Predicting 5-year risk of kidney transplant failure: A prediction instrument using data available at 1 year posttransplantation

Background Accurate prediction of kidney transplant failure remains imperfect. The objective of this study was to develop and validate risk scores predicting 5-year transplant failure, based on data available 12 months posttransplantation. Study Design Development and then independent multicenter validation of risk scores predicting death-censored and overall transplant failure. Setting & Participants Outcomes of kidney transplant recipients (n = 651) alive with transplant function 12 months posttransplantation in Birmingham, United Kingdom, were used to develop models predicting transplant failure risk 5 years posttransplantation. The resulting risk scores were evaluated for prognostic utility (discrimination, calibration, and risk reclassification) in independent cohorts from Tours, France (n = 736); Leeds, United Kingdom (n = 787); and Halifax, Canada (n = 475). Predictors Weighted regression coefficients for baseline and 12-month demographic and clinical predictor characteristics. Outcomes Death-censored and overall transplant failure 5 years posttransplantation. Measurements Baseline data and time to transplant failure. Results Following model development, variables included in separate scores for death-censored and overall transplant failure included recipient age, sex, and race; acute rejection; transplant function; serum albumin level; and proteinuria. In the validation cohorts, these scores showed good to excellent discrimination for death-censored transplant failure (C statistics, 0.78-0.90) and moderate to good discrimination for overall transplant failure (C statistics, 0.75-0.81). Both scores demonstrated good calibration (Hosmer-Lemeshow P > 0.05 in all cohorts). Compared with estimated glomerular filtration rate in isolation, application of the scores resulted in statistically significant and clinically relevant risk reclassification for death-censored transplant failure (net reclassification improvement [NRI], 36.1%-83.0%; all P < 0.001) and overall transplant failure (NRI, 38.7%-53.5%; all P < 0.001). Compared with the previously described US Renal Data System-based risk calculator, significant and relevant risk reclassification for overall transplant failure was seen (NRI, 30.0%; P < 0.001). Limitations Validation is required in further populations. Conclusions These validated risk scores may be of prognostic utility in kidney transplantation, accurately identifying at-risk transplants, and informing clinicians and patients.