Patient, disease, and therapy-related factors that influence discontinuation of disease-modifying antirheumatic drugs: a population-based incidence cohort of patients with rheumatoid arthritis.

OBJECTIVE A major challenge in management of rheumatoid arthritis (RA) is prediction of longterm response to disease-modifying antirheumatic drug (DMARD) treatment. Our objective was to identify the predictors of DMARD discontinuation in an incidence cohort of patients with RA followed continuously from their incidence date. METHODS Members of a population-based incidence cohort of Rochester, Minnesota, residents aged > or = 18 years diagnosed with RA (by 1987 American College of Rheumatology criteria) from January 1, 1955, to January 1, 1995, were followed longitudinally through their complete medical records until January 1, 2001. Detailed drug exposure data were collected on all DMARD and glucocorticoid regimens. Subjects were considered exposed to a DMARD if duration of use was > or = 30 days. Time to discontinuation of DMARD was estimated using survival analysis techniques. Andersen-Gill models with multiple events per patient were used to assess the influence of demographics, calendar time, comorbidities, disease characteristics [disease duration, rheumatoid factor (RF), erythrocyte sedimentation rate (ESR), joint counts, radiographic changes, nodules, RA complications], and therapy characteristics (DMARD use, singly or in combination, glucocorticoid use, first or subsequent regimen, effect of previous therapy) on time from DMARD initiation to discontinuation. RESULTS The study population comprised 345 DMARD-treated patients (73% female) with mean age of 53.1 years and mean followup 15.4 years. Median time taking any DMARD was 16.0 months for the first, and 17.9 months for all regimens. Methotrexate (MTX) had the longest time to discontinuation, with a median of 30.3 months without folate, and 61.7 months with folate supplementation. Among the various disease characteristics examined, only higher ESR at DMARD initiation was significantly associated with a shorter time taking DMARD [hazard ratio (HR) 1.05 per 10 mm/h increase, 95% CI 1.02, 1.08]. In multivariable Andersen-Gill models considering all DMARD regimens, hydroxychloroquine use (HR 0.77, 95% CI 0.64, 0.92) and MTX use (HR with folate 0.39, 95% CI 0.30, 0.51; HR without folate 0.51, 95% CI 0.39, 0.67) were significantly associated with longer time to DMARD discontinuation, whereas prior MTX use (HR 1.96, 95% CI 1.57, 2.45) was associated with shorter time to DMARD discontinuation, after adjusting for age, sex, calendar year, Charlson comorbidity index, disease duration, and ESR at DMARD initiation. Disease duration was negatively associated with time to DMARD discontinuation; each 10 year increase in disease duration corresponded to a 14% decrease in the risk of discontinuation (HR 0.86, 95% CI 0.75, 0.98). CONCLUSION Longer RA disease duration does not appear to increase the risk of DMARD discontinuation. However, high disease activity (as assessed by ESR) is associated with a higher likelihood of discontinuing DMARD. MTX failure may identify a subgroup of patients who are less likely to respond to other DMARD and therefore could be considered as candidates for biological therapies.