247. Predicting the 1-year risk of kidney failure in patients with ANCA-associated vasculitis

Background: The risk of kidney failure varies substantially amongst patients with anti-neutrophil cytoplasm antibody (ANCA)-associated vasculitis. We derived a simple instrument to estimate the 1-year risk of kidney failure in patients with ANCA-associated vasculitis. Methods: Analysis of individual patient data from 7 international randomized controlled trials conducted by the European Vasculitis Study Group covering the spectrum of severity in ANCA-associated vasculitis. Candidate predictors included age, sex, ANCA type, creatinine, estimated glomerular filtration rate, and use of dialysis. The outcome of interest was kidney failure by 1-year post-enrollment. Candidate predictors were included in iteratively fit logistic regression models with parameters that did not improve fit dropped. Models were internally validated using bootstrapping and externally validated in an 8th randomized controlled trial. Alternative models were compared using discrimination and calibration statistics and decision curve analysis. Results: Kidney failure occurred in 54 of 786 participants in the derivation data. Serum creatinine, fit with a cubic spline, and need for dialysis at baseline resulted in the simplest, most parsimonious model with a C-statistic of 0.919 and Brier score of 0.050 in the derivation data set and a C-statistic and Brier score of 0.876 and 0.102 respectively in the validation data. A simpler model using only serum creatinine categorized as 300 to 500 µmol/L and >500 µmol/L or receiving dialysis performed similarly to the best model for patients up to an 11% risk and discrimination and calibration statistics in the external validation data were similar to the best fitting model (C-statistic 0.830 and Brier score 0.108). Conclusions: A single serum creatinine is a useful estimator of the risk of kidney failure at one year in patients with ANCA-associated vasculitis. Even simple categorization of serum creatinine can reliably identify patients at low risk of kidney failure adequately. Disclosures: The authors have no relevant financial disclosure