OBJECTIVE To compare the safety and efficacy of didanosine with that of continued zidovudine therapy in persons with human immunodeficiency virus (HIV) infection who had received zidovudine for at least 6 months and had CD4 cell counts of 200 to 500 CD4 cells/mm3. METHODS Double-blind, randomized controlled trial. METHODS 10 Canadian university-affiliated specialty clinics. METHODS 246 patients were assigned to receive standard doses of either zidovudine or didanosine. METHODS The primary clinical end point was the occurrence of a new, previously undiagnosed acquired immunodeficiency syndrome (AIDS)-defining illness or death. RESULTS 245 of 246 patients were eligible (118 receiving didanosine and 127 receiving zidovudine). Sixty-six percent were asymptomatic, 30% had AIDS-related complex, and 4% had AIDS. The median baseline CD4 count was 320 cells/mm3. The median previous duration of zidovudine therapy was 471 days. Nine new AIDS-defining illnesses developed during the study; all but one were in the zidovudine group (relative risk, 7.9 [95% CI, 1.0 to 63.3; P = 0.02]). A change to didanosine led to a statistically significant increase in CD4 counts by week 2 that persisted until the end of the study at week 48 (P < or = 0.01). Viral sensitivity studies (done in 102 patients) showed that 28% of the zidovudine group and 21% of the didanosine group had high-level in vitro resistance to zidovudine (50% inhibitory concentration greater than 0.8 microM) at baseline (P = 0.49). Only one patient in the didanosine group developed high-level resistance to zidovudine during the study. In the zidovudine group, the cumulative probability of developing high-level resistance to zidovudine was 59% at 1 year (P = 0.01). Abdominal pain, leukopenia, and neutropenia were more frequent in the zidovudine group, and hyperuricemia was more frequent in the didanosine group (P < 0.05). CONCLUSIONS In clinically stable patients with 200 to 500 CD4 cells/mm3 who had tolerated zidovudine for at least 6 months, a change to didanosine led to a decrease in the rate of disease progression, a sustained increase in CD4 counts, and a decrease in the chances of developing high-level resistance to zidovudine. Both drugs were generally well tolerated.