In curatively resected gastric cancer, the incidence of distant relapse is as high as 30%. Although the most important factor contributing to the local control of the tumor is the microscopic tumor-free margin of the surgical resection, the occurrence of distant metastases is in many cases due to preoperative or perioperative tumor cell dissemination. In addition to the established TNM staging system, disseminated tumor cells may serve as independent prognostic factors influencing patient outcome after curative surgery. Basically, in gastric cancer three compartments have been identified in which single tumor cells may be shed: lymph nodes, peritoneal cavity, and bone marrow. Assessment of resected regional lymph nodes with monoclonal antibodies directed against cytokeratin antigens leads to an upstaging in comparison with conventional histology. Nodal micrometastases detected by immunohistochemistry result in an upstaging of up to 36% of patients. However, their prognostic significance remains controversial. Local dissemination of tumor cells in the peritoneal cavity determines the outcome in advanced gastric cancer and diffuse-type carcinoma. Patients with negative peritoneal washings seem to have a more favorable prognosis. Moreover, with the use of these diagnostic tools, patient subpopulations may be identified which profit from intraperitoneal therapy regimens. Diffuse hematogenous tumor cell dissemination into the bone marrow has been shown to be a prognostic factor in several studies. In our own population of 180 gastric cancer patients, bone marrow cells were screened immunohistochemically with a monoclonal antibody directed against cytokeratin 18 (CK18). In 95 patients (53%), CK2-posititve cells were detected. In a multivariate analysis, the independence of the presence of three or more disseminated tumor cells per 10(6) mononuclear cells was proven to be a prognostic factor in patients with intestinal-type tumors, pT1/2 status, and pN0 status. In conclusion, the TNM status only partially reflects the actual extent of systemic disease in patients with resected gastric cancer. The assessment of minimal residual disease is valuable in estimating the prognosis in many patients. In the future, staging systems will have to not only include TNM data but also provide specific information on biological properties of residual cancer cells in order to establish more exact prognostic estimates and provide patients with an individually tailored multimodal treatment.