Two Thyroid Stimulating Hormone assays correlated in clinical practice show disagreement in subclinical hypothyroidism patients

Abstract

Thyroid Stimulating Hormone (TSH) levels are related to the pituitary gland's ability to detect thyroid hormone concentration. Many studies have analyzed the correlation between TSH and T4, demonstrating a complex system correlation. This complex system may vary among different TSH levels and patients. ObjectivesThe main purpose of this study is to assess the correlation and agreement of serum TSH measured with two assays in different settings. Design & methodsWe evaluated healthy individuals as well as subclinical or overt hypothyroid patients. Eighty participants had TSH levels measured by Cobas Roche Elecsys 600 (Roche Diagnostics) and Abbott Architect I 2000 (Abbott Diagnostics). The TSH methods correlations were established with Pearson's correlation, and the strength of the agreement was determined by the McBride scale. The paired Student's t-test was applied to evaluate TSH values from both methods. The one-sample t-test was used to evaluate the difference between TSH values. The agreement was also assessed by a Bland-Altman plot. A regression analysis was applied to the correlation between TSH and T4. ResultsThere was a significant difference in TSH values measured by the two methods (p<0.01). Our results demonstrated a poor correlation for TSH in the euthyroid (r: 0.888, p<0.01) and the subclinical hypothyroid (r:0.886, p<0.01) range. The Bland-Altman plot demonstrates that the majority of the TSH values fell between the lines of equality. There were few differences in the values in the normal upper range and slightly above that range (from a TSH: 3.25 to 6.36mUI/L). The level of correlation between TSH assays remains high in all scenarios for age (r≥0.951), BMI (r≥0.962), anti-TPO antibodies (r: 0.977) or levothyroxine use (r: 0.970). ConclusionsTSH measurement is essential to access thyroid function. Although the overall agreement between the methods is substantial, there was a poor agreement in the normal upper range and close above. The disagreement observed reinforces the difficulty in using different assays in clinical practice. The better correlation with fT4 and the reference range used by Cobas assay allowed the best clinical performance.