A comparison of 13 different immunometric assay kits for gonadotropins: implications for clinical investigation.

To determine whether the new immunometric (sandwich) assays for gonadotropins achieve the alleged improvements over RIAs, we compared 13 commercial kits for LH and FSH with our established and validated polyclonal RIAs. Although the kits claimed detection limits as low as 0.05 IU/L, when sensitivity was tested with a uniform hormone standard (the Second International Reference Preparation of human menopausal gonadotropin urinary standard) and the requirement that the limit must be determined from a detectable standard point rather than the variance around zero, only 4 kits surpassed the detectability achieved by the existing LH and FSH RIAs. Seven of the 10 LH kits tested displayed greater than 10% cross-reactivity with either the alpha- or LH beta-subunit. The relationship between the kit standard and the Second International Reference Preparation of human menopausal gonadotropin uniform standard in each kit was nonlinear, so that attempts to compare results between assays with simple multiplication factors are inaccurate. Only 2 assays were designed to eliminate a hook effect. The following conclusions were reached. 1) Most available gonadotropin immunometric assays do not yet offer features not already available in existing validated polyclonal research assays. 2) Conversion of data from one assay to another is not straightforward. 3) Many of the manufacturers' claims do not appear justified. 4) Determination of the detection limit and other immunometric assay characteristics requires standardization of criteria. We propose the following minimum criteria for validating gonadotropin immunometric assays: 1) a uniform definition of the detection limit based on the lowest distinguishable standard concentration, 2) determination of the standard concentration at which a hook occurs, 3) determination of cross-reactivity to subunits as well as intact gonadotropins, and 4) establishment of an adequate normative data base.