The role of markers of cytomegalovirus (CMV) infection, such as the isolation of CMV, the presence of CMV antigenemia, or detection of viral DNA by polymerase chain reaction (PCR) assay, as predictors of subsequent CMV disease has been examined in recent studies. We studied the value of performing surveillance cultures of blood, urine, and throat specimens in a cohort of 156 liver transplant recipients who had participated in clinical trials and had received ganciclovir only for documented CMV disease. Cultures of urine and throat specimens for detection of CMV were performed every week, and cultures of blood specimens were performed every other week for the first 2 months after transplantation, then monthly for 6 months. Eighty-nine (57%) of 156 patients developed CMV infection, 41 (46%) of whom developed clinical CMV disease (36 had organ involvement and five had CMV syndrome). Fifty (32%) of 156 patients had positive blood cultures, 35 (22%) had positive urine cultures, and 41 (26%) had positive throat cultures. The positive and negative predictive values of surveillance cultures for predicting CMV disease were as follows: blood cultures, 46% and 83%, respectively; urine cultures, 26% and 74%, respectively; and throat cultures, 32% and 76%, respectively. These data indicate that such cultures are not useful in predicting CMV disease in liver transplant recipients. Future studies should examine the value of alternative markers, such as CMV antigenemia or the detection of viral DNA by PCR, for predicting CMV disease in this setting.