Graft Survival

Graft survival may be presented in two different ways, each valid in its own right, but yielding quite different answers. One of the problems in the field is that many graft survival analyses do not make clear which is being used. It is possible to understand the problem using the 1000-patient example. Assume that 100 patients have died with their transplant still functioning at the time of death and 80 have lost their grafts from rejection and other causes. If one counts death as a cause of graft loss, then 180 of the 1000 patients have lost their grafts; if one ignores death as a cause, then only the 80 have lost their grafts. The latter calculation is best described as death-censored graft survival, whereas the inclusive analysis is described as patient and graft survival. When the analysis is not specified, the findings can be misleading and may make no sense. For example, in the illustrative analysis, the actual patient survival would be 90% (100 deaths in 1000 patients) and the death-censored graft survival would be 92% (80 grafts lost in 1000 transplants), so it might be concluded that 2% or 20 kidneys somehow survive in the absence of the recipient! Both approaches are of course relevant to different situations—on the one hand to understand what causes kidneys to fail, and separately to understand what causes patients to die—but there must be methodologic clarity.

Randomized Controlled Trials

Randomized controlled trials (RCTs) are a conventional way of determining differences between two or more therapies and are used in transplantation especially to evaluate new pharmaceutical products, which require testing in randomized studies before they can be registered for use. The majority of RCTs therefore compare two drug regimens—one with and one without the new agent. Two critical problems compromise the value of RCTs in renal transplantation: First, the relatively small number of transplants performed in any one center means that it is hard to mount studies with sufficient power to detect significant differences; and second, good early surrogate markers for long-term outcomes are lacking. In cardiovascular studies, in contrast, it is often possible to randomize many thousands of patients and detect differences in meaningful outcomes other than death; and in cancer studies it is possible to use short-term evidence of tumor response as a surrogate for longer-term outcomes. In an earlier era of renal transplantation, relatively small studies could be highly informative. An example of an RCT from the 1980s is the Oxford cyclosporine trial,⁸ in which azathioprine and prednisolone could be compared with cyclosporine with sufficient power to detect a difference with less than 100 patients randomized to each group. This was because the survival rates in the control arm were susceptible to large improvements—5-year graft and patient survival improved by 20% from a baseline around 47% to 67% with the addition of cyclosporine. Today the results of standard of care therapy are so high that recruitment of sufficient patients to detect either graft or patient survival differences at 1 or 2 years is not only extremely expensive but, unless the trials are very large, likely to founder on small random variations in such issues as surgical complications.

In the 1990s the introduction of the composite end point controlled the required size of RCTs. The composite end point is an acceptable indicator of benefit for the U.S. Food and Drug Administration (FDA) and other regulatory authorities to register a drug for therapeutic use. The composite end point includes patient death, graft loss, acute rejection episodes, and loss to follow-up (the last index conservatively defaults the patients who are lost to follow-up as having failed the therapy). Analysis is by intention to treat—in other words, for outcome analysis the patient is assigned as randomized whether or not the patient ever received the assigned treatment. Three key studies confirming the benefit of mycophenolate mofetil (MMF) were conducted using this approach and recruited around 500 patients allocated randomly into three study arms.^9,10

Moving forward another 10 years, the largest transplant RCT undertaken to date (the Symphony trial¹¹) required more than 1500 patients to be randomized, and even then was powered only for a surrogate marker of long-term outcome, differences in renal function. It is now accepted that transplant RCTs cannot be sufficiently powered to detect differences in the “hard” end points of patient and graft survival and must all be powered for a composite end point and surrogates of long-term outcomes. Which short-term surrogates provide good predictors of long-term outcomes has yet to be determined. Acute cellular rejection was, in the 1990s, a short-term measure of importance for long-term results, but perhaps not when current immunosuppressive regimens are used.¹²

Cohort Studies

Some clinical issues are not susceptible to testing by RCT. For example an RCT design will never be able to test whether cancer is more prevalent in patients on dialysis compared with patients after transplantation, first because randomization to dialysis or transplantation is ethically unachievable, and second because the numbers required to power such a study would exceed global transplant capacity. However, it is still important to understand whether the risks change between dialysis and transplantation and thus to understand whether there is any action that can increase or decrease death from cancer after transplantation. An example of an analysis addressing these questions is the Australian cancer study in which data from the Australia and New Zealand Dialysis and Transplant (ANZDATA) registry on a cohort of patients who were first treated for ESRD between 1982 and 2003 were linked to data from the national cancer registries.¹³ It is critical to ensure the selection of the cohort is appropriate to the question being asked. All too often a cohort study is selected from a convenience sample and thus has critical systematic flaws. A typical convenience sample is “all patients being followed by a particular transplant unit.” The analysis becomes an assessment only of patients still alive and still coming to the transplant unit for follow-up. The cohort has been distorted, the flaw being that the patients not being followed, because of death or graft failure, may be crucial to the analysis.

