[Quality control of radioimmunoassay by bivariate analysis]

Automatic calculation of control charts for precision and accuracy of radioimmunoassay was reported by Faure, et al. Duplicate control samples independently measured was assumed to have a bivariate normal distribution. In this case they assumed that the correlation coefficient between each value of the pairs of control samples is zero. Our experience using this method revealed that a considerable number of assayed samples distributed outside the calculated control limits in case of "accuracy control". It was considered that this happened because in radioimmunoassay the between-assays precision is usually larger than the within-an-assay precision and there is a significant correlation between values of duplicates. We also found equal probability density did not make a true circle but a long circle. Therefore in the present paper we proposed for control charts of radioimmunoassay an equal probability long circle calculated by bivariate analysis of Mahalanobis' generalized distance. It was found that a Mahalanobis' long circle could explain the density distribution of radioimmunoassay with a reasonable percent of samples outside the calculated control limits. What happened here can be interpreted by a large between-assays variability shown by some commercial kits. This automatic calculation method could be applied not only for quality control but also for evaluation and comparison of radioimmunoassay system or commercial kits. Control survey could also be analyzed by such a method.