Thyroid Markers

Thyroid Stimulating Hormone

TSH

A pituitary hormone that tells the thyroid how much hormone to make. It moves in the opposite direction to thyroid output, so when thyroid hormone falls, TSH rises to push harder. Why that inversion happens, and why the axis amplifies small changes, is in Thyroid, in the Body Guide.

If you are young, the range printed next to your result is probably too wide at the top. Reference ranges are built from the whole adult population, and TSH climbs across life. A range built from people aged 18 to 90 therefore sets its upper limit high enough to cover an 80-year-old. In people screened for thyroid disease and found not to have it, the upper limit at age 20 to 29 sits at about 3.6, and by 80 and over it reaches about 7.5. A survey of 111 NHS labs found that among those running the most commonly used analyser the reported upper limits spanned 4.0 to 6.0, so the spread turns up between labs using the same method. The figure your lab prints changes who gets treated. Across those labs, every 1 unit higher on the reported upper limit went with about 0.5% fewer people on levothyroxine.

One caution before you do arithmetic with any of that. TSH is not a standardised measurement. Assays are traceable to an international reference material, but that material does not behave the same way as real patient samples on every platform, so an age-banded figure derived on one analyser is not a number you can hold your own result against. What the age data establishes is the direction and the rough size of the drift, not a threshold you can apply to yourself.

Three things move this number before it reaches the lab

The clock, and not the meal. TSH peaks overnight and is highest first thing, then falls by roughly 0.5 by mid-morning. That matters more than it sounds. In one small paired study, 15 of the 20 people who met the criteria for subclinical hypothyroidism on a fasting sample were back inside the range 2 hours after eating. The same group then ran the obvious follow-up, sampling on one day after breakfast and on another while still fasting, and the fall was the same either way. So the effect people attribute to eating is the time of day. Neither study was randomised and both were small. The practical point holds anyway. Sample early, sample consistently, and do not compare an 8am result with an 11am one. The lower end of the range is stable through the day and across life. It is the top that moves.

Energy deficit. Dieting hard, training through a deficit or being unwell shifts the whole panel as an adaptation and not as disease. TSH drifts down, T3 falls, reverse T3 rises. The Free T3 entry sets that pattern out, because it is the one most often mistaken for thyroid disease in people who train.

Body weight in the other direction. Carrying more weight raises TSH without anything being wrong with the gland, and it comes back down as weight is lost. A European thyroid guideline published in 2025 says this outright and warns that a TSH-based label of subclinical hypothyroidism should be treated carefully in anyone with obesity. A large share of people who buy a thyroid panel have a raised body mass index, so this is one of the most common reasons a TSH comes back mildly high in someone with a perfectly normal thyroid.

Biotin. TSH is one of the markers biotin affects, and on this test it reads falsely low while free T4 and free T3 read falsely high and the antibodies can be dragged up too. That combination looks like an overactive thyroid. People have been started on antithyroid drugs on the strength of it, and newborns treated for Graves disease they did not have. Biotin, at the top of this guide, has the detail. If you take levothyroxine, take the dose after the draw and not before.

Where the numbers come from

NICE commits to one figure. Levothyroxine is considered where TSH reaches 10 mIU/L on 2 separate occasions 3 months apart, and a 6-month trial may be considered below that in symptomatic adults under 65 whose TSH is above the reference range. For the lower boundary it says "above the reference range" instead of naming a number, because that figure belongs to whichever analyser ran the sample.

That guidance was published in 2019, updated in 2023 on biotin, and reviewed again in October 2025. The review was triggered by the falling cost of liothyronine and nothing else, it concluded the recommendations would not change, and it did not re-examine TSH thresholds, subclinical hypothyroidism or antibody testing at all. So the thresholds stand, and the evidence behind them is 7 years old.

How old the guidance actually is

The background to everything on this page is that the guidance UK labs and GPs work from is not new, and some of it is very old.

