We assessed the safety and surrogate markers' effect of acemannan as an adjunctive to antiretroviral therapy among patients with advanced HIV disease receiving zidovudine (ZDV) or didanosine (ddI) in a randomized, double-blind, placebo-controlled trial of acemannan (400 mg orally four times daily). Eligible patients of either sex had CD4 counts of 50-300/microl twice within 1 month of study entry and had received 26 months of antiretroviral treatment (ZDV or ddI) at a stable dose for the month before entry. CD4 counts were made every 4 weeks for 48 weeks. P24 antigen was measured at entry and every 12 weeks thereafter. Sequential quantitative lymphocyte cultures for HIV and ZDV pharmacokinetics were performed in a subset of patients. Sixty-three patients were randomized. All were males (mean age 39 years). The mean baseline CD4 counts were 165 and 147/microl in the placebo and acemannan groups, respectively; 90 percent of the patients were receiving ZDV at entry. Six patients in the acemannan group and five in the placebo group developed AIDS-defining illnesses. There was no statistically significant difference between the groups at 48 weeks with regard to the absolute change or rate of decline at CD4 count. Among ZDV-treated patients, the median rates of CD4 change (ACD4) in the initial 16 weeks were - 121 and - 120 cells per year in the placebo and acemannan groups, respectively ( p = 0.45), ACD4 from week 16 to 48 was 0 and - 61 cells per year in the acemannan and placebo groups (p = .11), respectively. There was no statistical difference between groups with regard to adverse events, p24 antigen, quantitative virology, or pharmacokinetics. Twenty-four patients, 11 receiving placebo and 13 receiving acemannan, discontinued study therapy prematurely, none due to serious adverse reactions. Our results demonstrate that acemannan at an oral daily dose of 1600 mg does not prevent the decline in CD4 count characteristic of progressive HIV disease. Acemannan showed no significant effect on p24 antigen and quantitative virology. Acemannan was well tolerated and showed no significant pharmacokinetic interaction with ZDV.