The usual method for calculation of the "sensitivity" of thyrotropin immunometric assays is multireplicate analysis of the zero analyte standard. Although this is a statistically valid estimate of the scatter likely to be found in the response variable, it is unrelated to normal analytical practice (usually analysis in duplicate) and estimates intra-assay errors only. This study was designed to assess the analytical performance of 10 immunometric assays used routinely for measurement of thyrotropin in human serum. Response data from each assay were accumulated to provide (a) an estimate of "sensitivity" from multireplicate analysis and (b) an estimate of "minimum detection limit," relating directly to errors associated with routine performance and derived from a minimum of 500 duplicate analyses. We conclude that the "minimum detection limit" should be promoted as a more meaningful measure of assay performance at low analyte concentrations than the "sensitivity" derived from multireplicate analysis of the zero-analyte standard.