The UK Guidelines for the Use of Thyroid Function Tests were published in July 2006 and have never been replaced. They are 20 years old, and NHS lab handbooks still cite them as the reference for how these tests should be used and reported. NICE's thyroid guideline is from 2019, updated in October 2023 to add a recommendation on biotin supplements affecting test results, and its October 2025 review looked only at the cost of liothyronine. The American hypothyroidism treatment guideline is from 2014 and the hyperthyroidism one from 2016. The European guideline on subclinical hypothyroidism is from 2013. The joint statement people quote as having opened the door to combination treatment is from 2021, and it is a specification for a trial that has still not reported.

Two current documents do exist. A European guideline on levothyroxine treatment published in July 2025 is the most recent thing in this field, and its strongest recommendation, carrying the highest grade of evidence in the whole document, is that once TSH is inside the reference range, elaborate dosing regimens and small adjustments made to chase better wellbeing or a lower body weight are not useful and are not advised. It also says to use a more relaxed TSH range in the over-70s, to monitor pituitary-caused hypothyroidism with free T4 and not TSH, to dose against lean body mass, to retest before changing a dose when a previously stable result looks odd, to stay on the same brand and recheck 6 weeks after any switch, and to treat a requirement above 1.5 to 1.7 micrograms per kilogram a day in someone with a thyroid still in place as a reason to look at adherence and absorption before going higher. A second European guideline published in 2026 covers interference in thyroid blood tests specifically, and it comes with a plain-English summary written for patients rather than clinicians.

None of that makes the older guidance wrong. It means a guideline is a snapshot of what was settled when its evidence review closed, that several of those reviews closed a decade or more ago, and that the studies have kept arriving since. Where this guide and a guideline disagree, the disagreement is named and both sides are given.

There is a gap behind that. The large placebo-controlled trials finding no symptom benefit from levothyroxine in mild subclinical hypothyroidism were run in people over 65, and one of them in people over 80. Smaller trials in younger and mixed-age adults exist and none was big enough to answer the question. No adequately powered trial has been done in symptomatic adults under 65, nothing of that kind is registered, and the cautious position applied to younger people is extrapolated down from a different population. That cuts both ways.

What a borderline result calls for

Thyroid antibodies and a repeat. A mildly raised TSH on its own is a number. The same number with antibodies present is a direction of travel, and long-term follow-up shows the combination predicts future hypothyroidism far more strongly than either alone.

The way those figures are usually quoted is misleading, so here they are properly. In the long-running British community study, a raised TSH alone raised the odds of being hypothyroid 20 years later about 8-fold in women and 44-fold in men, antibodies alone about 8-fold in women and 25-fold in men, and the 2 together 38-fold in women and 173-fold in men. Read quickly, that says men are at far greater risk. They are not. Hypothyroidism developed in about 3.5 women per 1,000 each year and about 0.6 men per 1,000, so the male multipliers are large because the male starting point is small. The male estimates also rest on very few cases and the ranges around them are wide.

NICE says not to repeat antibody testing, and for most people the answer does not change. A negative today is not a negative for life though. People do develop them later, and the postpartum year is the most common time for it.

One thing antibodies do not predict is heart disease. A pooled analysis of 14 population cohorts covering 100,250 adults found no association between being TPO antibody positive and coronary events, coronary death, stroke or stroke death, and the same absence held within people who had subclinical hypothyroidism. Antibodies say something about the future of the gland. They say nothing about your arteries, and anyone selling them as a cardiovascular marker is going beyond the evidence.

The drift with age is real, and smaller than it is usually made to sound

TSH rises across life in people who never develop thyroid disease. The obvious reading is accumulating damage, and that appears to be wrong. Following antibody-negative people over more than a decade, TSH rose while free T4 did not move, and the largest rises were in those who started lowest, which is the opposite of creeping thyroid failure. Some of that last part is a statistical artefact and not a discovery, because whenever anything is measured twice the low values tend to come back higher.

The largest pooled analysis of thyroid function with age, covering 137,488 people across 31 cohorts, changes the picture in 3 ways. The rise is small, adding up to about 0.6 across the whole span from 18 to 100 in women and about 1.0 in men. It depends on iodine supply, with older age going with higher TSH in iodine-sufficient populations and lower TSH where iodine is short. And free T4 does move over time after all, upward where iodine is sufficient and downward where it is not. Most people were stable.

