Comparing methods for monitoring serum creatinine to predict late renal allograft failure

• Few studies have systematically investigated what changes in chronic renal allograft function best predict subsequent graft failure, when these changes occur, and whether they occur soon enough to allow possible intervention. We collected serum creatinine values (mean, 183 ± 75 values/patient) measured over a maximum follow-up of 22 years in 101 consecutive renal transplant recipients (excluding creatinine levels from periods of acute rejection). We determined the dates of first decline in inverse creatinine (Δ1/Cr; < -20%, -30%, -40%, -50%, and -70%), declines in estimated creatinine clearance (CCr; <55, 45, 35, 25, and 15 mL/min), and declines in measured slope of 1/Cr over time. We used time-dependent covariates in Cox proportional hazards analyses to determine the relative effect of each renal function parameter on outcomes while adjusting for other risk factors. The best predictor of subsequent graft failure was Δ1/Cr. Δ1/Cr less than -40% first occurred at a median of 1.28 years after transplantation in 73 patients, and 67 patients went on to have graft failure a median of 3.28 years after Δ1/Cr less than -40%. The independent relative risk for graft failure attributable to Δ1/Cr less than -40% was 5.91 (95% confidence interval, 3.25 to 10.8; P < 0.0001). A decline in CCr, eg, less than 45 mL/min, also was a strong predictor of subsequent graft failure. Conversely, declines in allograft function estimated from slopes of 1/Cr were poor predictors of graft failure. In analysis limited to patients followed up for 2.5 years or less, Δ1/Cr continued to predict graft failure, suggesting that Δ1/Cr will be a useful predictor in populations with shorter follow-up. If confirmed in other populations, eg, patients treated with calcineurin inhibitors, this simple marker of chronic allograft dysfunction may prove to be a practical tool for defining patients at high risk for late graft failure.