Development Of Risk Prediction Model For Osteonecrosis In Patients With Systemic Lupus Erythematosus

Objective:The aim of this study was to explore the clinical characteristics and predictors for the development of avascular necrosis(AVN)in patients with systemic lupus erythematosus(SLE)and to establish a clinical prediction model based on COX regression analysis.Methods:This study was based on the platform of Chinese SLE treatment and research group(CSTAR),where SLE patients have been registered for regular follow-ups and monitoring of disease courses.Patients who were registered within 2 years after the diagnosis of SLE were included,and a follow-up of more than 2 years was required for sufficient observation of AVN.Standardized electronic data extraction platform was used to document the clinical manifestations,laboratory test results,organ damage and treatment strategies of the patients.And electronic questionnaires were used as supplement.Variables were initially screened via univariate COX regression.In combination with clinical background,accessibility and stability,candidate variables were further selected for Lasso regression analysis.Multivariate COX regression was used to establish the final prediction model.C-Index and Calibration plot were used to internally validate the discrimination power and accuracy.Nomogram was used to illustrate the model for clinical use.Results:A total of 1807 patients were included and 100(5.5%)patients developed AVN after registration.Variables with statistical significance after univariate COX regression included maximum daily dose of corticosteroid within the first year of SLE diagnosis,existing cerebrovascular event,lowered white blood cell count or lymphocyte count,cranial nerve disorder in SLEDAI.cerebrovascular accident in SLEDAI,and vasculitis in SLEDAI(p<0.05).In consideration of clinical accessibility and stability,candidate variables were included in Lasso regression for final selection.Multivariate COX regression model was established including variables of maximum daily dose of corticosteroid within the first year of SLE diagnosis,existing lowered white blood cell count or lymphocyte count,cerebrovascular event,serositis,antiphospholipid antibodies,and vasculitis in SLEDAI to predict the risk of AVN among SLE patients in 2 years and 3 years respectively.This model(C-Index=0.646)was illustrated with Nomogram for clinical use.Enlargement of sample size and external validation are required for future extensive use.Conclusion:A variety of risk factors can increase the risk of AVN in SLE patients.Based on demographic information,clinical features and laboratory test results,a COX regression model was established to predict the risk of AVN among SLE patients.