Background. Rapid quantifiable diagnostic techniques for the diagnosis of cytomegalovirus (CMV) infection may predict patients at risk of CMV pneumonitis and allow preemptive antiviral treatment. Methods. Using CMV antigenemia as a prospective surveillance technique for CMV infection, we compared the outcome of preemptive treatment (PT) with ganciclovir, 10 mg/kg/day for 21 days directed by 'high levels' of CMV antigenemia (PT group, n=19), with the outcome in a group of historical controls (n = 18) treated with ganciclovir when CMV illness occurred. Greater than 50 antigen-positive cells per 2 x 105 polymorphonuclear leukocytes was considered to be high- level antigenemia. Results. Nine of the 18 controls developed high-level CMV antigenemia at a median of 38 days (range: 13-65 days) and 5 of the 9 developed CMV disease. Ten of the 19 PT group had high levels of CMV antigenemia detected at a median of 47 days (range: 20-63 days) and were given ganciclovir; none developed CMV disease. There was a significantly lower incidence of CMV disease in the PT group in comparison to controls (0 of 19 vs. 5 of 18: P=0.019). Conclusion. We have reduced the incidence of CMV disease using preemptive treatment, and because of a 100% negative predictive value, we omitted unnecessary antiviral prophylaxis for many at-risk patients.