In a prospective longitudinal study, detection of cytomegalovirus (CMV) DNA in plasma (plasma polymerase chain reaction [PCR]) was compared with PCR of CMV DNA in peripheral blood leukocytes (PBL PCR), the CMV pp65 antigenemia assay, and viral cultures from blood, urine, and throat of 29 patients, 14 of whom received pp65 antigenemia-guided early ganciclovir treatment and 15 of whom received ganciclovir at engraftment. Among 328 blood samples tested by all methods, PBL PCR was the most sensitive test, followed by the pp65 antigenemia assay, plasma PCR, and viremia. In the 14 patients who received pp65 antigenemia-guided early treatment, the incidence of PBL PCR, pp65 antigenemia, plasma PCR, and viremia before day 100 was 79%, 79%, 71%, and 27%, respectively, with a median day of onset of day 32, 42, 45, and 51, respectively. Nine patients (64%) became positive by PBL PCR, pp65 antigenemia, and plasma PCR. Of 15 patients who were treated with ganciclovir at engraftment, 12 (80%) became positive by PBL PCR, plasma PCR, and/or pp65 antigenemia while receiving ganciclovir; 3 (20%) had breakthrough infection with all three methods, including 2 with high-grade antigenemia (more than three positive cells in duplicate staining); none of these patients subsequently developed positive CMV cultures or disease. In 49 specimens, PBL PCR and/or pp65 antigenemia assay could not be performed because of insufficient neutrophil counts. In conclusion, the sensitivity of plasma PCR is significantly lower than that of PBL PCR but similar to that of the pp65 antigenemia assay. Plasma PCR may be particularly useful in clinical situations in which a less sensitive and possibly more specific assay is warranted or in which leukocyte counts are inadequate to perform cell-based assays.