Background: Distribution of TSH levels is not normal. This is due to physiological changes that cause temporary increases in TSH during physiological events. Several methods are used to normalize the distribution when defining normal limits.
Objective and hypotheses: To compare the normal limits defined by three normalization methods vs non-normalized distribution based on a large cohort with no known thyroidal illness.
Method: Data were collected from a computerized data base of the Clalit health services in Jerusalem, Israel. Exclusion criteria were positive anti-thyroid antibodies and treatment with any drug. TSH values were normalized with the Hoffmann, Tukey and Tukey followed by tural log transformation (NLT) methods. The lower normal limits (LNL) i.e. the 2.5th percentile and the upper normal limits (UNL), i.e. the 97.5th percentile were defined. The clinical relevance of the limits was tested by calculating the mean FT3 and mean FT4 for results of TSH below, within and above the limits for each method.
Results: We report the results of the 610 age group, based on 1450 subjects, as a representative example. According to the non-normalized, Hoffman, Tukey and Tukey followed by NLT, the limits were 0.986.87, 0.954.31, 0.975.55 and 0.975.55 IU/l respectively, i.e. maximal change from non-normalized data occurred for the UNL by the Hoffman method (−33%) and for the LNL also by the Hoffman method (+4%). There was no difference in average FT3 or FT4 between patients with TSH within, below or above the normal range for all four methods.
Conclusion: i) As expected, normalization methods alter mainly the UNL but the difference between methods is more than 30%. ii) In individuals without thyroid illness, thyroid hormone values are stable over a wide range of TSH levels thus questioning the value of TSH as a screening test.
20 - 22 Sep 2014
European Society for Paediatric Endocrinology