Cross-sectional stroke risk identification in rheumatoid arthritis: integrating traditional and disease-specific factors.

Abstract

Patients with rheumatoid arthritis (RA) face significantly elevated stroke risk, yet existing cardiovascular risk assessment tools perform poorly in this population. This study aimed to develop and validate a cross-sectional stroke risk identification model integrating traditional and RA-specific clinical characteristics. A two-step modeling approach was employed using NHANES 2011-2020 data (n = 1,366) and a Beijing Tiantan Hospital RA cohort (n = 774). LASSO regression identified traditional stroke risk factors to establish a basic model, which was then externally validated in the RA cohort. RA-specific factors DAS28-CRP score, anti-CCP antibody, rheumatoid factor, methotrexate use, and disease duration, were incorporated to develop an enhanced model. Eleven traditional risk factors were identified, with neutrophil count, hypertension, and coronary heart disease showing the strongest associations. The basic model achieved an AUC of 0.712 (95%CI: 0.665-0.759), with external validation AUC of 0.716 (95%CI: 0.674-0.758). After incorporating five RA-specific indicators, the enhanced model AUC improved to 0.829 (95%CI: 0.794-0.864, P < 0.001), with sensitivity 70.0%, specificity 81.6%, and accuracy 78.9%. Both likelihood ratio test (χ²=137.26, P < 0.001) and DeLong test (Z = -5.751, P < 0.001) confirmed superiority over the external validation model. The enhanced model also outperformed the Framingham Risk Score (AUC 0.611; DeLong Z = 8.210, P < 0.001). Integration of RA-specific factors significantly improved stroke risk identification, demonstrating superior discrimination over traditional cardiovascular risk assessment tools. This model provides a practical tool for stroke risk stratification and identification of high-risk individuals among RA patients, warranting further prospective validation.