BACKGROUND: Qualitative polymerase chain reaction (PCR) for the identification of cytomegalovirus (CMV) infection has a low predictive value for the identification of CMV pneumonia. This study prospectively evaluated the application of a quantitative PCR Enzyme-Linked Immuno-Sorbent Assay (ELISA) assay in 9 lung- and 18 heart-transplant recipients who did not receive ganciclovir prophylaxis. METHODS: DNA was collected from peripheral blood polymorphonuclear leucocytes (PMNL) posttransplantation. Oligonucleotide primers for the glycoprotein B gene (149 bp) were used in a PCR ELISA assay using an internal standard for quantitation. CMV disease was defined as histological evidence of end organ damage. RESULTS: The median level CMV genome equivalents in patients with CMV disease was 2665/2 x 10(5) PMNL (range 1,200 to 61,606) compared to 100 x 10(5) PMNL (range 20 to 855) with infection but no CMV disease (p = 0.036). All patients with CMV disease had genome equivalents levels of >1200/2 x 10(5) PMNL. A cut-off level of 1,200 PMNL had a positive predictive value for CMV disease of 100% and a negative predictive value of 100%. The first detection of levels of CMV genome equivalents above a level of 1200/2 x 10(5) PMNL was at a median of 58 days (range 47 to 147) posttransplant. CONCLUSIONS: Quantitative PCR assays for the diagnosis of CMV infection may predict patients at risk of CMV disease and thereby direct preemptive treatment to high-risk patients.