Seventy-four HIV-infected homosexual males belonging to CDC groups IIB, III, and IVC2 were treated with increasing doses of zidovudine within the Multicentre Canadian AZT Trial. Following a 3 week observation period, consenting volunteers received 600 mg/day for 18 weeks, 900 mg/day for 9 weeks, and 1,200 mg/day for 9 weeks. This was followed by a 6 week washout period after which zidovudine was restarted at 1,200 mg/day for 18 weeks. Patients were followed for a total of 63 weeks. Every 3 weeks they underwent a full clinical and laboratory assessment. For the purpose of this analysis, subjects were divided according to the mean initial platelet value (≥150,000 or <150,000/L) into normals (n = 57) and thrombocytopenics (n = 12), respectively. Analysis of variance was used to compare the mean platelet values at each zidovudine dose. All comparisons were made against baseline values. Zidovudine increased platelet counts in normal and thrombocytopenic subjects. The magnitude of this effect varied depending on the baseline platelet count. Among normals, the platelet count increased from 241,000 ± 45,000/L at baseline to 261,000 ± 51,000/L (p < 0.01) while receiving 600 mg/day of zidovudine. This effect was self-limited, reaching a peak by week 3. The platelet count decreased to baseline values despite increasing the dose of zidovudine to 900 or 1,200 mg/day. The platelet count further decreased to 218,000 ± 43,000/L during the washout phase (washout vs. 1,200 mg, p < 0.01). The platelet count increased to 238,000 ± 50,000/L upon reinstitution of zidovudine at 1,200 mg/day (reinstitution vs. washout; p < 0.01). Among thrombocytopenics, the platelet count increased from 81,000 ± 37,000/L at baseline to 129,000 ± 51,000/L (p < 0.01) while receiving 600 mg/day of zidovudine. This effect reached a peak by week 9 and was sustained for 36 weeks with counts of 117,000 ± 43,000 and 119,000 ± 41,000/L while on 900 and 1,200 mg/day of zidovudine, respectively (900 mg/day vs. baseline, p < 0.01 and 1,200 mg/day vs. baseline, p < 0.01). This effect was not found to be dose related. The platelet count decreased to 101,000 ± 34,000/L during the washout phase (washout vs. 1,200 mg/day, p < 0.07). The platelet count increased once again upon reinstitution of zidovudine at 1,200 mg/day (reinstitution vs. washout p < 0.01). We conclude that zidovudine increases the platelet count in HIV-infected individuals. This effect is drug related but dose independent within the studied dose range. Furthermore, the effect of zidovudine on platelets appears to be more marked and sustained among thrombocytopenic individuals. Based on our findings, HIV-associated thrombocytopenia should be considered a relative indication for zidovudine therapy.