Objective. Circulating antibodies that bind to human endothelial cells cultured in vitro have been detected in a variety of diseases, including Behçet's disease. In this disorder the reported prevalence of AECA has varied widely. One likely source of variability is the ELISA assay itself, in which differing conditions and reagents have been used in different reports. Methods. We have re-examined the frequency of AECA in 132 Turkish Behçet's patients and 50 healthy Turkish controls, comparing several different methods of preparing the target endothelial cells. Human umbilical vein endothelial cells (HUVEC) were used either: 1) fresh and non-treated, 2) fixed, or 3) TNFα-stimulated. All stages of the procedures were performed at room temperature. Results. In Behçet's patients, using fresh, non-treated HUVEC, 17 of 130 (13.1%) and 9 of 132 (6.8%) sera were positive for IgG- and IgM-AECA, respectively. However, among 50 normal controls, 2 (4.0%) had IgG-positive and 4 (8.0%) had IgM-positive ELISAs under the same conditions. The difference in the frequency of positives between patients and controls was not statistically significant. Fixed HUVEC and TNFα-treated HUVEC gave similar results as well. When group means were examined, only the mean for IgG-AECA determined with TNFα-stimulated HUVEC reached statistical significance. Conclusion. The discrepancy between our data and earlier reports in the literature probably reflects the methodological differences alluded to, and highlights the difficulties in interpreting ELISA assays for AECA.