The Relationship Between TSH and Free T 4 in a Large Population Is Complex and Nonlinear and Differs by Age and Sex

Transcription

1 ORIGINAL Endocrine ARTICLE Research The Relationship Between TSH and Free T 4 in a Large Population Is Complex and Nonlinear and Differs by Age and Sex Narelle C. Hadlow, Karen M. Rothacker, Robert Wardrop, Suzanne J. Brown, Ee Mun Lim, and John P. Walsh Department of Clinical Biochemistry (N.C.H., K.M.R., R.W., E.M.L.), PathWest Laboratory Medicine, Queen Elizabeth II Medical Centre, Nedlands, Western Australia 6009, Australia; Western Diagnostic Pathology (N.C.H.), Myaree, Western Australia 6154, Australia; School of Pathology and Laboratory Medicine (N.C.H.) and School of Medicine and Pharmacology (J.P.W.), The University of Western Australia, Crawley, Western Australia 6009, Australia; and Department of Endocrinology and Diabetes (K.M.R., S.J.B., E.M.L, J.P.W.), Sir Charles Gairdner Hospital, Nedlands Western Australia 6009, Australia Context: The relationship between TSH and T 4 is thought to be inverse log-linear, but recent studies have challenged this. There are limited data regarding age and sex differences in the TSH-T 4 relationship. Objective: The purpose of this study was to evaluate the TSH free T 4 relationship in a large sample. Methods: In a cross-sectional, retrospective study, we analyzed TSH and free T 4 results from subjects collected over 12 years by a single laboratory. For each free T 4 value (in picomoles per liter), the median TSH was calculated and analyzed by sex and age (in 20-year bands). Results: The relationship between log TSH and free T 4 was nonlinear. Mathematical modeling confirmed that it was described by 2 sigmoid curves with inflexion points at free T 4 concentrations of 7 and 21 pmol/l. For free T 4 within the reference range (10 20 pmol/l), median TSH was higher in men than in women (P.001) and increased across age bands with the highest values in those 80 years and older (P.001). In contrast, in overt hypothyroidism (n 4403), TSH was lower in older age groups than in those aged years (P.001). Conclusions: The TSH free T 4 relationship is not inverse log-linear but can be described by 2 overlapping negative sigmoid curves. At physiological free T 4 concentrations, TSH is higher in men and in older people, whereas the TSH response to hypothyroidism is more robust in younger people. These results advance understanding of the TSH free T 4 relationship, which is central to thyroid pathophysiology and laboratory diagnosis of thyroid disease. (J Clin Endocrinol Metab 98: , 2013) In healthy individuals, homeostatic mechanisms ensure that circulating concentrations of TSH and T 4 are tightly regulated. Repeated measures of TSH and T 4 in healthy subjects (reflecting intraindividual variation) fall within a narrower range than that for a population (interindividual variation) (1, 2), leading to the concept that each person has a unique set point for hypothalamo-pituitary-thyroid (HPT) axis function (1 3). The HPT set ISSN Print X ISSN Online Printed in U.S.A. Copyright 2013 by The Endocrine Society Received December 17, Accepted May 7, First Published Online June 17, 2013 point reflects, in part, the characteristics of each individual s negative feedback loop, including the responsiveness of thyroid follicular cells to TSH stimulation and the sensitivity of TRH neurons and thyrotropes to thyroid hormone feedback. A key feature of the TSH-T 4 relationship is that small changes in circulating T 4 concentrations can result in large relative changes in TSH, although the absolute changes in Abbreviations: ED ID, equilibrium dialysis isotope dilution; HPT, hypothalamo-pituitarythyroid; LC-MS/MS, liquid chromatography tandem mass spectrometry; LQ, lower quartile; UQ, upper quartile jcem.endojournals.org J Clin Endocrinol Metab, July 2013, 98(7): doi: /jc

2 doi: /jc jcem.endojournals.org 2937 TSH differ widely among individuals, depending on their HPT set points (4). In studies of hypothyroid individuals treated with varying doses of thyroxine, the relationship between serum TSH and T 4 appeared hyperbolic when plotted on linear graph coordinates (5), whereas when TSH was plotted on a logarithmic (log) scale, it was approximately linear (4, 6). Similar results have been reported from studies of thyroid hormone loading in euthyroid subjects (2, 3, 7). In 1990, Spencer et al (7) reported an inverse, linear relationship between log TSH and free T 4 index in a cross-sectional analysis of 505 ambulatory individuals with a wide range of thyroid function tests; this has since become a generally accepted tenet of thyroid physiology. Perhaps surprisingly, the log-linear relationship between TSH and T 4 has not been reassessed in large epidemiological studies until recently. In an analysis of 2 datasets comprising 3223 and 6605 patients referred for thyroid function testing, Hoermann