3.3.4 Subclinical hyperthyroidism

Subclinical hyperthyroidism is defined biochemically as the association of a low serum thyroid-stimulating hormone (TSH) value with normal circulating concentrations of free thyroxine (T₄) and free triiodothyronine (T₃). The biochemical diagnosis of subclinical hyperthyroidism is dependent upon the use of sensitive assays for TSH able to distinguish normal values found in euthyroid people from reduced values, so our understanding of this topic has accumulated in recent years since such assays became widely available. An expert panel has recently classified patients with subclinical hyperthyroidism into two groups (1): (1) those with low but detectable serum TSH (0.1–0.4 mU/l) and (2) those with undetectable serum TSH (<0.1 mU/l) reflecting the fact that studies of this condition largely divide people into these categories and that the likely consequences reflect the biochemical severity of the condition.

The biochemical finding of low serum TSH in association with normal serum thyroid hormone concentrations may reflect an underlying thyroid disorder (Box 3.3.4.1) but low serum TSH often reflects other nonthyroidal illnesses or their treatment. Furthermore, in those with underlying thyroid disease, the cause of TSH suppression may be exogenous, i.e. reflecting thyroid hormone therapy, or endogenous, i.e. reflecting a degree of autonomous thyroid function.

People who have been treated for Graves’ hyperthyroidism with antithyroid drugs, partial thyroidectomy, or radio-iodine may have suppression of serum TSH concentrations for months or occasionally years after restoration of a clinically euthyroid state and return of serum T₄ and T₃ concentrations to the reference range. In those who have received drug therapy alone for Graves’ hyperthyroidism, suppression of TSH may serve as an early marker for relapse. The finding of persistent suppression of TSH may reflect the long period of recovery of pituitary thyrotrophs after removal of thyroid hormone excess, or may reflect a degree of persistent thyroid autonomy since suppression of serum TSH may be more common in those with persistent thyroid-stimulating autoantibodies and such individuals have higher circulating thyroid hormone concentrations, albeit within the reference range. Suppression of serum TSH is also common in those with Graves’ ophthalmopathy but absent clinical and biochemical features of overt hyperthyroidism. In individuals with symptoms or signs suggestive of thyroid eye disease, the presence of suppression of serum TSH, together with the presence of thyroid autoantibodies, lends supporting evidence to the diagnosis of Graves’ ophthalmopathy. A further group with a high prevalence of subclinical hyperthyroidism is that with thyroid enlargement. Up to 75% of patients with a nodular goitre have suppression of TSH (with normal serum T₄ and T₃) reflecting autonomous function of one or more thyroid nodules. Imaging studies, typically with ^99mTc, reveal the presence of one or more ‘hot’ nodules with suppression of uptake of isotope into surrounding areas of the thyroid.

The most common group with circulating TSH concentrations below the normal range is that prescribed thyroxine replacement therapy. One study of patients treated in primary care for hypothyroidism revealed reduction in TSH in 25%, this finding being most common in those prescribed thyroxine in doses of 150 μg/day or more (2). Similar findings have been reported in a large population-based study in the USA (3).

Nonthyroidal illness and therapy with a variety of drugs represent the other major associations with a biochemical diagnosis of subclinical hyperthyroidism. Illness itself may be associated with suppression of TSH through ill-defined effects upon the hypothalamic–pituitary axis, the inflammatory cytokine interleukin-6 being one factor implicated in the pathogenesis of the changes in tests of thyroid function found in hospital inpatients. A variety of pharmacological agents, particularly glucocorticoids and dopamine, are associated with low circulating concentrations of TSH, probably through direct inhibitory effects of these drugs on hypothalamic secretion of thyrotropin-releasing hormone and/or pituitary secretion of TSH. Iodine-containing compounds, most notably the antiarrhythmic drug amiodarone, may cause TSH suppression, as may anticonvulsants such as phenytoin. Suppression of TSH is also common in the first trimester of pregnancy, probably due to a rise in circulating human chorionic gonadotropin (which itself has a thyroid-stimulating effect). In nonthyroidal illness, the serum TSH value is typically low but detectable.

There have been several large population-based studies of the prevalence of subclinical hyperthyroidism and results vary depending upon the demographic features of the group examined, as well as the inclusion or exclusion of those with known thyroid disease or taking thyroid hormone therapy and the assay for TSH employed (see Chapter 3.1.7). The large NHANES III study in the USA (4) showed that endogenous subclinical hyperthyroidism is more common in women, in older people, and in black people compared with white people, with an overall prevalence in adults of 3.2% (cut-off serum TSH <0.4 mU/l) or 0.7% with undetectable TSH (TSH <0.1 mU/l). We recently conducted a prevalence study of almost 6000 community-dwelling people in the UK without known thyroid disease and aged more than 65 years. We found subclinical hyperthyroidism with low but detectable serum TSH in 2.1% and undetectable TSH in 0.7% (5). Exogenous subclinical hyperthyroidism is found in 10–30% of those prescribed thyroid hormones. In the Colorado study of more than 25 000 people, a serum TSH of <0.3 mU/l was found in 0.9% of those not taking thyroid hormones but 20.7% of those taking such medications (3).

