Tumor relapse occurs frequently in patients with ductal pancreatic head cancer despite the absence of residual tumor detectable at primary surgery. Therefore it has to be assumed that current tumor staging procedures fail to detect minimal amounts of disseminated tumor cells present in secondary organs, which might be the seed for subsequent metastatic relapse. We evaluated lymph nodes from 18 patients without overt metastases who had undergone radical tumor resection (R0 resection). Lymph nodes judged as "tumor-free" by routine histopathology were further examined for the presence of single tumor cells using immunohistochemistry with the antiepithelial monoclonal antibody Ber-EP4. Sixteen of 37 "tumor-free" lymph nodes (43.2%), obtained from 13 of 18 patients (72.2%), displayed Ber-EP4+ tumor cells. Twelve of these 18 patients presented at pT2 stage. Nine of 12 patients (75%) staged as pN0 had these cells. Two of six pN1 patients had no Ber-EP4+ in histopathologically tumor-free lymph nodes. Using multivariate Cox's regression analysis, the presence of Ber-EP4+ cells in "tumor-free" lymph nodes was an independent factor for a significantly reduced relapse-free survival (p = 0.006) and overall survival (p = 0.01). Remarkably, all patients who were restaged as lymph node negative by both histopathology and immunohistochemistry survived the observation period without recurrence. The frequent occurrence of disseminated tumor cells in patients with pancreatic cancer and their prognostic impact support the need for a refined staging system of excised lymph nodes, which should include immunohistochemical examination. Thus patients with a minimal residual tumor load who might benefit from an adjuvant therapy could be selected.