation as a reference cohort. The three outcomes in LYFT (patient survival without a prior to transplantation, patient survival with a functioning transplant, and patient survival after transplant failure) 59

76 are estimated using separate adjusted Cox proportional hazards regression models. Three methods were examined to estimate the lifetimes of particular candidates in each time period using the survival probabilities from the Cox models: 1) expected lifetime (calculated as area under the survival curve), 2) truncated expected lifetime (calculated as area under the survival curve limited to a specific time period, e.g. 10 years after transplantation), and median lifetime or half-life (calculated as the time at which half the population has died). The expected lifetime method was abandoned because it could not directly estimate patient and graft survivals after the study end of follow-up (15 years) at which point many patients/grafts were still alive. The truncated expected lifetime method was able to be estimated with the data, but was rejected because it was believed that the loss of LYFT calculated after the truncated time would be differential in patients/ grafts with shorter versus longer expected lifetimes and thus bias LYFT away from the youngest and healthiest patients. As a result, the median lifetime method was selected for final calculations of LYFT. Median lifetimes were able to be calculated up to 15 years for 99% of candidates without a transplant and 72% of candidates with a transplant. Estimates of LYFT in the remaining populations were calculated using extrapolation The kidney donor profile index In 2009, expanded criteria donors accounted for 16% of deceased donors in the US population. 92 The use of a dichotomous classification for donor quality allows patients the opportunity to accept organs of lower quality in exchange for shorter waiting times to transplantation, and also allows for patient counselling about the risk of graft failure for a 60

77 given donor. Due to the limitations of a dichotomous classification, there was overlap in quality of kidney in each category (ECD/ SCD). In 2009, Rao et al 93 developed a new and continuous score (the kidney donor risk index (KDRI_Rao)) to determine deceased donor kidney quality. The KDRI_Rao is the relative hazard of graft failure for a given deceased donor kidney compared to a reference deceased donor kidney (brain dead donor, aged 40-years, non-african American race, serum creatinine =1.0, no history of hypertension or diabetes, cause of death not cereberovascular disease, height=170 cm, weight=80 kg, hepatitis C negative, 2 human leukocyte antigen (HLA) B mismatches 2, 1 HLA DR mismatch with a cold ischemic time of 20 hours). The index was developed using data from all (N=69,440) adult, first kidney-only deceased donor ABO compatible transplant recipients between in the Scientific Registry of Transplant Recipients (United States). The KDRI_Rao was modeled using a multivariate Cox regression stepwise deletion model. The baseline model included all recipient and deceased donor data available in the data set. The OPTN has amended the KDRI_Rao to the KDRI_Median (KDRI), such that the reference deceased donor is the median quality donor. The KDRI is calculated as the KDRI_Rao using only donor factors divided by the median KDRI_Rao of the previous year. The KDRI is commonly reported as the Kidney Donor Profile Index (KDPI), a direct 1:1 mapping of the KDRI to the ranked percentile of deceased donors (i.e. the median donor would have KDRI=1.00 and KDPI=50%). A donor with a KDRI of Human leukocyte antigens (HLA) are proteins found on the surface of cells of the body that are inherited genetically. HLAs are used to determine the biologic compatibility between a donor and recipient. In transplantation, 3 different loci are generally examined: A, B and Dr. Each of these loci has two antigens, representing a possibility of up to 6 different HLA mismatches per donor-recipient pair. 61

78 would have a 70% increased relative risk of graft failure compared to the median donor. Similarly, a donor with a KDPI of 80% would have a higher likelihood of graft failure compared to 80% of the donor population. The KDPI has higher predictive ability compared to ECD, and is more discriminative because it is continuous and not dichotomous. This allows for the identification of higher and lower quality ECD kidneys; allowing for more efficient allocation of marginal kidneys Dialysis time The third component accounted for in KAS is dialysis time. In current deceased donor kidney allocation in the United States, patient waiting time has been calculated as the time since referral for transplantation (i.e. placement on the waiting list). Waiting time is an objective measure, and the major tiebreaker for allocation based on justice. So as not to disadvantage patients who are referred for transplantation long after dialysis start, the time since the start of dialysis treatment (DT) is included as a modified component of the proposed allocation score Criticisms of KAS and LYFT The primary advantage of including LYFT in deceased donor kidney allocation is the increased number of life years that would be gained with the same kidneys transplanted into different people. 91 DT was included in KAS to offset the probable inequity in access to transplantation among patients with poorer expected survival after transplantation, and 62

79 the KDPI was included to calculate the life expectancy of deceased donor kidneys, to ensure that the healthiest kidneys would be allocated to transplant candidates with the longest life expectancies. Despite the utilitarian advantages of KAS, the allocation score has been strongly criticized. KAS has been judged to be complex and non-transparent, making it difficult for patients and caregivers to predict how long patients will be required to wait on dialysis for transplantation, and which patients will be transplanted next. 37, 90, 94, 95 One author predicts that KAS is more likely to change which candidates are transplanted, and not the quality of organ that a given candidate will receive. 29 Also, it is unclear whether the transplant community should be responsible for making decisions about the allocation of a public resource. 95 In addition to these valid concerns, the most widespread criticism about KAS has focused on the LYFT component. LYFT predicts population but not individual life years from 29, 91 transplant. The LYFT models have been validated neither externally 90 nor prospectively 95 and their predictive power is low. Specifically, the ability of the LYFT models to discriminate between all candidates with different survival probabilities is poor (c-statistic 3 : ), 37, 90, 91, 94, 95 although LYFT is good at differentiating between candidates with the shortest and longest survival probabilities (c-statistic ). 91 Hippen et al 95 predicts that the long-term use of KAS would result in a homogeneous candidate waiting list, where LYFT will be forced to choose between similar patients, 3 The c-statistic (or concordance statistic) is a measure of model discrimination. Traditionally, the c-statistic is used to determine the ability of a statistical model to assign a higher predicted value to a subject who experienced a dichotomous outcome compared to a subject who did not experience the outcome. In survival analysis, the c-statistic is extended to represent the ability of the model to assign higher predicted survival times to a subject with longer survival compared to a subject with shorter survival. In this example, the c-statistic is the probability of the model predicting a greater survival time in the subject with the longer survival time among all possible pairs in the model. 63

80 potentially encouraging gaming. That is, manipulating the waiting list to leverage access to transplantation for a patient (e.g. multiply wait-listing patients, wait-listing candidates prior to transplant suitability). The quality and definitions of the data elements in LYFT are limited because they are from administrative data. 90, 95 Also, the addition of variables after the inclusion of age in the models only marginally improves prediction, but adds to model complexity. 29 The LYFT models are dependent on the length of follow-up. Long time horizons are needed to show the added benefit of transplantation in the young and healthy, but longer follow-up biases against the older and sicker. 90 In addition to the analytic exceptions to LYFT are some moral concerns. First, for nonrenal organs survival is the primary goal of transplantation, whereas for kidney transplant candidates quality of life is of preeminent importance. 37 Allocating kidneys based on life years gained from transplantation as opposed to gain in quality of life may not be appropriate. Second, each year after transplantation is given the same value in the LYFT models, but the first year of life after transplantation may not be equivalent to the 20 th year of life after transplantation. 95 Incorporating discounting into the measurement of LYFT may be valuable. Third, it is unclear how prioritizing kidneys for the young and healthy through LYFT will impact living donation. 37, 90, 95 Fourth, LYFT does not account for patient preferences and autonomy over what quality of kidneys patients would be willing to accept. 90 Fifth, LYFT focuses on only one side of increasing utility, maximizing life years post transplantation; but there is no discussion about minimizing harm (e.g. inequity in transplantation, decreased post-transplant survival and increased waiting list deaths for certain patient groups)

81 As a result of many harsh reviews of LYFT, the KAS was not supported as presented in St Louis, and the needs and wants of the community were discussed. Many in the community believe that a national, transparent and comprehensible system, able to incorporate change based on new data and outcomes is necessary. 29, 94 Some suggestions to improve allocation have included wider geographic sharing of organs, 32 allocating the youngest kidneys to the youngest candidates, 32, 90 allowing candidates to indicate an acceptable range of kidney quality, 90, 94 allocating the lowest 20% quality kidneys (using KDPI) to the highest 20% candidates (using LYFT), 37 using estimated survival time instead of LYFT as a measure of transplant life expectancy, 94 and limiting regression models for survival (such as LYFT) to three to four factors Proposed national allocation policy After the public forum in 2009, and feedback from transplant professionals, patients, donor families and the public, the OPTN Kidney Transplant Committee developed a new deceased donor kidney allocation proposal. This proposal was released in February 2011 with a request for comments. The three key limitations to the current allocation system stressed by the committee at this time were: high discard rates of poorer quality (ECD) kidneys, variability in access to transplantation by different candidate subgroups, and lack of matching the life expectancies of deceased donor kidneys to transplant candidates. Age matching was a recommendation after the previous review and was incorporated into the new proposal

82 The 2011 proposed allocation first calculated the donor quality using the KDPI. If the KDPI 20% (i.e. the highest quality kidneys), then the kidney was allocated with priority to candidates with the highest 20% expected post-transplant survival (EPTS) 4. Kidneys with KDPI > 20% were allocated with priority to candidates aged within 15 years of the kidney age (e.g. a 50 year old kidney would be allocated to a year old recipient). Priority ranking within the two aforementioned priority groups was proposed to be similar to current allocation priorities. The 2011 proposed allocation was for kidney-only transplants and did not alter priority for multi-organ (e.g. kidney-pancreas) transplantations. The OPTNKC performed simulations to determine if under the 2011 proposal certain groups would have disadvantaged access to transplantation compared to current allocation. 35 Candidates were not differentially disadvantaged by blood group, or by race. However, fewer older candidates with diabetes and high sensitization were simulated to receive transplantation. A simplified alternative proposal dropped the special allocation by KDPI < 20%, and had all deceased donor kidneys being allocated by age matching (within 15 years). This alternative proposal showed similar results to the 2011 proposal, but was rejected for being less modifiable. It is believed that allocation priorities will need to be adjusted with constantly updated data and the mixed 2011 proposal allocating by life expectancies and by age matching was determined to be more flexible for future amendment. 4 The expected post-transplant survival is calculated using 4 variables: candidate age, duration of dialysis exposure (DT), diabetes mellitus as a comorbidity or cause of ESRD and whether candidate was a prior organ donor. 66

