Background/Purpose: Cardiovascular (CV) risk in rheumatoid arthritis (RA) is increased. The CV risk algorithms for the  general population may underestimate the risk of cardiovascular disease (CVD) in RA patients¹. The objective of this study was to test the performance of the original and updated SCORE risk algorithm for the prediction of the 10-year risk of CVD in RA patients.

Methods: Data from the Nijmegen early RA inception cohort (n=1017) were used. The systematic coronary risk evaluation algorithm (SCORE) recalibrated for the Dutch population was used as a basis for the new model. This algorithm was recalibrated in this cohort (SCORE recalibrated) by first adjusting the weights of only the risk factors that were included in the original SCORE (age, gender, smoking, systolic blood pressure and total cholesterol:HDL ratio) and in the following this step the original SCORE was updated (SCORE updated) by including other risk factors as assessed in univariate and multivariate Cox proportional hazard regression analysis (significant at p-value <0.2). Predictive performance was assessed by the area under the receiver operating characteristic (ROC) curve and by comparing the observed versus expected number of CV events using Hosmer-Lemeshov tests and calibration plots.

Results: During follow-up, 144 patients had a CV event. In addition to the risk factors of the original SCORE, the updated model included BMI, diabetes, hypertension and high disease activity (DAS28>5.1) at baseline. Areas under the ROC curve were 0.72 (95% CI; 0.68-0.77, 0.71 (0.67-0.75), 0.75 (0.71-0.79), for the original, the recalibrated and the updated SCORE respectively (figure 1A), indicating moderate to good discrimination. The agreement between the observed and the predicted CV events was poor for the original score which underestimated CV risk at low and middle observed risk levels, which was confirmed in the Hosmer and Lemeshow (H-L) test that indicated poor model fit (p< 0.001). The updated SCORE still showed some underestimation (figure 1B) in the low-middle risk groups (≤ 20% CV risk) and overestimation in the highest risk groups, but overall model fit and predictive accuracy improved (H-L test p=1.0).

Conclusion: CV risk predictions according to the updated SCORE risk algorithm were more accurate compared to the original SCORE risk algorithm, particularly in the low and intermediate CV risk groups. The next step will be to externally validate these results in a different cohort.

Figure 1. Panel A: ROC-curves for the different SCORE algorithms. Panel B:  Calibration plot of the updated SCORE. Depicted are the observed events versus predicted events (observed versus predicted probabilities).

References

1.     Crowson CS, Matteson EL, Roger VL, Therneau TM, Gabriel SE. Usefulness of Risk Scores to Estimate the Risk of Cardiovascular Disease in Patients With Rheumatoid Arthritis. Am J Cardiol. 2012.3