| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This study was undertaken at Western Hospital, Melbourne, Australia. The hospital services >600 000 people in Melbourne's industrial west. It provides complete pediatric, adult medicine and surgery, obstetrics and gynecology, psychiatric, and geriatric services to both inpatients and ambulatory outpatients. It draws patients from the surrounding general practitioner-treated pool. There are no large private hospitals in the area and no areas of large personal wealth. As a result the hospital patients uniformly represent the regional population. The ethnic composition of the population is ~70% Caucasian and 30% Asian. There is a negligible negroid component and there are a few aboriginal Australians.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
• patients with thyroid diseases (including those patients being treated for thyroid disease, as for instance with 1-T4)
• patients suspected of having thyroid or pituitary disease
• patients with psychoses or dementia
• patients being investigated for infertility
• patients taking drugs known to interfere with thyroid gland function
For every patient tested, the clinician's request form was scrutinized and, where necessary to find further information, the patient medical record was consulted.
Patient age and ambulatory/inpatient status were noted, and, for women, whether or not they were pregnant. Full blood examination (FBE) and renal function test (RFT) data were accessed and categorized into within or outside reference intervals for age.
Work reported elsewhere demonstrates that all the results from those patients who neither had, nor were suspected of having, thyroid disease were homogenous (3). A process similar to that prescribed by the IFCC for the selection of individuals for the production of reference ranges (4) was followed. The results could not be distinguished statistically on the basis of being in any of the nonthyroid Medicare categories listed above. Moreover, there was no difference between inpatients and ambulatory patients, nor between patients whose FBE, RFT, or both were or were not within the reference intervals, or alternatively were unknown. Hence, for this latter study, all results except those from patients with thyroid disease were included. After the patients with thyroid disease were excluded, 754 sets of measurable TSH and FT4 results in patients between ages 1 and 101 years remained for analysis. None of the patients was excluded from this study on the basis of any other medications being taken, even though several commonly used medications have been reported to alter TFT.
Patients were sorted into five-year age groups: ages 1–4, 5–9, 10–14, and so on until the final group, age 94 and older. Patients <1 year of age were excluded because they have different, and constantly differing, reference ranges compared with older individuals (5). The one patient found in the group >100 years of age, a woman age 101 years, was included in the >94 years of age group.
Statistical calculations were performed with the Stata package (Release 3: Computing Resource Centre, 1992). Arithmetic means for FT4 and geometric means for TSH, with 95% confidence intervals (CI), were calculated for each age group, first for males and females, second for inpatients and ambulatory patients, and finally in total, for each of TSH and FT4. To provide a perspective of the result spread, the 5th and 95th percentile values for the total result in each age group were also calculated.
Two-way ANOVA among the FT4 and TSH results was undertaken for age (as a continuous variable), gender, and patient location, and analysis of covariance (ANOCOVA) for age with gender and for age with patient location. The Stata program accommodates designs such as found in the data here, with imbalance, even to the extent of missing cells. CI overlaps were noted and Student's t-tests were performed post hoc where indicated among the results found in the age group data, with age as an arbitrarily categorical variable. Linear regression analysis was performed where needed.
This work was performed in accordance with the appropriate standards of the Western Hospital Research and Ethics Committee.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). Adjacent values are formally defined as three-halves of the
interquartile range rolled back to where data persist; they run the
length of the whisker. Outside values are more extreme than the
adjacent values.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Age-dependent reference intervals have previously been used for TSH (8), but more recent work has shown that although there are minor variations in both TSH and FT4 values as between younger and older adults, the limits of the variation lie within the accepted reference intervals (TSH 0.5–5.0 mU/L and FT4 9–24 pmol/L) (Ciba Corning Diagnostics, assay kit insert data) and should not result in diagnostic algorithms needing modification on an age-dependent basis (9). This cross-sectional study confirms and redefines these findings.
tsh
In the population studied here TSH does not vary with age
(ANOCOVA, F >0.885). Nelson et al. (10)
studied the complex FT4-TSH interplay in infants, children,
and younger adults to age 45 years. They found that an adult plateau in
TSH concentrations is reached at age 31 years. Our numbers in the age
groups <20 years are too few to compare with Nelson's work but
thereafter would suggest that the adult plateau has been reached by as
early as 20 years of age.
