Data concerning the specificity of cytokeratin 20 (CK20) as a reverse transcriptase-polymerase chain reaction analysis (RT-PCR) marker to detect disseminated tumor cells in blood are conflicting. Underlying causes for these discrepancies need to be determined to clarify the significance of CK20 detection. Because differences in RT-PCR assays and blood sample handling may be important, their influence on CK20 detection was studied. Using a series of healthy donor blood samples spiked with colon tumor cells, the authors compared the sensitivities of two conventional PCRs with different primer sets and a quantitative LightCycler PCR (Roche Diagnostics GmbH, Penzberg, Germany). Additionally, the influence of sample collection and preparation on assay specificity was studied by examining CK20 expression in the mononuclear cell fraction (MNC) of the first and the second aliquot of blood drawn from healthy donors and in the granulocyte cell fraction. At the concentration of one spiked tumor cell/mL blood, the CK20 detection frequency varied from 17% and 67% for the conventional to 78% for the LightCycler PCR. In the unspiked samples, CK20 was detected in 0% and 8% of the conventional and in 11% of the LightCycler PCR tests. Quantitative analysis revealed that CK20 was expressed at a high level in the granulocyte samples. The results demonstrate that differences in assay sensitivity and sample handling influence CK20 detection in blood.