Performance of direct estradiol immunoassays with human male serum samples.

BACKGROUND Steroid immunoassays originally required solvent extraction, chromatography, and structurally authentic tracers to avoid interference from steroid cross-reactivity and matrix effects. The demand for steroid assays has driven assay simplification, bypassing this triplet of validity criteria to allow use of unextracted serum, which has introduced bias and nonspecificity at low steroid concentrations. We aimed to evaluate the performance of commercial direct estradiol (E2) immunoassays relative to the reference method of LC-MS and compared serum E2 measurements from each assay with biomarkers of estrogen action. METHODS We measured serum E2 in duplicate using 5 commercial direct immunoassays and LC-MS in a nested cohort of 101 healthy, asymptomatic men >40 years old from the Healthy Man Study. For each immunoassay, we evaluated the detectability and distribution of serum E2 measurements, CV, and bias (relative to LC-MS) by Passing-Bablok regression and deviance plots. RESULTS Three assays detected E2 in all samples, whereas E2 was detected in only 53% and 72% of samples by 2 other assays. All 5 assays had positive biases, ranging from 6% to 74%, throughout their ranges. CVs were lower with 4 immunoassays than with LC-MS. LC-MS, but none of the direct immunoassays, correlated with serum testosterone and sex steroid-binding globulin. CONCLUSIONS The positive bias of direct E2 immunoassays throughout their working range reflects the nonspecific effects of steroid cross-reactivity and/or matrix interference arising from the violation of the triplet validity criteria for steroid immunoassay.