Background Patients with rheumatoid arthritis (RA) have an increased risk of cardiovascular disease (CVD), which is not accurately predicted by risk calculators designed for the general population.
Objectives To develop a risk score calculator that will accurately predict 10 year CVD risk in patients with RA.
Methods The study population included RA patient cohorts from 8 rheumatology centers in 7 countries (UK, Norway, Netherlands, USA, Sweden, Greece and South Africa). In all cases, data had been collected prospectively on CV outcomes (myocardial infarction, revascularization, angina, stroke, transient ischemic attack, peripheral vascular disease and CV death) using standardized definitions. CV risk factors (smoking status, hypertension/blood pressure, lipids, diabetes mellitus, body mass index, use of antihypertensive or antilipemic medications) and RA characteristics (duration, seropositivity [rheumatoid factor-RF and/or anti-citrullinated protein antibody-ACPA], activity score [DAS28] and acute phase reactants) using the same definitions at all centers had also been collected at baseline for each cohort. Cox models stratified by center were used to develop a CVD risk calculator considering traditional CV risk factors and RA characteristics. Model performance was assessed using measures of discrimination and calibration.
Results In total 3176 RA patients who did not have known CVD at entry in their respective cohorts were included (mean age: 55 [SD: 14] years, 73% female). During a mean follow-up of 7.8 years (24733 person years), 314 patients developed CVD. The multivariable risk score modeling revealed 2 possible models including either seropositivity or DAS28 along with age, sex, current smoking, presence of hypertension, and ratio of total cholesterol to high-density lipoprotein (see table). Both 10-fold cross validation and multiple imputation analyses confirmed these findings with little change to the estimated coefficients. Both models demonstrated good discrimination (c-statistic: 0.76 and 0.74) and calibration (observed/predicted ratio: 1.00; 95% confidence interval: 0.89, 1.12). The ATACC-RA risk score (mean: 11.5%, SD 14.1%) showed significantly improved discrimination compared to either Framingham (c-statistic: 0.71, p<0.001) or SCORE (c-statistic: 0.72, p<0.001) risk algorithms.
Conclusions Development of an RA-specific CVD risk score calculator with metric properties superior to traditional CV risk scoring methods is feasible by pooling resources from many centers. Further development including external validation is underway.
Acknowledgements Sherine Gabriel, Cynthia Crowson, George Kitas, Karen Douglas, Anne Grete Semb, Silvia Rollefstad, Eirik Ikdahl, Piet Van Riel, Elke Arts, Jaap Fransen, Solbritt Rantapää-Dahlqvist, Solveig Wållberg-Jonsson, Lena Innala, George Karpouzas, Petros P. Sfikakis, Evi Zampeli, Patrick H. Dessein, Linda Tsang, Miguel A. Gonzalez-Gay, Alfonso Corrales, Hani El-Gabalawy, Carol Hitchon, Virginia Pascual Ramos, Irazú Contreras Yáñez, Daniel Solomon, Katherine Liao.
Disclosure of Interest None declared
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.