The Thyroid and its Diseases
	6A -- Assay of Thyroid Hormones and Related Substances 6B -- Clinical Strategies in the Testing of Thyroid Function Revised by Jim Stockigt, 28 September 2004 6C -- Ultrasonography of the Thyroid 6D -- Fine Needle Aspiration Biopsy of the Thyroid Gland 6E -- Evaluation of Thyroid Function in Health and Disease	Downloads

1. WHO SHOULD BE TESTED FOR THYROID DYSFUNCTION?

The clinical manifestations of thyrotoxicosis (hyperthyroidism), or hypothyroidism are so diverse that diagnosis based on clinical features alone lacks both sensitivity and specificity. Hence, reliance is placed on measurements of circulating thyroid hormones and thyroid stimulating hormone (TSH) to confirm or rule out thyroid dysfunction (TD).

Guidelines from the American College of Physicians in 1998 (1, 2) for the detection of TD now recommend routine measurement of TSH in women over 50, the group most likely to have overt or subclinical thyroid dysfunction. It should be noted that testing is advocated at the time of presentation for medical care, ie a case finding strategy, rather than screening of the whole population group. This approach depends on the application of TSH assays sufficiently sensitive to detect the earliest stages of either thyrotoxicosis or primary hypothyroidism. A normal serum TSH value in ambulatory patients without associated disease or pituitary dysfunction has a high negative predictive value in ruling out both primary hypothyroidism and thyrotoxicosis (1, 2), which suggests that free T4 need only be routinely estimated if TSH is abnormal. However, it should be noted that this ”TSH first” strategy of thyroid function testing has definite limitations (see below). After a normal TSH value, re-testing is probably not required for 5 years (3). Recent American Thyroid Association guidelines (4) extend these indications to recommend that all adults have their serum TSH concentrations measured, beginning at age 35 and every 5 years thereafter. The benefits from this approach, particularly in males, remain to be demonstrated.

In some groups (table 1), known to be at increased risk of TD, there is a case for routine testing even in the absence of suggestive clinical features.

Almost all developed countries now have routine neonatal screening programs for congenital hypothyroidism using heel prick filter paper blood spots (see chapter 15). The value of such programs is clear (19), but it is notable that neonatal screening is not yet routine in numerous developing countries where the prevalence of neonatal hypothyroidism may be high, often associated with iodine deficiency (20). In terms of clear benefit from allocation of health care resources in developing countries, the establishment of neonatal screening (20a) should probably take precedence over routine testing of adults.

Based on recent evidence of significant intellectual impairment in the offspring of women who were even mildly hypothyroid early in pregnancy (21), testing of thyroid function may soon become standard practice as early as possible in pregnancy, or in women who intend to become pregnant. Early diagnosis and treatment are crucial, because foetal brain development in the first trimester is influenced by maternal thyroid function and follow-up studies indicate significant intellectual impairment in the offspring of women who were mildly hypothyroid during pregnancy (22,22a)The case for routine testing of thyroid function early in pregnancy would be further reinforced by confirmation of the finding of about 4-fold increase in fetal death in women found to have serum TSH >6 mU/l in the second trimester of pregnancy (23). The arguments for and against screening for autoimmune thyroid diseases (5) and postpartum TD in particular (24) have been debated in detail.

An increased prevalence of features of thyroid autoimmunity has been reported in infertile women, particularly those with endometriosis (24a), and in women who have recurrent miscarriages (24b), but benefits from correction of subclinical thyroid dysfunction have yet to be demonstrated.

1.1. The basis for a case-finding strategy

Routine laboratory testing of particular population groups becomes well founded if a testing strategy satisfies the following criteria:

1. An abnormality cannot be identified in a reliable and timely way by standard clinical assessment.
2. Dysfunction is sufficiently common to justify routine testing, either by case finding or by population screening.
3. There are adverse consequences of failure to identify dysfunction, including the possibility of progression towards more severe disease.
4. The laboratory test method is cost-effective and sufficiently sensitive and specific to identify those at risk of adverse consequences.
5. There are no major adverse consequences of testing
6. Treatment is safe and effective and prevents some or all of the adverse consequences.
7. Abnormal findings from screening or case-finding can be adequately followed-up to ensure an appropriate clinical response. An early detection program may be of little value if this last requirement cannot be met.

