Validation of Leiden Score in Predicting Progression of Rheumatoid Arthritis in Undifferentiated Arthritis in Indian Population

BACKGROUND: Leiden Score, is a very useful tool for predicting future development of rheumatoid arthritis (RA), among undifferentiated arthritis (UA) patients. This score has been validated in various western studies but rarely among south east Asian patients. AIMS: To validate the Leiden early arthritis prediction rule in an Indian cohort of patients for predicting rheumatoid arthritis (RA) in undifferentiated arthritis (UA) patients and to formulate any simpler version of prediction score taking only clinical variables of original Leiden prediction rule. SUBJECTS AND METHODS: In a group comparative longitudinal study model, 58 patients with early symmetrical polyarthritis were enrolled and baseline evaluation was done according to Leiden prediction rule and then 3 monthly. After 1 year, Leiden prediction score and chance of evolving into RA were calculated. Patients were divided into two groups: Those who developed RA and who did not. They were selected on random sampling process. Tender joint count (TJC), duration of morning stiffness, and duration of arthritis were selected as clinical variables for linear discriminant analysis with disease outcome being the dependent variable. Discriminant scores (D) for each patient was calculated. A receiver operating characteristic (ROC) curve was constructed with the discriminant score and compared with Leiden prediction score. RESULTS: About 54% (27/50) of patients were diagnosed with RA and 46% (23/50) developed other rheumatologic condition or viral inflammatory arthritis or remained undifferentiated or attained complete remission. None of the patients with UA, who scored the regression coefficients 4 or less progressed to RA, and those who scored 7 or more, almost certainly progressed to RA. Unstandardized canonical discriminant coefficients for TJC (T), duration of morning stiffness (M), and duration of arthritis (A) were calculated. ROC curve was plotted with the formula: D = 0.164 × T + 0.066 × M + 0.012 × A − 2.838. Area under curve (AUC) at 95% confidence interval for our discriminant function was 0.845 (standard error [SE] 0.054). In comparison, AUC of Leiden prediction score was 0.897 (SE 0.043). CONCLUSIONS: Leiden prediction rule is highly applicable to UA patients to predict progression of RA in Indian patients and larger multi-center study with larger cohorts is needed to validate the formulation we derived to predict RA.