AB0529 PULMONARY HYPERTENSION (PAH) IN SYSTEMIC LUPUS ERYTHEMATOSUS (SLE) IS NOT DEFINED BY AUTO-ANTIBODY AND CLINICAL VARIABLES.

Abstract

Background PAH is a well-recognized but infrequent manifestation of SLE. Often insidious in onset, PAH may have progressed for a while, before symptoms manifest. Prediction of clinical risk of PAH may have clinical benefit during the follow-up of a patient with SLE. Objectives The aim of the study was to develop a prediction algorithm for PAH in patients with SLE using clinical and auto-antibody variables. Methods The study included patients from the INSPIRE cohort which is, a national multi-center study enrolling patients with disease duration of less than three years. Lupus patients with symptomatic PAH confirmed by a transthoracic 2D ECHO right ventricular systolic pressure of more than 40 mm Hg were enrolled as cases. Controls were age, sex and date of enrolment matched patients from the same cohort without any clinical evidence of PAH. We excluded patients with underlying known heart disease and critical illness. The clinical variables and antibody data was retrieved from records. Outcome of death if any was noted. Using supervised learning, a logistic regression model was built for the prediction of SLE-PAH. The final Prediction model included one clinical variable (RP), sixteen autoantibodies (Sm-RNP, Smith, SS-A, Ro-52, SS-B, SCL-70, PM-SCL, Centromere, Jo1, PCNA, Nucleosome, dsDNA, histone, Ribosomal P and AMA, anticardiolipin antibody) and complement level. To ensure accuracy of the model, a confusion matrix using python was performed. The proportion of right predictions (ρ) is a measure of accuracy of the model. If the model is good, ρ must be close to 1. Results A total of 69 patients with symptomatic PAH, confirmed by 2D echo-cardiography were enrolled as cases. Raynaud phenomenon (RP), Interstitial lung disease (ILD) and ischemic stroke were significantly higher in cases than control (p= 0.026,0.001, 0.05 respectively). There was no difference in the prevalence of individual auto-antibodies between the groups except high prevalence of anti-cardiolipin antibody in cases (39% vs 23% p= 0.048). Mortality rate was higher in cases (17% vs 0.06%, p=0.003) during a follow up of 3 years with all the deaths occurring within one year of PAH diagnosis. Ten repetitions of the confusion matrix analysis using Python for this prediction model (both derivation and validation) yielded the following ρ scores 0.60, 0.47, 0.57, 0.63, 0.50, 0.60, 0.57, 0.43, 0.50, 0.57. which exhibited wide variability. Conclusion A prediction model using RP and commonly assayed autoantibodies failed to be informative for PAH in Indian lupus patients.