From age 20 to 80 years our findings probably concur with those of Hershman et al. (11). They compared geometric mean TSH (and 95% CI) in 216 persons >70 years of age with 211 of their offspring, ages 40–60 years. They do not give any more detail as to the age composition. Their results and ours for the older group are 1.24 (0.29–5.40) and 1.32 (1.15–1.52), and for the younger 1.45 (0.54–3.90) and 1.41 (1.17–1.68) mU/L, respectively. The results are statistically indistinguishable, although it should be noted that our greater subject numbers produce much narrower CIs. When age is treated as a continuous variable, we find no difference between the results of men and women (ANOCOVA, F >0.883), nor is there any interaction between age and gender (F >0.846). In the 40–44 years age group there is a statistically significant difference (mean ± SD 0.99 ± 2.11 vs 2.31 ± 2.67, P <0.039), but this is considered to be a sampling phenomenon only. No difference is seen in any of the other categorical age groupings.
Visual inspection of the graphed TSH data suggests that from age 80 years and up, TSH concentrations decline, and although the TSH mean (±SD) from age 80 years (0.88 ± 3.32 mU/L) does not differ from the mean found through the 20–79 years of age adult plateau (1.61 ± 2.23 mU/L), the difference approaches significance (P <0.079). This is partly in agreement with Hershman et al. (11), whose older women had slightly lower TSH values than their younger offspring (1.21 vs 1.52 mU/L, P <0.05). Beyond age 90 years, the subject numbers in our age groups are smaller than in the groups up to age 90 years, and are thus less reliable.
Klee and Hay (8) mention "studies of 448 euthyroid subjects [who] did not show significant age differences in [TSH in] adults between 20 and 80 years... " in their review of TSH in the diagnosis of thyroid disease. Our data agree and extend the age limit to 90 years. Although our octogenarians' TSH values trend downwards, they are still unequivocally within the reference interval and do not give cause for concern when using a TSH-based diagnostic algorithm.
No difference was seen between the TSH results of inpatients and those of the ambulatory outpatients (ANOCOVA, F >0.294), nor was there any interaction between age and patient location (F >0.298). Although not absolutely contiguous, the boundaries of the in/outpatient groups are probably very close to those of the acute vs chronic illness groups. Because all these results are from hospital patients, a minimum background of chronic disease could be expected, or at least perceived illness by the referring community practitioner, in the outpatient group. Acute disease, or an acute exacerbation of chronic disease, will be found in the inpatients. No differences between the two groups' TSH results is in accord with the primary study (3), in which the finding that a patient had an abnormal FBE or RFT, indicating disturbance in those basic body systems, did not distinguish among TSH results.
t4
At the two extremes of the adult range of our study population,
ages 20–24 years vs 85–89 years, the mean FT4 ± SD
differs (12.96 ± 2.36 vs 15.84 ± 3.48 pmol/L, respectively,
P <0.0003), but both means are unequivocally within the
reference interval, 9–24 pmol/L.
Szabolics et al. (12) found that FT4 ranges in their euthyroid geriatric (>60 years of age) and healthy young (ages 20–40 years) groups were indistinguishable. They noted that although FT4 did apparently independently decrease with increasing age and increase with severity of illness, the mean values were still clearly within the reference interval. Higher FT4 values with paradoxically nonsuppressed TSH occurred more frequently with increasing severity in the clinical disease status. They caution that their conclusions should be restricted to "hospitalized chronic geriatric patients." By contrast, our results relate to a free-range human population, albeit half with acute illness justifying hospital admission.
By contrast with Szabolics et al., we see the TSH in the same age groups trending, appropriately, downwards, while remaining within the reference interval, as noted above. This relation reflects the expected response in the healthy pituitary/thyroid axis and suggests that the trend seen in the data we present is a real phenomenon in the aging human. It is not affected by hospital admission for acute illness or the acute exacerbation of chronic illness, or by the possibility of concurrent dementia (data available on request.) It partly confirms the findings of van Coevorden et al. (13) who compared only eight "healthy... men" 67–84 years of age with eight "normal... men" 20–27 years of age, but did assay TSH at 15-min intervals over 24 h. They found that the 24-h mean TSH concentration in the older men was ~50% lower than that in the younger (0.78 ± 0.37 vs 1.43 ± 0.41 mU/L, P <0.01). In their eight elderly men, "basal T4 levels were normal."