Meta-Analyses

Some areas of transplantation are rich in both RCTs and nonrandomized studies of a particular intervention—for example, interleukin-2 (IL-2) receptor blockade for induction in renal transplantation. Because the studies have produced variable results, which one should be believed? This issue is approached by meta-analysis of available studies and exemplified by the work of the International Cochrane Collaboration.¹⁴ An example of a Cochrane review is shown in Figure 109-3.¹⁵ These types of analysis combine the results from patients who have undergone the same interventions from all similar studies, which thus gives much greater statistical power than each study alone can provide. The methodology requires searching for all studies of the question and especially capturing unpublished studies. The problem of publication bias plagues the field; studies with a positive finding for the intervention are more likely to be published than negative studies. The compulsory preregistration of all trials of a drug by commercial entities in a public database such as www.clintrials.gov is designed to avoid this publication bias, and it is now widely accepted that nondisclosure of data by commercial sponsors of clinical trials is neither scientifically acceptable nor tolerated by the regulators.

Can a case report provide valuable information about outcomes after transplantation? There are some instances of rare events where the knowledge from a case report has proved important. One example is the early anecdotal reports of hepatitis B transmission from hepatitis B surface antigen (HBsAg)–negative donors to recipients of their livers.¹⁶ A few case reports led to a cohort analysis, which in turn led to an understanding that hepatitis B can be transmitted from a donor who tests negative for HBsAg but positive for hepatitis B core antibody. The surveillance and vigilance strategy for the safety of cells, tissues, and organs of human origin is based on this concept—identifying and publicizing the rare cases so that the problems can be recognized (see www.notifylibrary.org).

Expert Opinion Guidelines

The final type of analysis that is pervasive in the transplantation literature is expert opinion, sometimes leading to clinical practice guidelines. Only the major questions have answers in RCTs, meta-analyses, or cohort studies. The clinician is faced with unanswered questions in day-to-day clinical decision making. Guidelines rate the available evidence on particular questions from good to nonexistent. An example of a widely disseminated set of guidelines is the KDIGO guidelines for the care of a transplant recipient,¹⁷ which clearly set out the limits of knowledge in the area. Much of the guidance provided on how to maximize outcomes for particular patients is based on the combined wisdom and experience of clinical practitioners and not on RCT or other data analyses.

Qualitative Studies

Some questions are not amenable to quantitative data collection yet are of great importance to the field. What causes potential living donors to decide to donate? What do physicians consider most important in assessment of living donors? Why does a particular community not donate many organs after death? These questions are amenable to qualitative research and can be assessed and measured, not to produce proof related to a particular question, but to generate testable hypotheses. Qualitative studies create insights into behaviors and attitudes and do not assume that there is a single reality to be discovered; people have different views on organ donation, and any system of measurement that expects everyone to be driven by the same perspectives is going to be disappointed. A guide to the approach of qualitative studies in transplantation has been published to assist scientists familiar mostly with quantitative studies.¹⁸

Registries

The methodology of registry data collection varies from voluntary to compulsory and from automated to manual. The completeness of data is related to the amount collected, the frequency with which it is updated, and the methods used. National transplantation registries, such as the ANZDATA registry in Australia and New Zealand¹⁹ or the similar Malaysian Dialysis and Transplant Registry, are often voluntary. Some, such as the North American UNOS (United Network for Organ Sharing), database are mandatory, in that financial penalties are threatened for failure to fill in critical data elements.²⁰ The largest international registry is the Collaborative Transplant Study run from Heidelberg, Germany, which uses voluntary data collection on the repeatedly stressed basis that all sequential transplants are entered and followed.⁴ Donor registries tend to be complete because they capture all data at the time of organ donation as a part of the process of donation.²¹

Related posts:

Management of Refractory Heart Failure Membranous Nephropathy Complications of Peritoneal Dialysis Immunologic Principles in Kidney Transplantation Gastroenterology and Nutrition in Chronic Kidney Disease Imaging

Stay updated, free articles. Join our Telegram channel

Join