One hundred thirty-one patients with rheumatoid arthritis treated with either azathioprine sodium (n = 37, 102.7 +/- 32.9 mg/d) or methotrexate sodium (n = 94, 8.4 +/- 3.0 mg/wk) were followed up for 38 +/- 23.3 months to determine the nature, frequency, and potential predictors of "major" toxic reactions. Thirty-one methotrexate-treated patients (33%) and 11 patients (30%) receiving azathioprine experienced a major toxic reaction during the study period. With the case-control method, no predictors of major toxic reactions secondary to azathioprine therapy were found. Sex, drug dosage, response to prior slow-acting antirheumatic drug therapy, concurrent use of salicylates, and age did not predict major toxic reactions secondary to methotrexate treatment, but the methotrexate-treated patients who experienced a major toxic reaction had a significantly greater mean level of blood urea nitrogen at the time of their reaction compared with the control group. Life-table analysis suggested toxic reactions posed a greater risk of treatment termination in methotrexate-treated patients compared with the lack or loss of efficacy. This trend was not apparent in the azathioprine group. The majority of patients in each treatment group (79 for methotrexate and 29 for azathioprine) experienced one or more "minor" toxic reactions during the follow-up period.