354 patients with multiple sclerosis were randomised to receive either azathioprine 2·5 mg/kg daily or placebo in a double-masked trial. During follow-up of at least 3 years only small differences emerged between the groups. After 3 years the mean deterioration in Kurtzke disability score was 0·62 in the azathioprine group and 0·80 in the placebo group, a difference of 0·18 (95% confidence intervals [CI] -0·15 to +0·52) and in the ambulation index it was 0·84 and 1·25, respectively, difference 0·41 (95% CI 0·03 to 0·80). After 3 years there had been slightly fewer relapses in the azathioprine group (average 2·2) than in the placebo group (average 2·5) but the difference of 0·3 (95% CI -0·2 to +0·9) was not significant. Although the results favour a small beneficial effect from azathioprine the benefit is so small that the use of azathioprine cannot be generally recommended for most patients with multiple sclerosis. Analysis of subgroups (by sex, age, severity, rate of progression, HLA status, relapsing or progressive course) has not revealed any that have shown clear clinical benefit.