Analytical performances achieved by the ‘Aria II’ automatic system in an interlaboratory survey triiodothyronine and thyroxine assays

During the last decade radioimmunological techniques have become a very useful diagnostic tool and many assays are frequently performed in clinical laboratories; this widespread use has prompted numerous attempts to partially or completely automate radioimmunoassay (RIA). Recently, a fully automatic system (the Aria II*), developed by Becton-Dickinson [1], has become widely used in laboratories which assay triiodothyronine (T3) and thyroxine (T4). Automatic RIA systems must be evaluated in terms of their practicability (speed, cost and technical skill requirements) and also, and most importantly, in terms of their reliability (accuracy and precision). Results produced in a single laboratory have been used to evaluate the Aria II system [-2 and 3-1; data gathered in inter-laboratory surveys can also provide a useful means of comparing the performances of different analytical systems [4 and 5]. This paper describes an inter-laboratory approach--the accuracy and precision of the Aria II system has been estimated using data collected during a national external quality control survey (EQCS) for T3 and T4, which has been operated by the authors since January 1980.?


Introduction
During the last decade radioimmunological techniques have become a very useful diagnostic tool and many assays are frequently performed in clinical laboratories; this widespread use has prompted numerous attempts to partially or completely automate radioimmunoassay (RIA). Recently, a fully automatic system (the Aria II*), developed by Becton-Dickinson [1], has become widely used in laboratories which assay triiodothyronine (T3) and thyroxine (T4).
Automatic RIA systems must be evaluated in terms of their practicability (speed, cost and technical skill requirements) and also, and most importantly, in terms of their reliability (accuracy and precision). Results produced in a single laboratory have been used to evaluate the Aria II system [-2 and 3-1; data gathered in inter-laboratory surveys can also provide a useful means of comparing the performances of different analytical systems [4 and 5]. This paper describes an inter-laboratory approach--the accuracy and precision of the Aria II system has been estimated using data collected during a national external quality control survey (EQCS) for T3 and T4, which has been operated by the authors since January 1980.?

Materials and methods
Main EQCS data Details of the EQCS organization and data processing have been reported elsewhere [6]. The   The inset table reports the mean readings by the Aria II system (computed from the regression line) corresponding to four assigned T3 levels. Significance of the difference from the reference value (p < 0.01%; indicated by an asterisk) was estimated from the confidence limits of the regression line. these findings strongly support the use of the consensus mean as reference value.

Results and discussion
Figures and 2 compare all the results obtained using the Aria II with the consensus means of the EQCS samples for T3 and T respectively. It can be seen that T3 measurements produced by the automatic system are affected by a significant positive bias, which ranges from 17 in the low concentration values to 7 in the high values. As far as the T 4 assay is concerned, the Aria II results appear to be slightly biased (from 3 to 3 in the low and in the high T 4 values respectively). However, these small biases, even when statistically significant as mean values, are of low analytical relevance. The Aria II accuracy has been also directly evaluated from recovery tests; table reports the mean results obtained by laboratories using the Aria II system in added samples, together with the spectrophotometrically measured concentrations of T3 and T. These latter findings confirm that the automatic system overestimates T3 concentrations, but that it assays T4 without a significant bias.
The mean imprecision of the Aria II system achieved in the EQCS pools sent as hidden replicates during the whole period is shown in table 2; for comparison the table also shows the mean imprecision achieved by the six 'kit' methods more frequently used in the EQCS (by six to 20 participants). From these findings it is evident that the Aria II system attains a between-laboratory, between-batch precision which is similar, or better, than that obtained by the more precise method used in the EQCS. This good precision could, however, be expected if one considers the better standardization of an automatic analytical system compared with manual ones.