Reverse transcriptase polymerase chain reaction (RT-PCR) is increasingly used to detect small numbers of circulating tumour cells, though the clinical benefit remains controversial. The largest single contributing factor to the controversy of its value is the different approaches to sample processing. The aim of this study was to compare the sensitivity and reproducibility of RT-PCR for the detection of tumour cells after four commonly used different methods of sample processing. Using RT-PCR, one tumour cell spiked in 2 ml of whole blood was detected after analysis of separated mononuclear cell RNA, whole blood total or poly-A+ RNA. No false positives were identified with any method. However, the reproducibility of tumour cell detection was reduced after isolation of the mononuclear cell fraction. Only analysis of poly-A+ RNA had a sensitivity of 100% in all the cell spiking experiments. In patient blood samples, analysis of poly-A+ RNA increased the number of blood samples positive for tyrosine hydroxylase (TH) mRNA compared with those positive after analysis of total RNA. This may reflect high levels of cDNA reducing the efficiency of the PCR. Isolation of poly-A+ RNA increases the sensitivity and reproducibility of tumour cell detection in peripheral blood.