OBJECTIVE To evaluate the safety, tolerability, and acceptability of zidovudine used for prevention of human immunodeficiency virus (HIV) transmission to health care workers. METHODS Prospective study between January 1989 and December 1992 at a suburban New York City medical center of health care workers who had percutaneous or permucosal exposures to blood or body fluids of HIV-infected patients. The zidovudine regimen offered was 200 mg orally every 4 hours for 42 days. RESULTS Only 60 (53%) of 113 health care workers chose to take zidovudine, and only 21 (35%) of these health care workers completed the recommended 42-day course of therapy. Men were more likely than women to choose zidovudine therapy (P = .06), and a greater proportion of men completed the full 42 days of zidovudine therapy (P = .09). The most frequent reason for stopping treatment prematurely was clinical adverse reactions (n = 18 [30%]). Overall, clinical adverse events occurred in 44 (73%) of those taking zidovudine, with the most frequent events being nausea (n = 28 [47%]), headache (n = 21 [35%]), and fatigue (n = 18 [30%]). Selected laboratory parameters (hemoglobin level, hematocrit, alanine aminotransferase level, and white blood cell, granulocyte, and platelet counts) were compared at baseline and at week 4. Of interest, platelet counts were significantly more likely to increase by week 4 than to remain the same or decrease (P = .035). This observation may shed light on the mechanism of platelet elevation seen in zidovudine-treated patients infected with HIV. None of the laboratory changes was considered clinically significant. Of the 42 health care workers followed up for 3 months or longer (range, 3 to 32 months), none had undergone HIV antibody seroconversion. CONCLUSIONS Zidovudine therapy is poorly accepted and tolerated by health care workers, at least in the dosage regimen used in this study. Alternative prophylactic regimens are needed.