SAT0098 Patient and disease characteristics that predict switching from a tnf inhibitor to another biologic or targeted synthetic dmard in patients with ra in clinical practice

Background Specific patient demographics or disease characteristics may be associated with switching from one therapy to another for patients with RA. Identifying these predictive factors may help inform prospective treatment decisions. Objectives To examine factors predicting switching among patients with RA from a TNF inhibitor (TNFi) to a subsequent biologic (b)DMARD (TNFi or non-TNFi [abatacept, tocilizumab, rituximab]) or the targeted synthetic (ts)DMARD tofacitinib. Methods This analysis included patients aged ≥18 years, who were enrolled in a large sequential RA registry established in October 2001 and who initiated a TNFi on/after 1 January 2005 and had ≥24 months’ follow-up. Switch was defined as discontinuation of a TNFi and initiation of another bDMARD or tofacitinib within 6 months. Of TNFi initiations, 67% were randomly selected as a prediction dataset and used to develop the final model; 33% were considered in the validation dataset. Logistic regression modelling was used to predict switching status; baseline demographics (age, sex, race), patient attributes (smoking status, BMI, work status) and clinical characteristics (RF and anti-cyclic citrullinated protein status, erosive disease, history of co-morbidities, prior and current treatment, disease activity, patient-reported pain, fatigue, morning stiffness) were considered. Goodness-of-fit statistics were used to assess model fit and receiver operating characteristic curves (area under the curve [AUC]) to validate the model. Results Among 6909 eligible TNFi initiations, there were 1343 switchers (prediction dataset: 4623 TNFi initiations, including 898 switchers). Compared with non-switchers, switchers were younger, had a shorter duration of RA and higher baseline mean CDAI score. Fewer switchers were positive for erosive disease or on combination therapy with MTX, but more were on monotherapy or combination therapy with a non-MTX DMARD. After investigation of several models, the best-fit model (Table) to predict switching from a TNFi yielded an AUC=0.705 (sensitivity=81%; false positive rate=49%). Conclusions The model identified in this analysis revealed that factors including age, duration of RA, CDAI, history of co-morbid conditions, prior treatment and year of TNFi initiation predicted switching from a TNFi to another bDMARD or tsDMARD. Disclosure of Interest L. Harrold Shareholder of: Corrona, LLC, Grant/research support from: Pfizer, Consultant for: Roche, Bristol-Myers Squibb, Employee of: Corrona, LLC, University of Massachusetts Medical School, H. Litman Employee of: Corrona, LLC, S. Connolly Shareholder of: Bristol-Myers Squibb, Employee of: Bristol-Myers Squibb, E. Alemao Shareholder of: Bristol-Myers Squibb, Employee of: Bristol-Myers Squibb, S. Kelly Shareholder of: Bristol-Myers Squibb, Employee of: Bristol-Myers Squibb, S. Rebello Employee of: Corrona, LLC, T. Blachley: None declared, J. Kremer Shareholder of: Corrona, LLC, Grant/research support from: AbbVie, Bristol-Myers Squibb, Genentech, Lilly, Novartis, Pfizer, Employee of: Corrona, LLC, Speakers bureau: Genentech (non-branded talks only)