While the first five of the above criteria are reasonably firmly established, there is at best only weak evidence to support the final two requirements. The strength of evidence for benefit from intervention in subclinical thyroid dysfunction has recently been reviewed (24c, 24d). Despite reasonable evidence to link subclinical thyroid dysfunction to potential adverse consequences, long-term studies will be required to establish whether intervention improves mortality or quality of life.

1.2. Sensitivity and accuracy of clinical assessment

Studies of unselected patients evaluated by primary care physicians show that clinical acumen alone lacks both sensitivity and specificity in detecting previously undiagnosed TD. In two Scandinavian studies of over 3000 unselected patients who were assessed by both clinical and laboratory criteria, a thyroid disorder was not suspected by primary care physicians in over 90% of those who tested positive, even when clinical features were apparent in retrospect (25, 26). Furthermore, in up to one-third of patients evaluated for suspected thyroid dysfunction by specialists, laboratory results lead to revision of the clinical assessment (27). Systematic comparison of the standard clinical features of hypothyroidism with laboratory tests (28) showed that clinical assessment identified only about 40% with overt hypothyroidism and that classical signs were present only in patients with severe overt hypothyroidism. During the post-partum period, TD could not be reliably identified from clinical features (6).

There is some recent controversy on the relative value of clinical and laboratory evaluation of thyroid function. O'Reilly (29) has expressed the view that the clinical criteria are being side-lined, while biochemical assessments lack specificity. However, this author considered TSH and T4 individually, rather than in the feedback relationship between trophic hormone and target gland product that is fundamental to endocrine diagnosis (see below). In response to his position, several authorities emphasised the lack of sensitivity of clinical features in detecting TD, while conceding that optimal assessment requires both clinical and laboratory input (30, 31). Both overt thyrotoxicosis and hypothyroidism can have important consequences before the usual clinical features are obvious, and clinicians often fail to recognize diagnostic features even when they are present (25, 26). Thus, it may be no more valid to wait for recognition of clinical features before considering a diagnosis of thyroid dysfunction, than to rely on thirst and polyuria for diagnosis of diabetes.

Clinical evaluation remains of central importance to assess the severity of TD, evaluate discordant results, establish the specific cause of TD and monitor the response to treatment. There is little doubt that repeated laboratory confirmation of normal thyroid function is wasteful; several strategies have been suggested to improve cost-effectiveness (32, 33).

1.3. Prevalence

In considering the prevalence of thyroid dysfunction, a distinction needs to be made between so-called subclinical and overt abnormalities; a distinction that is based on laboratory rather than clinical criteria. The term 'subclinical' has been used when the serum concentration of TSH is persistently abnormal, while the concentrations of T4 and T3 remain within their reference intervals. There is a trend to replace the term 'subclinical hypothyroidism' with the designation 'mild thyroid failure' (33a) because of accumulating evidence that TSH excess alone is accompanied by tissue evidence of thyroid hormone deficiency (see below).

In the progressive development of TD, abnormal values for serum TSH generally occur before there is a diagnostic abnormality of serum T4, because of non-linearity of the negative feedback relationship between serum T4 and release of TSH from the anterior pituitary. For a two-fold change in serum T4 up or down from the setpoint for that individual, the serum TSH will normally change up to 100-fold in the reverse direction (34, 35). Thus, TSH may be recognisably abnormal months or years before there is a diagnostic change in the serum concentrations of T4 or T3.

The more widespread the testing of thyroid function in the absence of suggestive clinical features, the greater the proportion of abnormal results that indicate mild TD in which only TSH is abnormal. In evaluating serum TSH, typically defined with a normal reference interval of about 0.4-4.0 mU/l, it is important to note that normal values are logarithmically distributed, with mean and median values at 1.0-1.5 mU/l (36, 37). While values of 2-4 mU/l lie within the reference range, the likelihood of eventual hypothyroidism may increase progressively for values above 2 mU/l, especially if thyroid peroxidase antibody is positive (38).

The Whickham study, first reported in 1977 from an iodine replete region in Northern England (39), showed a prevalence of 1.9 -2.7 % overt thyrotoxicosis and 1.4 -1.9% overt hypothyroidism in women, with progressive increase with age; prevalence in males was much lower. Estimates of 'subclinical' dysfunction were 4-5 fold higher, with about 10% of women over 50 showing an increase in serum TSH, again with progressive increase with age (39). The subsequent 20 year follow up of this cohort showed an incidence of spontaneous hypothyroidism of about 3.5/1000 subjects/year in females and 0.6/1000/year in males (38). The equivalent figures for thyrotoxicosis were 0.8/1000/year and <0.1/1000/year. In women, the likelihood of developing hypothyroidism increased with age and was augmented about 8-fold if either thyroid peroxidase antibody (TPOAb) was positive, or if serum TSH had been increased in the initial study; the risk rose almost 40-fold if both were abnormal (38).

