Detection of rare disseminated tumor cells identifies head and neck cancer patients at risk of treatment failure.

This study was designed to help establish the most appropriate samples and tests to detect disseminated tumor cells (DTCs) for head and neck cancer patients. Samples of bone marrow (BM) and central venous blood (CVB), collected preoperatively, and BM and peripheral venous blood, collected 3 months transcription postoperatively, were screened for the presence of DTCs using immunocytochemistry (ICC) with a pan-cytokeratin antibody and E48 reverse transcriptase-PCR. The molecular approach was also applied to intraoperative CVB. The concordance between the molecular and ICC tests applied to preoperative samples, measured by Cohen's kappa, was not uniformly good, which likely reflected sampling errors, heterogeneity of E48 antigen expression, or stochastic effects. However, testing samples of BM and CVB preoperatively with the molecular or ICC approaches gave results that predicted disease-free survival and distant-metastases-free survival. Application of a single preoperative test predicted development of distant metastases, and the prediction could be improved by combining information derived from testing both CVB and BM. Applying two tests to the same sample of blood or BM served to validate the findings from a single assay but did not improve the prediction. Testing the intraoperative sample of CVB was also a sensitive predictor of distant metastases, but testing BM or blood 3 months postsurgery was not useful. These findings suggest that detection of DTCs pre- or intraoperatively indicates a high risk of local and distant recurrence and reduced survival in head and neck cancer.