The longevity findings are real but thinner than they are usually presented. People living into their late 90s carry higher TSH than younger comparison groups, with a genetic component behind it that shows up in their children as well. In people in their late 80s, a higher TSH went with lower mortality over the following few years. Against that, the pooled work above found that any changing pattern of thyroid function, in either direction, went with higher mortality than a stable one. Those measure different things, levels against trajectories, so they do not settle each other. What can be said is that the set point appears to move with age instead of the gland simply failing, and that a higher TSH in an older person is not automatically a problem.

When the number is simply wrong

Biotin is the common one and it is covered above. Three others are rare, and all of them produce a raised TSH in someone who feels fine, which is the exact result consumer testing generates most often.

Some people carry TSH bound up in a large complex with an antibody, called macro-TSH. It is biologically inactive and clears slowly, so it accumulates and reads high. In a series of consecutive people with subclinical hypothyroidism it accounted for about 1 in 400, and across a run of 124,000 unselected TSH tests it turned up about 1 in 21,000 times. In both series people affected had often been started on levothyroxine they did not need. Heterophile antibodies, which are antibodies that bind the test's own reagents, do the same thing. And antibodies against thyroid hormone itself can push a free T4 or free T3 result up, with the direction depending on how the assay is built, so the same blood gives a high free T4 on some platforms and a normal one on others.

None of this is common and none of it is worth worrying about in the abstract. What is useful is the trigger. Interference was usually spotted because the numbers did not fit each other, not because of symptoms. A high TSH alongside a normal or high free T4, or a strikingly abnormal result in someone who feels entirely well, is a result to question before anything is done about it. The move is a repeat, and where possible a repeat run on a different platform. A European guideline published in 2026 deals with exactly this problem and says the first step is a proper medication and supplement history, biotin especially.

What this test cannot tell you

TSH does not describe a lump. Thyroid function is usually normal in thyroid cancer, so a normal TSH is not reassurance about a nodule and a thyroid panel is the wrong test to order for one. That question is answered by ultrasound and, if needed, a needle sample.

Pregnancy, where the guidance has just moved

This moved in 2025 and 2026, and UK and US guidelines are mostly aligned. The fixed trimester cut-offs of 2.5 and 3.0 went in 2017, when the American guideline moved to lab-specific thresholds with a default of 4.0, and universal screening was already not recommended then. What the 2026 version adds is to separate the threshold for making the diagnosis from the target to treat to, to narrow the list of reasons to test so that age, weight and number of previous pregnancies no longer count on their own, and to say a mildly raised TSH should be repeated within 3 weeks before anything is treated. The UK obstetric guideline published the year before moved in the same direction, and both keep a treatment target below 2.5 for women already on levothyroxine. UK specialists have since argued in print that the 2.5 target is unevidenced, pointing out that a TSH between 2.5 and 4.0 with a normal free T4 has not been shown to raise miscarriage risk, and that overtreatment carries its own harms including lower childhood IQ and more attention problems at ages 9 and 10. Overtreatment being named as a harm, and not only undertreatment, is the change worth registering.

Units

The UK reports TSH in mIU/L and the US in µIU/mL. They are the same number, so no conversion is needed.

Free T4

Free thyroxine

The unbound, active fraction of thyroxine, the main hormone the thyroid produces. T4 is largely a reservoir, converted to the more active T3 out in the tissues.

What "free" means here, and where the estimate breaks

Almost all thyroxine in blood is bound to carrier proteins. Three of them do the work. Thyroxine-binding globulin carries roughly three quarters of it, transthyretin about a fifth and albumin the small remainder, which leaves around 0.03% of your total T4 circulating free. Those proportions are round approximations and reference books disagree at the margins. That tiny free fraction is the part that can enter a cell, and it is what this test is trying to measure. T3 is carried by the same 3 proteins in quite different proportions, so the split above does not transfer to it.