83 4.3.5 American allocation summary The United States Department of Justice determined that allocation of organs for transplantation based on age alone was discriminatory. 96 Therefore, the 2011 proposal has since been revised. The latest version of the new allocation system, which eliminates age matching within 15 years, was approved in 2013 by the OPTN. This approved allocation maintains the proposed top 20% KDPI to top 20% EPTS, and broadens the geographic allocation of kidneys. For example, the most highly sensitized patients will now be offered kidneys from the national pool, and allocation for the lowest quality expanded criteria donor kidneys (KDPI > 0.85) will now be regional, as opposed to local. The SRTR/OPTN are currently disseminating information about the new allocation system, and enhancing their data collection. The new allocation model is expected to be implemented in December Summary The growth in the number of patients with ESRD who would benefit from transplantation is remarkable. Despite ongoing efforts of programs to increase donation, and the expanded criteria for donors, the increase in deceased and living donation in the last decade has been minimal, and innovation is needed to address the widening gap between the demand for and supply of kidneys for transplantation. A consequence of long waiting lists for kidney transplantation, is that current allocation systems favour justice by heavily prioritizing patients who have been waiting the longest for transplantation. This practice may lead to discrepancies in donor kidney and candidate life expectancies (i.e. organ waste, or candidate need for repeat transplantation) (more detail provided in Chapter 5). 67

84 The use of age matching to increase utility in these justice-based systems has been explored in Europe and the United States. The Eurotransplant Senior program which began allocating donor kidneys aged 65 years to candidates aged 65 years in 1999 has been successful and is ongoing. In the United States, allocation of donor kidneys by age alone was rejected by the Department of Justice, despite public and medical support to include age matching in allocation algorithms. The new allocation system that will be implemented in the United States by the end of 2014 will maintain a component of age matching that will preferentially allocate the highest quality kidneys to the highest quality candidates (donor age and candidate age are the greatest predictors of donor kidney and candidate quality respectively). However, the exclusion of the 15-year age matching component from the 2011 proposal allows young candidates to elect to receive the lowest quality (i.e. oldest) kidneys. The potential impact of the policy allowing old to young transplantation in the United States will be explored in Chapter 7. 68

85 5 The type and quantity of inefficiency in donor-candidate age matching: measuring the area between the curves 5.1 Introduction Transplantation is the optimal treatment for patients with kidney failure, but the number of organs is not sufficient to transplant all patients who could benefit. Therefore, allocation rules dictate the method by which transplant candidates are offered kidneys for transplantation. Allocation rules with the intent of maximizing the use of the scarce resource would prioritize transplantation for the youngest and healthiest candidates, while a justice based system would prioritize candidates with the longest wait times on dialysis. The current allocation systems prioritize justice in candidate access to transplantation in this way by heavily weighting candidate time on the transplant waiting list. This justicebased allocation allows the possibility of system inefficiencies (i.e. discordance in the expected survival times of the transplant recipient and the deceased donor kidney). In contrast, allocation of organs by matching life expectancies of the donor kidney and recipient is an opportunity to decrease these inefficiencies (Figure 5.1). A perfectly efficient system would allocate deceased donor kidneys to candidates such that the transplanted patient would outlive the kidney graft by one day 37. This ideal system would maximize the benefit to the population of transplant candidates by minimizing: 1) the loss of potential kidney transplant function that occurs when the recipient dies before the time the graft would otherwise fail (organ waste) and 2) the return to dialysis and need for repeat transplantation that occurs when the deceased donor 69

86 kidney fails prior to recipient death and the recipient returns to the wait list to compete for a subsequent transplant (Figure 5.2). For example, transplantation of a 20-year old deceased donor kidney to a patient aged > 60 years would result in a loss of potential kidney transplant function because the anticipated life expectancy of the transplanted organ is greater than that of the transplant recipient 29. Alternatively, if a 60-year old deceased donor kidney is transplanted into a 20-year old recipient, the recipient s life expectancy will exceed that of the deceased donor kidney and the recipient will require another transplant. Repeat transplantation is a significant burden on the Canadian organ supply, accounting for 11% of kidney transplantation. 3 Therefore, matching the estimated life expectancies of transplant candidates and deceased donor kidneys has the potential to reduce inefficiencies in allocation. 29 Accurately predicting post-kidney transplant recipient and deceased donor kidney life expectancies prior to transplantation is an imperfect science, in large part because life expectancies are dependent on post-transplant factors, which are unpredictable prior to transplantation (e.g. the recipient s immunologic response to the transplanted kidney, post-transplant disease development (e.g. diabetes, cancer), procedural factors (e.g. time between organ retrieval and transplantation)). Pre-transplant information available to estimate candidate life expectancy includes patient case-mix; and similar information available to estimate deceased donor kidney life expectancy includes donor factors (e.g. cause of death, hypertension, kidney function). Recipient and donor ages are the most important predictors of post-kidney transplant recipient and deceased donor kidney survival 30, 97 and have the advantage of being objectively measurable and transparent 70

87 factors for both patients and medical personnel. In contrast, the presence and severity of other factors (e.g. diabetes) may be more difficult to characterize. The inclusion of variables in addition to recipient and deceased donor kidney age, in models to estimate recipient and deceased donor kidney survival, has been proposed but rejected for being only marginally superior to models including donor and recipient ages alone, and too complex for patients and policy makers. 31, 32 For these reasons, donor and candidate ages are reasonable proxies for recipient and deceased donor kidney life expectancies. 95 In the United States, a proposed allocation strategy incorporated age matching (± 15 years) into allocation rules to exploit the existing differences in deceased donor kidney and recipient survival. 89 This strategy eliminates extreme age mismatches thereby increasing the system s utility. Two disadvantages of this system are 1) the distribution of deceased donor kidney ages is younger than the distribution of candidate ages, and therefore older candidates have reduced access to transplantation, 35 and 2) the life expectancies of candidates and deceased donor kidneys of similar ages (i.e. ± 15 years) may not be equivalent, and therefore candidates may be disadvantaged because they are not eligible to be allocated a deceased donor kidney that has adequate graft survival for them. For example, a 55 year old candidate may be expected to live 15 years with a transplant, but a 55 year old deceased donor kidney may be expected to survive for only 7 years. In this example, the 55 year old candidate may be disadvantaged if they are not eligible to receive a younger deceased donor kidney (e.g. a deceased donor kidney aged 39 years with 15 years of life expectancy). This example suggests that direct age matching may not be satisfactory. 71

88 In order to select appropriate age matching cut-points for allocation, it is imperative to measure the inefficiency that occurs for each possible donor-recipient pair. Therefore, using deceased donor kidney and recipient ages as surrogates for life expectancies, the objectives of this paper are to determine for each recipient age group 1) the type of inefficiency (organ waste or return to dialysis), and 2) the quantity of inefficiency, that occur when recipients are transplanted with deceased donor kidneys of different ages. The type and quantity of these inefficiencies will also be used to compare the utility of different allocation systems. 5.2 Methods Study population and data source Adult (age > 18 years) recipients of deceased donor kidney-only transplantation captured in the Canadian Organ Replacement Register between January 1, 1995 and December 31, 2008 were included in the study. Pediatric recipients were excluded from the study because they a priori receive prioritization for allocation due to the impact of dialysis on their neurologic and physical development. Patients transplanted in Quebec were not available at the time of data request and were not included. For each recipient, only the first kidney transplantation that occurred during the study period was included. Recipients were excluded from the study if they had missing data on either recipient or donor age at transplantation. 72

89 Population level characteristics of recipients and deceased donor kidneys were represented as means ± standard deviations (or medians and first and third quartiles) for continuous variables, and proportions for categorical variables Analytical methods Selection of age strata For each of recipient and deceased donor kidney age, dummy variables were assigned for the following categories: age <19 (pediatric, deceased donor kidney age only), 19-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, and 65 years. A Cox proportional hazards regression analysis was performed to compare the patient survival of recipients of different ages adjusted for deceased donor kidney age, and a similar analysis was performed to compare the graft survival of deceased donor kidneys of different ages adjusted for recipient age. Age groups were collapsed to form final recipient and donor age strata for the analysis by taking into consideration overlap between survival curves by recipient and donor ages respectively, as well as the limited sample size in some age strata Estimating deceased donor kidney and recipient survival The two primary dependent variables in this study were recipient survival and deceased donor kidney survival (graft survival). Recipient survival refers to the probability that a transplant recipient is still alive at a given time after transplantation; and graft survival refers to the probability that the deceased donor kidney is still functioning at a given time after transplantation. Each dependent variable is time to event, and time was counted 73

90 from the date of transplantation until the outcome of interest (recipient death or deceased donor kidney failure (i.e. either recipient return to dialysis, or repeat transplantation)), or the end of follow-up (December 31, 2011). The primary outcome of recipient survival was neither censored at deceased donor kidney failure nor repeat transplantation because the outcome of interest was absolute patient survival after kidney transplantation, independent of the recipient s type of renal replacement therapy. In contrast, for the outcome of graft survival, recipient death was treated as a censoring event Area under the survival curve The area under a survival curve represents the expected (or average) survival time for subjects. 98 For the outcome of recipient survival, the area under the survival curve (mean survival) was calculated using the Kaplan-Meier product limit method with stratification by recipient age strata. Similarly, the area under the graft survival curve was calculated with stratification by donor kidney age. The area below the recipient and graft survival curves were provided by SAS 9.4 (Carey, N.C) using Irwin s restricted mean, truncated at 120 months (10 years). 99 Additional covariates were not included in the models because recipient and graft survival curves were directly compared within the same population. In the models, the inter-cluster correlation between deceased donor kidneys from the same donor (maximum 2 kidneys per donor) was accounted for using the robust 100, 101 sandwich covariance matrix Area between survival curves When two populations are compared, the surface area between their corresponding survival curves is the average duration of life gained in the population with the greatest survival relative to the population with the least survival 98. For example, the average life 74