The natural history of subclinical hyperthyroidism depends upon its cause and severity (in terms of the degree of reduction of serum TSH below the reference range). Some patients with TSH suppression associated with Graves’ hyperthyroidism or nodular goitre will progress to overt hyperthyroidism, although the incidence is relatively low at around 1–3% per year. Most of these patients with an underlying thyroid disorder demonstrate complete suppression of serum TSH concentrations to below the limit of assay sensitivity (rather than TSH values below normal but still detectable) and when compared as a group with normal controls display higher mean circulating concentrations of T₄ and T₃, consistent with a minor degree of thyroid hormone excess. In contrast, those in whom low serum TSH values reflect nonthyroidal illness or drug therapy often have low but detectable serum TSH concentrations, as well as serum T₃ (and less frequently T₄ values) below the reference range, and in these patients the biochemical abnormality often disappears after recovery from illness or cessation of drug therapy. A recent large study demonstrated that serum TSH below 0.35 mU/l returns to normal in more than one-half of patients after a follow-up period of 5 years (6). We screened an ambulatory population of people aged over 60 years recruited in primary care and found that TSH values below normal were present in 6.3% of women and 5.5% of men (undetectable TSH in 1.5% of women and 1.4% of men). The follow-up of those patients with low serum TSH showed that of those with low but detectable TSH at initial testing, TSH had returned to normal in 76% of them at 1 year, compared with the group with undetectable TSH in whom 88% still had undetectable TSH at 1 year (7). A 10-year follow-up of the same group showed that only 4.3% of those with low serum TSH developed overt hyperthyroidism (8).

The often transient nature of the biochemical abnormality in those in whom reduction in TSH is associated with illness or drug therapy suggests that in this group the diagnosis of subclinical hyperthyroidism is of little consequence in terms of long-term effects. The potential consequences of subclinical hyperthyroidism are therefore probably confined to those in whom suppression of TSH reflects a minor degree of thyroid hormone excess, the major patient groups being those with thyroid autonomy due to the presence of a nodular goitre or Graves’ disease and those receiving thyroxine replacement therapy. Unsurprisingly, any association with specific symptoms and signs is weak, although palpitation may be more frequent. We have reported a lack of association with cognitive dysfunction or symptoms of depression (9). Research has largely focused upon the effects of subclinical hyperthyroidism on bone metabolism and upon the heart, findings highlighting possible adverse effects upon these tissues and leading to debate about treatment of this biochemically defined condition.

Overt hyperthyroidism is associated with an increase in bone turnover and a net loss of bone, while effective treatment is associated with restoration of bone metabolism to normal and an increase in bone mineral density (BMD). The increasing recognition of subclinical hyperthyroidism as a frequent biochemical finding has prompted many studies of the effect of more minor degrees of thyroid hormone excess upon bone metabolism and hence upon risk of osteoporotic fracture. The results of these studies have proved conflicting and their interpretation difficult because of the relatively small numbers of patients investigated, their heterogeneous nature (including those with previous goitre or hyperthyroidism), and relatively poor matching with controls in terms of factors which may modify BMD. Our own early studies of patients with thyroid disease and carefully matched controls suggested that subclinical hyperthyroidism secondary to T₄ therapy (even when administered in TSH-suppressive doses long-term) is not associated with significant reductions in BMD, in contrast to the situation in those with a past history of overt hyperthyroidism (especially postmenopausal women) in whom small reductions in BMD compared with controls are evident (regardless of the need for subsequent T₄ replacement therapy) (10, 11). Two meta-analyses of cross-sectional studies of BMD in those with subclinical hyperthyroidism secondary to long-term therapy with thyroxine have demonstrated small but significant effects upon BMD of T₄ given in doses associated with reduction in TSH, but only in postmenopausal women (12, 13).

Evidence that such changes in BMD are translated into an increase in risk of osteoporotic fracture, especially fracture of the femur, is so far relatively poor. T₄ usage was not associated with fracture in a case–control study of patients admitted to hospital with hip fracture (14). An important population-based study has examined prospectively risk factors for the later incidence of fracture of the femur in postmenopausal women. Thyroxine prescription was identified as a risk factor for fracture (RR 1.6; 95% CI 1.1 to 2.3) but this relative risk was no longer significant when a previous history of overt hyperthyroidism was taken into account, previous hyperthyroidism itself being associated with a relative risk of 1.8 (95% CI 1.2 to 2.6) (15). Even fewer studies have evaluated endogenous subclinical hyperthyroidism and fracture risk. In a study of 686 women aged over 65 years with low serum TSH levels, a three- to fourfold increased risk of hip or vertebral fracture was found after adjustment for previous hyperthyroidism and thyroxine prescription when compared with individuals with normal TSH levels (16).

