BACKGROUND: CMV DNA screening may be a useful adjunct to serologic tests in distinguishing potentially infectious blood donations from those that are "CMV-safe." However, there is currently no consensus on the optimal assay method for accurate detection of CMV DNA in donors. STUDY DESIGN AND METHODS: A blinded multicenter evaluation of seven CMV PCR assays was performed by five laboratories by using coded sets of analytical controls and donor blood samples. RESULTS: Five assays displayed sufficient sensitivity for donor screening, as judged by consistent detection of a minimum of 25 CMV genome equivalents (geq) in analytical controls constructed to contain from 1 to 100 CMV geq in background DNA from 250,000 cells, while the other two assays displayed inadequate sensitivity. Three sensitive assays, two based on nested PCR directed at the UL93 and UL32 regions of the CMV genome and another test (Monitor Assay, Roche), did not detect CMV DNA in samples from any of 20 pedigreed CMV-seronegative, Western blot-negative (S-/WB-) donors. Two other assays based on nested PCR occasionally detected CMV DNA in S-WB- samples, and one sensitive nested PCR assay directed at UL123 detected CMV DNA in a large proportion (85%) of S-WB- samples. CONCLUSION: Seven CMV PCR assays currently used for research and/or diagnostic applications displayed marked variations in sensitivity, specificity, and reproducibility when applied to coded analytical and clinical control samples containing cellular DNA from the equivalent of 250,000 WBCs. These results will be useful in the selection of assays with performance characteristics appropriate to donor screening objectives. They may also help explain discrepant findings from previous studies that used PCR to determine CMV DNA prevalence in seronegative and seropositive blood donors.