91 gained for a study treatment group relative to a control. Similar to Meier-Kriesche et al, 102 we extended this argument to compare population level recipient and graft survival after kidney transplantation. We quantified the area between the recipient and deceased donor kidney survival curves as either the expected time a recipient would be required to return to dialysis after deceased donor kidney failure (i.e. when recipient survival > deceased donor kidney survival) or the expected loss of potential kidney function after the recipient died (i.e. when deceased donor kidney survival > recipient survival). The area between the curves was calculated as the difference in the areas under the two survival curves. Positive values for area between the curves represented the recipient outliving the deceased donor kidney, while negative values represented the recipient dying with residual deceased donor kidney function Allocation utility Each individual in the study was assigned a type (organ waste or return to dialysis) and quantity of inefficiency equivalent to the area between the curves corresponding to the intersection of their recipient and donor age strata. These inefficiencies were totalled over all recipients to calculate the summative utility of a given allocation system. This was done for each recipient based on the actual deceased donor kidney they received (actual allocation), as well as by assuming the same deceased donor pool had been allocated to the same recipient pool under an alternative allocation rule. In this study we examined a rule that ranked each recipient and donor by age and allocated deceased donor kidneys to recipients by matching their rank order (rank allocation matching). For example, the 75

92 youngest deceased donor kidney was allocated to the youngest recipient, and the oldest deceased donor kidney to the oldest recipient. 5.3 Results There were N=6 443 patients who received deceased donor kidney transplantation between 1995 and Of these, N=19 were missing data on deceased donor kidney age and a further N=300 recipients were aged < 19 years. These recipients were excluded from the analyses leaving a final population of N=6 124 patients. The mean potential follow-up of study patients was 10 years. Recipients were mostly Caucasian, male, and without diabetes (Table 5.1). Recipients were not highly sensitized (i.e. not estimated to be tissue incompatible with a high percentage of deceased organ donors as measured by the peak panel reactive antibody >30% (8%)), and perfect genetic tissue matching (HLA mismatch=0) was infrequent. The median (q1, q3) waiting time for transplantation on dialysis was 2.6 years (1.4, 4.1). The distribution of recipient ages was normal and centred at 50 years of age, while the distribution of donor ages was bimodal with peaks near 19 and 50 years of age (Figure 5.4). Cox proportional hazards models for life expectancy outcomes by recipient and deceased donor kidney ages are shown in Figure 5.5. Final age strata (and the corresponding sample sizes) for area-between-the-curve analyses, determined by collapsing age strata with overlapping survival curves from Cox proportional hazards models were: recipient ages (19-39, 40-49, 50-54, 55-59, 60) and deceased donor kidney ages (< 35, 35-49, 50) (Table 5.2). 76

93 Table 5.3 shows for each strata the number of months of recipient and deceased donor kidney survival (maximum 120 months=10 years), and the difference in area (inefficiency) for each age combination. Boxes where the recipient outlived the deceased donor kidney are shaded in red and the difference in survival represents a return to dialysis and need for repeat transplantation. Boxes shaded in blue show age combinations where the deceased donor kidney outlives the recipient and the difference in survival represents the amount of organ waste. Boxes with no shading represent deceased donor kidney and recipient age strata with roughly equivalent survival, and negligible differences between the curves. Recipients aged less than 40 years outlived deceased donor kidneys of all ages; recipients aged outlived older deceased donor kidneys, but had no difference in survival compared to younger donor kidneys. In contrast, recipients aged 55 years and over died before their donor kidney failed with the youngest deceased donor kidneys, but had no difference in survival with the oldest donor kidneys. The magnitude of inefficiency increased as the discrepancy between deceased donor kidney and recipient ages increased Summative utility The type and quantity of inefficiency for the current allocation system and the rank matching allocation system are shown in Table 5.4. The summative utility for the actual allocation, which treats the two types of inefficiencies equally, was years (11 months per person), and for the rank matching allocation was years (8 months per person). This resulted in a difference in summative utility between the two systems of years (or 3 months per person). 77

94 5.4 Interpretation This study provides a method, using area between survival curves, to categorize and quantify the inefficiency that results when there is a difference in the life expectancy between a transplant recipient and their transplanted kidney. We found that adult recipients younger than 40 years outlived deceased donor kidneys of all ages and required repeat transplantation. Recipient aged outlived donor kidneys that were young relative to them, and had similar survival to older donor kidneys. In contrast, recipients aged 55 years were most likely to die with a younger functioning donor kidney, but had similar life expectancy to older donor kidneys (Table 5.3). For each recipient age, the amount of inefficiency increased with more discrepant deceased donor kidney age. When comparing an age-rank based allocation system compared to actual allocation we found that we could eliminate close to 500 years of wasted graft function, and more than 800 years of dialysis (Table 5.4). This would necessarily result in increased recipient and graft survival. Several studies have suggested alternative methods to incorporate age matching into allocation rules to exploit the differences in deceased donor kidney and recipient survival that arise in the current justice favoured systems. 35, 103 These studies have estimated improvements in efficiency by calculating changes in recipient and deceased donor kidney survival. The new American proposal, which prioritizes the top 20% quality kidneys to the 20% of candidates with the greatest expected survival, may result in an increase in patient survival with a functioning kidney graft (i.e. improvement in utility) as a result of transplanting a greater proportion of younger and healthier people; but their 78

95 increased efficiency comes at the expense of equitable access. 35 Ross et al 103 propose a similar allocation structure that redistributes the youngest deceased donor kidneys to the youngest recipients, but maintains equity in access to transplantation by restricting candidate age groups to receive a proportional number of deceased donor kidneys. Although both systems avoid extreme age mismatches, they do not account for inefficiencies that occur for non-extreme ages, and may disadvantage moderate aged well candidates from receiving deceased donor kidneys that meet their survival needs. The measure of inefficiency in this study is unique because it simultaneously characterizes the type (organ waste and the need for repeat transplantation), as well as the magnitude of the difference in deceased donor kidney and recipient life expectancies. In addition, it does not assume a strategy for age matching (i.e. rank order allocation) but allows for the evaluation of efficiency under different age constraints. Depending on the importance of inefficiency type and magnitude to the decision maker, the measure of inefficiency in this study can be used to suggest appropriate and inappropriate utility based cut-points for allocation using age matching. For example, to avoid organ waste all deceased donor kidneys could be allocated to the youngest candidates, but this would be offset by an increase in need for repeat transplantation in those young candidates who receive the oldest deceased donor kidneys. Alternatively, allocating all deceased donor kidneys to the oldest candidates would minimize the need for repeat transplantation in these candidates, at the expense of increased organ waste. Ideally, these inefficiencies could be used in tandem with the candidate and deceased donor kidney age distributions to identify equitable utility-based age cut-points for 79

96 balanced allocation. In addition, the summative utilities calculated under different proposed allocation systems could be used to compare the efficiency of different allocation strategies. The calculation of inefficiency in this paper is limited by restricting the follow-up time to ten years. However, the additional area between the curves at the age extremes would likely inflate the estimated inefficiencies, as the curves tend to separate over time, not reduce them. In addition, in the model for deceased donor kidney life expectancy, recipient death was treated as a non-informative censoring variable. The inclusion of recipient death as a competing risk was considered, but declined. The purpose of the model was to estimate deceased donor kidney failure from an allocation (i.e. decision maker) perspective. From this vantage, a recipient death was considered to be independent of the deceased donor kidney life expectancy that we try to predict pretransplant (i.e. we assume a prioi in some cases that the deceased donor kidney will outlive the recipient). In contrast, when considering the outcome of graft failure from the recipient s perspective, recipient death could be considered a competing risk for deceased donor kidney death because it precludes the recipient from realizing the outcome of graft failure without death. Furthermore, the most accurate way to estimate deceased donor kidney survival would be in recipients with the greatest expected survival. This was not done because it would not allow for an interaction for the same aged deceased donor kidney in different aged recipients; and also deceased donor kidneys of the same age transplanted into different candidates may differ in quality due to confounding by 80

97 indication. Ideally, this would be accounted for in a paired kidney analysis, in which the two kidneys from one deceased donor would be allocated to candidates of different ages. The estimated inefficiencies in the study model can be used to compare between similar Canadian populations and over time, but should be recalculated to evaluate inefficiencies in different populations (e.g. other countries). Although the numbers from Table 5.3 are not generalizable outside of the study population, the methods are commutable. An advantage of this measure of inefficiency is its reliance on donor and recipient ages alone. As the age distributions of the candidate and donor pool change over time, utilitybased allocation could be altered simply by changing the age cut-points. The inclusion of other important factors in models to estimate life expectancy, such as diabetes and cardiovascular disease in candidates, may slightly increase the predictive ability of these models, but remains a limited improvement due to the inability of administrative data on these reported dichotomous variables to accurately account for the spectrum and severity of these comorbidities. Given the impact of diabetes on recipient life expectancy, it would be advantageous to recreate this analysis, in future work, by stratifying on diabetes as a comorbidity or cause of renal failure. In conclusion, the study provides a method of measuring the type and quantity of inefficiency for different recipient and deceased donor kidney ages. These measures can be used in tandem with information about the distribution of deceased donor kidney and candidate ages to inform the selection of cut-points for age matching in deceased donor kidney allocation. 81

98 Figure 5.1 Panel A shows possible differences in predicted patient and deceased donor kidney survival between donor kidneys and their recipients under current allocation rules. Panel B shows possible differences in predicted patient and deceased donor kidney survival for the same aged donor kidneys and recipients using allocation by life expectancy matching. Figure 5.2 A description of the inefficiencies that occur when the same recipient is transplanted with deceased donor kidney grafts with differential expected survival.the dark grey bars represent the two types of inefficiency that occur when the recipient and the donor kidney have unequal survival. 82

