The thyroid sets the body's metabolic pace. Too little hormone and everything slows — fatigue, cold intolerance, weight gain, constipation, slowed thought. Too much and everything speeds up — palpitations, heat intolerance, weight loss, tremor, anxiety. Both are common, both fall disproportionately on women, both are mostly autoimmune, and both are among the most satisfying conditions in medicine to treat, because the abnormality is measurable and the correction is precise.
In the United States, NHANES III measured thyroid function in 17,353 people and found hypothyroidism in 4.6 percent of the population — 0.3 percent overt and 4.3 percent subclinical [1] [2]. That ratio, fourteen mild cases for every obvious one, is the shape of the whole subject: most thyroid disease is found by a blood test in someone who has not clearly complained of anything.
Which raises the question this review is built around: why is a single hormone measurement — TSH — enough to screen an entire endocrine axis? The answer is a piece of control-systems biology worth understanding, and it is the centerpiece.
Start here: the loop
The hypothalamus releases TRH, which drives the pituitary to release TSH (thyroid-stimulating hormone), which drives the thyroid to make thyroxine (T4) and the more active triiodothyronine (T3). T4 is largely a prohormone, converted to T3 in peripheral tissues, and T3 acts on nuclear receptors to regulate transcription across essentially every tissue [3] [4].
The loop closes because the pituitary senses circulating thyroid hormone and turns TSH down when it is plentiful. The TSH receptor on thyroid cells, cloned in 1989 [5], is the point at which the signal is received — and, as it turns out, the autoantigen at the centre of Graves' disease [6].
The effects of getting this wrong are systemic. Thyroid hormone acts directly on the heart, altering rate, contractility and vascular resistance [7] [8]; on bone, where excess accelerates loss [9] [10] [11]; and on mood and cognition [12].
The two common diseases are autoimmune.
Hashimoto's thyroiditis is lymphocytic destruction of the gland producing hypothyroidism, typically with antibodies to thyroid peroxidase [13] [14] [15].
Graves' disease is the mirror image and mechanistically stranger: autoantibodies that stimulate the TSH receptor, driving the thyroid to overproduce hormone with no regard for the feedback loop [16] [6]. It is unusual among autoimmune diseases in that the antibody is an agonist rather than a destroyer.
Both share genetic susceptibility with the wider autoimmune family [17]. Thyroiditis also arises from viruses [18], from amiodarone [19], and increasingly from cancer immunotherapy, where checkpoint inhibitors precipitate it in a substantial minority [20].
Pillar 1: measurement and diagnosis
Why TSH is the first test
Because the pituitary is an amplifier. That is not a metaphor — it is a measured relationship, and the centerpiece quantifies it.
The practical consequence: TSH moves out of its reference range before free T4 moves out of its own. A high TSH with a normal free T4 is the earliest biochemical signature of a failing thyroid, and a suppressed TSH with normal hormone levels the earliest signature of an overactive one. Those two states are what "subclinical" means.
Free T4, and when T3 matters
Free T4 confirms and grades what TSH suggests, and separates overt from subclinical disease. T3 is measured chiefly in suspected hyperthyroidism, where it can be raised while T4 is not.
There is one important situation where TSH alone misleads: central hypothyroidism, in which the pituitary itself is the problem. TSH is then inappropriately normal or low despite low free T4, and the whole logic inverts — which is why pituitary disease requires its own diagnostic approach [21].
Antibodies
Anti-thyroid peroxidase (anti-TPO) antibodies mark autoimmune thyroiditis and predict progression from subclinical to overt hypothyroidism; they were measured across the US population alongside TSH in NHANES [1]. TSH receptor antibodies (TRAb) establish Graves' disease as the cause of hyperthyroidism, and independently predict eye involvement [22].
Imaging
Ultrasound assesses gland structure and nodules. In hyperthyroidism, radioiodine uptake scanning distinguishes causes that look identical on blood tests — a diffusely overactive gland in Graves', a single hot nodule, or a thyroiditis releasing preformed hormone with low uptake, which matters because the last needs no antithyroid drug at all [23].
Centerpiece: a simple simulatable model of the feedback loop
Two measured quantities, from two different studies, explain both why TSH is such a good screening test and why its reference range is a blunt instrument for any individual person.
