OBJECTIVE To evaluate the effect of zidovudine on human immunodeficiency virus type 1 (HIV-1)-associated central nervous system infection in Centers for Disease Control and Prevention stage II or III disease. METHODS In an open-ended trial, patients received 500 mg of zidovudine twice a day for 12 months. Lumbar punctures, neurological, neuropsychological, and neuroradiological examinations were repeatedly performed during the trial period and were compared with pretrial values. In 11 patients post-trial neurological follow-up of 10 to 20 months was performed. METHODS Initially, 14 volunteers with stage II or III disease and intrathecal synthesis of HIV-1-specific antibodies were enrolled. Additionally, patients had slight neuropsychological disturbance or brain atrophy unrelated to other agents than HIV-1. Two patients dropped out because of poor compliance. METHODS Intrathecal and systemic immune and virological responses, cognitive performance, and brain images were repeatedly monitored. RESULTS After 6 weeks of zidovudine therapy, initial low-grade pleocytosis and elevated levels of beta 2-microglobulin, both in cerebrospinal fluid and in serum samples, declined. Intrathecal HIV-1 antibody synthesis could no longer be detected in half of the patients after 12 months of zidovudine therapy. Patients with defective cognition transiently improved cognitive speed and flexibility after 6 months of therapy. Slight atrophic brain changes, however, remained unchanged. CONCLUSIONS Zidovudine reduces intrathecal immuno-activation and transiently improves cognitive functioning in HIV-1-infected subjects who show evidence of central nervous system involvement by HIV-1 but are otherwise asymptomatic.