Measuring the free fraction directly is hard, so most labs estimate it with an immunoassay that assumes the carrier proteins are behaving normally. When the international standardisation body compared 17 ways of doing it in 2010, 15 immunoassays and 2 mass spectrometry procedures, only 2 came within 10% of the reference method and the other 15 read low, the worst by 42%. Recalibrating them closed most of the average gap and left a residue that is specific to the individual sample and not a fixed offset, which is the harder problem. Recalibration fixes the average and not your result. That is why the range on your report belongs to the method that produced it, and why a free T4 from one lab should not be set against one from another.

How the assay is built matters too, and this is new. Where someone carries antibodies against thyroid hormone itself, the one-step assays used on some of the most common analysers report a falsely high free T4 while two-step assays on other platforms agree with the reference method on the same blood. The same sample, 2 different answers, and the reason is assay architecture.

Anything that changes thyroxine-binding globulin changes the total without changing your thyroid status. Oestrogen raises it, and oral oestrogen far more than transdermal, because the liver sees the whole dose first. It is the same story as SHBG with testosterone and cortisol-binding globulin with cortisol, and the free measurement is meant to see past it.

The estimate fails outright in pregnancy. Free T4 immunoassays are unreliable there, and non-pregnant ranges do not apply anyway because free T4 falls as pregnancy progresses. TSH is the more dependable measure. Ovarian stimulation for fertility treatment does something similar, lowering free T4 and raising TSH in women whose thyroid was working normally, because binding protein production climbs faster than the axis can compensate. A panel taken mid-cycle will read oddly for reasons that have nothing to do with your thyroid.

This marker is what decides whether you get called subclinical

A raised TSH with a low free T4 is primary hypothyroidism. A raised TSH with a normal free T4 is subclinical hypothyroidism. The second one usually comes with reassurance and a repeat in a year.

The reassurance rests on trials showing levothyroxine does not improve symptoms in mild cases. Those trials are sound and that finding is real. The drug not working is not the same as the finding not mattering, and the 2 have been quietly collapsed into each other.

What the cohort evidence shows is that a raised TSH tracks with more coronary heart disease and more heart failure, and that the risk climbs with the number instead of switching on at a line. Between 4.5 and about 7 there is nothing. Above 7 the estimates start to lift but only coronary death reaches statistical significance, with coronary events and heart failure pointing the same way without getting there. At 10 and above it is clear on all of them. The heart failure relationship is also a J, with a suppressed TSH below 0.1 carrying about as much risk as a TSH of 10, so this is not a one-directional worry.

Age is where it gets contested, and the guide is not going to pretend otherwise. Grouping studies by the age of the people in them, a raised TSH predicted ischaemic heart disease and death in the younger groups and predicted nothing in the over-65s. The single largest pooled analysis of the same question, working with individual participant data instead of study averages, found no significant difference by age at all. Those 2 disagree and the second is the stronger design. For stroke the age pattern survives better, with a roughly threefold risk in people aged 18 to 49, nothing over 65, and a wide range around the young figure because it rests on few events.

Our own position, stated as ours. A raised TSH in a young person is worth taking seriously instead of filing, in the way a raised fasting glucose is worth taking seriously, because the direction of travel matters more than the label. That is a view about what is worth attention and not a claim that the age effect is settled, because it is not. What the evidence does not support in any reading is reaching for levothyroxine. What it points at is the thyroid antibodies, iron and ferritin, iodine intake, energy availability, sleep and training load, because those are the things that are actually movable.

TSH does not always see the problem

If the pituitary is the thing that has failed, TSH is the wrong marker to follow, because it is the signal that is broken. The 2025 European levothyroxine guideline is direct about it and says replacement in that situation should be monitored with free T4 and not TSH, as its own separate recommendation. The reverse trap also exists. A pituitary tumour producing TSH gives a persistently raised TSH alongside a normal or high free T4, and there is a published case of that being treated as ordinary hypothyroidism with escalating levothyroxine for 20 years. This is the practical argument for having free T4 on the panel at all and not TSH alone.

