Highly sensitive assays of serum-thyrotropin in the diagnosis of hyperthyroidism: assessment of performance and reference values.

The performance of three different highly sensitive immuno-assays for thyrotropin in serum (S-TSH) was compared. Specimens from 803 patients with suspected hyperthroidism or other thyroid disease were used. Hyperthyroid patients were identified and characterized by determining serum thyroid hormones and by clinical observation. The performance of the methods in discriminating between hyperthyroidism and euthyroidism was described in terms of diagnostic sensitivity and specificity. The performance was also assessed by systematically varying the reference values. If patients with treated hyperthroidism were eliminated, the three S-TSH assays performed equally well at a suggested lower reference value of 0.4 mU/l. A diagnostic efficiency of more than 0.95 was achieved at the observed prevalence of disease (0.12). A graphical model which allows an adjustment of reference values to favour sensitivity or specificity is presented. Reference values were also calculated from the euthyroid subjects using parametric and non-parametric approaches. The specificity, when using the recommended discrimination level, suggests that patients with low S-TSH values should be subjected to further investigation by biochemical or other methods.