The aim of the present study, a multicentre trial of didanosine (ddI) compassionate use, was to identify factors associated with a better outcome in patients given ddI monotherapy. Enrolled were 1047 HIV-positive patients intolerant of and/or unresponsive to zidovudine (ZDV) therapy, with CD4+ cell counts of < 200/microliter or AIDS. Didanosine was given at a dose of 250 mg b.i.d. (patients > or = 60 kg) or 167 mg b.i.d. (patients < 60 kg). Clinical examinations and laboratory tests were performed every two months. Endpoints included death, the occurrence of a new AIDS-defining disease, or permanent discontinuation of ddI for a severe adverse event. At entry, the median CD41 cell count was 47/microliter and the median duration of prior ZDV treatment 19 months; 446 patients (43%) were classified as having AIDS. Severe toxicity occurred in 143 subjects (14%); the frequency of pancreatitis was very low (0.2%). The benefit in terms of CD4+ cell counts was greater for patients whose counts exceeded 100/microliter at entry and remained at this level until month 12 in those patients still receiving treatment. Death and/or new AIDS-defining events were observed in 374 cases (36%) over a median follow-up of eight months. AIDS dementia was observed in 11 patients (1%). Multivariate analysis of survival without disease progression showed that the factors associated with a worse outcome include the severity of immunodepression, a diagnosis of AIDS at entry, and a history of both intolerance of and clinical resistance to ZDV. Surprisingly, the patients who had received previous prolonged treatment with ZDV had a better outcome. In conclusion, severely immunodepressed patients previously administered long-term monotherapy may receive a short-term benefit from being switched to another antiretroviral drug.