et al (8) found that the relationship between log TSH and free T 4 was, in fact, not linear; rather it was a complex relationship with at least 3 distinct segments. Similar results were reported by Clark et al (9) in a cross-sectional analysis of 5117 individuals in whom the TSH free T 4 relationship was not linear but was best described as a fourth-order polynomial. Both these studies had limitations: the study by Hoermann et al included a large proportion of hospitalized patients (41%) in whom nonthyroidal illness might have affected the TSH-T 4 relationship and was conducted in an area of mild iodine deficiency, whereas the study of Clark et al was restricted to individuals aged 65 years or older. Although HPT axis set points are, at least in part, genetically determined (10, 11), several lines of evidence suggest that aging is associated with alterations in set points. In cross-sectional studies of individuals thought to be free of thyroid disease, TSH concentrations increased with age (12 14), and in 2 cohort studies the age-related TSH increase was not accompanied by a fall in free T 4, suggesting an alteration in TSH set point (15, 16). In contrast to the TSH increase observed with healthy aging, older people who develop hypothyroidism exhibit a lesser TSH response to low T 4 levels than do younger people (17 19). The biological basis of these age-related alterations in HPT axis function is not well understood. Sex differences in HPT axis function have been described. Women have a greater TSH response to exogenous TRH than men (20), and both endogenous and exogenous estrogen enhance the TSH response to TRH (20 22). Nonetheless, 24-hour TSH secretion profiles reportedly do not differ between the sexes apart from a positive correlation between age and TSH secretion observed in women but not in men (23). In the study of Clark et al (9), however, the TSH free T 4 relationship differed significantly between older men and women. Thus, the extent and significance of any sexual dimorphism in HPT function and whether this is affected by aging remain unclear. In view of the conflicting data regarding the relationship between TSH and T 4 and the uncertainty over the modulating effects of age and sex, we examined the TSH free T 4 relationship in a cross-sectional study of a large population. Materials and Methods We extracted thyroid function tests with concurrent TSH and free T 4 results performed over a 12-year period (January 1, 2000 December 31, 2011) by Western Diagnostic Pathology (Myaree, Western Australia, Australia). Western Diagnostic Pathology is a private laboratory providing statewide pathology services. Western Australia is an iodine-sufficient region (16) with a predominantly white population. All results were deidentified before analysis, and because the study fits the criteria of an audit, institutional ethics approval and informed consent were not required. TSH and free T 4 measurement and quality control All samples were collected into serum separator tubes (Becton Dickinson, North Ryde, New South Wales, Australia). Serum TSH and free T 4 concentrations were measured on an ADVIA Centaur analyzer (Siemens Ltd, Bayswater, Victoria, Australia; Siemens Diagnostics Division, Tarrytown, New York). The TSH assay is a 2-site sandwich immunoassay using direct chemiluminometric technology, whereas the free T 4 assay is a competitive, direct chemiluminescent immunoassay. The free T 4 assay was reformulated by Siemens in 2008, but there was no clinically significant trend in median free T 4 over the study period. The laboratory participates in ongoing internal and external quality assurance programs under the accreditation of the National Association of Testing Authorities, Australia and the Royal College of Pathologists of Australasia. Interassay percentage coefficients of variation over a typical 6-month period for free T 4 at 9, 18.8, and 38.5 pmol/l were 5.6%, 4.3%, and 4.2%, respectively, and for TSH levels of 0.06, 0.21, 4.8, and 18.1 mu/l were 9.5%, 7.1%, 5.2%, and 5%, respectively. The limit of detection for TSH is mu/l and for free T 4 is 1.3 pmol/l. TSH results of greater than 150 mu/l were diluted according to the manufacturer s procedure. The TSH assay is standardized to the Second International Reference Preparation (World Health Organization 80/558). Laboratory reference ranges (derived from data extraction and consensus between local chemical pathologists and endocrinologists) are 0.4 to 4.0 mu/l for TSH and 10 to 20 pmol/l for free T 4. Free T 4 concentrations are routinely reported as whole numbers in picomoles per liter; to convert to nanograms per deciliter, multiply by Exclusions The initial dataset contained records. We excluded records (15.2%) from patients thought likely to have factors confounding the physiological HPT axis response or resulting in analytical interference with thyroid function testing.

