Background/Purpose: Rheumatoid arthritis (RA) patients are at high risk of cardiovascular disease (CVD). Risk algorithms for the general population lack precision when applied to RA patients and validated RA-specific CVD prediction models are missing. Risk age estimations are recommended as adjuncts to assessment of absolute 10-year risk of fatal CVD events. Two risk age models based on the Systematic Coronary Risk Evaluation (SCORE) algorithm have been developed; the cardiovascular risk age and the vascular age. However, the performance of these models has not been compared. Using longitudinal data on CVD events in RA patients, we aimed to compare the discriminative ability of cardiovascular risk age and vascular age among RA patients and in subgroups of RA patients based on disease characteristics.

Methods: Patients with RA were included from an international consortium, aged 30-70 years at baseline. Those with prior CVD, diabetes and/or users of lipid-lowering and/or antihypertensive therapy at baseline were excluded. Cardiovascular risk age was estimated based on chronologic age, smoking status, total cholesterol and systolic blood pressure at baseline. Vascular age was derived from the 10-year risk of CVD according to the SCORE algorithm, with or without high density lipoprotein cholesterol, using the equations for low and high risk countries. Performance of each risk age model in predicting CVD events was assessed using the concordance index. 

Results: Among the1867 RA patients included, 74% were female, median (inter-quartile range) age and disease duration were 52.0 (44.0, 59.9) and 0.6 (0.1, 6.4) years, 72.5% were rheumatoid factor positive, 24.7% were using glucocorticoids and 10.3% were using biologics at baseline. Overall, 144 CVD events occurred and median follow-up time was 5.0 (2.6, 9.3) years. Median difference between estimated risk age and chronologic age was 4.0 to 6.7 years, depending on the specific risk age model applied. Overall, the C-index across risk models ranged from 0.71 to 0.73 with standard errors of 0.03. Across prediction models, the lowest observed concordance was found among women and in glucocorticoid users and in those with new-onset disease (≤1 year). Additional analyses including RA patients on cardio preventive therapy yielded slightly lower c-indexes. Since SCORE was developed for use in Europe, we performed analyses on European RA patients, which yielded similar results. The trend of reduced concordance among women, glucocorticoid users and RA patients with short disease duration was preserved in these additional analyses.

Conclusion:  The cardiovascular risk age and vascular age models have comparable performance in predicting CVD in RA patients. Sex, disease duration and/or glucocorticoid treatment may influence the performance of risk age estimations.