The risk of transfusion-transmitted viral infections may be estimated by several methods, but only prospective studies of transfusion recipients can directly measure the incidence, with associated 95% upper confidence bound, of these infections. From 1989 through 1995, 764 recipients of allogeneic or autologous red blood cell transfusions were enrolled; 486 (64%) provided both pretransfusion and 6-month follow-up specimens. Both specimens were tested for anti-HBc, anti-HCV, anti-HTLV-I and anti-HIV, with appropriate confirmatory testing. Thirty-nine (8.0%) subjects had seroprevalent anti-HBc, 19 (3.9%) subjects had seroprevalent anti-HCV, three (0.6%) subjects had seroprevalent anti-HTLV-I/II, and one (0.2%) subject had seroprevalent anti-HIV. There were no seroconversions for any agent among the 34 patients who received only autologous blood, and no confirmed seroconversions for anti-HTLV-I or anti-HIV among all subjects. There were three seroconversions for anti-HBc (incidence 1.04 x 10(-3); 95% confidence interval (CI) 2.15 x 10(-4), 3.05 x 10(-3) per allogeneic unit transfused), and two confirmed seroconversions for HCV (incidence 6.94 x 10(-4); 95% CI 8.34 x 10(-5), 2.51 x 10(-3) per allogeneic unit transfused). One of the two anti-HCV seroconversions occurred in March 1994, after the institution of HCV EIA 2.0 screening of donated blood. Transfusion-associated seroconversions to hepatitis B and C markers were observed at low rates in the early 1990s despite testing donors for markers of both viruses, whereas seroconversions to HTLV-I or HIV were less than 1.04 x 10(3) per allogeneic unit transfused, based upon the upper 95% confidence interval of the zero incidence in this study.