The safety, tolerability, and antiviral activity of atevirdine (ATV), a nonnucleoside reverse transcriptase inhibitor, were studied in a phase I/II clinical trial (ACTG 187) of patients with CD4 counts < or =500/mm3. In all, 34 HIV-1-infected patients were randomized to receive ATV for 12 weeks in doses chosen to achieve one of three serum trough levels: 5 to 13 microM, 14 to 22 microM, or 23 to 31 microM. Rash was the most common adverse event, with a grade 3 or 4 rash occurring in 4 patients. No significant change from baseline in HIV-1 plasma RNA mean copy number was detected at week 4 (+0.09 log10 copies/ml; p = .30). However, some evidence indicated moderate antiviral activity at week 4, based on median changes in CD4 count (+23/mm3; p = .05), and viral peripheral blood mononuclear cell (PBMC) titer (-0.68 log10) copies/ml; p = .03). In addition, 2 of 4 patients with detectable baseline serum p24 antigen showed declines of >50%. HIV-1 resistance to ATV was detected in 41% of patients and was most commonly associated with RT mutations K103N and Y181C. In contrast, the Y181C mutation was not detected in ATV-resistant isolates obtained from patients enrolled in ACTG 199, a study of ATV given in combination with zidovudine. Under the conditions of this study, ATV failed to demonstrate significant antiretroviral activity. However, transient in vivo activity might have been obscured by rapid development of resistance coupled with inadequate sampling at early time points following initiation of ATV therapy.