99 Figure 5.3 Left panel: Area between the recipient and graft survival curves represents the time a patient would be required to return to dialysis and await repeat transplantation. Right panel: area between the recipient and graft survival curves represents the loss of potential kidney function. The vertical distance between each pair represents the difference in predicted patient and deceased donor kidney survival. 83

100 Figure 5.4. Histograms showing the distribution of recipient (top panel) and deceased donor kidney (bottom panel) ages 84

101 Probability of Remaining Event Free Figure 5.5. Left panel: Recipient survival by recipient age categories. Right panel: Graft survival by donor age categories. Recipient survival Graft survival < Time from transplantation (years) Time from transplantation (years) 85

102 Table 5.1 Recipient and donor characteristics Recipient N=6 124 Age mean (std); median (q1,q3) 50 (12.9); 50 (41,60) min=19 max 84 Female sex (%) (45) Race Caucasian Other Cause of End-Stage Renal Disease Diabetes Other Peak panel reactive antibody % >80 Missing* Time on dialysis prior to transplantation (years) Median (q1,q3) Province of transplantation Alberta British Columbia Saskatchewan Manitoba Ontario Nova Scotia Year of transplantation Donor (66) (34) 695 (11) (89) (76) 796 (16) 274 (6) 104 (2) (1.8, 5.8) (17) 826 (14) 295 (5) 268 (4) (48) 757 (12) (37) (33) (30) Age 40 (16.9); 43 (26,53) min 0 max 86 Cause of death: cerebrovascular disease 600 (10) Donation after cardiocirculatory death 79 (1) 86

103 Table 5.2. Study sample size by recipient and donor kidney age categories Sample Size (N) Recipient age (years) Deceased donor kidney age (years) <

104 Table 5.3. Area under and between recipient survival and graft survival curves Deceased donor kidney age (years) < Recipient age (years) R Survival Graft Survival Δ Survival 9.15 (0.08) a 8.28 (0.13) 0.87 (0.15) 9.47 (0.09) 8.05 (0.15) 1.42 (0.17) 9.17 (0.10) 7.06 (0.20) 2.11 (0.22) R Survival 9.02 (0.09) 8.97 (0.11) 9.04 (0.11) Graft Survival 8.83 (0.11) 8.39 (0.14) 7.91 (0.17) Δ Survival 0.19 (0.14) 0.58 (0.18) 1.13 (0.20) R Survival 8.84 (0.15) 8.73 (0.17) 8.33 (0.17) Graft Survival 8.62 (0.16) 8.39 (0.17) 7.71 (0.21) Δ Survival 0.22 (0.22) 0.34 (0.24) 0.62 (0.27) R Survival 8.63 (0.17) 8.28 (0.20) 8.24 (0.15) Graft Survival Δ Survival 9.70 (0.12) (0.21) 8.36 (0.17) (0.26) 8.47 (0.17) (0.23) 60 R Survival 7.81 (0.16) 7.60 (0.16) 7.39 (0.14) Graft Survival 9.03 (0.13) 8.81 (0.14) 7.61 (0.12) Δ Survival (0.21) (0.21) -0.22(0.18) R= recipient; = difference in recipient and graft a Data presented as mean (standard error) White boxes represent non-significant difference in area between recipient and graft survival curves p>0.05; Red boxes represent donor and recipient age strata where the life expectancy of the recipient was greater than that of the donor kidny (p<0.05); Blue boxes represent donor and recipient age strata where the life expectancy of the donor kidney was greater than that of the recipient (p<0.05) Table 5.4. The summative utility of two allocation systems Actual allocation Rank matching allocation Column difference Organ waste 1975 years 1480 years 495 years (negative inefficiency) (4m/person) (3m/person) (1m/ person) Return to dialysis 3291 years 2457 years 833 years (positive inefficiency) (7m/ person) (5m/ person) (2m/person) Row totals 5266 years 3937 years 1328 y (11m per person) (8m per person) (3 m per person) 88

105 6 Defining equitable-based utility cut-points for donorcandidate age matching 6.1 Introduction Deceased donor kidney and candidate ages The supply of deceased donor kidneys is limited, thus decisions need to be made about which patients to prioritize for transplantation, and which deceased donor kidneys to allocate to which candidates. Donor kidney age and candidate age have been suggested as components for allocation in different countries with varying utilization. Children treated with chronic dialysis have impaired growth and neurological development, and are therefore, prioritized for transplantation in most countries (i.e. greater medical need for treatment with transplantation). Based on the principle that increased (but potentially unequal) access to treatment in patients with greater need is equitable, maximizing access to transplantation in pediatric patients can be considered just. In addition, because deceased donor kidney graft function is not maintained indefinitely, allocating the highest quality deceased donor kidneys (i.e. kidneys from younger deceased donors with the longest expected survival) to pediatric candidates increases utility because it reduces the duration of time that these patients will need a different form of ESRD treatment (i.e. repeat transplantation or dialysis) after failure of their first transplant. Similarly, young adults have the longest life expectancies and prioritizing these young adult ESRD patients for transplantation with young donor kidneys would maximize their survival with a functioning transplant, and reduce or eliminate the potential wasted graft 89

106 function that occurs when these higher quality kidneys are transplanted into older candidates with shorter expected survival. However, because young adults are not subject to the same developmental problems experienced by children, the same equity considerations do not apply. Preferential allocation of young donor kidneys to young adult candidates may therefore be inequitable as a policy in isolation. In order to maintain equitable access to deceased donor kidney transplantation among all adult candidates, additional allocation constraints are essential to ensure that efforts to match organ survival with patient survival do not decrease elderly access to transplantation. Specifically, equity would be preserved by an organ allocation policy that increases the utility from the available organ supply by minimizing differences in patient and donor kidney age, without altering the age distribution of the population transplanted. One possibility to offset the advantaged access to young donor kidneys for young adult candidates is to prioritize older candidates for another group of deceased donor kidneys. These kidneys are expanded criteria donors (ECD) kidneys. An ECD kidney is a physiologically marginal kidney, differentiated by an increased risk of graft failure. The widely used American definition of an ECD kidney includes donor kidneys aged 60 years or aged years with specified comorbidities. 69 There is no equivalent Canadian definition for deceased donor kidneys at increased risk of graft failure, and ECD kidneys in Canada are generally defined as kidneys from deceased donors aged 60 years. 90

107 Deceased donor kidneys are usually transplanted if they are expected to increase the candidate s survival relative to remaining on dialysis. Despite providing shorter expected graft survival, transplantation with older deceased donor kidneys (i.e. ECD kidneys) provides a survival advantage relative to remaining on dialysis for older and less healthy candidates, as well as candidates living in regions with long expected waiting times for transplantation. 16 However, older deceased donor kidneys provide reduced long-term graft and patient survival for all candidates relative to transplantation with younger deceased donor kidneys. 70 For example, although a 60-year old kidney is expected to increase the life expectancy of a 65-year old candidate compared to remaining on dialysis, these kidneys may be associated with a shorter life expectancy for the same candidate compared to transplantation with a 40-year old donor kidney. Transplantation of these lower quality kidneys may in part be justified if they can be transplanted more rapidly than younger donor kidneys because the risks of continued dialysis exposure while waiting for a transplant are particularly high in older candidates (i.e. candidates may die waiting for a higher quality kidney, or their condition may deteriorate rendering them ineligible for transplantation). For example, transplantation of a 60-year old donor kidney to a 50-year old candidate with diabetes may be justified when the waiting time is less than the waiting time for a 40-year old donor kidney because the risk of death on dialysis in older ESRD patients with diabetes is high. However, if the 60-year old donor kidney was transplanted under a circumstance with equal or similar waiting time to the 40-year old kidney, the candidate may be done an injustice, even though their life expectancy was still increased compared to treatment with dialysis. This injustice is 91

108 currently accepted for some patient groups because there simply are not enough young donor kidneys for all candidates Considerations for allocation by donor and candidate age matching In age matching, young donor kidneys are allocated to young candidates to provide increased survival with graft function. However, older donor kidneys may not provide adequate graft survival to non-young candidates. Therefore, a criticism of donorcandidate age matching is that it will advantage younger candidates, but adversely affect older candidates by reducing their access to higher quality kidneys. 104 Importantly, in the Eurotransplant Senior Program deceased donor kidneys aged 65 years are preferentially allocated to candidates aged 65 years, and these older kidneys have been used safely to increase transplantation in elderly candidates. 76 These successes in Europe challenge the notion that old to old age matching necessarily leads to inferior outcomes in older candidates. Despite this information, there is a reluctance to use older kidneys, particularly in the United States where older kidneys are discarded at a high rate (i.e.40%). 81 Importantly, it is not clear if the lower post-transplantation life expectancy that is associated with older donor age is due to the age of the kidney alone, or perhaps attributable to unmeasured confounding by indication. For example, candidates with poorer expected outcomes (e.g. high comorbid burden or older age) may have a higher likelihood of being transplanted with lower quality (older) kidneys

109 Finally, the impact of donor-candidate age matching on decreased access to transplantation in elderly candidates requires consideration. For example, age matching proposed in the American system increases recipient and graft survival, but does so at the expense of transplanting fewer candidates aged > 60 years Restricting access of older patients to young donors kidneys may be problematic if similar limits on access to younger recipients are not put into place. For example, the failure to restrict younger candidate access to ECD kidneys, despite that fact that ECDs will likely not provide younger candidates with a sufficient duration of graft function to obviate the need to return to dialysis after ECD failure and/or repeat transplantation. Reese and Caplan 104 suggest that despite decreased access to elderly transplantation, the proposed allocation is not inequitable by the fair innings argument. The fair innings argument as applied in this context maintains that young people who develop ESRD are worse off than older patients developing ESRD because they have experienced fewer healthy life years (i.e. they deserve a chance to be old, whereas older patients have already had the chance to be 103, 106 young). The implications of donor-candidate age matching depend on how strict the matching criteria are. The strictest matching might fix a minimal difference between donor kidney age and candidate age of one, five or ten years, compared to a looser or more fluid policy of age matching that is proportional by donor and candidate ages. The choice of age matching strategy may accentuate or diminish inequities in access to transplantation by age. For example, a proportional age-matching strategy that allocates by decile (i.e. the youngest 10% of donor kidneys to the youngest 10% of candidates, up to the oldest 10% 93

