Background The application of statistical clustering methods to various clinical datasets has revealed unexpected patterns of disease that may have practical implications. There are advantages to categorising patients with rheumatoid arthritis (RA) into distinct subsets for treatment and prognostication.
Objectives To examine the cluster pattern of joint deformity of an inception cohort of RA patients.
Methods All patients fulfilled the 1987 ACR criteria for RA, had ≤2 years disease duration at recruitment and were followed up from enrolment till 2014. Correlation of joint deformity symptoms at last study visit among the eight joints were analysed using eigen vectors from principal component analysis. K-means clustering analysis was applied to classify the RA patient subgroups based on distribution of joint deformity involvement. Joint deformity pattern was analysed at the last visit and correlated with socioeconomic, extra-articular features (EAF), DAS28 and HAQ scores, rheumatoid factor (RF), anti-citrullinated peptide antibody (ACPA) status and treatment.
Results The mean age of the RA cohort was 50.4 years, 75.5% were Chinese and mean follow-up period was 88.4 months. Correlation analysis showed that the joint deformity involvement could be classified into three categories: proximal interphalangeal/metacarpophalangeal joint (PIP/MCP), shoulder/elbow/ankle (S/E/A) and wrist/knee/metatasophalangeal (W/K/MTP). Cluster analysis showed four patterns of joint deformity: mild involvement of PIP/MCP joints (group I), dominant involvement of S/E/A and W/K/MTP (group II), W/K/MTP involvement (group III) and dominant PIP/MCP/S/E/A/W/K/MTP involvement (group IV). There was no significant difference in mean age, disease duration, work status, occupation, smoking status, prevalence of RF and ACPA, DAS28 and HAQ scores among the four groups. Group IV patients had a significantly longer symptom duration prior to diagnosis, received higher number of non-biologic DMARDs and were more likely to have radiographic erosions (p<0.05). Unlike other joints, wrist deformity was significantly correlated with prior wrist tenderness or swelling (34.5% of patients).
Conclusions Cluster analysis showed four subgroups of RA patients based on pattern of joint deformity. The presence of wrist synovitis at baseline may predict wrist deformity. Factors other than socioeconomic, RF, ACPA need to be explored to explain the clustering of joint deformity in our inception cohort.
Acknowledgements We thank Ms Safiyya Mohamed Ali for her kind assistance.
Disclosure of Interest None declared