How to Define Boolean Low Disease Activity in Rheumatoid Arthritis: Experience from a Large Real-World Cohort

Abstract Objectives: To propose Boolean-defined low disease activity (LDA) and to test its utility in rheumatoid arthritis (RA). Methods: We used data from a longitudinal academic clinical database of RA in Peking University First Hospital over a decade. The initial proposal of Boolean-defined LDA with ascending thresholds from 2 to 5 in steps of 1. Agreement and residual swollen joint count (SJC) pattern with the index-based (Simplified Disease Activity Index [SDAI] and Clinical Disease Activity Index [CDAI]) LDA was analyzed. To confirm discovery, we randomly classified RA patients in a 3:2 ratio into either analysis cohort or validation cohort.Results: In total 4,881 visits of 672 patients were included in the analysis cohort. Of these visits, the frequencies of achieving LDA were 71.9% (SDAI), 73.6% (CDAI), 52.8% (Boolean-LDA2), 65.2% (Boolean-LDA3), 73.5% (Boolean-LDA4), and 80.7% (Boolean-LDA5). High consistency and similar SJC pattern with SDAI-LDA or CDAI-LDA were observed in Boolean-LDA3 (kappa=0.796, 0.771). Further analysis found meeting SDAI-LDA but not Boolean-LDA3 were largely attributable to higher patient’s global assessment (PGA) scores (62.9%). In further modification of Boolean-LDA3, best agreement with SDAI-LDA or CDAI-LDA was reached when evaluator’s global assessment (EGA) replaced PGA with cutoff of 3.0 (kappa=0.851, 0.825), rather than exclusively increasing the PGA cut-offs. These findings were further replicated in randomly-generated validation cohort of 449 patients with 3,306 clinic visits. Conclusions: Using cut-off of 3 to Boolean-LDA with a substitute EGA for PGA provides highest consistency and similar residual SJC pattern with index-based LDA. This may deserve considering in clinical practice.