110 of donor kidneys to the oldest 10% of candidates) would eliminate inequity in access to transplantation by candidate age, and would inherently account for any changes in donor and candidate ages that occur over time. In addition, proportional age-matching as described would be equitable even in the presence of current differences in donor kidney and candidate ages. In contrast, an allocation strategy that requires matching of donor kidneys and candidates within one year of age, would result in a recipient population with an age distribution similar to the donor kidney age distribution, but not necessarily the same as the candidate age distribution. In the current environment, donor kidneys are younger than candidates and therefore a stricter age-matching allocation strategy would disadvantage older candidates. Existing differences in international clinical practice surrounding discard of older donor kidneys may make the policy of age matching more attractive in different countries. For example, in Eurotransplant the discard rate of older donor kidneys is lower than in the United States, and this may be an accepted clinical practice because the intention is that these older donor kidneys will be used to increase transplantation in older patients. In an organ poor environment, the goal of donor-candidate age matching is to increase utility by reducing organ waste and increasing recipient survival with a functioning graft. In order to maintain equity in access to transplantation across adult age subgroups while age matching, the limited number of young donors preferentially allocated to young adult candidates should be offset by the additional and expeditious transplantation of older adult candidates with kidneys from older donors. Ideally, the addition of donor-candidate 94

111 age matching in allocation algorithms would improve recipient outcomes, without changing which patients are transplanted The Canadian perspective In Canada, healthcare is administered provincially; although there is general agreement on some principles of organ allocation, there is still variation in allocation policies across provinces. In 2006, the Canadian Council for Donation and Transplantation (CCDT) convened a forum to develop recommendations for the allocation of deceased donor kidneys that could be incorporated in each province. An underlying premise of the forum was that organs be allocated fairly, considering both equitable access and optimal outcomes for transplantation. 107 As part of these recommendations, deceased donor age and candidate age were considered as key determinants of allocation policy; and agespecific recommendations for kidneys from deceased donors aged < 60 years (Canadian definition of standard criteria donor) and deceased donor aged 60 years (Canadian definition of expanded criteria donor) were provided. 107 The CCDT age specific recommendations for the allocation of standard criteria donor kidneys included prioritization for: 1) pediatric candidates (high priority), 2) young standard criteria donors to pediatric candidates (high priority), and 3) young standard criteria donors (young donors) to young adult candidates (medium priority). 107 The CCDT recommended that ECD kidneys should be allocated by waiting time, preferentially to candidates aged 60 years or candidates aged < 60 years with significant comorbidity

112 In Canada, the allocation of deceased donor kidneys across provinces is aligned with the principles put forth by the CCDT recommendations, but the constructs of young donor kidney and young adult recipient are not explicitly defined, and the age cut-point for expanded criteria donors is empirical. Therefore, there is no uniformly accepted formula for donor-candidate age matching across Canada. As such, the objective of this paper is to qualify and quantify equitable utility-based definitions for age cut-points for the allocation of deceased donor kidneys to transplant candidates using age matching in Canada. 6.2 Definitions Nominal definitions of young and old The designation of a candidate as young or old can be practically dictated by the age (i.e. life expectancy) of the donor kidney. When the life expectancy of the candidate is longer than that of the donor kidney, the candidate is considered young; and when the life expectancy of the candidate is shorter than that of the donor kidney the candidate is considered old. For example, a candidate aged 50 years with a life expectancy of fifteen years may be designated old in relation to a donor kidney aged 18 years with a life expectancy of twenty years, or young relative to a donor kidney aged 55 years with a life expectancy of ten years. Similarly, whether a donor is considered young or old depends on the age of the intended recipient. For example, a donor kidney aged 50 years with a hypothetical life expectancy of ten years could be young in relation to a candidate aged 70 years with a life 96

113 expectancy of five years, or old in relation to a candidate aged 25 years with a life expectancy of twenty years. It is possible that some donor kidneys may always have a shorter life expectancy relative to any aged candidate and may always be considered old. In practice, if the life expectancy of donor kidneys changed, a previously defined young donor, could be considered old, or vice versa Operational definitions of young and old Young donor kidneys and candidates The number of deceased donor kidneys available for transplantation is greatly exceeded by the number of patients who would benefit from transplantation (candidates). Therefore, the total life expectancy of candidates is greater than that provided by the pool of deceased donor kidneys available for transplantation. There is a group of adult transplant candidates who will always be considered young because their life expectancy is greater than that provided by even the highest quality kidney (i.e. youngest adult donor kidneys). These candidates are young adult candidates. In the absence of a living donor, the best treatment for young adult candidates is transplantation with the youngest adult deceased donor kidneys (i.e. young donor kidneys). In order to maintain equity for older candidates, a limit on the number of young donor kidneys preferentially allocated to young recipients is required. The maximum donor age cut-point for young adult candidates should be informed by the supply of ECD kidneys that can be preferentially allocated to older candidates. As previously stated, older donor kidneys are associated with shorter patient and graft life 97

114 expectancy relative to younger donor kidneys, and transplantation of these ECD kidneys may only be justified in older candidates in the context of rapid transplantation with minimal dialysis exposure. If the number of young donor kidneys allocated to young adult candidates is too large, the number of older donor kidneys remaining for non-young candidates is reduced, and the waiting time for transplantation of older candidates is increased, limiting the benefit of transplantation from ECD kidneys. Therefore, the upper age cut-point for young donor kidneys should be defined by the available supply of ECD kidneys. Importantly, the time that donor kidneys of different ages become available is not known in advance. In practice, balancing the number of young donor kidneys preferentially allocated to young candidates, with the number of ECD kidneys preferentially allocated to older candidates, is based on the assumption that there is no difference in the time of availability for younger and older donor kidneys. If the discrepancy in the supply of young donor kidneys and ECD kidneys changes over time, one candidate age group may have to wait longer for kidney transplantation. Therefore, any implemented age-matching policy will require ongoing evaluation to assess the impact of variations in donor kidney availability by age Older donor kidneys and candidates The definition of an ECD kidney was operationalized by a donor age above which the deceased donor kidney does not provide a lifetime of graft function in younger adult 98

115 candidates, but does provide a lifetime of graft function in an older group of candidates (older candidates). That is, an older candidate is defined by a candidate age above which an ECD kidney provides an expected lifetime of graft function. In addition, an ECD kidney should not decrease the expected life span of the older candidate Summary The designation of a candidate/donor as young or old is in reference to the age of the respective donor/candidate. To determine cut-points to define young and old candidates and donors both the utility (i.e. the years of graft function that a given aged donor kidney will provide a given aged candidate) and equity (i.e. the allocation of donor kidneys fairly to candidates of different age groups) of age matching should be considered. The following section will outline the study method used to quantify the utility and equity components of age matching in Canada Quantifying ages of donors and candidates for age matching Utility component of donor and candidate age cut-points The following analyses were performed with Canadian recipients captured in the CORR database, as described in Chapter 5. Pediatric candidates receive the highest priority for deceased donor kidney allocation, and are thus excluded from the utility components of the definitions below. 99

116 To calculate the difference in donor and candidate life expectancy, we used the areabetween-the-curve methodology from Chapter 5. Using the operational definition for young adult candidate above, we determined the age cut-point for a young adult candidate from the results in Table 5.3. That is, a young adult candidate was determined to be years of age; an age range where recipients were expected to live as long or longer than deceased donor kidneys of all ages. Based on this definition of a young adult candidate, we determined a maximum utility age cut-point for a young donor by calculating the time back on dialysis (i.e. the area between patient and graft survival curves) for young adult recipients with deceased donor kidneys of different ages (i.e. donor age <35 years, years, years, years, years, years and 60 years). Among candidates who do not meet the criteria to be considered young (i.e. age 55 years), the subset of oldest patients (with the shortest life expectancy) will derive a lifetime of transplant function even when transplanted with a kidney with the lowest expected post-transplant survival (i.e. ECD). Finally, a third intermediate group of patients who are considered old with respect to young donor kidneys, but are considered young relative to ECD kidneys can be defined. In order to determine the age cut-point below which candidates will be considered young or old in relation to ECD kidneys, we divided older recipients into two age groups: and 60 years. (The number of age groups here was restricted due to constraints with sample size). Using the methods from Chapter 5, we calculated the area between the 100

117 unadjusted patient and graft survival curves for these recipient age groups with donor age cut-points (possible ECD kidney age threshold) of years and 60 years. Among candidates who we determined would always be considered old relative to deceased donor kidneys of any age (i.e. older candidates), we determined the degree to which patient survival was negatively impacted by older donor age in a multivariate Cox proportional hazards regression model adjusted for the following covariates: recipient sex, cause of ESRD, peak panel reactive antibody titre, duration of dialysis prior to transplantation, and year of transplantation Equity component of donor and candidate age cut-points In the previous section we outlined methods to determine a cut-point for ECD kidneys that would provide the oldest candidates with a lifetime of graft function, without reducing patient survival. To determine an acceptable age cut-point for young donors we examined the distribution of deceased donor kidney and recipient ages from transplants recorded in the last three years of available CORR data ( ). Specifically, we first determined the number of deceased donor kidneys that met our utility definition for ECD kidneys and could reasonably be prioritized for transplantation in older candidates. We then used this number to determine the maximum number and age limit for young donor kidneys that could be prioritized for transplantation in young recipients. This approach to defining the 101