A population study in Colorado (40), of over 25, 000 individuals of mean age 56 years, 56% of whom were female, showed TSH excess in 9.5 %, with a 2.2 % prevalence of suppressed TSH; over half the group with suppressed TSH were taking thyroid medication. In women, the prevalence of TSH excess increased progressively from 4% at age 18-24 to 20% over age 74 (40).

The National Health and Nutrition Examination Survey (NHANES III) (40a), found hypothyroidism in 4.6% of the US population (0.3% overt and 4.3% subclinical) and hyperthyroidism in 1.3% (0.5% overt), with increasing prevalence with age in both females and males (figure 1). Abnormalities were more common in females than in males. The prevalence of positive thyroid peroxidase was clearly associated with both hyper- and hypothyroidism, but there were important ethnic differences in antibody prevalence.

Figure 1. Percentage of the US population with abnormal serum TSH concentrations as a function of age. The disease-free population excludes those who reported thyroid disease, goiter or thyroid-related medications; the reference population excluded, in addition, those who had positive thyroid autoantibodies, or were taking medications that can influence thyroid function. Note the much higher prevalence of TSH abnormalities in the total population, than in the reference population. (from reference 24a)

The prevalence of thyroid dysfunction is also high when younger women are tested during the post-partum period (41). In an Australian study, 11.5% of women showed at least one laboratory abnormality of thyroid function when tested 6 months after delivery (42). TSH values were increased in 6%, of whom almost 90% showed positive TPOAb, indicating an autoimmune abnormality. In about half the women with initial TSH elevation who had not been treated, TSH was still increased 30 months after delivery (6), consistent with several studies that show an increased prevalence of late hypothyroidism after postpartum dysfunction (41). These findings clearly demonstrate that women with postpartum hypothyroidism should be followed to allow early detection of later overt hypothyroidism (6, 41). It remains uncertain whether long-term treatment is justified on the basis of the initial finding (6,41), but there are clearly important implications of untreated maternal hypothyroidism for fetal development in subsequent pregnancies (22,22a).

Prevalence data from one region do not necessarily apply in other populations, because of differences such as ethnic predispositionor variations in iodine intake. For example, in Hong Kong, where iodine intake is marginally deficient, only 1.2% of Chinese women aged over 60 years had serum TSH values > 5mU/l, with a comparable prevalence of suppressed values indicating possible thyrotoxicosis (43). Several European studies (44, 45) have compared the effect of various levels of iodine intake on the prevalence of thyroid over- and underfunction. Hypothyroidism is generally more common with abundant iodine intake, while goitre and subclinical thyrotoxicosis are more common with low iodine intake (44, 45). These regional differences may influence the choice of diagnostic test and target population. For example, in an iodine replete environment, emphasis could be placed on testing younger or pregnant women for subclinical hypothyroidism by measurement of TSH and peroxidase antibody, whereas in an iodine-deficient region there might be additional emphasis on early detection of thyroid autonomy and thyrotoxicosisin older people, using a highly sensitive TSH assay.

1.4. Adverse consequences of thyroid dysfunction

Before reliable tests of thyroid function became widely available, severe thyrotoxicosis and hypothyroidism were sometimes life-threatening disorders, but thyroid storm or myxoedema coma now have become uncommon, a change that is probably the result of widespread diagnostic testing in clinical practise. Elderly patients with previously unrecognised overt primary hypothyroidism probably have much to gain from a case-finding approach at the time of presentation for medical care. For example, numerous complex management issues may arise when elderly patients with uncorrected severe hypothyroidism are admitted to hospital for emergency procedures (46). Factors that adversely influence outcome include abnormal sensitivity to anaesthetic agents, sedatives, analgesics and narcotics, as well as anemia, hypoventilation, hyponatremia and impaired temperature regulation (46).