3 2938 Hadlow et al Complex Relationship Between TSH and Free T 4 J Clin Endocrinol Metab, July 2013, 98(7): Table 1. Demographics and Thyroid Function in the Cohort Thyroxine Treatment Status Untreated All Treated Untreated P Value: Treated vs Untreated Men Women P Value: Men vs Women No. of subjects (21) (79) (28) (72) (% of total) Mean age (SD), y 51.6 (18.9) 55.5 (16.9) 50.6 (19.3) (19.6) 49.5 (19.0).001 Median TSH 3.0 (1.0, 4.8) 2.0 (0.6, 4.0) 3.5 (1.2, 4.9) (1.4, 5.0) 3.3 (1.1, 4.9).001 (LQ, UQ), mu/l Mean free T4 (SD), pmol/l 14.2 (4.0) 15.8 (3.9) 13.8 (3.9) (3.9) 13.7 (3.9).001 To convert free T 4 to nanograms per deciliter, multiply by These included hospitalized patients, pregnant women, patients aged younger than 1 year, patients receiving specialist endocrine, surgical, or medical care, patients having records with an unknown time of collection or collection times outside usual office hours, or patients with unknown age. Using laboratory records going back to 1985, we excluded patients with certain HPT axis disorders (treated Graves disease or thyrotoxicosis, multinodular goiter, thyroid cancer, partial or total thyroidectomy, and hypopituitarism) and those with a history of treatment with radioiodine, antithyroid drugs, lithium, antiepileptic drugs, amiodarone, or liothyronine. Patients treated with thyroxine were not excluded but were analyzed separately so that the TSH free T 4 relationship could be compared in treated and untreated patients. The remaining records were from subjects; the first record of each individual was retained for further analysis and subsequent records were excluded. Statistical analysis The TSH free T 4 relationship was examined by collating all TSH values corresponding to each integer free T 4 value and calculating the median TSH (with upper quartiles [UQs]and lower quartiles [LQs]) for that free T 4 concentration. The median was chosen as the focus for analysis of TSH, it being a more statistically robust measure than the mean. Other variables were summarized by mean (SD) if continuous and symmetrically distributed or count (percentage) if categorical. The functional relationship between TSH and free T 4 demonstrated nonlinear properties even after log transformation of TSH. To capture the curvature of the log TSH free T 4 relationship, we used a nonlinear quantile regression approach, implementing parameters that would match a negative sigmoid curve in 2 stages. The mathematical relationship proposed is as that in a Venegas equation (24), with a sign change on the independent variable: ln TSH A 1 e B C freet4 /D. Here, the parameters A and A B correspond to the lower and upper asymptotes, the sections of the curve that appear to plateau. The parameter C corresponds to an inflection point, whereas D determines the horizontal scale of the curve, with larger values reflecting a more elongated appearance. We applied the methodology separately to data for which free T 4 was 12 pmol/l and 12 pmol/l. This value was chosen as a suitable split point, being located midway along the section of the median TSH curve that showed the least change. The TSH free T 4 relationship was evaluated for the effects of age, sex, and thyroxine therapy using quantile regression methods (25), with the natural log transform of TSH as the response. In untreated individuals, full regression models were adjusted for sex (unless separate models for sex were fitted) and stratified by age in 20-year bands (1 19, 20 39, 40 59, 60 79, and 80 years), with the TSH response further clarified by segmenting the free T 4 domain into 3 regions: below, within, and above the reference range. For the analysis of hypothyroid subjects, overt hypothyroidism was defined as free T 4 10 pmol/l with TSH 4.0 mu/l and severe hypothyroidism as free T 4 5 pmol/l with TSH 4.0 mu/l. When multiple comparisons were made between groups, a Bonferroni correction to the significance level was made such that.05/number of comparisons. All analysis was completed in the R statistical computing environment, version R (26). Results Demographics After exclusions, the study sample consisted of individuals, including (21%) who were taking thyroxine therapy. Demographics are shown in Table 1. Most subjects were women (75%), and women comprised 85% of those receiving thyroxine replacement and 72% of those not receiving therapy. On average, women were younger than men (50.8 vs 54.1 years, P.001), whereas patients taking thyroxine were older than those not receiving therapy (55.5 vs 50.6 years, P.001). As expected for a population with a wide range of thyroid function, including those with overt and subclinical thyroid dysfunction, the median TSH was in the upper part of the reference range at 3.0 mu/l (LQ, 1.0; UQ, 4.8). The distribution of TSH was non-gaussian. Relationship of TSH to free T 4 By quantile regression analysis, there was a significant inverse relationship between median log TSH and free T 4 (P.001). Figure 1 shows the median TSH (with LQs and UQs) corresponding to each free T 4 concentration (in picomoles per liter) for the subjects who were not taking thyroxine treatment. TSH is plotted on a log scale and free T 4 on a linear scale. If the TSH free T 4 relation-

