The optimal pathological assessment of sentinel nodes (SLNs) in breast cancer is a matter of debate. Currently, multilevel histological evaluation and immunohistochemistry (IHC) are recommended, but alternative RT-PCR procedures have been developed. To assess the reliability of these different procedures, we devised a step-sectioning protocol at 100 micron-intervals of 74 SLNs using methacarn fixation. mRNA was extracted from sections collected from levels 4 to 5. Mammaglobin, CEA and CK19 were used for RT-PCR. mRNA extraction was successful in 69 SLNs. Of these, 7 showed macrometastases (>2mm), 2 showed micrometastases (<2 mm) and 7 showed isolated tumour cells (ITC) by IHC. RT-PCR was positive for the three markers in 6 of 7 macrometastases and in 1 of 2 micrometastases. In the 2 RT-PCR negative cases, metastases were detected only on sections distant from those analysed by RT-PCR. CEA and/or CK19 were positive by RT-PCR in 3 of 7 ITC and in 23 morphologically negative SLNs. In conclusion, the main goal of our study was to show that the use of alternate sections of the same sample for different procedures is the key reason for the discrepancies between molecular and morphological analyses of SLN. We believe that only prospective studies with quantitative mRNA analysis of specific metastatic markers on the whole lymph node can elucidate the utility of molecular assessments of SLN.