In addition to the benefits of early identification of undiagnosed overt TD, widespread diagnostic testing can be justified on the basis of evidence that mild or 'subclinical' dysfunction has adverse consequences. The issues that identify the clinical importance of subclinical TD are summarised in table 2; many of these adverse effects relate to the cardiovascular system, as recently reviewed ( 47, 48).

There is some evidence that mild thyroid abnormalities may influence mortality. A survey of the relationship between serum TSH and all-cause and cardiovascular mortality over a 10 year period in individuals over 60, showed that the group with serum TSH below 0.5 mU/l, had a significantly increased mortality, apparently due to cardiovascular disease (49), although an increased serum TSH was not associated with excess mortality (49). It is not yet known whether suppression of TSH to the levels (<0.03 mU/l) typical of thyrotoxicosis would be associated with a more marked effect on mortality.

1.4.1 Thyrotoxicosis

1.4.1.1 Progression of subclinical to overt thyrotoxicosis

Follow up studies suggest that spontaneous progression to overt thyrotoxicosis is uncommon and that subnormal detectable levels of TSH in the range 0.05-0.4 mU/l frequently return to normal within one year ( 65). By contrast, undetectable TSH values normalize infrequently, but the chance of spontaneous progression to overt thyrotoxicosis appears to be < 10% per year ( 56,65).

1.4.1.2 Cardiac effects

From the Framingham study it was found that undetected subclinical thyrotoxicosis, defined only by suppression of TSH, carried a three-fold increased risk of atrial fibrillation within 10 years ( 51). As yet, there is no study that shows that treatment given on the basis of low TSH alone, modifies this risk, although it is clear that survival is adversely affected by atrial fibrillation ( 66). Other cardiac consequences of subclinical thyrotoxicosis have been documented in patients receiving suppressive
doses of T4 for differentiated thyroid cancer. Prevalence of atrial premature beats is increased, as is left ventricular mass index, associated with higher values for fractional shortening and velocity of shortening ( 52). Impaired left ventricular ejection fraction and reduced exercise capacity have been documented in this group and can be alleviated by beta blockade ( 53).

1.4.1.3 Iodine-induced thyrotoxicosis

Undiagnosed subclinical thyrotoxicosis due to autonomous nodular thyroid disease, a condition especially prevalent in iodine deficient regions, carries the risk of progression to severe overt thyrotoxicosis after iodine exposure ( 67). An Australian study from a region that is not known to be iodine deficient, was suggestive of recent iodine exposure, most often from radiologic contrast agents, in up to 25% of elderly thyrotoxic patients ( 50). Prior knowledge of subnormal serum TSH suggests thyroid autonomy in this high risk group in whom iodine exposure should be avoided because of the risk of severe iatrogenic thyrotoxicosis.

1.4.1.4 Osteoporosis

While longstanding overt thyrotoxicosis is an important risk factor for osteoporosis, the association has until recently been less clear for subclinical thyrotoxicosis. Overt thyrotoxicosis is associated with reduced bone density, predominantly affecting cortical rather than trabecular bone, so that the femoral neck is more affected than the lumbar spine, associated with an increase in fracture rate ( 68). In postmenopausal women, either treated with T4 sufficient to suppress TSH, or with subclinical thyrotoxicosis due to multinodular goitre, there is a decrease in mean bone density and an increase in urinary excretion of pyridinoline cross links ( 69). It has now been shown that women over 65 with suppressed TSH are at increased risk of hip and vertebral fractures ( 55). Notably, in a recent controlled trial of suppressive T4 treatment for multinodular goitre, TSH suppression without clear excess of serum T4 or T3 was shown to result in a mean 3.6% decrease in lumbar spine density within 2 years ( 54).

1.4.2 Hypothyroidism

1.4.2.1 Progression of subclinical (or mild) hypothyroidism to overt deficiency

The benchmark study of thyroid epidemiology from Wickham, UK,(38), showed that the likelihood of overt hypothyroidism after 20 years was directly related to the initial serum TSH, even when that value was in the range 2-4 mU/l, within the upper reference interval. Huber et al (59) followed 82 Swiss women with subclinical hypothyroidism, with normal free T4 and serum TSH > 4 mU/l, for a mean of 9.2 years (figure 2). About half of their cohort had had previous ablative treatment for Graves’ disease. The cumulative incidence of overt hypothyroidism, defined as low free T4 with TSH > 20 mU/l, was directly related to the initial serum TSH, with 55% of women with initial serum TSH >6 mU/l progressing to overt hypothyroidism. Progression was not uniform, and over half of the cohort showed no deterioration of thyroid function. Positive microsomal antibodies increased the likelihood of progression in both studies.