4 doi: /jc jcem.endojournals.org 2939 Figure 1. Relationship between TSH and free T4 in subjects not receiving thyroxine treatment. The median TSH for each free T4 integer value is represented ( ). The UQs and LQs for TSH are shown by the dashed lines. The solid curve is derived from quantile regression analysis of all data points with the relationship described by 3.5 ln TSH freet4 /1.0 for free T4 values 12 pmol/l 1 e 5.3 and ln TSH freet4 /3.0 for free T4 1 e concentrations 12 pmol/l. TSH is represented on a logarithmic scale and free T4 on a linear scale. Dotted horizontal and vertical lines mark the TSH reference range ( mu/l) and free T4 reference range (10 20 pmol/l), respectively. To convert free T4 to nanograms per deciliter, multiply by ship were inverse log-linear, this plot would be a straight line; instead it resembles 2 negative sigmoid curves that merge in a central threshold region corresponding to the lower part of the reference range for free T 4. Mathematical modeling confirmed that the relationship can be described by 2 sigmoid curves with inflection points corresponding to free T 4 concentrations of 7.0 (95% confidence interval, ) and 20.6 (95% confidence interval, ) pmol/l, respectively. For free T 4 concentrations of 12 pmol/l, the relationship was ln TSH e 7.0 freet4 /1.0, whereas for free T 4 levels 12 pmol/l, it was ln TSH e 20.6 freet4 /3.0. Further information on the fitted parameters is provided in Supplemental Table 1 (published on The Endocrine Society s Journals Online web site at Effect of thyroxine treatment The median TSH was lower in patients receiving thyroxine treatment than in untreated individuals (2.0 vs 3.5 mu/l, P.001) (Table 1). The relationship between TSH Figure 2. Relationship between TSH and free T4 in thyroxine-treated and untreated subjects. The median TSH for each free T4 integer value for treated ( ) and untreated subjects (Œ) is represented. UQs and LQs for TSH are shown for each group. Note that the logarithmic scale for TSH visually exaggerates differences in median TSH at high free T4 levels and that actual differences are small. Dotted horizontal and vertical lines mark the TSH reference range ( mu/l) and free T4 reference range (10 20 pmol/l), respectively. and free T 4 appeared to be similar in subjects receiving thyroxine therapy and untreated individuals except at extremely low free T 4 values (Figure 2); the latter results need to be interpreted with caution because treated individuals may not have been at steady state. For the same reason, the following sex and age group analyses included only individuals who were not receiving thyroxine treatment. Sex differences in untreated subjects In individuals not receiving thyroxine treatment, median TSH was higher in men than in women (3.8 vs 3.3 mu/l, P.001 unadjusted and after adjustment for age), whereas the sex difference in mean free T 4 was small (14.0 pmol/l in men vs 13.7 in women, P.001) (Table 1). Figure 3A shows the median TSH for men and women plotted according to free T 4 values. Within the free T 4 reference range (10 20 pmol/l), median TSH was significantly higher in men than in women at every free T 4 value except 10 pmol/l (P.001, where significance is set at.0045). The differences were most pronounced for free T 4 values in the upper part of the reference range (14 20 pmol/l), whereas the absolute TSH difference between sexes ranged from 0.6 to 1.2 mu/l. Age effects in untreated subjects Figure 3B shows the TSH free T 4 relationship in untreated subjects in 20-year age bands. For individuals with free T 4 concentrations within the reference range (10 20 pmol/l), there was an age-related increase in median TSH across adult age groups (all P.001 and less than

