Performance of the Expanded Cardiovascular Risk Prediction Score for Rheumatoid Arthritis: An External Validation in the BRASS Registry

BACKGROUND: Cardiovascular (CV) disease occurs more frequently in rheumatoid arthritis (RA) than in the general population and is difficult to predict. The expanded risk score for RA (ERS-RA) has demonstrated good discrimination and calibration in prior studies, but additional external validation studies are needed to understand clinical utility. METHODS: Data from a single center RA registry was integrated with electronic health records (EHR) and linked to Medicare/Medicaid administrative claims data. A less restrictive CV outcome was defined as an event identified in any of the three data sources (questionnaires/EHR/claims), and a more restrictive outcome included events in EHR and claims. External validation of the ERS-RA examined model discrimination and calibration. RESULTS: The longitudinal RA registry has 1,332 participants aged 20-80 years without prior evidence of CVD, followed for a mean of 6.8 years. The mean age was 54 years and 84% were female. We observed 164 (12.3%) and 69 (5.2%) participants with a CV event with the less and more restrictive CV definitions, respectively. The mean ERS-RA predicted score was 7.1%, and 424 (32%) participants had a predictive score ≥7.5%. We observed c-statistics of 0.66 (95% CI 0.62-0.71; less restrictive) and 0.80 (95% CI 0.76-0.85; more restrictive). The ERS-RA also demonstrated good calibration, with better performance for the more restrictive CV event definition. CONCLUSIONS: The ERS-RA demonstrated good discrimination and calibration for CV risk prediction in an external validation RA cohort. It also highlights the importance of using multiple sources of data to find CV events in a non-CV registry.