It is not raining in Bethesda today. I have considerable data of varying relevance to verify this: I have seen no raindrops; traffic is relatively good; the ground is dry; the humidity is 42%; I am wearing brown shoes. Based on prevailing regional weather patterns, I can guess tomorrow's weather with reasonable accuracy. However, as the amount of time from today's date increases, my accuracy decreases such that neither I nor any computer-assisted mathematical simulation can predict the weather one year from now.
We can completely characterize very complex conditions and remain helpless to predict them, even for common occurrences like rain. This realization was formalized by Lorenz and subsequently by Yorke as Chaos theory (1,2) whereby the reliability of forecasts in dynamical systems deteriorates exponentially over time due to a sensitive dependence on complete (unachievable) characterization of initial conditions. Similarly, despite our understanding of the factors that produce meteorological events, we still cannot actually make it rain anywhere. Meteorology is an observational rather than an interventional science.
In this issue of the American Journal of Transplantation, Roussey-Kesler and colleagues report on 10 kidney transplant patients that were found to be operationally tolerant (3). That is to say that the patients had stable allograft function without taking exogenous immunosuppressants or displaying a phenotype of immunoincompetence. Descriptions of this type are fascinating in their defiance of standard immune responsiveness. Indeed, they inspire us by clearly demonstrating that tolerance is a biological possibility, at least transiently. They also taunt us by documenting variability among tolerant patients that is more striking than any apparent similarity. Viewed at face value, these reports aid our understanding of immune tolerance. However, it is important to recognize that tolerance, particularly when it arises from standard immunosuppressive regimens, remains an observational science.
Immune responsiveness, even to a nominal antigen, is the product of many interdependent forces that lead to spectral (not binary) responses. Given the incalculable variables that influence immunity, particularly in response to an array of alloantigens, it is impossible to predict future immune responses with precision, and this too deteriorates exponentially with time. Yet when we remove a patient from all immunosuppression without a clinical indication, we are in effect trying to predict distant outcomes. Herein lies the peril in reporting these data. We must clearly distinguish them as observational not interventional. The authors of the present study make no claims to the contrary, but it is important that the general readership, and more importantly, the lay readership, recognize that this is a descriptive study and not an indication that an era of tolerance is upon us.
It is equally critical to recognize that the existence of operationally tolerant patients does not a priori mean we can induce tolerance in anyone. Given enough observations, the occurrence of even the most remote circumstances is as numerically certain as the mean. There is extraordinary danger in selling spontaneous tolerance as some mystical state that can be prospectively achieved. Rather in this setting it should be recognized for what it is: an unusual circumstance that is randomly inevitable on rare occasion due to the complexity of immunity. It is at present governed by probability, not practitioners.
The complexity of interpreting tolerance studies is compounded by the recognition that tolerance is not necessarily a durable state. The concept of an individual that is both immunocompetent and incapable of ever rejecting an organ is flawed. The same logic would hold that non-transplanted persons are incapable of ever developing autoimmunity. A functional immune system works because of its adaptability, and likewise fails when rigid. Immunocompetence, by definition, requires the ability to respond adaptively, and this will always hold some potential for heterogenous allo- or auto-immunity. Tolerance, operational or otherwise, is a current diagnosis, not a permanent one.
As clinicians, we take precautions with our patients based on the probability and severity of potential adverse events. It is likely, based on reports of complete immunosuppressive withdrawal like these, that rejection late after transplantation does not occur immediately upon drug
withdrawal (3,4). Its likelihood increases, but its speed of development decreases, with time. Some patients go for months or years before the conditions, remote from the trauma of surgery and the chemotactic effects of reperfusion, conspire for a clinical rejection. Yet when it does occur, it is clearly adverse. Many tolerant patients (20% in the current study and approximately 40% of those referenced in the bibliography) reject with continued follow-up. In this series, another 20% have apparently initiated an immune response in the form of alloantibody, and the adverse effects of alloantibody are becoming increasingly apparent (5). At times, patients die before the conditions for rejection are met, and these are patients that we define as ‘truly’ tolerant. It is likely, though, that their outlier condition is simply one that any bell-shaped curve mandates. With this in mind, it is safe to say that tolerance warrants as many precautions as does immunosuppression.
If spontaneous tolerance is to be viewed as a mere statistical aberrancy, what are we to do with these patients? At the very least, we should counsel them that they should be closely monitored and actively evaluated (usually by biopsy) if indicators of dysfunction (increased creatinine, proteinuria, alloantibody) are detected. This is easier said than done for patients many of whom are off immunosuppression due to noncompliance. A therapeutic hands-off approach may be warranted in some cases since we understand the effects of immunosuppression on tolerance far less well than we appreciate its effects on rejection. However, given the high graft failure rates in tolerant patients followed for a reasonable length of time, it is appropriate to suggest a trial of immunosuppression dosed to avoid side effects as a prudent safety net. Admittedly, clinicians who have cared for tolerant patients are the first to recognize that most tolerant patients will not accept this recommendation. Still, patients cannot be allowed to think that rejection is impossible. This is especially the case for patients on low-dose immunosuppression and monotherapy approaches. There is no evidence that these patients are anything other than efficiently immunosuppressed.
This is not to say that many critical discoveries cannot be derived from trying to understand aberrancies. Even if tolerance is not a clinical strategy, it does not mean that it has nothing to teach us. There may indeed be immune or genetic backgrounds that facilitate hyporesponsiveness; indeed pathways that could be therapeutically manipulated could be unearthed. Nevertheless, science and medicine at times have diverging goals and in the realm of clinical transplantation, the health of the individual patient (not the group) should always need to take precedent.
In the end, the quest to ‘get people off immunosuppression’ must be driven by patient outcome rather than practitioner ideology. We must understand the forces that produce immunity and tolerance, manipulate them safely and rationally, and make recommendations based on the natural history of the disease; for transplanted organs, this is rejection. Only after the successful implementation of this strategy has begun to validate our forecasts should we imply relative safety in immunosuppressive-free transplantation. Until then, we must avoid the temptation to present ourselves as ‘rainmakers’ (6).