To assess changes in HIV RNA and their relationship to disease progression. Delta was a randomized double-blind trial comparing zidovudine (ZDV) monotherapy with ZDV plus didanosine (ddI) or ZDV plus zalcitabine (ddC). Participants had AIDS (with CD4 cell counts above 50 x 10(6)/l), AIDS-related complex, or were asymptomatic with CD4 cell counts below 350 x 10(6)/l. The trial included both ZDV-naive and ZDV-experienced participants. A total of 1280 participants in the Delta trial whose serum samples had been stored at -70 degrees C and who had a minimum of one sample taken before the start of treatment and at least one later sample. HIV-1 RNA quantification was performed using the nucleic acid sequence-based amplification HIV-1 RNA quantitative assay with a cut-off of 800 copies/ml. Reductions in HIV RNA by treatment group were consistent with the clinical results; in ZDV-naive participants the maximum median fall occurred at 4 weeks for all three groups (ZDV, 0.54 log10 copies/ml; ZDV-ddI, 1.38 log10 copies/ml; ZDV-ddC, 1.31 log10 copies/ml). On average the reductions were smaller in ZDV-experienced participants but the difference between the monotherapy and combination arms was very similar in ZDV-naive and experienced participants. Baseline HIV RNA levels, adjusted for CD4 cell counts were highly predictive of time to virological response (HIV RNA < 800 copies/ml); HIV RNA nadirs achieved were predictive of survival. Viral load rebound following response was independent of treatment group and previous ZDV therapy. Virological changes in response to treatment are of value in assessing prognosis and the activity of new therapies; in particular, there is a strong association between the minimum HIV RNA achieved in the first 16 weeks and subsequent clinical response. CD4 cell counts are independently predictive of response.