5 2940 Hadlow et al Complex Relationship Between TSH and Free T 4 J Clin Endocrinol Metab, July 2013, 98(7): less than.0125). This differential age response in TSH was also evident in 878 individuals with severe hypothyroidism, defined as free T 4 5 pmol/l with elevated TSH (Figure 4B). Figure 3. TSH and free T4 relationship analyzed by sex and age in subjects who were not taking thyroxine treatment. A, Median TSH for each free T4 integer value for men (f) and women ( ). B, Results for 20-year age bands in adults. TSH and free T4 values are plotted on linear (rather than logarithmic) scales to improve visualization of true differences in median TSH values between the groups. In A, the UQs and LQs for TSH are shown by the dashed lines. In B, the youngest age band (1 19 years) is not depicted because there were insufficient data to calculate median TSH for each free T4 value reliably; for clarity, interquartile ranges are not shown. Dotted horizontal and vertical lines mark the TSH reference range ( mu/l) and free T4 reference range (10 20 pmol/l), respectively..0125), with the highest TSH values being in the oldest subjects ( 80 years) (P.001) (Figure 3B and Table 2). This age-related increase was present in both adult men and women (all P.001 and less than.00625). Age and TSH in untreated hypothyroidism Of 5591 individuals not receiving thyroxine treatment who had free T 4 levels 10 pmol/l, 4403 (79%) had elevated TSH concentrations, indicating overt hypothyroidism. In these subjects, there was a significant quadratic pattern in median TSH with advancing age (P.001 and P.003 for linear and quadratic terms, respectively). Figure 4A shows the median TSH in hypothyroid subjects plotted for 20-year age bands. Using adults aged 20 to 39 years as the reference group, median TSH was significantly lower in older age groups than in the younger adults (P.001 unadjusted and after adjustment for free T 4, and Discussion This study clearly demonstrates that the relationship between log TSH and free T 4 is nonlinear. Although a change in slope is to be expected at extremes of free T 4 and TSH (when TSH secretion is maximally stimulated or suppressed), lack of linearity is not confined to those regions of the plot but is clearly apparent across the physiological range of free T 4 and TSH. In fact, the relationship is best described by 2 negative sigmoid curves, with a central threshold area corresponding to the lower part of the free T 4 reference range. This finding suggests that the TSH response to changes in free T 4 is amplified when free T 4 levels are less than normal than when they are within the reference range. Our results are consistent with 2 recent, smaller cross-sectional studies, each of which reported a complex, nonlinear relationship between log TSH and free T 4 (8, 9). Our results (and those of Hoermann et al [8] and Clark et al [9]) differ from previous studies that reported a simple log-linear relationship between TSH and T 4 (7, 27, 28), probably because of the more sophisticated statistical analysis used in the more recent studies. The TSH free T 4 relationship differed significantly between men and women, particularly in individuals with free T 4 values within the reference range, in whom the Table 2. Median TSH in Subjects Not Receiving Thyroxine Treatment with Free T 4 Concentrations Within the Reference Range (10 20 pmol/l) Median TSH, mu/l 1 19 y y y y >80 y P Value All 2.8 (1.2, 4.5) 2.0 (0.8, 4.3) 3.6 (1.3, 4.9) 3.9 (1.6, 5.0) 4.1 (1.6, 5.4).001 a n Men 3.7 (1.8, 4.7) 2.6 (1.1, 4.5) 3.7 (1.4, 4.9) 4.0 (1.7, 5.1) 4.3 (2.2, 5.6).001 a n Women 2.3 (1.0, 4.4) 1.9 (0.8, 4.2) 3.6 (1.3, 4.9) 3.9 (1.6, 5.0) 4.0 (1.4, 5.3).001 a n Median TSH (LQ, UQ) are given for each age band in the entire cohort and for men and women separately. Number of subjects (n) for each age band and sex is indicated. a P value indicates a significant quadratic relationship across age groups (P.001 for both quadratic and linear terms). When the 1- to 19-year age group was excluded from analysis, there was a significant linear trend in median TSH across adult age groups (P.001).

6 doi: /jc jcem.endojournals.org 2941 Figure 4. Median TSH in 4403 subjects with overt hypothyroidism (A) and 878 subjects with severe hypothyroidism (B) according to age. Overt hypothyroidism was defined as free T4 10 pmol/l with TSH 4.0 mu/l and severe hypothyroidism as free T4 5 pmol/l with TSH 4.0 mu/l. Error bars indicate LQs and UQs for TSH. Asterisks indicate significant differences compared with the reference group of young adults aged 20 to 39 years at median TSH for any given free T 4 value was higher in men than in women. Consistent with our data, Clark et al (9) also reported that median TSH was higher in men than in women for free T 4 levels in the upper reference range and higher, but, in contrast to our data, they found that for free T 4 levels less than the mid reference range, TSH was higher in women than in men. This difference between studies is unexplained: it might reflect the more restricted age group studied by Clark et al (all participants aged 65 years or older) or different approaches to analyzing the data. The sexual dimorphism in HPT axis function found in our study is consistent with the effects of sex hormones on pituitary function in animal models. In