We evaluated the analytical performance of six commercial direct radioimmunoassays of 17 beta-estradiol, those from Radioassay Systems Laboratories, IRE Medgenix, Biotecx Laboratories, Farmos Spectria, International CIS, and Diagnostic Products Corp. The mean value for estradiol (E2) and the within- and between-run CVs were determined for serum pools and control materials, measured in seven to 10 runs with each method. Mean values for E2 in pooled sera deviated by 75% to 350% from the means measured with our routinely used extraction method. Between-run CVs ranged from 4% to 14% for the direct assays as compared with 10% for the extraction assay. We also investigated, for two of the direct assays, the effect of extraction with diethyl ether before radioimmunoassay, with respect to improvement in the agreement with our extraction method. All of the assays were easy to perform and results were obtained within 4-5 h, but we conclude that matrix effects may be expected in direct assays of estradiol.