TSH Reference Intervals: Their Importance and Complexity

The initial concept of a reference range or interval seems deceptively simple. It involves gathering a sample of healthy individuals, representative of a larger population, and defining a mean and range for a particular analyte based on accepted statistical methodology. ¹ The definition of a reference interval extends from the 2.5th to the 97.5th percentile, thus encompassing the central 95th centile estimates of the values obtained in the healthy sample of the population. Based on that determination, the implication is that the values for a particular individual are “normal or representative” and fall within that interval, or are not normal or not representative and do not fall within that reference interval. The implication can be extended to conclude that a disease is present or that there is a need for monitoring or treatment. However, a value outside of the reference interval does not, of necessity, indicate that treatment is indicated.

TSH reference intervals are established for healthy individuals without a diagnosis of thyroid disease, abnormalities in their thyroid gland, or thyroid peroxidase antibodies. ^2,3 Individuals with interfering medications or taking thyroid hormone are also excluded from contributing to the reference interval. Evidence has accrued that other factors may also affect the bounds of the TSH reference interval. Such factors may include age, ethnicity, sex, iodine status, and pregnancy. ⁴ Perhaps the best explored example outside of the pregnant state is age, with the upper limit of the reference interval extending to higher TSH values with advancing age, most noticeable in those over 80 years. ^3,5 Another example is the lower bounds of the reference interval being lower in certain racial or ethnic groups, such as in African Americans. ^2,6 It is also clear that the range of TSH levels characteristic of a particular individual is far narrower than the range of TSH values characteristic of a population, implying that tighter personalized reference intervals likely exist. ^7,8

The concept of using cut-off values to define a reference interval may fail to recognize that the underpinnings of thyroid disease may, in fact, be better understood as a continuum. This can perhaps be illustrated by the fact that even within the bounds of normal TSH values there are trends for increased cardiovascular events, worsening lipid profile, and worse metabolic outcomes. ⁹ If the inference is made that a disease is present based on a TSH value higher than the upper limit of the normal range, this might be subclinical or overt thyroid dysfunction, and then treatment might be pursued. ¹⁰ Such a decision to treat has consequences, including those of undertreatment and overtreatment, and this decision is best reached on a patient-by-patient basis. Once a decision is made to intervene and treat “thyroid disease” in someone with a TSH value outside the reference interval, it is generally accepted that the goal of treatment is to return the TSH to within the reference interval. For example, an elevated TSH would be lowered to within the reference interval by levothyroxine treatment. Sufficient data are currently lacking from prospective studies to provide evidence for cut offs for initiation of levothyroxine therapy which are associated with improved outcomes. ¹¹

The negative log-linear relationship between TSH and free thyroxine is well-established. This relationship illustrates that free thyroxine values also impact TSH values, as a result of a hypothalamic-pituitary-thyroid axis set point phenomenon. ¹² In addition, the relationship between TSH and free thyroxine, based on the constituent TSH-free thyroxine pairs, is different in levothyroxine-treated individuals, compared with non-treated individuals. ¹³

An article of great interest in this month’s Thyroid introduces another nuance to the complexities of TSH reference intervals. ¹⁴ Kuś and colleagues assembled three different cohorts of individuals from the Netherlands and Norway whom they studied to gain insight into the impact of genetic factors on TSH reference intervals. They first studied a disease-free cohort of almost 7000 individuals using the Rotterdam study participants in whom 59 genetic variants were used to assign a polygenic score. They determined quartiles for the polygenic score and calculated specific reference ranges for TSH for each quartile. The TSH reference ranges differed significantly between the polygenic score quartiles.

These genetically-informed TSH reference intervals were then compared with standard population-based reference intervals. The results were further validated in a second independent cohort of 3800 patients from the Nijmegen Biomedical Study. Notably the polygenic score was a more powerful predictor of TSH values than free thyroxine values or any other nongenetic factors such as age or body mass index. The polygenic score explained 9.2–11.1% of the total variance in TSH concentrations, compared with 2.4–2.7% of the variance being explained by free thyroxine values. When examining free thyroxine values, the investigators found that at the same free thyroxine concentration, individuals with a higher polygenic score had a higher TSH concentration. This confirms a hypothalamic-pituitary-thyroid axis set point effect, manifesting as upward shift in TSH concentrations for a particular free thyroxine value, with increasing polygenic scores.

The classifications of subclinical thyroid disease established by use of population-based reference intervals were not confirmed when genetically-based reference ranges were utilized. In fact, up to 24.7–30.1% of individuals classified as having mild thyroid dysfunction were reclassified as being euthyroid when the genetically-determined reference interval was applied. This is not a trivial difference and has significant consequences for the monitoring and treatment plan offered to the individual patient. The third cohort (Trøndelag Health Study [HUNT] cohort) was used to explore levothyroxine treatment implications. Individuals whose polygenic scores fell within the higher quartiles were found to be more likely to be prescribed levothyroxine treatment than individuals whose scores were in the lower quartiles. This has considerable implications for clinical decision-making if the higher TSH values in those with higher polygenic scores represents a different TSH setpoint, rather than the higher TSH value being indicative of thyroid disease. This raises valid concerns about potential over-treatment with levothyroxine.

These data illustrate the clumsiness of our current paradigms for diagnosis and treatment of thyroid disorders and open the way for refinements in our approach towards management of thyroid disease. Furthermore, incorporation of these particular genetic variants may be just the “tip of the iceberg” with additional genetic variants not studied having further impact on TSH reference intervals. As we explore the impact of genetic factors and establish sound approaches to implementing this knowledge, we may be moving towards more personalized treatment of thyroid dysfunction. Future questions that will require prospectively-conducted research include whether diagnosis and treatment based on more accurate reference intervals may improve the spectrum of patient outcomes, including improved health-related quality of life and reduced morbidity and mortality from cardiovascular, metabolic, and neurocognitive dysfunction.