The aim of this study was to determine whether flow cytometry (FACS) could detect spiked or circulating colorectal cancer cells. A flow cytometric assay was developed and its sensitivity compared with the reverse transcription polymerase-chain reaction (RT-PCR), using carcinoembryonic antigen (CEA) and cytokeratin (CK) 20 mRNA as target markers. Sensitivity limits for RT-PCR and flow cytometry (FACS) were established using spiked blood, and pre-operative blood samples from 20 colorectal cancer patients and 16 healthy no-cancer controls were analysed for circulating tumour cells (CTC) using both methods. Blood samples for FACS analysis were immuno-magnetically enriched using ferrofluid particles. CTC were defined as positive for pan-cytokeratin and negative for CD45 pan-leucocyte antigen (CK+/CD45- events). There was a significant (P < 0.0001) correlation between the number of spiked cancer cells and their recovery using FACS. The lowest detectable concentration was 20 spiked cancer cells in 14 ml blood for both RT-PCR and FACS. A positive FACS result significantly (P < 0.05) concurred with a positive RT-PCR result in spiked blood. The number of CK+/CD45- events detected in the blood of colorectal cancer patients was not significantly greater (P = 0.07) than in blood taken from 'no cancer' controls and furthermore there was no concordance (P = 1) between RT-PCR and FACS positivity in cancer patients' blood. FACS detection of tumour cells was feasible in vitro, and correlated with RT-PCR. However, its sensitivity in vivo was poor and did not correlate with RT-PCR detection of CTC. Uncertainties about antigen expression on normal circulating cells and about CTC phenotype need to be resolved, before FACS can be developed for detection of tumour cells within the circulation.