Figure 2 Kaplan Meier estimates of the cumulative incidence of overt hypothyroidism in women with subclinical hypothyroidism (initial serum TSH > 4mU/l) as a function of initial serum TSH, thyroid secretory reserve in response to oral TRH and detectable microsomal antibody. Serum TSH appears to be the strongest of these predictors. (from reference 59)

1.4.2.2 Atherosclerosis

Subclinical hypothyroidism, or mild thyroid failure, was shown to be an independent risk factor for both myocardial infarction and radiologically-visible aortic atherosclerosis in a recent study of Dutch women over 55 years of age (60). This effect was independent of body mass index, total and HDL cholesterol, blood pressure and smoking status. The attributable risk for subclinical hypothyroidism was comparable to that for each of the major risk factors, hypercholesterolemia, hypertension, smoking and diabetes mellitus. The association was slightly stronger when subclinical hypothyroidism was associated with TPOAb, but thyroid autoimmunity itself was not an independent risk factor.

1.4.2.3 Vascular compliance

The finding of impaired flow-mediated, endothelium-dependent vasodilatation even in subjects with borderline hypothyroidism or high-normal serum TSH values is of potential importance (61). Baseline artery diameter and forearm flow were comparable, but flow mediated vasodilatation during the period of reactive hyperemia was significantly impaired even in the group with serum TSH of 2-4 mU/l, compared with the group with serum TSH 0.4-2 mU/l. The difference could not be attributed to a difference in maximal nitrate-induced vasodilatation, age, sex, hypertension, diabetes, smoking, serum cholesterol, or levels of total T3 and T4. This finding suggests that even minor deviation from an individual's pituitary-thyroid set point may be associated with alteration in vasodilatory response. There is no known direct action of TSH that would account for this effect.

1.4.2.4 Cardiac function

From echocardiographic studies, there is evidence that mild thyroid failure can significantly increase systemic vascular resistance and impair cardiac systolic and diastolic function, as demonstrated by decreased flow velocity across the aortic and mitral valves (48). These changes, which were associated with reduced cardiorespiratory work capacity during maximal exercise, were reversed by T4 treatment sufficient to normalize serum TSH (48). Impairment of both diastolic and systolic function was demonstrable by echocardiography in a subclinically hypothyroid group of patients with TSH in the range 4-12 mU/l (62). Thyroxine treatment sufficient to normalise TSH to a mean of 1.3 mU/l. for 6 months was associated with improvement in myocardial contractility (62). It remains to be established how these reversible abnormalities relate to cardiovascular prognosis.

1.4.2.5 Lipids

Overt hypothyroidism is associated with a well known increase in the serum cholesterol concentration and correction of overt hypothyroidism results in a decrease in total and LDL cholesterol, apolipoprotein A1, apo B and apo E, while serum triglyceride concentrations may also decrease (70). A defect in receptor-mediated LDL catabolism, similar to that seen in familial hypercholesterolaemia, has been described in severe overt hypothyroidism (71), but there is no evidence to support such an abnormality in mild thyroid failure.

The Colorado study of over 25,000 subjects showed a continuous graded increase in serum cholesterol over a range of serum TSH values from <0.3 to >60 mU/l (40). However, there is still no consensus that mild thyroid failure has an adverse effect on plasma lipids, or that T4 treatment sufficient to normalize isolated TSH excess has a beneficial effect. A recent meta analysissuggests that T4 treatment of subjects with mild thyroid failure does lower the mean total and LDL cholesterol, and is without effect on HDL cholesterol or triglyceride (63).
In a prospective double-blind, placebo-controlled trial of thyroxine in subclinical hypothyroidism in which the response was carefully monitored with TSH, Meier et al (63a) reported that the decrease in LDL cholesterol was more pronounced with higher initial TSH levels >12 mU/l or with elevated baseline LDL concentration.

It remains uncertain whether the serum concentration of the highly atherogenic Lp(a) particle is increased in overt hypothyroidism and whether T4 treatment sufficient to normalize TSH has a favourable influence. Serum concentrations of Lp(a) have been found to be increased in overt hypothyroidism with normalization after treatment in some studies (72, 73 ) while others fail to confirm this finding (63a,74,75).

