The potential advantage of using quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) methodology to detect metastasis in sentinel lymph nodes (SLNs) of breast cancer (BC) patients was evaluated in this prospective study. We measured the expression of relevant gene transcripts in SLNs using an innovative algorithm and compared the results of single-marker assays versus multi-marker assays with conventional histological detection methods. SLNs from women aged ≥18 years diagnosed with unilateral BC were examined by haematoxylin-eosin staining and immunohistochemistry and analysed for transcripts of several relevant genes using qRT-PCR (learning group). Four candidate panels of expressed transcript combinations with high sensitivity and specificity were selected for further investigation. The candidate panels were then validated using SLNs from a second group of BC patients (validation group). In the learning group, 74/314 SLN sections from 150 patients were positive for metastasis by histology. The transcripts analysed showed the following individual sensitivities/specificities: cytokeratin 19 (CK19) 94.6%/97.9%; mammaglobin 1 (MGB1) 82.4%/91.7%; mammaglobin 2 (MGB2) 82.4%/96.7%; carcinoembryonic antigen (CEA) 71.6%/97.5%; EPCAM (epithelial cell adhesion molecule) 91.9%/97.1%; and NY-BR-1 82.4%/93.8%. The optimal panel based on the predefined criteria comprised four markers: CK19, MGB1, EPCAM, and NY-BR-1, of which ≥2 had to be positive (95.9% sensitivity, 95.0% specificity, 85.5% positive predictive value (PPV), and 98.7% negative predictive value (NPV)). Overall concordance with histology was 95.2%. In the validation group, 84/315 SLN sections from 235 patients were histologically positive, and panel sensitivity, specificity and overall accuracy were 88.1, 95.2 and 93.3%, respectively, at the SLN section level. In conclusion, molecular staging using expression patterns of relevant transcripts in SLNs could serve as a useful complement to standard diagnostic work-up in BC patients. The proposed flexible multi-parametric approach does not improve the overall accuracy compared with the single-marker approach. However, it overcomes several limitations of the previously reported molecular assays for SLN diagnosis.