Three nucleoside analogues, zidovudine (AZT), didanosine (ddI), and zalcitabine (ddC), are approved for use in the treatment of patients with HIV infection. This retrospective study compares the 3 drugs and examines the overall utility of antiretroviral therapy by way of comparisons to a no treatment (No Rx) group in patients with advanced HIV disease. Patients with advanced HIV disease were enrolled in didanosine (August 1989-December 1990) or zalcitabine (October 1990-February 1992) expanded access programmes; continued on zidovudine treatment despite fulfilling criteria for zidovudine failure or intolerance; or maintained on no antiretroviral treatment. Statistical analysis revealed that patients on nucleoside analogue therapy had fewer opportunistic infections (P = 0.001) than those who received no antiretroviral treatment. The Kaplan-Meier 12-month estimate of survival was significantly longer among patients who switched from zidovudine to zalcitabine but not among patients who switched to didanosine, when compared to the other 2 groups (P = 0.05).