Differential influence of Clinical Disease Activity Index components based on disease state in rheumatoid arthritis patients: real-world results from the Ontario Best Practices Research Initiative.

Abstract

The Clinical Disease Activity Index (CDAI) is routinely used in clinical care when treating-to-target RA patients. Previous validation studies have looked at CDAI's overall performance; this analysis aimed at evaluating its properties by disease state and identifying drivers of variance. RA patients enrolled in the OBRI registry, with available follow-up of ≥6 months were included. Construct validity of CDAI was assessed with principal component analysis; internal consistency with Cronbach's alpha (α); correlational validity with Spearman's rho (ρ); agreement in disease state classification with the kappa statistic. Stratification by disease states was performed. CDAI correlation with DAS28 was strong when CDAI>10 (ρ=0.79), moderate when CDAI≤10 (ρ=0.56) or 2.8<CDAI≤10 (ρ=0.44), and very weak when CDAI≤2.8 (ρ=0.02). Classification agreement of low disease activity (LDA)/remission between CDAI (≤10) and DAS28 (≤3.2) or HAQ-DI (≤2) was fair-to-moderate, while poor-to-fair for remission. For CDAI>10, CDAI was able to be reduced to a single component with patient global assessment (PtGA) having the lowest loading. When CDAI≤10, two distinct components were identified: (1) PtGA and physician global assessment; (2) SJC28 and TJC28. Moderate levels (α=0.71) of internal consistency were observed when CDAI>10 but low when CDAI≤10 (α=0.23), 2.8<CDAI≤10 (α=-0.42) or CDAI≤2.8 (α=-0.35). PtGA showed lowest correlation with remaining CDAI components. CDAI and DAS28 correlated well in moderate/high disease activity and poorly in LDA/remission. PtGA correlated weakly with other CDAI components and had a stronger influence on CDAI in LDA/remission. Thus, careful interpretation of PtGA is necessary when making treatment decisions, particularly in patients in CDAI LDA who are non-remitters.