The amplifier. A study of 9,519 thyroid function tests in 4,064 patients estimated the feedback slope directly, using what it calls "the physiological log-linear relationship between fT4 and TSH": feedback inhibition causes a 0.1345 decrease in log TSH for each 1 pmol/L increase in free T4 [24].
Log-linear means TSH responds exponentially to a linear change in hormone. A fall in free T4 of d pmol/L multiplies TSH by 10^(0.1345·d). Across a 10 pmol/L span of free T4 — a modest move, comfortably inside the width of a typical reference interval — TSH changes about twenty-two-fold. That is the amplification, computed from the measured slope, and it is the entire reason one test screens the axis.
The catch. A longitudinal study sampled 16 healthy men monthly for a year and found that each had a narrow personal set point: individual 95 percent confidence intervals were about half the width of the group's, with an index of individuality of 0.49 for TSH and 0.54 to 0.59 for the hormones [25]. Those two statements are the same finding in different currencies, and the script asserts they agree — a check that the model is reading the source correctly.
A low index of individuality means the population reference range is insensitive to changes that matter for a person. The authors state the consequence plainly: "a test result within laboratory reference limits is not necessarily normal for an individual." Independently, the feedback study found that 21.9 percent of all possible fT4–TSH combinations lying within the current reference ranges were nonetheless abnormal when judged against the physiological relationship [24].
That is the teaching point, and it has two halves that pull in opposite directions. The loop's amplification makes TSH an excellent population screen — it is the reason a single cheap test catches thyroid disease early. The same loop, combined with individual set points, makes a TSH inside the reference range weak evidence that a particular person is euthyroid. Andersen and colleagues drew the sharper conclusion: because the diagnosis of subclinical disease depends on where a patient's own set point sits within the laboratory range, the line between subclinical and overt disease is to a considerable extent arbitrary [25]. Broader analyses of variation within the reference range reach compatible conclusions [26].
Three honest limits. The line's intercept and the reference intervals drawn are conventional display choices, not measurements — assay-specific ranges vary, and only the slope and the indices of individuality are measured quantities. The log-linear form is an approximation that flattens at the extremes. And the individuality study followed 16 healthy men for a year, which is a small and unrepresentative sample for a claim of this reach.
Pillar 2: treatment
Hypothyroidism: replace the hormone, titrate to the loop
Levothyroxine is synthetic T4, taken once daily. The dose is adjusted to bring TSH into range — which is an unusually clean piece of clinical feedback control, since the body's own sensor reports whether the replacement is right.
Two practicalities matter more than they sound. Absorption is easily disrupted — by food, by other medications, and by gastrointestinal conditions including coeliac disease and Helicobacter pylori gastritis, all of which raise the required dose [27]. And the loop is slow: TSH takes about six weeks to re-equilibrate after a dose change, so testing sooner produces misleading numbers.
The T4-versus-T4/T3 debate persists. Levothyroxine alone relies on peripheral conversion of T4 to T3, and a minority of adequately replaced patients report continuing symptoms. Whether adding T3 helps them is unresolved.
Hyperthyroidism: three definitive options, no obviously best one
Antithyroid drugs block hormone synthesis. Methimazole and carbimazole are first-line; a randomised comparison found methimazole 30 mg daily normalised free T4 in 96.5 percent of patients at 12 weeks against 78.3 percent for propylthiouracil 300 mg daily and 86.2 percent for methimazole 15 mg [28]. Propylthiouracil is reserved for specific situations — chiefly the first trimester of pregnancy — because of rare severe hepatotoxicity [23]. Agranulocytosis is the rare shared hazard, and every patient must be told to seek help for fever or sore throat.
Radioactive iodine is taken up selectively by thyroid tissue and destroys it. It is effective and well characterised over decades of follow-up [29] [30]. Its predictable consequence is hypothyroidism — which is a conversion of one disease into another, easier one, and is best framed that way rather than as a failure.
Surgery — total or near-total thyroidectomy — is definitive and fast, at the cost of an operation with specific risks to the recurrent laryngeal nerves and parathyroid glands. Meta-analysis supports its efficacy in Graves' disease [31], and outcomes depend substantially on surgeon volume [32].