Covariance analysis suggests that there may be gender-related variation in FT4 (F <0.041), and interaction is demonstrated between age and gender (F <0.016). An explanation is found in gender-isolated linear regression studies of FT4 with age. Among males there is no statistically significant trend, but among females FT4 is age related (FT4 = 0.035age + 12.087, P <0.0001, adjusted R2 = 0.056). The difference in FT4 seen between the two ends of the age spectrum thus appears limited to women, although, as noted above, the variation identified remains definitely within the reference interval. The spread of results among men is also unequivocally within that interval but is wider and thereby eliminates any gender differences when FT4 results are compared by categorical age groups.
There was no difference seen between the inpatients' results and those of ambulatory patients for FT4, except in the 25–29 years of age group (15.31 ± 3.41 vs 12.26 ± 1.90 pmol/L, P <0.003). One such event, again, can be attributed to the effect of a sampling phenomenon. The ANOCOVA result reflects the effect of the difference in the 25–29 years of age group (F <0.022) when patient location alone is examined, but there is no interaction between age and patient location overall (F >0.216).
synthesis
The primary study from which these data were drawn a posteriori
found that a TSH-based thyroid disease screen is broadly efficacious in
ruling out thyroid disease (3). It also touched
on the question of ruling in thyroid disease. This current
study into the effect of age on TSH and T4 concentrations
did not examine the question of the relation between age and the
ability of TSH to rule in thyroid disease; i.e., we did not determine
how many false-negative results might occur in a population of patients
who truly have thyroid disease when TSH alone is used to screen them
and age-related reference intervals are not used. In the primary study,
2.7% of abnormal FT4 results (from 6.4 to 29.5 pmol/L)
were missed because the patient had a TSH within reference intervals,
but those abnormal FT4 results were only marginally outside
their reference interval and no age-related trend was found among them
on subsequent analysis.
The answer to the diagnostic conundrum of the false negative may never be known. An accepted method for finding patients with preclinical thyroid disease involves testing for thyroid antibodies, but this is not generally accepted as a part of an initial thyroid screen, except in an algorithm-triggered mode, and with the fiscal climate in a permanently contractionary mood, it is highly unlikely that having ousted FT4, we can replace it with even more expensive immunologic tests looking for subclinical disease.
The symptoms of thyroid disease are, however, admittedly vague and protean, and the disease process may be well advanced before it is brought to clinical notice. Clinicians' abilities to discern the presence of the disease have been poor (3)(14), but they do need to maintain a high suspicion for the diagnosis and accept that there is now a reliable, inexpensive, and highly robust screening method to exclude thyroid disease when it does not exist. In selected cases where there is discord between initial testing results and clinical suspicion it may become appropriate to add the antibody tests.
This study confirms earlier findings that the pituitary gland is not only an exquisitely sensitive monitor of thyroid status but that it is also an extremely stable monitor into late adult life. It shows that the effect of intercurrent illness, which becomes more likely as we age, does not appear to interfere with TSH values so sufficiently as to render them diagnostically useless. In particular, the occasional case of the sick euthyroid syndrome with depressed TSH values must be considered, but those cases do not negate the usefulness of TSH as the primary screen for disease; they merely necessitate the further assay of FT4. And, having been tested in the challenging arena of an acute general hospital's patient population and found to be robust, TSH-based algorithms can be used safely across the range of primary healthcare.
Diagnostic algorithms for the exclusion of thyroid gland disease, based on TSH, can confidently be used over the full range of adult life. Age-dependent TSH reference intervals should be abandoned.
|
|
|
Acknowledgments |
|---|
|
Footnotes |
|---|
1 Nonstandard abbreviations: TFT, thyroid function testing; TSH, thyrotropin; FT4, free thyroxine; T4, thyroxine; FBE, full blood examination; RFT, renal function test; CI, confidence interval; and ANOCOVA, analysis of covariance. 
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
. Refer to the
Tables to find the ages allocated to each age group.