To predict the probability of long-term viral suppression during treatment with zidovudine and lamivudine, human immunodeficiency virus type 1 (HIV-1) RNA values were retrospectively pooled for 1083 patients from six randomized, double-blind clinical trials. All analyses of HIV-1 RNA were obtained using the Roche Amplicor assay or its earlier prototype. Time to loss of response was evaluated by Kaplan-Meier analysis; Cox proportional hazards models were used to assess the influence of baseline variables. Among 523 patients with < or = 6 months of prior zidovudine treatment, the probability of HIV-1 RNA suppression below 400 copies/ml at 48 weeks was 71% in those with baseline HIV-1 RNA < 5000 copies/ml, but only 14% in those with HIV-1 RNA between 50,000 and 200,000 copies/ml. Among 560 patients with > 6 months of prior zidovudine treatment, the rates of sustained viral suppression were lower, but also significantly associated with the baseline HIV-1 RNA. Multivariate analyses showed no independent effect of CD4 cell count, age, sex, race, or CDC disease stage on the probability of sustained HIV-1 RNA suppression. When patients with < or = 6 months of prior therapy were stratified based on the magnitude of HIV-1 RNA nadir achieved during treatment, those who reached a nadir of < 400 copies/ml retained this response for significantly longer time periods than the ones who only achieved partial viral suppression. In conclusion, baseline HIV-1 RNA levels and the duration of prior zidovudine therapy strongly predict the antiretroviral efficacy of zidovudine/lamivudine. The baseline parameters should influence the choice of the antiretroviral regimen.