To better understand the biologic meaning and potential clinical utility of p24 antigen measurements in human immunodeficiency virus (HIV) infection, p24 antigen and antibody and HIV RNA were quantitated in parallel. Specimens (n = 311) were analyzed from 74 participants in a zidovudine treatment study. Parallel antigen and RNA measurements revealed the frequent occurrence of two types of discordant results. First, p24 antigen was often not detected in samples with high antibody levels even when > 10(6) RNA copies/mL were present. Second, in specimens in which p24 antigen was detected, the concentration was greater than expected on the basis of HIV RNA values. These results suggest that optimal use of serum p24 antigen values will require consideration of both specific antibody levels and non-virion associated antigen.