Poor predictive value of serum creatinine for renal allograft loss

Both medical care and pharmaceutical development have led to an increase in expected graft and patient survival for patients who undergo renal transplantation. From a research perspective, it has become increasingly difficult to study the efficacy of new therapies using traditional 'hard' endpoints. In reaction to this dilemma, the transplant community has sought a surrogate endpoint. A natural candidate for a surrogate marker for graft loss that has been proposed is renal function (serum creatinine or calculated GFR levels). Using data from the USRDS, we conducted a retrospective evaluation of transplant data from 1988 to 1999 to quantify the predictive value of renal function for the outcomes of graft loss, death-censored graft loss, and patient death. Renal function along with the change in renal function demonstrated a high relative risk for ultimate graft survival and graft loss (odds ratio = 2.2 for an increase of 1 mg/dL). However, the predictive value as measured by the area under the receiver operating characteristic curve (AUC) for this criteria was poor (0.627). These findings held true for the slope of creatinine and formulations of GFR. While renal function is a strong risk factor and highly correlated with graft failure, the utility of renal function as a predictive tool for graft loss is limited.