rats, serum TSH concentrations are higher in gonad-intact males than in females or castrated males, and testosterone treatment of castrated males increases TSH back to normal male levels (29, 30). In comparison with intact males, female rats and estrogen-treated males have a higher density of T 3 receptors and higher type II 5 -deiodinase activity in the pituitary, suggesting that estrogen augments the negative feedback effect of thyroid hormone on TSH secretion (29). Data from humans are very limited, but in one study, pharmacological doses of estrogen reduced TSH secretion in men (31). The TSH free T 4 relationship also differed among age groups, with an age-related increase in median TSH at free T 4 concentrations within the reference range. This result is consistent with cross-sectional studies reporting an agerelated increase in TSH concentration (12, 13, 32) and 2 longitudinal studies in which the age-related TSH increase was not accompanied by a fall in free T 4 concentration (15, 16). The mechanisms underlying this change are not clear, but possibilities include a higher prevalence of occult thyroid disease in older people (33) or changes in the HPT set point, deiodinase activity, or TSH bioactivity with aging (14, 16). Survival bias is another possible explanation because higher TSH concentrations are associated with extreme longevity (32); increased survival of those with high-normal TSH levels would lead to an apparent increase in TSH with age. Our study shows for the first time that within the reference range for free T 4, older subjects have significantly higher TSH concentrations than younger subjects for the same free T 4 concentration. This finding provides strong support for the view that the age-related increase in TSH does not arise primarily from occult thyroid disease but rather from an age-related change in HPT set point or reduced TSH bioactivity. It also suggests that older people are more likely to have subclinical hypothyroidism than are younger people with the same free T 4 concentration, which may not be appropriate (14). In contrast to the higher TSH concentrations observed in older subjects with free T 4 levels in the reference range, the TSH response to hypothyroidism was more robust in younger than in older subjects, consistent with previous, smaller studies (17 19). This finding was particularly evident in individuals with free T 4 concentrations of 5 pmol/l, which corresponds to a flattening of the log TSH free T 4 curve. It appears, therefore, that the effects of aging on thyrotrope function differ among pathophysiological conditions: when free T 4 concentrations are normal, immunoreactive TSH concentrations are higher in older people, whereas when thyrotropes are maximally stimulated by hypothyroidism, the TSH response is attenuated in older age groups. The mechanism underlying this difference is unknown. This study is important for several reasons. First, it advances understanding of the TSH-T 4 relationship, which is central to thyroid physiology and pathophysiology and laboratory diagnosis of thyroid disease. Second, it confirms and extends previous studies on sex differences in HPT axis function and changes in the function of that axis with aging. Third, it has important implications for thyroid research, because the relationship between log TSH and T 4 has been assumed to be linear in diverse studies, including assessment of free thyroxine assay performance (34, 35), assessment of pituitary resistance to thyroid hormone (27), diagnosis of central hypopituitarism (28), mathematical modeling of the TSH-T 4 relationship (36), and studies of the HPT axis set point (2).

7 2942 Hadlow et al Complex Relationship Between TSH and Free T 4 J Clin Endocrinol Metab, July 2013, 98(7): Strengths of this study include the very large size of the data set which, with more than subjects, (including more than 4000 with overt hypothyroidism) provides unrivaled statistical power to study the TSH free T 4 relationship. We deliberately did not restrict the sample to a healthy population because inclusion of subjects with thyroid disease was important to allow for assessment of the TSH free T 4 relationship over a broad range of thyroid states. Women were overrepresented, reflecting the female bias in thyroid disease and greater readiness to order thyroid function tests in women, but the number of men remains substantial. A limitation is that we did not have access to potentially relevant data such as ethnicity, body mass, diabetes, glucocorticoid treatment, or tobacco smoking; the latter, in particular, has been shown to affect the TSH free T 4 relationship (9). Our study is cross-sectional, and it is uncertain how far the results apply to individuals whose thyroid