This is the steadier one, and that makes it the better marker to track

Free T4 barely moves within a day, does not swing with the seasons, and is not pushed around by eating. TSH does all three. Measured week to week in the same healthy person, free T4 varies by around 5% while TSH varies by around 20%.

That gap matters, because thyroid hormones run on a personal set point. Your own level oscillates in a band roughly half the width of the printed population range, and where that band sits is about 65% inherited. Two people can both sit at 14 pmol/L and only one of them is where they normally live. Two caveats on that. The half-width finding comes from a small study of healthy men sampled monthly for a year, and measured a calculated index and not free T4 itself. And the narrow band is a statement about months, not decades.

Two practical things follow. A single in-range free T4 tells you very little about you specifically, but 2 or 3 readings taken when you are well will map your own band, and after that every future result has something to be compared against. And a repeat has to move by a fair margin before it counts as a real change and not noise. One published set puts that at about 16% for free T4, 18% for free T3 and 55% for TSH, which is why a TSH that has "gone up a bit" between 2 tests usually has not. Those figures carry the imprecision of the analyser they were worked out on, so they are indicative and not universal.

Higher inside the range is not better

In large cohorts of people with normal thyroid function, sitting at the top of the free T4 range instead of the bottom tracks with shorter life expectancy, more atrial fibrillation and more hip fractures. Highest against lowest third, life expectancy at 50 was about 3 and a half years shorter. For atrial fibrillation the top quartile carried about 45% more risk, and TSH showed no association at all, which makes free T4 the sharper marker of the 2. The fracture signal runs at about 22% more risk per standard deviation of free T4.

Two things have to be said alongside that. The mortality association strengthens sharply with age and the absolute risk only becomes material above about 70, while under 50 the pattern is a U and not an absence. And a systematic genetic analysis found little evidence that normal-range thyroid function causes outcomes outside the thyroid, with fracture among the ones it found nothing for, although it did flag atrial fibrillation. For the rest, reverse causation or confounding is the likelier reading. So this tracks with, and there is no basis for concluding that it causes. Nothing here is a reason to try to lower a normal free T4.

A rising free T4 with a falling T3 is not a good sign

It is the signature of energy deficit. Cut energy availability far enough and free T4 rises while T3 falls and reverse T3 climbs, because conversion is being diverted and not because the thyroid is working harder. Someone dieting hard can watch free T4 drift up and read it as a win. The Free T3 entry sets out the full pattern and the thresholds it happens at.

Biotin. Free T4 is one of the markers biotin affects, and on this test it reads falsely high while TSH reads falsely low. Biotin, at the top of this guide, has the detail.

If you take levothyroxine

Absorption is where most unexplained results come from, and the standard advice has just been tested properly. Take it at a consistent time, either 60 minutes before breakfast or at bedtime at least 3 hours after the evening meal. Consistency matters more than which of the 2 you pick. In a 2026 randomised trial, taking it with breakfast and raising the dose by 15% held TSH just as steadily as taking it fasting, and most people preferred it and chose to carry on that way. That is a real option and not a licence to stop bothering, because it needs a deliberate dose change and a repeat test to confirm it worked.

Coffee cuts absorption, so leave an hour. Calcium and iron need about 4 hours, and so do soy and high-fibre food. Magnesium has now been tested for the first time. Magnesium aspartate cut thyroxine absorption by about 12%, less than iron or calcium but real, and magnesium citrate moved it less, so separate either of them too. Grapefruit juice appears on most avoidance lists and should not, because when it was put to a randomised test it did not substantially affect absorption.

The gut conditions that raise the dose you need are better documented than the internet suggests, and the stomach matters more than the bowel. Helicobacter pylori and atrophic gastritis raise the requirement by 22 to 34%. In the same work, adding a proton pump inhibitor raised TSH in every one of the 10 people tested and needed a 37% dose increase to correct. Ulcerative colitis raises it by about 26%, measured while the disease was quiet. Coeliac disease raises it by about half in people still eating gluten, and the extra requirement disappears on a gluten-free diet without changing the dose, so a dose that keeps climbing is a recognised prompt to look for it.