118 limits of prioritization of young recipients based on the availability of suitable organs to be prioritized for older recipients may be considered an equity-based ceiling. Finally, we determined any longitudinal change in the supply of deceased donor kidneys and the age of recipients. This was done to inform the potential stability of the age cutpoint selected over time Recommendations and hypothetical redistribution of deceased donor kidneys to recipients The utility and equity analyses above were combined to put forth recommendations of age cut-points for donor and candidate age matching. Using the ages of deceased donor kidneys and recipients of actual transplanted donor-recipient pairs from in the data set, we performed a hypothetical redistribution of these deceased donor kidneys to the same group of recipients based on the age matching recommendations in this study, specifically: deceased donor kidneys aged <35 years to candidates aged 19 to 54 years; deceased donor kidneys aged 55 years to candidates aged 60 years; and deceased donor kidneys aged years to all candidates on the list (see Section 6.3.2). The redistribution did not consider any discrepancy in age distribution that may exist between transplanted recipients and candidates on the waiting list. All analyses were performed using SAS 9.4, Carey, N.C. 102

119 6.3 Results Donor-candidate age matching cut-points Young adult candidate age cut-point Using the operational definition for young adult candidate above and the results from Table 5.3, the age range for a young adult candidate was defined as years; 55 years is the age below which young adult recipients lived as long, or longer, than deceased donor kidneys of all ages ECD kidney and older candidate age cut-points Donor age cut-points < 50 years were not considered for older donors because these donors kidneys survived as long or longer than older adult recipients (Table 5.3). When non-young adult recipients were further stratified by age, donors aged 55 years did not provide adequate graft survival to recipients aged years, but did provide adequate graft survival for recipients aged 60 years (Table 6.1). In addition, recipients aged 60 years had similar patient survival when transplanted with deceased donor kidneys aged years and aged 60 years [HR (95% CI): 0.85 (0.60, 1.21) reference = donor age 60 years]. Therefore, donors aged 55 years meet the utility definition for ECD kidneys, when transplanted into recipients aged 60 years (older candidates). Table 6.2 summarizes the designation of candidates as young or old relative to deceased donor kidneys of different ages. 103

120 Young donor age cut-point Donors aged < 45 years offer superior graft survival for recipients aged < 55 years (Table 6.1). Therefore, from a utility perspective the cut-point for young donor age could be any age less than 45 years. To maintain equity in access to transplantation by candidate age, we must offset the number of young donors prioritized for young, by the number of mature donors prioritized for mature recipients. From , there were N=462 kidneys aged 55 (29%), and this matched the number of deceased donor kidneys that were 33 years and younger. Therefore, to maintain equity in distribution of deceased donor kidneys, and given that kidney aged < 35 are also prioritized for pediatric candidates we selected a young donor age cut-point of 35 years which is consistent for our utility requirement that the young donor age be less than 45 years. Importantly, the age distribution of deceased donor kidneys was not constant by study year (p<0.0001) (Figure 6.1) Recommendations and hypothetical redistribution of deceased donor kidneys to recipients The results of this study were presented as recommendations for allocation using age matching to the Canadian National Kidney Working group, a group comprised of transplant nephrologists and other health care workers directly involved in managing deceased donor kidney transplantation and informing deceased donor kidney allocation. There were no new recommendations presented with respect to pediatric transplant candidates, but consensus is that these candidates continue to be prioritized for transplantation with deceased donor kidneys aged < 35 years. The following age 104

121 matching cut-point recommendations for allocation of deceased donor kidneys to adult candidates were presented: 1. Deceased donor kidneys aged <35 years should be allocated to candidates aged 19 to 54 years. 2. Deceased donor kidneys aged 55 years should be allocated to candidates aged 60 years. 3. Deceased donor kidneys aged years should be allocated irrespective of age to all candidates on the list. Within each of these age recommendations, candidates continue to be selected for transplantation based on previous Canadian Council for Donation and Transplantation recommendations (i.e. biologic compatibility, sensitization, waiting time) Redistribution: a hypothetical example Using the ages of deceased donor kidneys and recipients of actual transplanted donorrecipient pairs from in the data, we performed a hypothetical redistribution of these kidneys to the same group of recipients under the age matching recommendations put forth above. That is, deceased donor kidneys aged < 35 years were first allocated to pediatric recipients, then to recipients aged 19-54, and deceased donor kidneys aged 55 years were allocated to recipients aged 60 years. The remainder of kidneys were divided between the remaining recipients in all groups (in practice these donor kidneys would be allocated to candidates with the same blood type, with the first offer to the candidate with the highest allocation points (e.g. longest waiting time)). This example ignored the time sensitive nature of the availability of deceased donor kidneys, and assumed all transplantation could occur at the same time. Following the recommendations, the redistribution of deceased donor kidneys to recipients using the defined age matching strategy resulted in a reduction of donor-recipient age mismatch, and therefore 105

122 improvements in allocation efficiency, without altering the transplant candidate population (Figure 6.2). 6.4 Interpretation This study presents equitable utility-based nominal and operational age definitions for young adult candidates, young donors, ECD donors and older candidates; and puts forth specific age recommendations for the use of donor/candidate age matching in the allocation of deceased donor kidneys. The current system proposed in the United States 35 (detailed in Chapter 4), increases allocation utility similar to this chapter s proposed age matching strategy by eliminating the extreme age mismatch that occurs when the highest quality kidneys are allocated to the lowest quality candidates. The US proposal does this by offering the top 20% quality kidneys to the 20% youngest and healthiest candidates. In contrast to the study proposal, the American model does not address allocation at the other extreme (i.e. young, healthy candidates have the option of accepting ECD kidneys in the United States). Unfortunately, the increase in utility in the United States comes at the expense of fewer patients aged > 60 years receiving deceased donor transplantation, 35 whereas in Canada the recommendations from this chapter are aimed at maintaining distributive justice by not changing which candidates get transplanted by age (i.e. the proportion of each candidate age group that is transplanted should remain constant), but only redistributing the same kidneys to similarly aged people more efficiently (i.e. by offsetting the number 106

123 of young donor kidneys that are prioritized to the young, with a similar number of older donor kidneys prioritized for older patients). This analysis has several limitations. First, this analysis was done at the population level, assuming that the age distribution of deceased donors and candidates is stable over time. However, the data show that donor age is increasing over time and therefore, the age cutpoints recommended here should be re-evaluated every few years as the ages of the donor and candidate populations changes. In addition, CORR does not provide information on the age distribution of wait-list candidates; therefore, we cannot determine if the age distribution of transplant candidates is changing over time, nor can we assess the risk of candidate death on the waiting list to inform our allocation. However, given the increasing age of the end-stage renal disease population (i.e. population on chronic dialysis or transplantation) 3 and the increasing age of transplant recipients over time in this study, we suspect that candidate age is also increasing over time and a re-evaluation of the recommendations will be necessary to ensure that the proportion of candidates from each age group being transplanted is consistent over time, not simply that the age of transplant recipients is remaining constant. The prioritization of young donors to young candidates and ECD donors to older recipients will increase utility in these groups, but it is unclear what the impact will be on candidates aged years who aren t directly prioritized. Second, the analysis was based on the assumption that different aged kidneys would be available to the prevalent candidate population at the same time. Unfortunately, the 107

124 availability of deceased donor kidneys is unpredictable, and the implementation of these recommendations may change the waiting time for individual patients (i.e. either more quickly or more slowly), as well as the selection of individuals that will be offered transplantation. These recommendations are meant to be implemented at the population level, and maintain equity as such. Importantly, the implications of small additional changes to waiting times are not known, although increased dialysis time portends to worse outcomes. 108 In Canada, it has been shown that there are no changes in outcomes with dialysis time up to four years, 109 and thus if the changes are small, the detriment may be negligible. The greatest harm of additional waiting time is among individuals with poor access to transplantation due to biologic or geographic factors, or older patients whose life expectancy is limited. In contrast, for younger candidates the benefit of receiving a younger kidney that would provide them with more years of graft function and prevent them from returning to dialysis to compete for another transplant would most likely outweigh a small change in waiting time. Third, the current recommendations are based on national data, but uptake of recommendations would occur regionally. Although, the given recommendations might be successful in large regions (e.g. Ontario) where the age distribution of donors and candidates may be closer to the study means, and time between deceased donors becoming available will be shorter, in smaller regions (e.g. Manitoba) the age distribution of donors and candidates may differ and deceased donation may be less frequent. Therefore, the specific allocation guidelines for age matching may require regional modification. 108

125 The acceptance and uptake of age matching allocation policies for deceased donation may differ internationally. For example, Eurotransplant aggressively uses older donor kidneys in order to transplant more older candidates, whereas in the United States, the reluctance to use older kidneys (i.e. high discard of ECD kidneys) may make an age matching proposal less attractive because there are fewer older donors available. In addition, the legal acceptability of age matching allocation may differ by region, as exemplified by the use of direct age matching in ESP in Europe, but the removal of direct age matching from the new American kidney allocation system. Based on our relative definitions of young and old, the number of young donor and ECD kidneys available for transplantation may depend on the case-mix of candidates in different regions. In certain countries where dialysis outcomes are poor or dialysis is less accessible (e.g. India where patients pay for dialysis treatment), it may not be possible for candidates to wait for deceased donor kidneys to whom they are age matched. In countries such as India, living donation may be the most reliable treatment option. In conclusion, including donor-candidate age matching in the allocation of deceased donor kidneys will increase utility, and the recommendations put forth in this chapter should do so while maintaining equitable access across candidate ages. These recommendations need to be reviewed every few years as the age of the donor and candidate populations change, and the implication of altering waiting times in individual candidates needs to be explored. Importantly, the implications of implementing these 109

126 recommendations may vary regionally, and individual regions need to take care when interpreting these recommendations within their populations. 110

127 Age Age Figure 6.1 The distribution of Canadian recipient age and deceased donor kidney age by year of transplantation. Recipient mean age (std): 44 (14.7) in 1995, 53 (15.1) in 2010; Donor kidney mean age (std): 36 (16.9) in 1995, 43 (17.0) in P-value for trends in recipient age and deceased donor kidney age over time p< Recipient Age Distribution Year of transplantation Donor Age Distribution Year of transplantation 111

