“Statistics are no substitute for judgment”.
The transplant community has long held a tradition of transparency in terms of reported outcomes. Individual transplant programs are held to high standards and must take responsibility for a track record of inferior outcomes. The Scientific Registry for Transplant Recipients (SRTR) is charged with providing organ- and program-specific reports for transplant centers in the United States. The fundamental goal of program-specific reports is to provide a tool for quality improvement by using complex statistical and analytical models to generate risk-adjusted expected outcomes (or industry standards). Although this is a favorable situation from a regulatory perspective, it is crucial that these comparative standards are based on current, reproducible and validated data that account for case-mix and geographic differences in donors and recipients that occur among transplant centers. Mitigating factors can justify inferior outcomes under specific circumstances. However, the typical center response when program-specific outcomes fall below expected risk-adjusted outcomes is the adoption of risk-averse behavior with a commensurate reduction in transplant activity, particularly with respect to ‘high-risk’ transplants. Transplantation has relied on extraordinary achievements that have led to countless lives saved but recent studies suggest that the above regulatory systems may represent a threat to innovation and is counterproductive to the underlying goal of helping people and taking every opportunity to transplant every suitable organ . It is for this reason that we require estimations of outcomes and that they must be as accurate and reliable as possible.
In this issue of the American Journal of Transplantation, Kasiske et al. present a model developed in conjunction with the SRTR and representation from the United Network for Organ Sharing (UNOS) Pancreas Transplant Committee for establishing expected graft and patient survival outcomes . Previously, expected survival outcomes were only provided for simultaneous pancreas and kidney transplantation because there was an inadequate volume of isolated pancreas transplants in the form of pancreas after kidney and pancreas transplant alone to accurately make this calculation. If this methodology is accepted by the transplant community, perhaps a similar approach could be applied to calculating expected patient and allograft survival outcomes for intestinal transplantation and pediatric renal transplantation, which are two other transplant categories that lack the necessary volume to accurately model and establish expected outcomes.
One of the important missing elements, upon which the authors comment, is that graft survival calculations rely entirely on reported graft failures. Unfortunately, there is currently no uniform accepted definition of pancreas allograft failure. There is consensus that either mortality, allograft pancreatectomy, pancreas retransplantation or absence of C-peptide production clearly represent a failed graft, but at what point recurrence of diabetes management should be considered a graft failure has not been clearly established and is currently one of the main projects upon which the UNOS Pancreas Transplant Committee is focusing. Controversial areas include whether the use of oral or subcutaneous hypoglycemic agents or any resumption of insulin, particularly in the presence of detectable serum C-peptide levels, should be considered allograft failure. Moreover, does periodic rather than daily scheduled administration of antidiabetic therapies represent allograft failure? Certainly, the pancreas islet transplant community would consider multiple infusions, oral/subcutaneous antidiabetic agent or insulin use with detectable C-peptide levels or absence of either hypoglycemia unawareness or extreme glycemic excursions to represent a functioning graft [3, 4].
In the absence of a clear definition of pancreas allograft failure or type of diabetes, complex statistical modeling to determine expected graft survival outcomes for isolated pancreas transplantation could become an exercise in futility. The present methodology relies upon individual center definitions of pancreas allograft failure, which, to paraphrase Justice Potter Stewarts definition of pornography, may be “hard to define, but I know it when I see it” . Be that as it may, the authors are to be congratulated on undertaking the difficult task of finding the best way to compare pancreas transplant outcomes among different centers, taking into consideration that outcomes differ depending on whether one is looking at apples and oranges, apples after oranges or apples alone.