The aim of this study was to assess whether the sentinel node concept could be applicable to clinically early carcinoma of the esophagus. We studied ten consecutive cT1N0 patients who underwent radical esophagectomy with regional lymph node dissection. On the day before surgery, 99m-Tc tin colloid was injected endoscopically around the primary tumor. Lymphoscintigraphy was also performed about three hours after injection. Immediately after surgery, the radioactivity of all dissected lymph nodes was measured with a hand-held gamma probe. The radioactivity and the metastatic status assessed by routine histopathologic examination were compared. A total of six patients had hot spots detected by lymphoscintigraphy, of which the detection rate was 60% (6 of 10). The ex vivo hot node detection rate was 90% (9 of 10). Three patients were found to have metastatic nodes. In one patient, sentinel node mapping failed to identify any hot spot or hot node. In the other two patients, the metastatic nodes did not correspond to hot nodes. The accuracy of hot node status was 77.8% (7 of 9), and the false-negative rate was 100% (2 of 2). The present study showed that radio-guided sentinel node detection is insufficiently reliable at present due to the high false-negative rate and low accuracy.