Comparison of method-related reference intervals for thyroid hormones: studies from a prospective reference population and a literature review.

Introduction Reference intervals are dependent on the reference population, the analytical methods and the way the data are handled statistically. Individual method-related differences have been studied but the comparative differences in reference intervals have not. Methods We studied a reference population of healthy adult subjects and measured free thyroxine and thyroid-stimulating hormone by the four most commonly used analytical platforms used in the UK. Subjects were excluded if they were > 65 years or had positive thyroid peroxidase antibodies. We also performed a systematic literature review of thyroid hormone reference interval studies in non-pregnant adults. Results In total, 303 subjects were recruited and 42 excluded. The central 95th centile values for thyroid-stimulating hormone (mIU/L) were Abbott Architect (0.51-3.67); Beckman Unicel DxI (0.57-3.60); Roche Cobas (0.60-4.31) and Siemens Advia Centaur XP (0.63-4.29). The 95th centile values for thyroxine (pmol/L) were Abbott Architect (10.6-15.5); Beckman Unicel DxI (7.9-13.0); Roche Cobas (12.5-19.6) and Siemens Advia Centaur XP (11.8-19.0). We identified 55 papers describing thyroid reference intervals in male and non-pregnant female adults. The values for upper and lower reference intervals by manufacturer varied but were not significantly different for thyroid-stimulating hormone but were for thyroxine. Discussion Our study demonstrates clearly that there are marked variations in the reference intervals for thyroid hormones between analytical platforms. There is an urgent need for standardization of thyroid hormone assays to permit transferability of results. Until then, guidelines will need to reflect this method-related difference.