function is changing over time. In that regard, a longitudinal study with a large number of subjects studied repeatedly would be informative. Arguably, a limitation of this study is that we used a commercial free T 4 assay rather than a reference method such as equilibrium dialysis isotope dilution (ED ID) liquid chromatography tandem mass spectrometry (LC-MS/ MS). It is recognized that free T 4 immunoassays can be subject to bias and interference, particularly in states associated with altered protein binding such as critical illness and pregnancy (37); accordingly we excluded samples from known pregnant women and those requested from hospital inpatients. However, good correlation has now been demonstrated between modern free T 4 immunoassays and ED ID LC-MS/MS in pregnant women (37), and it has been argued that current free T 4 assays have now addressed the protein issues associated with early assays (38). Two recent studies concluded that LC-MS/MS for free T 4 gives results superior to those for immunoassay (34, 35), but this conclusion was based largely on correlations between log TSH and free T 4 with the erroneous assumption that the relationship should be linear. Furthermore, ED ID LC MS/MS is currently regarded as too technically difficult, demanding, inconvenient, and expensive for routine laboratory use (39). Although it would be of interest to examine the TSH free T 4 relationship using such a method, this would clearly be impractical for a study as large as ours. Thus, it cannot be assumed that our results apply to free T 4 measured by ED ID LC MS/MS methods. In conclusion, this study shows that the TSH free T 4 relationship is not simple inverse log-linear as previously thought but can be described by 2 overlapping negative sigmoid curves. The relationship differs between sexes and age groups with higher TSH concentrations observed in men and in older adults under physiological conditions, but paradoxically a more robust TSH response to hypothyroidism in younger people. Acknowledgments We thank Diane Bleasdale (Librarian PathWest) for her invaluable assistance and the scientific staff of Western Diagnostic Pathology for analysis of the thyroid function tests, which contributed the data for this analysis. Address all correspondence and requests for reprints to: Clinical Professor John Walsh, Department of Endocrinology and Diabetes, First Floor C Block, Sir Charles Gairdner Hospital, Nedlands, Western Australia 6009, Australia. john.walsh@health.wa.gov.au. Disclosure Summary: The authors have nothing to disclose. References 1. Andersen S, Pedersen KM, Bruun NH, Laurberg P. Narrow individual variations in serum T 4 and T 3 in normal subjects: a clue to the understanding of subclinical thyroid disease. J Clin Endocrinol Metab. 2002;87: Benhadi N, Fliers E, Visser TJ, Reitsma JB, Wiersinga WM. Pilot study on the assessment of the setpoint of the hypothalamus-pituitary-thyroid axis in healthy volunteers. Eur J Endocrinol. 2010; 162: Meier CA, Maisey MN, Lowry A, Müller J, Smith MA. Interindividual differences in the pituitary-thyroid axis influence the interpretation of thyroid function tests. Clin Endocrinol (Oxf). 1993; 39: Wehmann RE, Nisula BC. Radioimmunoassay of human thyrotropin: analytical and clinical developments. Crit Rev Clin Lab Sci. 1984;20: Reichlin S, Utiger RD. Regulation of the pituitary-thyroid axis in man: relationship of TSH concentration to concentration of free and total thyroxine in plasma. J Clin Endocrinol Metab. 1967;27: Fish LH, Schwartz HL, Cavanaugh J, Steffes MW, Bantle JP, Oppenheimer JH. Replacement dose, metabolism, and bioavailability of levothyroxine in the treatment of hypothyroidism. Role of triiodothyronine in pituitary feedback in humans. N Engl J Med. 1987; 316: Spencer CA, LoPresti JS, Patel A, et al. Applications of a new chemiluminometric thyrotropin assay to subnormal measurement. J Clin Endocrinol Metab. 1990;70: Hoermann R, Eckl W, Hoermann C, Larisch R. Complex relationship between free thyroxine and TSH in the regulation of thyroid function. Eur J Endocrinol. 2010;162: Clark PM, Holder RL, Haque SM, Hobbs FD, Roberts LM, Franklyn JA. The relationship between serum TSH and free T4 in older people. J Clin Pathol. 2012;65: Walsh JP. Setpoints and susceptibility: do small differences in thyroid function really matter? Clin Endocrinol (Oxf). 2011;75: Panicker V, Wilson SG, Walsh JP, et al. A locus on chromosome 1p36 is associated with thyrotropin and thyroid function as identified by genome-wide association study. Am J Hum Genet. 2010; 87: Boucai L, Surks MI. Reference limits of serum TSH and free T4 are

