Letter to the Editor: Hoermann Response to Fitzgerald et al . (DOI: 10.1089/thy.2019.0535)

Dear Editor:

A recent debate has evaluated the primary test schema for assessing thyroid function, direct measurement of thyroid hormones, or the pituitary hormone thyrotropin (TSH).

Fitzgerald et al. reignited this debate when conducting a meta-analysis of 58 studies (1). When both parameters were available in the same subject, free thyroxine (fT4) concentrations associated more strongly than TSH levels with mortality and morbidity (1). This challenges the strong preference for TSH measurements in clinical practice guidelines (2). Others questioned these suggestions, maintaining that TSH gives an “integrated answer as to the adequacy of thyroid replacement therapy” (3).

This raises two questions. What is the physiological role of TSH? How can a controlling variable such as TSH be properly employed in statistical analysis including correlative studies? In normal physiology, TSH expression is interlocked with thyroid hormones, acting as a driver for attaining euthyroid circulating fT4 concentration. In homeostatic equilibrium, the two hormones deliver a composite expression of multivariate normality (the setpoint) (4). Using bivariate representation, resulting TSH–fT4 equilibria are clustered (4). This approximately halves their intrasubject variability, compared with between-subject variation. Thus, expression of normality is largely determined individually. Due to trait-like properties (intraclass correlation), correlations between TSH and fT4 as observed in a population are weakened within the reference range, strengthening only with increasing severity of hypothyroidism (1,4,5). Though TSH is valuable in diagnosing hypothyroidism, it becomes both uncorrelated from the underlying thyroid status when approaching the euthyroid range and increasingly variable between subjects. Thereby it loses discriminatory accuracy and prognostic value. In the euthyroid population, fT4 concentrations, but not TSH, correlated with cardiovascular risk and mortality in epidemiological studies. Fitzgerald et al.'s meta-analysis corroborates individual studies, providing additional arguments against relying primarily on TSH measurements (1).

Despite contrary evidence, it has been argued that TSH is optimally useful in reflecting the intrinsic pituitary assessment of the thyroid state. Although appreciating the accuracy and sensitivity of the pituitary response, its purpose and specificity warrant further consideration. Pituitary action through TSH is by physiological design anticipatory, implementing central thyroid hormone signaling and corrective action rather than reflecting post-hoc on corporeal thyroid hormone availability (5). Thyroid regulation includes TSH-feedforward control over preferential thyroidal triiodothyronine (T3) secretion (5). This is particularly evident in patients with subclinical hypothyroidism and sufficient intact thyroid tissue (5). Conversely, in athyreotic patients on levothyroxine replacement, deficient thyroxine deiodinization and absent T3 secretion result in generally low free triiodothyronine (fT3)–fT4 ratios, falling fT3 levels, and altered equilibria between TSH and thyroid hormones, compared with those in thyroid health (5). Consequently, the previous setpoint or “normal” TSH value is no longer a suitable reference point for corporeal euthyroidism on levothyroxine replacement. This was shown in experimental animals treated with various thyroid hormone combinations wherein tissue hypothyroidism persisted on levothyroxine monotherapy despite TSH normalization.

Far from providing a universal answer by itself, considering its conditional control function and adaptive relationship with thyroid hormones, interpretation of TSH requires reconciliation and complementation by thyroid hormones in various conditions. The debate should strengthen the need to translate advances in our understanding of system regulation into clinical studies to address unresolved issues in thyroidology.