Forty-eight patients who provided 2 consecutive blood samples that tested positive for cytomegalovirus DNA by polymerase chain reaction (PCR) were randomized to receive either full-dose ganciclovir (5 mg/kg intravenously [iv] twice daily) or half-dose ganciclovir (5 mg/kg iv once daily) plus half-dose foscarnet (90 mg/kg iv once daily) for 14 days. In the ganciclovir arm, 17 (71%) of 24 patients reached the primary end point of being CMV negative by PCR within 14 days of initiation of therapy, compared with 12 (50%) of 24 patients in the ganciclovir-plus-foscarnet arm (P = .12). Toxicity was greater in the combination-therapy arm. In patients who failed to reach the primary end point, baseline virus load was 0.77 log10 higher, the replication rate before therapy was faster (1.5 vs. 2.7 days), and the viral decay rate was slower (2.9 vs. 1.1 days) after therapy. Bivariable logistic regression models identified baseline virus load, bone-marrow transplantation, and doubling time and half-life of decay as the major factors affecting response to therapy within 14 days. This study did not support a synergistic effect of ganciclovir plus foscarnet in vivo.