Components of variance analysis of data produced in a national quality-control survey of radioimmunoassays of thyroxin, triiodothyronine, thyrotropin, prolactin, and progesterone.

Data collected in a collaborative survey for radioimmunoassays have been studied by using analysis of variance to estimate the within-kit (CVw.kit) and the between-kit (CVb.kit) components of the total variability (CVT). This analysis has been applied to the results for triiodothyronine, thyroxin, thyrotropin, prolactin, and progesterone produced by 80-150 laboratories that assayed blind, replicate samples. Total variability was lowest in the thyroxin assay (CVT = 10.9%), associated with a very close between-kit agreement (CVb.kit = 4.0%); in the triiodothyronine assay, on the other hand, the large between-kit component (CVb.kit = 10.1%) increased the total variability to 16.1%. In the prolactin assay the CVT of 19.3% included 17.5% CVw.kit and 8.1% CVb.kit. Assays for thyrotropin and progesterone involve analyses of two pools at different hormone concentrations. The CVb.kit component was very high in the low-concentration pool, both for thyrotropin (25.1%) and progesterone (45.2%); in the higher-concentration pool it decreased to 8.3% for thyrotropin but remained high (21.6%) for progesterone. Applying analysis of variance to the triiodothyronine and thyroxin data obtained by different laboratories using the same kit showed that most kits yielded significantly different measurements when used in different laboratories.