2. INTERPRETATION OF TSH AND T4 ASSAYS

2.1.The TSH-free T4 relationship

Initial testing for possible TD may be based on estimation of either free T4 or TSH (90 ), but definitive assessment of thyroid status requires both a sensitive serum TSH assay and a valid free T4 estimate, with interpretation based on the general relation between the two hormones. Because of the feedback relationship between tropic and target gland hormones, typical disease-related changes lead to diagonal deviations from the normal T4-TSH relationship. (figure 3 ).

Figure 3. The relationship between serum TSH and free T4 concentration is shown for normal subjects (N) and in the typical abnormalities of thyroid function: A, primary hypothyroidism ; B, central or pituitary-dependent hypothyroidism; C, thyrotoxicosis due to autonomy or abnormal stimulation of the gland; D, TSH-dependent thyrotoxicosis or thyroid hormone resistance. Note that linear changes in the concentration of T4 correspond to logarithmic changes in serum TSH.

2.2 The serum free T4-TSH setpoint

Small changes in serum T4 and T3 concentrations, within the normal range, alter serum TSH, indicating that the inverse feedback relationship between serum free T4 and TSH applies across their normal ranges as well as in disease states ( 91,92). A study in which normal subjects were given incremental doses of T3, demonstrated significant individual variation, independent of sex and age, in the setpoint of the pituitary-thyroid axis ( 93), suggesting that the TSH set-point for a particular serum free T4 or free T3 concentration may be an individual characteristic. Studies of monozygotic and dizygotic twin pairs also suggest that genetic factors influence the serum concentrations of total and free T4 within the normal range ( 94).

Andersen et al (95), in a study of normal subjects sampled monthly between 0900 and 1200 for a year, showed that individual references ranges for T4 and T3 were only about half the width of the population reference ranges, indicating that a test result within the population range is not necessarily normal for that individual. Serum TSH showed greater between-sample variation for each individual than serum T4 or T3. Based on the degree of individual variation, it was estimated that a normal serum TSH concentration needed to change by 0.2-1.6 mU/l to be confident of a serial change in thyroid status. Based on this analysis, they estimated that a single morning sample defined serum T4 and T3 to within 25%, and serum TSH only to within 50%. Since TSH shows diurnal variation and pulsatile secretion (96), random samples are likely to show even greater variation.

It has been suggested that some healthy elderly subjects have normal serum TSH concentrations despite having low serum free T4 values, attributed to resetting of the threshold for TSH inhibition (97), but the view that an increase in serum TSH in some normal subjects may reflect an alteration in central set-point, rather than subclinical hypothyroidism, remains unproven (98).

2.3 The TSH-T4 relationship: diagnostic assumptions

The precise diagnosis of thyroid dysfunction can generally be established from a single serum sample from the relationship shown in figure 2, subject to six key assumptions (table 3). It should be noted that only the last three of these assumptions can be validated in the laboratory; the first three are best verified clinically.

2.3.1 Steady-state conditions.

This first assumption should be questioned whenever anomalous results occur during associated illness, or with medications that perturb the pituitary-thyroid axis. The half lives of plasma TSH (approx. 1 hour) and plasma T4 (approx. 1 week) differ so widely that acute perturbationof the pituitary-thyroid axis will often result in transient non steady-state conditions. With its much shorter half-life, serum TSH deviates more rapidly from the steady state. Other common deviations from the steady state relate to short-term pulsatile or diurnal fluctuations in hormone secretion, responses to treatment and spontaneous evolution of disease, as can occur in subacute thyroiditis or postpartum thyroid dysfunction.

Figure 4. Measurement of serum T4, rather than serum TSH, is the more reliable single test of thyroid function when steady state conditions do not apply, as in the early phase of treatment for thyrotoxicosis or hypothyroidism. (from reference 77)

2.3.2. Normal tropic-target hormone relationship.

During treatment of prolonged thyrotoxicosis, TSH secretion may remain low for several months after serum free T4 becomes normal (99). Conversely, after severe prolonged hypothyroidism, or in some children treated for congenital hypothyroidism (107), TSH hypersecretion may persist despite normalization of serum T4. Serum TSH will then give an inaccurate indication of thyroid status, with the potential for over-treatment if this parameter alone is used to assess therapy.

2.3.3 Tissue responses proportional to the hormone concentration.

The active or free concentrations of T3 and T4 generally correlate well with clinical features, but in generalized thyroid hormone resistance, high serum free T3 and T4 concentrations are entirely appropriate to maintain the euthyroid state (109). The onset and offset of thyroid hormone action is slow, so that tissue responses may lag behind changes in serum concentrations of free T4 and T3. There is a notable lack of convenient, sensitive, specific, objective indices of thyroid hormone action, so that assessment remains predominantly clinical. Corroborative measurements that are useful, especially in following the response of individuals to therapy, include measurement of oxygen consumption (123), sex hormone binding globulin (124), angiotensin converting enzyme (125) and serum ferritin (126), as well as several indices of cardiac contractility ( 52,62).

2.3.4 The assay measurement reflects the active hormone(s).

TSH and iodothyronine assays make comparative, rather than absolute, measurements of hormone concentrations, based on the premise that samples and assay standards differ only in their concentration of analyte. This assumption breaks down if there is any other difference between a serum sample and assay standards that influences the measured parameter, as, for example, dissimilar protein binding of tracer (127), the presence of binding competitors (128), or possible nonspecific interference with enzymatic, fluorescent, or chemiluminescent detection systems (129, 130). Circulating T3 and T4 autoantibodies may invalidate immunoassays by sequestering the assay tracer (131), while heterophile mouse antibodies and rheumatoid factor can interfere with immunoglobulin aggregation, or with cross linking of the signal and capture antibodies (132, 133).

If the biologic activity of circulating immunoreactive TSH is increased or decreased, the normal relationship between measured serum TSH and free T4 may be altered. Secreted immunoreactive TSH is heterogeneous, due to differences in its three oligosaccharide side chains (111). In hypothalamic hypothyroidism, the secreted TSH has decreased bioactivity (111, 112), whereas activity may be enhanced in thyroid hormone resistance, primary hypothyroidism and in some TSH-producing tumors (134, 135).

2.3.5 The assay can reliably distinguish low from normal levels

Assay precision inevitably deteriorates as the limit of detection is approached; this characteristic is crucial in evaluating TSH assays that are used to distinguish thyrotoxicosis from normal (34,77). The lower working limit of a TSH assay should be defined in terms of its between-assay reproducibility, defined as functional sensitivity, rather than by the analytical sensitivity of individual assay runs (34,77).

2.3.6 Reference ranges are appropriate

Reference ranges (table 4) show little change with advancing age (136), except for a possible decline in normal serum T3 with age (120). Serum T3 is significantly higher in children (121) and probably also in young adults. TSH reference ranges should be established after logarithmic transformation to give the geometric mean and to define a realistic lower normal limit (35,36,37). Associated illness, nutritional changes, and medications frequently cause assay results to fall outside the normal reference ranges as defined in healthy subjects, so that it may be relevant to use wider than normal reference ranges in the face of associated illness ( 84).