The presence of isolated carcinoma cells detected immunocytochemically in bone marrow has been shown to be of prognostic relevance for cancer patients. Unfortunately, the immunocytochemical method (ICC) is laborious and depends on the subjective interpretation of the individual investigator. Therefore, an immunoassay was designed for detection of cytokeratin 19 (CK19). By analyzing blood samples from 52 healthy volunteers and 40 bone-marrow aspirates from control patients, a cut-off point of 250 pg/ml CK19 was determined. Application of this cut-off point enabled a specificity of 95% to be shown for bone marrow and of nearly 100% for venous blood. The assay detected 10 HT-29 colon-carcinoma cells among 5 x 10(6) peripheral-blood leukocytes. In comparison with controls, bone-marrow samples of cancer patients were found to have significantly elevated levels of CK19 (p < 0.05). In the analysis of 386 marrow aspirates of cancer patients, a significant concordance of ELISA and ICC was observed (chi 2 = 18.3; p < 0.001). Both procedures, nevertheless, differed in 147 (38%) samples, of which more than two thirds (101) were only ELISA-positive. The CK status detected by ELISA did not correlate with the TNM stage and the histological grading. The established immunoassay allowed sensitive and specific detection of disseminated epithelial tumor cells and appeared to be faster, less laborious and more objective than ICC. Follow-up studies are required to assess the prognostic relevance of this ELISA before it can be applied as a routine method for monitoring of minimal residual epithelial cancer.