Internal quality control of hormone determinations by RIA. Detection of clinically significant analytical errors of serum thyroxine determination.

An internal quality control system by which clinically significant analytical errors can be detected with high probability was developed for serum thyroxine (T4) determination by RIA. The mean concentration of the quality control samples (C1, C2, and C3) were 40.0, 101.8, and 156.7 nmol/l, and their total standard deviations (st) 2.5, 4.1, and 7.1 nmol/l, respectively, in stable analytical performance of the T4 assay. From the clinical point of view it was accepted that the results of C1 varied from 30 to 50 nmol/l at the most. When expressed in standard deviations, the same size of error was accepted for C2 and C3. It followed that the quality control system had to detect a systematic shift of 2.43st (delta SEc) and a 2.08 times increase in random error (delta REc) in order that the quality goals could be reached. With three control samples the combination of rules, 1: 3s/CS1: (1.0s; 2.7s), which was chosen for manual quality control, detected the delta SEc with about an 80% probability and the delta REc with about a 30% probability. The probability of false rejections was less than 1%. The probability of detecting the delta SEc was considered acceptable, but to detect the delta REc with a higher probability, the within-assay variation of each run was calculated from the results of duplicate patient samples. Although the whole run was not rejected because of systematic and/or random errors, individual samples were reanalysed if the results of their duplicate determinations differed from the criteria stated.(ABSTRACT TRUNCATED AT 250 WORDS)