Preliminary preclinical and clinical data suggest that granulocyte-macrophage colony-stimulating factor (GM-CSF) may decrease viral replication. Therefore, 105 individuals with AIDS who were receiving nucleoside analogue therapy were enrolled in a placebo-controlled, double-blind study and were randomized to receive either 125 microgram/m(2) of yeast-derived, GM-CSF (sargramostim) or placebo subcutaneously twice weekly for 6 months. Subjects were evaluated for toxicity and disease progression. A significant decrease in mean virus load (VL) was observed for the GM-CSF treatment group at 6 months (-0.07 log(10) vs. -0.60 log(10); P=.02). More subjects achieved human immunodeficiency virus (HIV)-RNA levels <500 copies/mL at >/=2 evaluations (2% on placebo vs. 11% on GM-CSF; P=.04). Genotypic analysis of 46 subjects demonstrated a lower frequency of zidovudine-resistant mutations among those receiving GM-CSF (80% vs. 50%; P=.04). No difference was observed in the incidence of opportunistic infections (OIs) through 6 months or survival, despite a higher risk for OI among GM-CSF recipients. GM-CSF reduced VL and limited the evolution of zidovudine-resistant genotypes, potentially providing adjunctive therapy in HIV disease.