Engineering

Cohort

Rheumatology

Proportional hazards model

Biochemistry

Rheumatoid arthritis

Biomarker

Mechanical engineering

Retrospective cohort study

Stroke (engine)

Physical therapy

Internal medicine

Confidence interval

Hazard ratio

Medicine

Chemistry

Disease

<h3>Background</h3> The ACC/AHA recommends preventive strategies for patients with a high-predicted risk of atherosclerotic cardiovascular disease (CVD). RA patients are at higher risk for CVD events, yet the role of systemic inflammation and the influence of traditional CVD risk factors are unclear with respect to risk prediction in RA. <h3>Objectives</h3> A simple and accurate algorithm for predicting CVD event risk that considers systemic inflammation might help risk assessment for RA patients and optimize preventive care. <h3>Methods</h3> We derived a U.S. cohort of RA patients by linking multi-biomarker disease activity (MBDA) test data to Medicare claims data. Patients had to have ≥1 year Medicare coverage prior to the date of their first MBDA test, which was designated as baseline. Exclusions were past MI, stroke, or cancer. Follow-up ended at the earliest of: 1) CVD event; 2) death; 3) loss of coverage; or 4) 12/31/2016, the latest date for which the Medicare claims database was evaluated. CVD events were defined as incident MI, stroke or fatal CVD event, and were identified using validated algorithms (PPV &gt;=80%). The leptin-adjusted MBDA score (Curtis et al., <i>Rheumatology</i> 2018) and its 12 individual protein biomarkers were evaluated as predictors of CVD events, as were other variables, including demographics, healthcare utilization, CVD-related comorbidities and medications, and RA-related features (e.g. DMARD/biologic use, glucocorticoid use). The cohort was randomly split 2:1 to use 2/3 of patients for training and 1/3 for validation. Cox proportional hazard regression with LASSO was used for variable selection based on minimization of 10-fold cross-validated error + 1 SE. Model calibration (observed vs. expected) and discrimination were assessed for predicted CVD events at 3 years. Analyses are ongoing; model performance results are reported for the cross-validated training data. <h3>Results</h3> A total of 26,261 eligible RA patients were analyzed; mean (SD) age 68.6 (10.2) years, 80.1% female, 72.6% white, 23% diabetes, 43% statin use, 56% methotrexate, 44% on biologics/tofacitinib, 55% steroids, and median (IQR) adjusted MBDA score 40 (32-49). A total of 477 CVD events occurred over mean (SD) follow-up time of 1.7 (1.2) years yielding a CVD incidence rate of 16.5 (95% CI 15.0-18.0)/1000py. The most important predictors in these LASSO-selected models were age, beta-blocker use, sex, diabetes, adjusted MBDA score and a subset of individual MBDA biomarkers. The best performing model had a cross-validated area under the receiver operator curve of 0.70 and good observed: expected prediction at 3 years (Figure). <h3>Conclusion</h3> Preliminary results from this approach suggest that a simple algorithm consisting of a limited number of protein biomarkers and clinical measures can provide an accurate method to predict short-term CVD risk in RA. <h3>Acknowledgement</h3> The work for this abstract was funded by an industry/academic collaboration between Myriad and University of Alabama at Birmingham. <h3>Disclosure of Interests</h3> Jeffrey Curtis: None declared, Fenglong Xie: None declared, Lang Chen: None declared, Eric H. Sasso Shareholder of: Myriad Genetics, Inc., Employee of: Crescendo Bioscience, Inc., Elena Hitraya Shareholder of: Myriad, Employee of: Myriad, Jerry Lanchbury Shareholder of: Myriad, Employee of: Myriad, Darl Flake Shareholder of: Myriad Genetics, Inc., Employee of: Myriad Genetics, Inc., Alexander Gutin Shareholder of: Myriad, Employee of: Myriad, Cheryl Chin Shareholder of: myriad stock, Employee of: myriad, Cynthia S. Crowson: None declared

AB0289 PREDICTION OF CARDIOVASCULAR EVENTS IN RHEUMATOID ARTHRITIS PATIENTS USING PROTEIN BIOMARKERS AND CLINICAL FACTORS