In order to test the qualities of the 2 assays, in the same laboratory and on the same tumors, a single-point dextran-coated charcoal radioligand binding assay (RLA-DCC) and Abbott enzyme immunoassay (EIA) were used for more than two years to perform estrogen receptor determinations on cytosols from 2,134 breast cancers. Statistical analysis of the data was performed according to the method of Passing-Bablock. The final regression curve between EIA (y) and RLA-DCC (x) was excellent y = 1.187 X fmol/mg of protein. However, from 1986 to 1988, a great variability was observed for this correlation. We report the study of this variability, which could be explained by several factors, especially calibration problems for the immunoassay kits and changes in our technical team. The binding assay appears to be more sensitive to the technicians' experience than the immunoassay. Technical points are discussed, particularly cytosol preparation and KCl presence or absence in the homogeneisation buffer. The conditions allowing for optimal correlation and routine determination fiability can therefore be defined.