Seung Min Jung1, Kyung-Su Park1, Ki-Jo Kim2. 1. Division of Rheumatology, Department of Internal Medicine, St. Vincent's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea. 2. Division of Rheumatology, Department of Internal Medicine, St. Vincent's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea. md21c@catholic.ac.kr.
Abstract
OBJECTIVES: The clinical manifestations and treatment outcome in patients with rheumatoid arthritis (RA) are heterogeneous. We classified RA patients into subgroups with distinct phenotypes through unsupervised clustering and evaluated the utility of this subclassification for evaluation of clinical outcome. METHODS: A total of 1,103 patients with RA were clustered in an unbiased manner using a k-means clustering method, based on their clinical and phenotypic profiles. Initiation of biological disease-modifying anti-rheumatic drugs (bDMARDs) was evaluated in the segregated clusters to investigate the differential clinical course of each cluster. RESULTS: Patients with RA were classified into four clusters, each with distinct phenotypes. The key features for subclassification were sex, smoking, hypertension, and dyslipidaemia. Cluster 1 consisted of male smokers, who were most likely to initiate bDMARDs by 30 months (p=0.04). Multivariate analysis revealed that overweight, smoking, erythrocyte sedimentation rate, autoantibodies of high titre, and disease activity were the independent predictors of bDMARD initiation at 30 months. Cluster 1 was the highest or the second highest for these independent predictors, suggesting that cluster 1 contained a high-risk group for early initiation of bDMARDs. CONCLUSIONS: The unsupervised clustering of RA patients demonstrated the feasibility of the novel subclassification with respect to predicting clinical outcome. Identifying high-risk patients by a combination of clinical parameters may be useful for the management of RA.
OBJECTIVES: The clinical manifestations and treatment outcome in patients with rheumatoid arthritis (RA) are heterogeneous. We classified RA patients into subgroups with distinct phenotypes through unsupervised clustering and evaluated the utility of this subclassification for evaluation of clinical outcome. METHODS: A total of 1,103 patients with RA were clustered in an unbiased manner using a k-means clustering method, based on their clinical and phenotypic profiles. Initiation of biological disease-modifying anti-rheumatic drugs (bDMARDs) was evaluated in the segregated clusters to investigate the differential clinical course of each cluster. RESULTS: Patients with RA were classified into four clusters, each with distinct phenotypes. The key features for subclassification were sex, smoking, hypertension, and dyslipidaemia. Cluster 1 consisted of male smokers, who were most likely to initiate bDMARDs by 30 months (p=0.04). Multivariate analysis revealed that overweight, smoking, erythrocyte sedimentation rate, autoantibodies of high titre, and disease activity were the independent predictors of bDMARD initiation at 30 months. Cluster 1 was the highest or the second highest for these independent predictors, suggesting that cluster 1 contained a high-risk group for early initiation of bDMARDs. CONCLUSIONS: The unsupervised clustering of RA patients demonstrated the feasibility of the novel subclassification with respect to predicting clinical outcome. Identifying high-risk patients by a combination of clinical parameters may be useful for the management of RA.