128 Figure 6.2 The distribution of deceased donor kidneys to recipients by age. Top panel shows how deceased donor kidneys from were allocated in practice to recipients. Bottom panel shows the redistribution of the same aged deceased donor kidneys to the same aged recipients following the recommendations for donor-candidate age matching in this chapter. 112

129 Table 6.1. Area under and between recipient survival and graft survival curves Donor age in years Young recipient age years Non-young recipient 55 years Recipient age years Recipient aged 60 years < (0.06) a 8.69 (0.10) 0.55 (0.12) 8.19 (0.12) 9.20 (0.08) (0.14) (0.14) 8.55 (0.21) 0.68 (0.25) 8.00 (0.26) 8.97 (0.18) (0.32) (0.11) 8.51 (0.15) 0.65 (0.19) 8.16 (0.19) 8.84 (0.14) (0.24) (0.10) 8.27 (0.18) 0.85 (0.21) 7.58 (0.22) 8.89 (0.16) (0.27) (0.10) 7.85 (0.17) 1.31 (0.20) 7.86 (0.18) 8.66 (0.17) (0.25) (0.13) 7.55 (0.15) 1.47 (0.20) 7.83 (0.20) 8.49 (0.20) (0.28) 8.29 (0.27) 7.96 (0.37) 0.33 (0.52) 7.55 (0.26) 8.78 (0.24) (0.35) (0.16) 7.22 (0.18) 1.62 (0.24) 7.63 (0.16) 7.88 (0.22) (0.27) 8.78 (0.19) 8.40 (0.28) 0.38 (0.34) 6.86 (0.20) 7.71 (0.23) (0.30) a Data presented as mean (standard error) White boxes represent non-significant difference in area between recipient and graft survival curves p>0.05; Red boxes represent donor and recipient age strata where the life expectancy of the recipient was greater than that of the donor kidny (p<0.05); Blue boxes represent donor and recipient age strata where the life expectancy of the donor kidney was greater than that of the recipient (p<0.05) 113

130 Table 6.2. The designation of transplant candidates as young or old relative to the age of the deceased donor kidney. Deceased donor kidney age Candidate age 35 years years 55 years years Young Young Young years Old Old Young 60 years Old Old Old 114

131 7 Lifetime of allograft function a new metric to inform the optimal use of expanded criteria donor kidney transplantation 7.1 Introduction The number of patients awaiting kidney transplantation in the United States (U.S.) recently eclipsed Despite these staggering numbers, a high number of deceased donor kidneys from older aged donors are discarded. For example in 2013, 46% of the N = kidneys recovered from deceased donors aged 65 years were not transplanted. 88 This may in part be related to the inclusion of transplant outcomes in hospital accreditations in the U.S. In the new U.S. kidney allocation scheme, in order to expedite placement and encourage utilization of expanded criteria donor (ECD) kidneys (i.e. donor kidneys with kidney donor profile index >85%), local allocation (allocation within an organ procurement organization (OPO)) will be by-passed and ECD kidneys will be offered at the regional level (allocation within nearby OPOs). 110, 111 However, despite evidence that only certain patient groups will benefit from transplantation of ECD kidneys, 16 and that patients who will not benefit continue to be listed and transplanted with ECD kidneys 112, the new allocation scheme does not restrict which wait-list candidates can receive ECD kidneys. This policy is in contrast to the European Senior Program (ESP) which restricts the transplantation of kidneys from deceased donors 65 years to recipients 65 years, 115

132 while still allowing patients 65 years the option of waiting for a kidney from a younger deceased donor. Previous work has shown that the discard of ECD kidneys in ESP is significantly lower than that in the United States. 113 This chapter reports a series of analyses to inform the optimal use (i.e. allocation) of ECD kidneys. The first set of analyses compared the use and outcomes of ECD kidneys in two transplant systems: ESP and the U.S. The purpose of these analyses was to demonstrate the impact of different policies on the utilization and outcomes of ECD kidney transplantation. The second set of analyses examined the consequences of continuing to allow any consenting wait-list candidate to undergo transplantation with an ECD kidney in the U.S. Under the new U.S. kidney allocation system, ECDs will be defined by a Kidney Donor Profile Index (KDPI) greater than 85%. The KDPI is a linear scale from % that transforms the relative risk of graft loss for any deceased donor kidney compared to that of a kidney from a donor aged 40 years, with 0% representing the longest projected survival and 100% representing the shortest survival. The KDPI is calculated based on donor age, height, ethnicity, history of hypertension, diabetes, cause of death, serum creatinine, hepatitis C status and donation after circulatory death status. 114 Given that KDPI is not utilized in the ESP, and because virtually all (>95%) deceased donors 65 years would be classified as ECD in the United States, we defined ECD based on deceased donor age 65 years to maintain consistency in all analyses. 116

133 7.2 Methods Comparison of the use and outcomes of ECD kidney transplantation in the Eurotransplant Senior Program and the United States Study population and data sources The study population included recipients of a first, kidney-only transplant from a deceased donor 65 years of age captured in Eurotransplant or the United States Renal Data System (USRDS). A significant issue limiting international comparisons of transplant outcomes (i.e. recipient death and graft failure) is lack of validated outcome assessment. 115 To overcome this limitation, we restricted the analysis to patients transplanted between January 1, 1999 to December 31, 2003 with follow up through April 30, 2005 because of the availability of rigorous outcome assessment in this cohort of Eurotransplant patients as part of a clinical study. 76 Transplant outcomes in the USRDS cohort are routinely validated Statistical analyses Donor and recipient characteristics were described using the mean ± standard deviation or median (and quartiles) for continuous variables, or frequencies and proportions for categorical variables; group differences were compared using the t-test, Kruskal Wallis, or Chi-square test as appropriate. Among the subset of recipients aged 65 years, we determined the time to all cause graft failure (i.e. patient death or graft failure), time to graft failure (i.e. censored at patient death), and time to death with a functioning graft (i.e. patient death censored at graft loss) 117

134 in ESP and U.S. patients using the Kaplan-Meier product limit method and compared group differences using the log-rank test. Separate Cox multivariate proportional hazards regression models were used to determine the relative hazard of graft loss from any cause, death censored graft failure, death with a functioning allograft, in ESP compared to U.S. patients after adjustment for differences in: donor characteristics (age, sex, history of diabetes, history of hypertension, cause of death (cerebrovascular accident versus other), recipient characteristics (age, sex, cause of end-stage renal disease, body mass index, duration of pre-transplant dialysis exposure, peak panel reactive antibody titre (measure of sensitization) and transplant characteristics (cold ischemic time, use of induction therapy (depleting antibody, non-depleting antibody, none), type of calcineurin inhibitor (tacrolimus, cyclosporine, sirolimus), and the use of mycophenolate mofetil or azathioprine). The proportional hazards assumption was tested using log-negative-log plots of the within group survivorship probabilities versus log-time in all models. The area under survival curves represents the average (or mean) survival. 98 We calculated the mean (standard error) patient survival and graft survival for ESP and the U.S. at five years using the Kaplan-Meier product limit method (methods outlined in Chapter 5). The difference (standard error) in mean patient survival and graft survival (i.e. area between survival curves) was then calculated to quantify the time back on dialysis (when patient survival exceeded graft survival) or potential lost graft function (when graft survival exceeded patient survival). 118

135 7.2.2 Analysis of recipient outcomes after ECD kidney transplantation in the United States Statistical analyses These analyses included recipients of a first, kidney-only transplantation from a deceased donor aged 65 years (ECD) captured in the USRDS between January 1, 1995 and December 31, 2010 with follow up through October 31, We first determined the distribution of ECD transplants as a function of recipient age with recipient age categorized as follows: years, years, years, years, years, and 70 years. In each of the recipient age groups, we calculated the mean patient survival and graft survival (i.e. area under the Kaplan-Meier survival curves), as well as the difference in these curves at ten years after transplantation using the methods described in Section To determine the extent to which the observed differences between patient survival and graft survival among recipients 60 years (i.e. age at which ECD kidneys outlive recipients in this cohort) was impacted by increased death after transplantation with an ECD kidney, we compared the patient survival of recipients aged 60 years transplanted with an ECD kidney, with that of similar aged recipients who received a kidney from a deceased donor < 65 years with a KDPI 60-69%, 70-79%, 80-85% and >85%, during the same time period using separate Cox multivariate regression analyses adjusted for recipient factors (sex, race, cause of ESRD, peak panel reactive antibody titre, body mass index, primary insurer, comorbidities (inability to ambulate, chronic obstructive 119

136 pulmonary disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, cancer and ischemic heart disease)), transplant factors (HLA mismatch, cold ischemic time) and delayed graft function. Among patients aged < 50 years who received an ECD kidney from a donor 65 years and suffered death censored allograft failure, we determined the proportion that were relisted for transplantation, their level of sensitization (measured by peak panel reactive antibody titre) at the time of repeat wait-listing, and the proportion subsequently retransplanted with a either a deceased or living donor. 7.3 Results Comparison of the use and outcomes of ECD kidney transplantation in the Eurotransplant Senior Program and the United States During the period the number of deceased donor kidney transplants from donors aged 65 years in ESP was 1 870/ (12%) compared to only 1 312/ (3%) in the entire U.S.. ESP recipients received kidneys from donors that were older, and more frequently had a history of hypertension compared to U.S. recipients (Table 7.1). More than 80% of ESP recipients were 65 years compared to only 34% of U.S. recipients. ESP recipients were less likely to have diabetic ESRD or be obese compared to U.S. recipients, but had a longer exposure to dialysis prior to transplantation. Because ESP excludes sensitized patients, 76 97% of ESP patients were not sensitized (i.e. peak panel reactive antibody < 5%). Compared to ESP recipients, U.S. recipients were treated nearly twice as frequently with depleting antibody induction therapy and tacrolimus, and 120

