BACKGROUND/AIMS The determination of neuron-specific enolase (NSE) is relatively frequently requested in the differential diagnosis of small-cell lung carcinoma and non-small-cell lung carcinoma. The individual results of different immunoassays are often not comparable, which has been confirmed by long-term external quality assessments. In this study, we assessed the possible sources of these differences. METHODS More than 3,000 NSE analyses were performed using seven different immunoassays: DELFIA (PerkinElmer), Elecsys 2010 or Modular Analytics E 170 (Roche), Kryptor (B.R.A.H.M.S.), the enzyme-linked immunosorbent assay DRG and three assays based on immunoradiometric assays (DiaSorin, Immunotech and Schering-CIS). The following parameters were evaluated: precision profile of the individual methods, linearity on dilution and modified recovery, comparability and discrimination of immunoassays, sensitivity, and specificity. RESULTS There were differences in the correlation of values of certain low-concentration specimens. Some assays correlate well while others do not (up to fivefold difference), especially in the case of controls prepared synthetically. Therefore, the current non-standardized preparation of controls is questionable in our opinion. In the cutoff range, the difference in the results of native samples did not exceed its double value. The variation in values >100 microg/l obtained with different assays is <40%. CONCLUSION Our results confirmed expected matrix interferences especially in the range of normal and cutoff NSE concentrations. Another source of discrepancies can be attributed to different antibody affinity to alphagamma- and gammagamma-enolase isoenzymes. Finally, improper settings of cutoff values also contribute to the different discrimination of the methods.