Graves' orbitopathy, which follows its own rules
Eye disease affects a substantial minority and is not simply a consequence of being hyperthyroid — it can precede, follow or run independently of the thyroid problem, driven by autoantibody-mediated inflammation of orbital tissues [33] [34] [35]. TSH receptor antibodies independently predict it [22].
Two management points are firmly grounded. Radioactive iodine can worsen orbitopathy, a risk demonstrated in a randomised comparison of treatments for hyperthyroidism, and corticosteroid cover mitigates it [36]. And selenium supplementation improved the course of mild orbitopathy in a randomised trial [37]. Severe disease has been treated with glucocorticoids, historically with additional immunosuppression [38], and European guidelines set out a staged approach [39].
Subclinical disease, where the honest answer is "often, don't"
This is where the review has to be careful, because the temptation to treat a number is strong and the evidence does not support it.
A systematic consensus review of 195 papers reached conclusions worth quoting for their restraint. Evidence that subclinical disease progresses was rated good; evidence that treatment prevents progression was inadequate. Evidence linking TSH above 10 mIU/L to raised cholesterol was fair; evidence of treatment benefit insufficient. Evidence linking TSH below 0.1 mIU/L to atrial fibrillation was good, but no data supported treating to prevent it. Its recommendation for the intermediate zones — TSH 0.1 to 0.45 or 4.5 to 10.0 mIU/L — was against routine treatment, and against population screening, with aggressive case finding reserved for pregnant women, women over 60, and others at high risk [40].
The associations themselves are real. Low TSH predicts atrial fibrillation in older people [41], subclinical hyperthyroidism is associated with coronary heart disease and mortality [42] [43], and subclinical hypothyroidism with stroke events in individual-participant analysis [44] and with atherosclerosis and myocardial infarction in a population cohort [45]. What is missing is not the association but the demonstration that treating it helps — an early trial of thyroxine in subclinical hypothyroidism [46] and subsequent European guidance [47] have narrowed rather than settled the question [48] [49].
Pregnancy is the exception, and here the case for acting is much stronger. Maternal thyroid deficiency during pregnancy was associated with impaired neuropsychological development in the child [50], subclinical hypothyroidism with adverse pregnancy outcomes [51], and thyroid autoantibodies with miscarriage and preterm birth [52]. Dedicated guidelines govern this setting [53] [54].
Pillar 3: what is unresolved
Whose TSH target, and what number. The individual set point finding [25] implies that a population reference range is the wrong tool for deciding whether a treated patient is adequately replaced, and that the same TSH means different things in different people. Nobody has an operational replacement for it.
Whether to treat mild disease. The evidence gap identified two decades ago [40] has narrowed slowly. Given that the mild category outnumbers the overt one roughly fourteen to one [1], the aggregate stakes are large even though the individual stakes are small.
The residual-symptom problem. A minority of patients on levothyroxine with normal TSH continue to feel unwell. Whether this reflects inadequate tissue T3, a set point mismatch, or unrelated illness is unresolved, and it is the most common source of dissatisfaction in an otherwise highly treatable disease.
Better orbitopathy therapy. The mechanistic understanding of orbital autoimmunity has advanced considerably [34] [33]; targeted biologic treatment for the eye disease is the most active area of thyroid therapeutics, and is more recent than this substrate reaches.
Dig deeper in lmmol
Hashimoto's and Graves' belong with the other organ-specific autoimmune diseases in this collection, and cluster with them genetically [17] — see rheumatoid arthritis, inflammatory bowel disease and multiple sclerosis. The contrast is instructive: those three need increasingly potent immunosuppression, while thyroid autoimmunity is usually managed by ignoring the immune process entirely and simply correcting the hormone — because the gland's output can be replaced or ablated in a way a joint, a bowel or a myelin sheath cannot. Thyroid hormone excess accelerates bone loss, which connects to osteoporosis [9] [11]; low TSH predicts atrial fibrillation and thyroid status tracks cardiovascular risk in older adults, connecting to heart failure [41] [43]; hypothyroidism is a standard reversible contributor to the picture assessed in depression [12]; and thyroid function interacts with metabolic liver disease, linking to fatty liver disease [55]. The full collection is at health.