To determine whether cytomegalovirus (CMV) antigenemiaguided ganciclovir treatment may be as effective, may require less treatment, and thus may cause less marrow toxicity than ganciclovir administered at engraftment, 226 marrow transplant recipients were randomized at engraftment to receive placebo (antigenemia-ganciclovir group) or ganciclovir (ganciclovir group) until day 100 in a double-blind study. In patients with antigenemia of 3 or more positive cells in 2 slides and/or viremia, study drug was discontinued and ganciclovir was started for at least 3 weeks or until negative CMV antigenemia and resumed only if antigenemia recurred. More patients in the antigenemia-ganciclovir group developed CMV disease before day 100 after transplantation compared with the ganciclovir group (14% v 2.7%, P = .002). Of the 16 patients with CMV disease before day 100 in the antigenemia-ganciclovir group, 10 (8.8%) had disease before or during the first episode of antigenemia and 6 (5.3%) developed disease after discontinuation of ganciclovir. Untreated low-grade antigenemia progressed to CMV disease in 19% of patients with grade 3–4 compared with 0% of patients with grade 0–2 acute graft-versus-host disease (P = .04). There was no significant difference in CMV disease by day 180 after transplantation and thereafter. CMV-related death, transplant survival, and neutropenia were not significantly different between the groups. In the ganciclovir group, more invasive fungal infections occurred (P = .03) and more ganciclovir was used (P <.0001). Thus, delaying the start of ganciclovir until highgrade antigenemia and discontinuing ganciclovir based on negative antigenemia results in more CMV disease by day 100 than ganciclovir administered at engraftment. However, ganciclovir at engraftment is associated with more early invasive fungal infections and more late CMV disease resulting in similar survival rates.