The HUNT study: long-term within-subject variation of thyroid stimulating hormone (TSH)

Data on the long-term within-subject biological variation (CV_I) of serum thyroid stimulating hormone (S-TSH) are scarce. In the EFLM Biological Variation Database, the longest observation period was one year. We estimated a coefficient of variation that included analytical variation (CV_I+A) using data from 16,976 individuals in the Trøndelag Health Study (HUNT), where S-TSH was measured on two occasions with an average interval of 10.6 (range 9.3-12.3) years. These individuals reported their health to be 'good' or 'very good' on both occasions and were not registered with any diagnoses or use of medications (according to Norwegian Prescribed Drug Registry) related to the thyroid. We used the software refineR to identify an assumed nonpathological subpopulation in the distribution of distances from each observation to the center of the bivariate distribution of the two S-TSH-values. From this subpopulation, individuals with a distance ≤ the 95 percentile in the distribution of distances were selected for estimation of CV_I+A. The difference in percent of the mean (DPM) was calculated for each individual, and CV_I+A as the standard deviation in the distribution of DPMs divided by 2^0.5. This method was robust against systematic bias between the two measurements. CV_I+A was 21-23% for different groups of age and sex. Accounting for CV_A would imply a CV_I 0.1-0.8% less than CV_I+A. Our estimates are well within the 12-29.3% range of CV_I reported from the seven studies in the meta-analysis of the EFLM Biological Variation Database.