Estimation of lot-to-lot relative potency in the preparation of radioimmunoassay calibrators

The aim of this study was to determine guidelines for estimating lotto-lot differences in the potency of calibrator materials or batches of standards for radioimmunoassays. Thirty one lots of standards for thirteen different analytes were compared to the previous lot for that analyte with the relative potency computed by nine different methods. Assays were performed manually. The nine different calculation methods included non-simultaneous fitting of pairs of standard curves, full or partial simultaneous fitting, and least squares or robust minimisation. The simultaneous methods were found superior to the non-simultaneous in minimising the variance of the relative potency estimates, while robust fitting procedures did not result in a lower variance than least-squares minimisation. The root mean square coefficient of variation for the simultaneous estimation of the relative potency by least squares was 6.1 %. On this basis, it is recommended that relative potency estimations in radioimmunoassay be based on at least eight independent pair-wise standard curve comparisons. Additional guidelines for preparing and comparing batches of standards are also given.