During a continuous 14-year observation period we prospectively recorded clinical data on all patients with rheumatoid arthritis (RA) attending an outpatient clinic. Six hundred seventy-one patients received 1017 new administrations of slow acting antirheumatic drugs during more than 2000 patient years of observation. The median time to discontinuation for intramuscular gold, auranofin, hydroxychloroquine or penicillamine was 2 years or less, but was 4.25 years for methotrexate (p = 0.008 vs all other drugs combined). Adverse reactions were a more common reason for discontinuation than efficacy, and both were less common in patients taking methotrexate (p less than 0.01). Neither disease duration, disease severity, or demographic factors were useful predictors of discontinuation. Since controlled clinical trials do not provide long-term outcome assessments, measurement of time to termination is a practical tool to estimate drug inefficacy.