Reproductive hormone reference intervals for healthy fertile young men: Evaluation of automated platform assays.

Context: Management of male infertility and/or androgen deficiency requires accurate hormonal measurements with valid reference intervals. Objective(s): The objective of this study was to develop a valid reference panel of blood samples from healthy eugonadal young men with verified normal reproductive function and to use this panel to evaluate the performance of seven fully automated, commercial multiplex immunoassay platforms used to measure serum total testosterone (T), LH, and FSH. Design(s): This was an observational study of consistency among seven different automated immunoassays for each of total T, LH, and FSH. Each method was implemented in two laboratories, with each repeating the analysis of the full reference panel samples twice. Serum T concentrations were also measured by gas chromatography/mass spectrometry (GC/MS), and serum inhibin B levels were determined by an ELISA. Setting(s): The study was performed at commercial, high-volume, clinical pathology laboratories. Participant(s): From 147 men screened, sera from 124 healthy, reproductively normal men (age, 21-35 yr) with normal sperm output were used as a reference panel. All laboratories selected for elite performance in the national immunoassay quality assurance program agreed to participate. Main Outcome Measure(s): For each of the 868 assays, descriptive statistics were calculated in the natural and log-transformed scales and were analyzed by nested, repeated measures ANOVA after log transformation. Reference intervals, defined as 95% confidence limits, were calculated using arithmetic (natural scale), geometric (log scale) and nonparametric methods. Result(s): Descriptive statistics and reference intervals for serum T, LH, and FSH differed widely and significantly between methods, but variation between laboratories for the same assay was negligible. All T methods showed significant differences in regression slope and intercept in deviance plots as well as in estimated reference ranges compared