Metastasis to cervical lymph nodes (LN) is significantly associated with the outcome of patients with oral cancer. To provide a useful method for the detection of micrometastases, we analyzed 115 LNs from 10 patients with oral cancer using real-time quantitative polymerase chain reaction (PCR) based on the expression of squamous cell carcinoma antigen (SCCA) and cytokeratin 13 (CK13). The sensitivity and quantification of this method were assessed by means of limited dilution of cultured oral cancer cells and a model of cervical LN-metastasis established by inoculating green fluorescent protein (GFP)-expressing cells into the tongue of nude mice. In both investigations, a few cancer cells were detected by real-time quantitative PCR, but not by conventional reverse transcription-PCR (RT-PCR). SCCA mRNA was detected at high levels in metastatic LNs. In contrast, 26 of the 30 control cervical LNs did not express the gene at all, and the rest showed fairly low levels. Of 108 histologically metastasis-negative LNs, 19 (17.6%) expressed SCCA mRNA levels higher than the cut-off value (1.0: mean expression of control LNs + 2SD). CK13 mRNA is not a suitable marker for the real-time PCR since it was detected frequently even in the control LNs. These findings suggest that genetic diagnosis by real-time quantitative PCR based on SCCA mRNA expression may be clinically useful for detecting occult tumor cells in cervical LNs.