Estimating 2-year risk of first relapse in ANCA-associated vasculitis: derivation of a risk-prediction model based on participants in the PEXIVAS trial.

OBJECTIVES: Individuals with antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) differ in their risk of relapse. Estimation of an individual's risk may inform decision-making around monitoring and treatment. METHODS: We performed a post hoc analysis of participants with AAV with an estimated glomerular filtration rate (eGFR) <50 mL/min/1.73 m2 or diffuse alveolar haemorrhage enrolled in a trial assessing the effects of plasma exchange and a reduced glucocorticoid dosing regimen who achieved disease remission. We estimated the risk of first relapse within 2 years of remission. We first identified candidate predictors from the literature. Candidate predictors measured in our data set and associated with relapse within 2 years of remission underwent further selection using least absolute shrinkage and selection operator regression and time-to-event analysis. Model stability was assessed with variable inclusion, multiple shrinkage methods, and VanderWeele's E. Model discrimination was assessed using concordance statistics, and calibration using the slope. RESULTS: We included data from 649 trial participants, of whom 100 experienced a relapse within 2 years of remission. The final risk estimation model included sex, use of oral cyclophosphamide or rituximab as induction agent, ANCA subtype, nonhaemorrhagic respiratory involvement or mucous membrane/eye involvement at presentation, and eGFR at remission. The optimism-corrected Harrell's area under the receiver-operator curve 0.709 (95% CI 0.707-0.710). The calibration slope was 0.99 (95% CI 0.65-1.34). CONCLUSIONS: Our model estimates the risk of first relapse for individuals with AAV within 2 years after achieving disease remission. This may inform monitoring and treatment decisions. The model will benefit from future external validation.