Below that tier the evidence thins fast. Crohn's disease is named in the reviews but rests on almost no dose data. Small intestinal bacterial overgrowth has case reports only. For irritable bowel syndrome there is nothing at all. And on increased intestinal permeability, the thing usually called leaky gut, there is no evidence either way, because levothyroxine absorption depends on stomach acidity and the surface of the small bowel and not on tight junctions. Increased permeability is real and measurable and has been reported in Hashimoto's, but that work is about autoimmunity and not about whether a tablet gets absorbed.

Two more things, both from the 2025 European guideline. Needing more than 1.5 to 1.7 micrograms per kilogram a day with a thyroid still in place is a recognised signal to look at adherence and absorption before raising the dose again. Where the thyroid has been removed the figure is 1.8 to 2.0. And switching from a tablet to a liquid or softgel at the same dose is a lever that sometimes works where TSH stays stubbornly high.

Losing weight changes the dose you need

Levothyroxine is dosed against lean body mass, so substantial weight loss makes a stable dose too big for the person taking it. With the weight loss now being achieved on GLP-1 drugs this has become a live problem and not a theoretical one. Nobody has yet measured thyroid function properly through a course of them, and the direction is so far only described in individual cases, including someone tipped into a genuinely overactive state on an unchanged dose. What has been measured is the monitoring, and it is not happening. People on levothyroxine starting a GLP-1 drug were re-tested no more often than people starting an SGLT-2 inhibitor, which produces far less weight loss. After bariatric surgery, which is the closest thing to good evidence here, the average dose change is small and the direction is not consistent from one person to the next. So it moves in both directions, and the thing to do is re-test and not assume.

Units

The UK reports free T4 in pmol/L and the US in ng/dL. Multiply ng/dL by about 12.9 to get pmol/L, so 1.2 ng/dL is about 15.5 pmol/L.

Free T3

Free triiodothyronine

The unbound, active fraction of T3, the hormone that actually acts on tissue. Most of it is not made by the thyroid at all, it is converted from T4 out in the body. How that conversion works, and why a tissue can set its own T3 level independently of the blood, is in Thyroid, in the Body Guide.

What "free" means here

Only about 0.3% of the T3 in your blood is unbound, and that sliver is what this test is after. The rest is held on carrier proteins, which the Body Guide explains. The same arrangement runs through free T4, and through free testosterone where SHBG does the carrying, so all 3 entries are measuring the small usable fraction and not the total.

Free T3 is estimated indirectly and the methods are not standardised against each other, so results do not transfer between labs. If you are tracking it, use the same lab every time. It is also the most interference-prone of the 3 thyroid markers by some distance. Across 124,000 tests on one platform, confirmed interference turned up about 7 times more often in free T3 than in TSH. That is a caution about the marker consumer panels promote hardest.

Biotin. Free T3 is one of the markers biotin affects, and on this test it reads falsely high while TSH reads falsely low. Biotin, at the top of this guide, has the detail.

T3 is a gauge of how much energy is coming in

This explains most results that look wrong in people who are otherwise well. T3 tracks energy availability in both directions. It falls in a deficit and it rises in a surplus.

Going down. Cut energy hard enough and T3 falls, reverse T3 rises, free T4 drifts up and TSH drifts down. Nothing is wrong with the thyroid. The body has throttled back.

Three things about that are usually got wrong. It is the energy and not the training. In a study that crossed 3 exercise conditions against 2 levels of energy availability, the deficit moved T3, free T3, T4 and reverse T3 while free T4 held steady, and the exercise moved nothing, with 1,300 kcal a day of hard riding changing no thyroid hormone at all as long as the energy was replaced. It is fast. In healthy young women, 5 days at a 55% deficit dropped T3 and TSH, raised reverse T3 by about a third and lifted free T4 by about 7%, with no change in fat mass, cortisol, growth hormone or binding protein. And there is a threshold instead of a slope, with T3 falling abruptly somewhere between 19 and 25 kcal per kg of lean mass per day and holding above that. Free T4 and reverse T3 shift at a lower threshold than T3 does, so the pattern arrives in stages.