1.4.2.6 Neurobehavioural effects

It is well known that overt TD may present with psychological or psychiatric symptomatology. A small retrospective study has shown a 2-3 fold increased frequency of previous depression in subjects with mild thyroid failure (64), while T4 treatment has been reported to improve neuropsychological responses in this group (76). If these findings can be confirmed by studies that include appropriate control groups, it would be justified to conclude that even mild thyroid failure has potentially reversible effects on neuropsychiatric status.

1.5. Effectiveness of laboratory assays for case finding and screening

1.5. TSH reference intervals

If the diagnosis and classification of patients with potential thyroid disease is to based on serum TSH, there should be consensus on the reference intervals for this parameter, but consensus has not yet been achieved, for a variety of epidemiological and semantic reasons. The upper limit of the “normal range” for serum TSH is dependent on the analytical method and on the study population that is selected (77), in particular, whether subjects with positive peroxidase antibody are excluded. It remains uncertain whether apparent ethnic differences reflect the prevalence of mild thyroid dysfunction in those populations (40a). Serum TSH values within the upper normal statistical reference range are associated with an increased chance of long-term hypothyroidism (38), but an arbitrary lowering of the upper limit will decrease the specificity and predictive value of serum TSH as a first-line test.

The sensitivity of current assays identifies a substantial population whose serum TSH values are statistically subnormal, in the range 0.1-0.45 mU/l; these subnormal-detectable values are clearly different from those typical of thyrotoxicosis. There is currently lack of consensus as to how such intermediate TSH values should be classified. The NHANES III study (40a) reserves the designation “subclinical hyperthyroidism” for serum TSH values <0.1 mU/l. By contrast, recent guidelines for the diagnosis and management of subclinical thyroid disease classify values below the lower normal limit of 0.45 mU/l as indicating subclinical hyperthyroidism (24c, 24d), while making more limited recommendations for the management of the intermediate group in the range 0.1-0.45 mU/l. Such a difference in classification will probably affect over 1% of the population. Since the gradation from normality to severe thyroid dysfunction is a continuum, studies of adverse outcomes or benefits from intervention will be critically dependent on precise uniform terminology.  

1.5.2 Choice of initial test

The definitive diagnosis of TD should always be made using the typical relationships between trophic hormone and target gland secretion that define endocrine dysfunction (35, 37). In contrast, case-finding studies generally begin with a single test. Previous discussion about the relative value of TSH and T4 measurements as single initial tests has swung in favour of TSH because of its superior sensitivity in detecting the earliest stages of TD (3). Thus, assessment of untreated subjects now often begins with measurement of TSH alone, with T4 and T3 assays added only if TSH is abnormal, or if an abnormality of TSH secretion is suspected. It is self-evident that serum TSH loses its diagnostic value when pituitary function is abnormal ( 78); serum T4 then becomes the front-line test.

In the absence of associated disease, a normal serum TSH concentration by a so-called third generation assay (a functional lower limit of sensitivity of about 0.01 mU/l), has high negative predictive value in ruling out primary hypothyroidism and thyrotoxicosis. Such immunometric assays, which use two antibodies against different epitopes of the TSH molecule, give a wide separation between the lower limit of the normal reference range at about 0.4 mU/l and the typical TSH values that are found in thyrotoxicosis. It remains uncertain whether values in the subnormal detectable range merit the designation “subclinical thyrotoxicosis”. , . While some subjects with borderline-low TSH values do progress to overt thyrotoxicosis, values in this range may also revert to normal ( 56,65) .

Danese et al (3) analysed the cost-effectiveness of TSH measurement, performed every 5 years from the age of 35 and concluded that for females, such a strategy compared favourably with other preventative medical practices (3). From age 35, the cost per quality-adjusted life year was about US$9000 for females and $22500 for males, but this cost dropped by half if testing was commenced from age 65. Cost effectiveness was highly dependent on the price of the TSH assay and was adversely affected by more frequent testing (3).