8 doi: /jc jcem.endojournals.org 2943 significantly influenced by race and age in an urban outpatient medical practice. Clin Endocrinol (Oxf). 2009;70: Kahapola-Arachchige KM, Hadlow N, Wardrop R, Lim EM, Walsh JP. Age-specific TSH reference ranges have minimal impact on the diagnosis of thyroid dysfunction. Clin Endocrinol (Oxf). 2012;77: Surks MI, Hollowell JG. Age-specific distribution of serum thyrotropin and antithyroid antibodies in the US population: implications for the prevalence of subclinical hypothyroidism. J Clin Endocrinol Metab. 2007;92: Waring AC, Arnold AM, Newman AB, Bùzková P, Hirsch C, Cappola AR. Longitudinal changes in thyroid function in the oldest old and survival: The Cardiovascular Health Study All-Stars Study. J Clin Endocrinol Metab. 2012;97: Bremner AP, Feddema P, Leedman PJ, et al. Age-related changes in thyroid function: a longitudinal study of a community-based cohort. J Clin Endocrinol Metab. 2012;97: Over R, Mannan S, Nsouli-Maktabi H, Burman KD, Jonklaas J. Age and the thyrotropin response to hypothyroxinemia. J Clin Endocrinol Metab. 2010;95: Wiener R, Utiger RD, Lew R, Emerson CH. Age, sex, and serum thyrotropin concentrations in primary hypothyroidism. Acta Endocrinol (Copenh). 1991;124: Carlé A, Laurberg P, Pedersen IB, et al. Age modifies the pituitary TSH response to thyroid failure. Thyroid. 2007;17: Faglia G, Beck-Peccoz P, Ferrari C, Ambrosi B, Spada A, Travaglini P. Enhanced plasma thyrotrophin response to thyrotrophin-releasing hormone following oestradiol administration in man. Clin Endocrinol (Oxf). 1973;2: Sanchez-Franco F, Garcia MD, Cacicedo L, Martin-Zurro A, Escobar del Rey F. Influence of sex phase of the menstrual cycle on thyrotropin (TSH) response to thyrotropin-releasing hormone (TRH). J Clin Endocrinol Metab. 1973;37: Mortimer CH, Besser GM, Goldie DJ, Hook J, McNeilly AS. The TSH, FSH and prolactin responses to continuous infusions of TRH and the effects of oestrogen administration in normal males. Clin Endocrinol (Oxf). 1974;3: Roelfsema F, Pereira AM, Veldhuis JD, et al. Thyrotropin secretion profiles are not different in men and women. J Clin Endocrinol Metab. 2009;94: Venegas JG, Harris RS, Simon BA. A comprehensive equation for the pulmonary pressure-volume curve. J Appl Physiol. 1998;84: Koenker R. quantreg: Quantile Regression. R package version Vienna, Austria: R Foundation for Statistical Computing R Core Team.R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; Ercan-Fang S, Schwartz HL, Mariash CN, Oppenheimer JH. Quantitative assessment of pituitary resistance to thyroid hormone from plots of the logarithm of thyrotropin versus serum free thyroxine index. J Clin Endocrinol Metab. 2000;85: Jostel A, Ryder WD, Shalet SM. The use of thyroid function tests in the diagnosis of hypopituitarism: definition and evaluation of the TSH Index. Clin Endocrinol (Oxf). 2009;71: Donda A, Reymond F, Rey F, Lemarchand-Béraud T. Sex steroids modulate the pituitary parameters involved in the regulation of TSH secretion in the rat. Acta Endocrinol (Copenh). 1990;122: Christianson D, Roti E, Vagenakis AG, Braverman LE. The sexrelated difference in serum thyrotropin concentration is androgen mediated. Endocrinology. 1981;108: Gross HA, Appleman MD Jr, Nicoloff JT. Effect of biologically active steroids on thyroid function in man. J Clin Endocrinol Metab. 1971;33: Atzmon G, Barzilai N, Hollowell JG, Surks MI, Gabriely I. Extreme longevity is associated with increased serum thyrotropin. J Clin Endocrinol Metab. 2009;94: Spencer CA, Hollowell JG, Kazarosyan M, Braverman LE. National Health and Nutrition Examination Survey III thyroid-stimulating hormone (TSH)-thyroperoxidase antibody relationships demonstrate that TSH upper reference limits may be skewed by occult thyroid dysfunction. J Clin Endocrinol Metab. 2007;92: van Deventer HE, Mendu DR, Remaley AT, Soldin SJ. Inverse loglinear relationship between thyroid-stimulating hormone and free thyroxine measured by direct analog immunoassay and tandem mass spectrometry. Clin Chem. 2011;57: Jonklaas J, Kahric-Janicic N, Soldin OP, Soldin SJ. Correlations of free thyroid hormones measured by tandem mass spectrometry and immunoassay with thyroid-stimulating hormone across 4 patient populations. Clin Chem. 2009;55: Leow MK. A mathematical model of pituitary-thyroid interaction to provide an insight into the nature of the thyrotropin-thyroid hormone relationship. J Theor Biol. 2007;248: Yue B, Rockwood AL, Sandrock T, La ulu SL, Kushnir MM, Meikle AW. Free thyroid hormones in serum by direct equilibrium dialysis and online solid-phase extraction liquid chromatography/tandem mass spectrometry. Clin Chem. 2008;54: Midgley JE, Christofides ND. Free thyroxine assays: no going back! Am J Obstet Gynecol. 2010;202:e Thienpont LM. A major step forward in the routine measurement of serum free thyroid hormones. Clin Chem. 2008;54: