Functional improvement after patients with rheumatoid arthritis start a new disease modifying antirheumatic drug (DMARD) associated with frequent changes in DMARD: the CORRONA database

Abstract

Objective: We examined the relationships of rheumatoid arthritis (RA), disease duration (DD), number of previous disease modifying antirheumatic drugs (DMARD), and frequency of DMARD changes, with regard to changes in function in patients with RA evaluated by modified Health Assessment Questionnaire (mHAQ) after the start of a new DMARD.

Methods: In total, 889 patients with active RA from the CORRONA database [patients had mHAQ>or=0.5 and/or Disease Activity Score 28-joint count (DAS28)>or=1.6] started a new DMARD (baseline) and had at least one followup visit 6-12 mo later. Change in mHAQ from baseline to followup visit was modeled using univariate/multivariate linear regression analysis. Due to colinearity, separate multivariate regression models were performed including/excluding the predictors disease duration, number of prior DMARD, and frequency of DMARD changes.

Results: Baseline age, mHAQ, erythrocyte sedimentation rate (ESR), DAS28, and number of prior DMARD differed across DD groups. The univariate linear regression model showed that higher baseline values of mHAQ, DAS28, swollen joint count (SJC), tender joint count (TJC), Clinical Disease Activity Index (CDAI), ESR, physician global assessment, prednisone use, and subsequent addition/discontinuation of DMARD were associated with improvement of the mHAQ at followup (p=0.05). Multivariate linear regression models showed that mHAQ improvement was associated with shorter DD, higher baseline mHAQ, addition of subsequent DMARD, and the DMARD frequency index (no. previous DMARD/yrs of DD) (p<0.05). Number of DMARD patients used previously was not associated with change of mHAQ in either model.

Conclusion: Our study demonstrates that in clinical rheumatologic practices, more frequent changes in DMARD are associated with greater improvement in function (by mHAQ). It does not support the idea that number of previous DMARD used predicts response. Indirectly, these data support the concept that DMARD should be changed if optimal responses are not achieved within a specified time.