There are some clinical situations in which assessment of thyroid function will give a high prevalence of abnormalities that cannot be interpreted with certainty. Notably, glucocorticoids and dopaminergic agents have a potent effect to suppress TSH secretion ( 79), while TSH is also frequently subnormal in starvation or caloric deprivation ( 80). Transient increases to above normal can occur in euthyroid subjects during recovery from critical illness ( 81). The finding that 33% of serum TSH values fell more than 2 SD from the geometric mean in acutely hospitalized patients, while 17 % of values were more than 3 SD from the mean value, suggests that an abnormal TSH value lacks diagnostic specificity in this setting ( 82). A serum free T4 estimate will generally follow from an abnormal TSH value, but during critical illness, free T4 estimates often show non-specific abnormalities ( 83) (see section 4.3, below). This lack of specificity is the basis for a recommendation against routine assessment of serum TSH and free T4 during acute critical illness in the absence of risk factors, or clinical features suggestive of a thyroid disorder ( 84).

Testing of thyroid function is appropriate in a wide range of psychiatric disorders, but diagnostic specificity is limited by a high prevalence of transient non-specific abnormalities at the time of acute psychiatric admission ( 85). Thus, it has been recommended that laboratory evaluation should be delayed for 2-3 weeks after acute presentation, unless there are specific risk factors for TD ( 86).

1.6. Potential adverse effects of testing

In terms of potential for adverse effects, there may be important differences between screening of unselected populations and the case-finding strategy that is now recommended for TD There are no reports of a "labelling effect" (i.e. perception of chronic illness in previously asymptomatic subjects), described in hypertension screening programs ( 87), when testing for TD is done at the time of presentation for medical care. Nevertheless, further attention needs to be given to the potential for unwarranted treatment based on false positive results, as well as the cost of follow-up investigations for perceived abnormalities that may not warrant treatment at any stage. . In particular, there is a need for clinical consensus as to how marginally abnormal results should be classified. Widespread laboratory testing will lead to an increase in the number of false positive results and the potential for a diagnostic method to give misleading variations from normal may not become known for some years until the full diversity of the non-diseased population is documented ( 88).

1.7 Safety and effectiveness of treatment

The benefits of early diagnosis and treatment are self-evident from the obvious decline in hospitalisation and mortality rates for severe TD over past decades. By contrast, the arguments for treating mild TD are less compelling, but the number of affected people is much larger.

The points in favour of treating mild TD relate directly to the adverse consequences listed in table 2, but for most of these potential adverse outcomes there is still a lack of long-term studies that show benefit. On the basis of potential benefit from simple straightforward treatment and freedom from adverse effects, the argument for active treatment is generally stronger for mild thyroid failure than for subclinical thyrotoxicosis. Conservative T4 therapy aimed at normalising TSH is simple, inexpensive and generally safe ( 89), although replacement may not be warranted in the extremely elderly. Where cardiovascular disease precludes full thyroid hormone replacement, detailed evaluation of the cardiac abnormality is appropriate. In contrast, treatment of subclinical thyrotoxicosis needs to be evaluated in relation to drug side-effects and potential for hypothyroidism.

1.8. Follow-up of abnormal results

If serum TSH is used as a single initial test for case-finding, a value outside the reference interval should lead to estimation of serum free T4 on the same sample, if possible without further request or recall of the patient. This requires an algorithm-based testing protocol, which should also include measurement of serum free T3 if TSH is suppressed, in order to identify T3 toxicosis. It may also be relevant to measure TPOAb if TSH is increased to define an autoimmune mechanism of hypothyroidism, particularly as a raised level of this antibody is associated with an increased likelihood of progression to overt hypothyroidism (38
For screening or case-finding to be effective, patients with unsuspected overt TD should be actively traced because they will benefit most from treatment and have the most to lose if the abnormal finding is ignored. For mild TD, a practitioner who has continuing contact with the patient should correlate the assay results with the clinical presentation and initiate the necessary follow-up. However, there is currently no consensus as to how an appropriate clinical response to abnormal laboratory findings can be assured

A transition towards identification of TD by laboratory measurement, rather than on clinical criteria ,modifies, but in no way diminishes the clinician's role. An abnormal laboratory finding must be placed in context, and an assessment needs to be made of the cause of the abnormality and its severity., The severity of thyroid dysfunction cannot be judged from the extent of the laboratory abnormality (89a), which indicates neither the duration of exposure, nor individual susceptibility. A result that may lead to lifelong replacement therapy should be discussed in detail with the patient. Whether a decision is made to treat or to observe, patient education is crucial in establishing effective compliance and rational cost-effective follow-up. Computer based programs can identify affected individuals, but these programs cannot replace direct involvement of a clinician. There may be potential medicolegal consequences of failure to respond to abnormal results, if widespread laboratory testing is initiated without an established follow-up plan.