Article Text

THU0107 Improving the Prediction of 10-Year Risk of Cardiovascular Disease using Traditional and Disease-Specific Predictors in Patients with Rheumatoid Arthritis
  1. M. Schulpen1,
  2. G.C. Puts1,
  3. E. Arts1,
  4. A.A. den Broeder2,
  5. C.D. Popa1,
  6. J. Fransen1
  1. 1Radboud University Medical Centre
  2. 2Maartenskliniek, Nijmegen, Netherlands


Background Rheumatoid arthritis (RA) patients have a decreased life-expectancy, partly due to an increased incidence of cardiovascular diseases (CVD) (1, 2). Identification of RA patients with high CVD-risk would enable initiation of preventive interventions (3). However, routinely used CVD prediction models, e.g. Framingham Risk Score (FRS) and Systematic Coronary Risk Evaluation score (SCORE), generally underestimate the increased risk in RA patients (4, 5).

Objectives To investigate whether RA-specific factors have an additional predictive value to traditional risk factors in predicting the occurrence of CVD over 10 years in RA patients.

Methods This study was performed using the database of the Nijmegen early RA inception cohort (4).Two Cox proportional hazards models were created to predict the occurrence of the first non-fatal or fatal cardiovascular event (CV-event) within 10 years after inclusion in the cohort. The CVD model included a combination of the traditional CVD predictors from FRS and SCORE, using backward selection. The RA-specific model, was an extension of the CVD model by adding predefined RA-specific predictors, followed by backward selection. Harrell's C-statistics were calculated to assess discriminative power. A likelihood ratio test was performed to compare the models. Logistic regression analyses were performed to make receiver operating characteristic (ROC) curves and calibration plots to gain insight into the accuracy of the models.

Results Of the 1028 patients included in the analyses, 141 had experienced a CV-event with a mean time-to-event of 4.9 years. The RA-specific model (C-statistic=0.75) had a slightly higher discriminative ability compared to the CVD model (C-statistic=0.73). The likelihood ratio test (χ2 df=2=12.645, p<0.01) showed that the RA-specific model had a significant better fit to the data compared to the CVD model. Calibration plots show that both models underestimated the true probability on experiencing a CV-event.

Conclusions RA-specific predictors slightly improved the performance of a model containing traditional CVD predictors. Doubt exists whether the additional value of RA-specific predictors outweighs the disadvantages of using a disease-specific model instead of a general one. Future studies should focus on identifying RA-specific factors that further improve the prediction of CV-events in RA patients.


  1. Kitas G et al. Cardiac involvement in rheumatoid disease. Clinical medicine (London, England). 2001 Jan-Feb;1(1):18-21.

  2. Radovits BJ et al. Excess mortality emerges after 10 years in an inception cohort of early rheumatoid arthritis. Arthritis care & research. 2010 Mar;62(3):362-70.

  3. Avina-Zubieta JA et al. Risk of incident cardiovascular events in patients with rheumatoid arthritis: a meta-analysis of observational studies. Annals of the rheumatic diseases. 2012 Sep;71(9):1524-9.

  4. Arts EE et al. Performance of four current risk algorithms in predicting cardiovascular events in patients with early rheumatoid arthritis. Annals of the rheumatic diseases. 2014 Jan 3. [Epub ahead of print].

  5. Crowson CS et al. Rheumatoid arthritis and cardiovascular disease. American heart journal. 2013 Oct;166(4):622-8 e1.

Disclosure of Interest None declared

Statistics from

Request Permissions

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.