Effect of counting errors on immunoassay precision.

Using mathematical analysis and computer simulation, we studied the effect of gamma scintillation counting error on two radioimmunoassays (RIAs) and an immunoradiometric assay (IRMA). To analyze the propagation of the counting errors into the estimation of analyte concentration, we empirically derived parameters for logit-log data-reduction models for assays of digoxin and triiodothyronine (RIAs) and ferritin (IRMA). The component of the analytical error attributable to counting variability, when expressed as a CV of the analyte concentration, decreased approximately linearly with the inverse of the square root of the maximum counts bound. Larger counting-error CVs were found at lower concentrations for both RIAs and the IRMA. Substantially smaller CVs for overall assay were found when the maximum counts bound progressively increased from 500 to 10,000 counts, but further increases in maximum bound counts resulted in little decrease in overall assay CV except when very low concentrations of analyte were being measured. Therefore, RIA and IRMA systems based in duplicate determinations having at least 10,000 maximum counts bound should have adequate precision, except possibly at very low concentrations.