Two protocols for the extraction of cytomegalovirus (CMV) DNA and two methods for the amplification of CMV DNA in dried blood spots were evaluated for the retrospective diagnosis of congenital CMV infection. During the period from 1996 to 2006, a urine screening program detected 76 congenitally infected neonates. Stored Guthrie cards with blood from 55 cases and 12 controls were tested. Two spots of dried blood were cut from each card and evaluated in two centers. CMV DNA was extracted from a whole single spot. Center 1 used phenol-chloroform extraction and ethanol precipitation followed by a conventional PCR. Center 2 used the NucliSens easyMAG automated DNA/RNA extraction platform (bioMérieux) followed by a real-time PCR. For evaluation of the extraction method, DNA extracted from each blood spot was evaluated by the amplification method used by the collaborating center. The sensitivities were 66% for center 1 and 73% for center 2. None of the controls were positive. A sensitivity as high as 82% could be obtained by combining the most sensitive extraction method (the phenol-chloroform procedure) with the most sensitive PCR method (real-time PCR). The detection rate was not influenced by the duration of storage of the spots. The sensitivity was higher with blood from congenitally infected cases due to a primary maternal CMV infection, regardless of the protocol used. However, the difference reached significance only for the least-sensitive protocol (P = 0.036).