The introduction of new regimens in the chemotherapy of inoperable non-small cell lung cancer (NSCLC) patients provides a useful extension of survival probability that may now justify the application of tumor markers for the disease monitoring. In a prospective study of 48 consecutive NSCLC patients with TNM stages IIIB/IV we compared changes in the serum levels of the cytokeratin 19 fragment CYFRA 21-1 with the clinical evaluations of response to therapy. CYFRA 21-1 levels were measured using the enzyme immunoassay of Boehringer, Mannheim (Germany). Clinical response to therapy was evaluated according to standard criteria of the WHO. For the assessment of response to therapy by changes in the marker levels the difference between two consecutive levels must exceed 30%. This value is based on the formula: Difference = 2 square root of 2 x CV (CV: inter-assay coefficient of variation of the marker test). CYFRA 21-1 was found to be elevated in 29/48 (60.4%) patients prior to therapy and in 10/48 (20.8%) patients at tumor progression. 91 evaluations have been recorded in these 39 patients. The overall concordance between changes in the marker levels and the clinical assessment was 59.3%. The decrease of CYFRA 21-1 levels at remission was rather low resulting in a concordance of only 42.9%, i.e. marker assays cannot replace the clinical restaging by imaging modalities. In contrast, changes in the marker levels at progression did exceed the required 30% in the majority of cases (64.7%). Most of discordant results (40.7%) could be explained by insufficient decrease or increase of CYFRA 21-1 levels or by extended lead-time. The most striking result was the detection of progressive disease by rising marker levels. Except one case, there was no false-positive elevation of CYFRA 21-1 levels. It is concluded that the detection of progressive disease by rising CYFRA 21-1 levels may avoid continuation of ineffective treatment.