While more large-scale prospective studies of the effects of T₄ treatment upon BMD and upon fracture risk are required to clarify the situation, it seems likely that subclinical hyperthyroidism secondary to mild thyroid hormone excess does result in a minor increase in bone loss and fracture risk. This risk is probably clinically significant only in postmenopausal women, reflecting an associated deleterious effect of oestrogen deficiency. Preliminary evidence suggests that adverse effects of thyroid hormone excess upon BMD can be reversed in this group by oestrogen replacement therapy, and perhaps by other agents such as bisphosphonates, but the role of such interventions in patients with subclinical hyperthyroidism has yet to be defined. One small study of postmenopausal women treated with methimazole to normalize serum TSH levels found that forearm BMD was increased after 2 years of treatment; however, another small prospective study in premenopausal women found no difference.

Subclinical hyperthyroidism has clear effects upon the cardiovascular system similar to overt hyperthyroidism, although these are less marked. These effects include an increase in nocturnal heart rate, shortening of the systolic time interval (a marker of left systolic ventricular function), and an increase in frequency of atrial premature beats. Left ventricular mass is typically increased. While the significance of these findings is unclear, more convincing evidence of harm has accrued from epidemiological studies examining atrial fibrillation and death from vascular diseases. In the Framingham cohort, a 3.1-fold increased relative risk for atrial fibrillation was evident after 10 years in elderly individuals (>60 years) with TSH levels of not more than 0.1 mU/l, whereas a relative risk of 1.6 was observed in those with low but detectable TSH (0.1–0.4 mU/l) (17). A large retrospective study demonstrated an adjusted relative risk of 2.8 for the finding of atrial fibrillation in individuals with low TSH levels when compared with those with normal TSH (18). The association between endogenous subclinical hyperthyroidism and atrial fibrillation is further supported by an important study showing an increased incidence of atrial fibrillation (approximately twofold) in elderly individuals followed for 13 years, which included evidence for an increased incidence in those with low, but detectable, TSH values (19). Furthermore, we performed a cross-sectional study of 5860 people aged 65 years and over and observed a higher prevalence of atrial fibrillation in participants with subclinical hyperthyroidism than in those with normal serum TSH (20). We also found that serum free T₄ was independently associated with the finding of atrial fibrillation by electrocardiogram, even in euthyroid individuals with normal free T₄ and TSH values (Fig. 3.3.4.1).

There is also evidence linking subclinical hyperthyroidism and mortality. We reported increased deaths from circulatory diseases (both cardiovascular and cerebrovascular) in association with low TSH in a 10-year study of 1191 individuals aged more than 60 years (8) (Fig. 3.3.4.2). This increased mortality occurred in the absence of increased deaths from other common causes. In very elderly people (>85 years), a Dutch study reported increased cardiovascular mortality during a 4-year follow-up in those with low levels (21). By contrast, another study of elderly people did not find increased mortality after adjustment for age and sex, despite positive findings for atrial fibrillation (19). Taken together, the evidence for an association of subclinical hyperthyroidism with mortality is less strong than for atrial fibrillation.

Again, preliminary evidence suggests that antithyroid treatment in those with endogenous subclinical hyperthyroidism may reverse adverse effects. A study of 10 patients examined the effect of 6 months of treatment with methimazole (median dose 20 mg daily) to restore TSH to normal and reported reduced heart rate and ectopic beats, and restoration of left ventricular mass index to that of euthyroid individuals (22). Another study of six women with subclinical hyperthyroidism and nodular goitre treated with radio-iodine to normalize serum TSH resulted in reduced heart rate and cardiac output. Unfortunately, no long-term studies have examined the effect of antithyroid drugs or radio-iodine treatment on the risk of atrial fibrillation or other clinically relevant endpoints such as mortality.

In patients taking thyroxine therapy, intervention is relatively simple, i.e. dose reduction followed by further biochemical testing to ensure that TSH has returned to the reference range. Whether this is a cost-effective exercise given the marked prevalence of T₄ prescription in the general population and so far inconclusive evidence for harm in the long term (with the possible exception of occurrence of atrial fibrillation) remains unclear. In patients starting T₄ therapy it appears appropriate to aim for biochemical as well as clinical euthyroidism, in line with both UK and US guidelines.

The question of intervention is even more complicated in those in whom suppression of TSH reflects autonomous thyroid function. Given the lack of substantive evidence that treatment of endogenous subclinical hyperthyroidism is beneficial, it is perhaps surprising that surveys of thyroid specialists in the USA and UK indicate that treatment is now regularly undertaken, typically with radio-iodine. Some clinicians argue that the evidence for harm associated with subclinical hyperthyroidism is sufficient to drive the need for treatment. As radio-iodine can induce hypothyroidism (with potential associated adverse outcomes) it would be preferable to await results of cost–benefit analyses before treatment is routinely offered. Meanwhile, treatment decisions might be tailored according to biochemistry or clinical factors. In light of the limited evidence linking low but detectable TSH concentrations with adverse outcomes, a consensus statement suggested that treatment should be considered only for patients with undetectable TSH (1). More contentious is the question of whether treatment should be confined to older people or those with atrial fibrillation, given the associations of subclinical hyperthyroidism with poor outcomes in these groups. This approach might be correct but it could be argued that young patients might eventually accrue the greatest benefit in terms of atrial fibrillation and fracture prevention and long-term survival. Further evidence is required to determine which is the correct approach.