137 only half as much with azathioprine. The incidence of delayed graft function (ESP: 32%; U.S. 34%) and primary non-function (ESP: 7%; U.S. 6%) were similar in ESP and U.S. recipients. Similar differences were observed among the subset of recipients aged 65 years (Table 7.2). Figure 7.1 shows Kaplan-Meier plots comparing transplant outcomes among the N = 1520 ESP and N=446 U.S. recipients who were 65 years of age at the time of kidney transplantation from a deceased donor 65 years. The lower all cause graft survival in ESP recipients was due to a higher incidence of death with a functioning graft in U.S. recipients. Table 7.3 shows the results of separate Cox multivariate regression analyses for the outcomes of graft loss from any cause (including death), graft failure, and death with a functioning graft (patient death censored at graft failure). Among transplant recipients 65 years of age, there was a lower risk of graft loss from any cause, and death with a functioning graft in ESP compared to U.S. recipients, but the risk of death censored graft loss was similar in both groups. Table 7.4 quantifies the difference in mean five-year patient survival and graft survival in ESP and U.S recipients aged 65 years. ESP recipients were more likely to outlive their kidneys requiring them to return to dialysis for an average of 5.2 months over the fiveyear follow up period. Among U.S. recipients, graft survival exceeded patient survival by 121

138 5.0 months, indicating that elderly U.S. recipients 65 years die with a functioning allograft after ECD transplantation Analysis of recipient outcomes after ECD kidney transplantation in the United States. Figure 7.2 shows the age distribution of N= U.S. kidney transplant recipients from a deceased donor 65 years during the period January 1, 1995 December 31, Figure 7.3 quantifies the difference (in months) between the average patient survival and average graft survival among different recipient age groups at 10 years after transplantation with a deceased donor 65 years. Among patients and years, the average patient survival exceeded the average death censored allograft survival. As a result year old recipients returned to dialysis for an average (SE) of 32 (4) months, while year old patients returned to dialysis for an average (SE) of 21 (3) months in the ten year time period after ECD transplantation. In contrast, among recipients aged years the average patient survival (85 months) and graft survival (83 months) over a ten period were nearly equivalent, while among recipients aged 60 years patient survival was lower than death censored allograft survival, indicating that on average ECD transplantation provided a lifetime of function for patients 60 years. Table 7.5 shows patient survival among recipients aged 60-64,65-69,and 70 years who received a kidney from a deceased donor < 65 years with a KDPI 60-69%, 70-79%, 80-85% and >86%, or a deceased donor aged 65 years. The unadjusted decrement in ten- 122

139 year mean patient survival with transplantation from a donor 65 years compared to a donor < 65 years with KDPI 60-69% was of 8.2 months, 7.3 months and 7.3 months respectively in recipients aged years, years and 70 years Disposition of recipients <50 year, with a failed ECD transplant There were N=726 (14%) recipients age 50 years who underwent transplantation from a donor aged 65 years in the U.S. from Of these, N=408 (56%) suffered graft failure with a median time of 3.1 months (q1-q3: ), and were forced to return to dialysis or undergo repeat transplantation. Among the graft failures, 196 (48%) were wait-listed with a median time of 7.7 months (q1-q3: months), and the majority of these (61%) were highly sensitized (peak panel reactive antibody > 30%). Thirty-six percent of these patients received a second transplant [N=121 received a deceased donor transplant with a median time of 25.4 months (q1-q3: months) from first graft failure; N=27 received a living donor transplant with a median time of 3.5 months (q1- q3: months) from first graft failure]. 7.4 Interpretation This study was designed to inform the expanded utilization of ECD kidneys in the United States. The international comparison with ESP patients demonstrated that more liberal use of ECD kidneys (i.e. older with a greater comorbid disease burden) in a restricted patient population 65 years provided similar graft survival to that achieved in a contemporaneous and similarly aged cohort of U.S. transplant recipients despite better 123

140 patient survival in the ESP cohort. The inferior patient survival in the U.S. could be related to unaccounted for differences in patient case-mix between European and American transplant recipients aged 65 years, or due to increased transplant related complications in the U.S. that are not manifest in ESP due to differences in clinical transplant practice. These findings suggest that international collaboration may be very useful in understanding the key determinants of adverse outcomes after ECD transplantation in elderly patients. Specific issues that should be examined include differences in candidate selection, wait-list management, organ preservation, organ allocation and early and late post transplant management that might impact patient survival. For example, it is notable that ESP allocates kidneys from donors aged 65 years to recipients aged 65 years locally or in a narrow geographic area to minimize cold ischemic time and this resulted in shorter cold times in ESP in our analysis. The new U.S. allocation policy to forgo local prioritization of ECD kidneys, may paradoxically increase cold ischemic times leading to increased adverse outcomes and resistance to accept ECD offers. The estimates of the average difference between patient and allograft survival in ESP and U.S. recipients provide insight into the extent to which ECD transplantation succeeds in providing elderly patients with a lifetime of allograft function. Over the five-year followup period, we found that on average ECD recipients in ESP would return to dialysis for a period of 5.2 months; while in U.S. recipients, death censored allograft survival exceeded patient survival by an average 5.0 months, indicating that in the U.S., ECDs provide elderly transplant recipients with an expected lifetime of allograft function. The study 124

141 estimates provide another metric confirming the clinical utility of ECD kidneys in elderly patients, and support expanded use of ECD in elderly patients 60 years. Although there is limited literature regarding what transplant outcomes would be acceptable to elderly transplant candidates, in our experience most elderly patients would hope to enjoy transplant function for the remainder of their lives. The approach of calculating the difference in patient versus graft failure used in this analysis may therefore be useful in counselling elderly patients regarding the anticipated outcomes after ECD transplantation, especially if coupled with information from previous work from Merion and colleagues demonstrating the benefit of ECD compared to continued wait-listing for a standard criteria donor kidney. 45 The second part of our study focused on determining the potential downside of continuing to allow any wait-list candidate to receive an ECD kidney in the new U.S. kidney allocation policy. These analyses demonstrated that transplantation of ECD kidneys provides patients aged < 50 years with an insufficient duration of transplant function and that these patients will have to either return to dialysis or undergo repeat transplantation. Our analysis also showed that fewer than half of these patients are relisted for transplantation, but are frequently sensitized and are unlikely to receive a repeat transplant. In contrast, among year old recipients, ECD kidney transplantation provided nearly equal patient and death censored allograft survival, while in patients 60 years death graft survival exceeded patient survival, indicating that the majority of ECD kidneys provide patients 60 years with a lifetime of allograft function. To what extent the survival of patients 60 years was shortened by transplantation with an ECD kidney 125

142 is difficult to determine in this observational study. It is, however, reassuring that we only found a 7-8 month difference in the average ten year patient survival between recipients of a transplant from a donor 65 years compared to similar aged recipients of a deceased donor kidney transplant from a donor < 65 years with KDPI 60-69%. Our findings challenge the appropriateness of continuing to allow any consenting patient to accept an ECD kidney, especially when there is evidence that patients who will not benefit from ECD continue to receive ECD transplants. Young transplant recipients have a high likelihood of returning to dialysis, and are highly sensitized when they do, making repeat transplantation unlikely. The added insult of an increase in candidates waiting for transplantation makes repeat transplantation even less likely for young failed transplant recipients. In addition, part of the benefit of ECD transplantation comes with the trade-off of shorter time on dialysis prior to transplantation. By allowing young patients to accept ECD kidneys, the benefit of ECD transplantation for all patients is reduced, and access to transplantation for older candidates becomes more inequitable. In earlier iterations of the kidney allocation policy a broad (15 year) age matching had been proposed for kidneys with a KDPI > 20%. 89, 104 However, this proposal was rejected by the U.S. Department of Justice on the grounds that the use of age alone to determine organ allocation was discriminatory. Therefore, the default policy to allow any consenting patient receive and ECD transplant was maintained. Our findings suggest that the European approach to allow elderly patients to opt out of ESP, while prohibiting the transplantation of ECD kidneys in younger candidates, who clearly will not benefit, might have been a better alternative. Although not specifically examined in our analysis, the allocation of ECD 126

143 kidneys to younger recipients also limits elderly patients from receiving these kidneys. Readers of our study should consider the inherent limitation of observational studies and that our findings may not be directly applicable to individual patients. For example, the comparison between ESP and U.S. recipients is confounded by unmeasured differences in patient case-mix, donor characteristics and dialysis and transplant services that differ between the regions. We attempted to mitigate these population differences by restricting our analyses to the same inclusion and outcomes dates, in the same aged patients. In addition, complete data to measure outcomes in ESP patients was only available until There may be variation in the acceptance criteria of donors and candidates over time in ESP, as well as improvements in ESP with familiarity of the policy, that may lead to differences in transplant outcomes. In the U.S. we examined the use of ECD kidneys in different recipient age groups, to determine whether ECD transplantation would be appropriate for all aged candidates. It is possible that confounding by indication exists in the allocation of these older donor kidneys. For example, it is possible that younger candidates who are transplanted with older kidneys are inherently sicker than their young candidate counterparts who do not receive these organs. However, even these potentially less well young candidates outlived their ECD kidneys; suggesting that the transplantation of ECD kidneys into any young candidates may be irresponsible. In summary, these analyses 1) provide robust evidence to encourage increased utilization of ECD kidneys in elderly transplant candidates 2) highlight the need for international collaboration to devise strategies to minimize the risk of death after ECD transplantation, 127

144 3) suggest the need to carefully evaluate the impact of regional sharing on cold ischemic time in ECD transplants, and 4) challenge the U.S. policy to allow any consenting patient to proceed with ECD transplantation. These observations may be useful in increasing the safe utilization of ECD kidneys in the United States Disclosure The data reported here have been supplied by the United States Renal Data System (USRDS). The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy or interpretation of the U.S. government. 128

145 Figure 7.1. Kaplan-Meier plots comparing transplant outcomes among the N = 1520 ESP and n=446 U.S. recipients who were 65 years of age at the time of kidney transplantation from a deceased donor 65 years. (P< for each comparison) 129

146 Frequency Figure 7.2. Distribution of age of recipients of kidney transplantation from ECD kidneys from >=70 Recipient age (years) 130

147 Figure 7.3. Ten year mean patient and death censored graft survival and difference between curves. 131