We conducted a randomized, double-blind comparison of prednisone and placebo (group I) v prednisone and azathioprine (1.5 mg/kg/day) (group II) as early treatment of extensive chronic graft-v-host disease (GVHD). Patients with platelet counts less than 100,000/microL were placed into therapy with prednisone alone (group III). All three groups received identical doses of prednisone (1 mg/kg every other day) and one double-strength trimethoprim-sulfamethoxazole (TMP-SMX) tablet twice daily. Between January 1980 and December 1983, 179 previously untreated patients were enrolled and 164 were evaluable. Patients randomized to group I (n = 63) and group II (n = 63) were well matched for prognostic factors; those placed into group III (n = 38) had more frequent acute GVHD and progressive onset of chronic GVHD. Median duration of therapy was 2 years. Complications included diabetes (5%), aseptic necrosis (5%) and infection. For groups I, II, and III, the respective incidence of infection was disseminated varicella, 11%, 24%, 34%; bacteremia, 6%, 11%, 34%; and interstitial pneumonia, 5%, 14%, 18%. Recurrent malignancy was the most frequent cause of death and did not differ significantly across the groups. Nonrelapse mortality, however, did differ: 21% in group I, 40% in group II, and 58% in group III (I v II, P = .003; I v III, P = .001). Forty patients in group I, 30 in group II, and 10 in group III survive with a minimum follow-up of 3.8 years. Karnofsky performance scores for 68 survivors are 90% to 100%, scores for seven survivors are 70% to 89% and scores for five survivors are less than 70%. Actuarial survival at 5 years after transplant is 61% in group I, 47% in group II, and 26% in group III (I v II, P = .03; I v III, P = .0001). Treatment with prednisone alone results in fewer infections and better survival than prednisone and azathioprine in standard-risk chronic GVHD. Treatment with prednisone alone is less effective in high-risk patients with thrombocytopenia, and other strategies are required.