The polymerase chain reaction was used to detect cytomegalovirus (CMV) in 91 formalin-fixed paraffinembedded needle biopsies from 38 liver transplant patients with allograft dysfunction. Thirty donor liver biopsies served as negative controls. PCR results were compared with light microscopy (LM), immunohistochemical staining (IH) for CMV early and late antigen, and clinical data. Primers to the major immediate early gene (MDE) and the viral DNA polymerase gene were duplex amplified. PCR product was reamplified with a nested primer set for the MIE and confirmed by electrophoretic mobilities and dot blotting. Primers for human β-hemoglobin were used as internal controls. Seventeen of 38 patients had clinical evidence of cytomegalovirus disease, 12 of these were EH-positive, 14 were LM-positive, 15 were duplex PCR-positive and 17 were nested PCR-positive. In addition, duplex PCR was positive in one patient without other evidence of CMV disease, while nested PCR was positive in 12 such patients. The sensitivity and negative predictive value of nested PCR was 100%—however, the specificities and positive predictive values were only 42.9 and 58.6%, respectively. The control group was completely negative by LM, IH, and duplex PCR, however, 6 of 30 patients were nested PCR-positive. The number of nested-positive, duplex-negative patients without CMV disease was significantly greater in the transplant group versus the control group (12/21 vs. 6/30, P<0.009). The incidence of IgG seropositivity was also significantly greater in the transplant group versus the controls (29/32 vs. 15/24, P<0.02). We conclude that nested PCR may be an overly sensitive technique for the detection of clinically relevant CMV disease. A negative nested PCR assay for CMV may, however, help rule-out symptomatic CMV infection in an individual case. Duplex PCR showed little advantage over LM, while IH was confirmatory but did not add any new information in this study.