Coming back up is slower than going down. Recovery is measured in months.

Going up. People carrying more body fat run higher T3, not lower. The arrow points from the fat to the hormone and not the other way round. Overfeeding for 3 weeks raises T3 production by about three quarters, between 71% and 82% depending on whether the extra food was carbohydrate, fat or protein, and T4 production does not change at all. Losing weight brings it back down. Nobody has shown that a higher T3 inside the normal range causes fat gain.

So T3 is reporting how much fuel is arriving and not setting how lean you are. If you are dieting, deep in a training block or coming off a long deficit, a low free T3 is the expected finding. It is not a diagnosis, it is not treated with thyroid hormone, and what it responds to is food.

What it is and is not good at

It is the most useful of the 3 for detecting an overactive thyroid, and it catches one pattern the others miss entirely, a raised T3 with a normal free T4.

It is a poor test for detecting an underactive thyroid on its own. In the usual sequence TSH rises first, free T4 falls next, and free T3 is held inside the range for some time because the body raises conversion efficiency when hormone is scarce. A normal free T3 does not rule out a developing problem.

The argument about whether the blood level means anything

The standing objection is that a blood level cannot report what is happening inside cells. There is a real case against that objection, and it has limits.

The case rests on tracer work showing blood T3 tracks tissue T3, and tracks whether T3 is actually switching genes on, in most tissues. Brain and pituitary are exceptions because they set their own supply locally, and so is the whole body during illness or starvation. That work is animal work. Nobody has measured tissue T3 in a living person with an underactive thyroid. And when the practical implication was tested in people with no thyroid gland left and therefore no T3 of their own, no thyroid measurement of any kind explained their quality of life or fatigue, including free T3, reverse T3 and the ratio between T4 and T3.

Two things follow. The question is open instead of closed in either direction. And almost everything known about it comes from people already on thyroid medication, who are not a healthy population, so none of it transfers cleanly to someone with an untreated normal TSH and free T4.

Treatment with T3

Levothyroxine alone is the standard everywhere and works for most people. It is T4, converted to T3 as the body needs it.

Combination treatment adds liothyronine, which is T3. NICE does not recommend it routinely. The position has softened, and in 2021 the American, British and European thyroid associations jointly agreed there was equipoise for a new trial. That document is often quoted as though it endorsed or rejected combination treatment. It did neither. It is a specification for how a future trial should be built, and 5 years on the trial it called for has not reported, though trials following its design are running.

The trials so far show no benefit on objective measures. What they also show is that quality of life improved on combination treatment and improved just as much on levothyroxine alone, which is a large placebo effect and not an absence of change. The preference studies run the other way. Pooling randomised trials in which patients tried both, 52% preferred a T3-containing regimen and 24% preferred levothyroxine alone, with the rest indifferent. The trials varied considerably and a quarter of people did prefer levothyroxine, so this is a roughly twofold lean and not a consensus. Both findings are real and they have not been reconciled.

The genetic explanation people reach for, a common variant in the enzyme that converts T4 to T3, was tested in nearly 19,000 treated patients and predicted neither worse wellbeing nor worse cognition, with no sign that carrying it changed how people did on levothyroxine. The same work found something else worth sitting with. Being on levothyroxine at all went with more tiredness, lower wellbeing and slower reaction times than not being on it. The residual symptom problem is real. The genetic explanation for it is what failed.

Two large datasets published in 2026 looked at what happens to people on T3-containing treatment and reached opposite conclusions, one finding less dementia and lower mortality and the other finding more psychiatric illness. Both are observational and both are open to the same objection from different directions, because the people who end up on T3 are both the most symptomatic and the most motivated. They do not settle anything and the trial that would has still not been run.

Liothyronine alone is not used for ordinary hypothyroidism. National guidance frames it as an option where symptoms persist despite an adequate trial of levothyroxine, and separately as preparation for radioiodine in thyroid cancer. NHS England guidance says it should only be started by an NHS consultant endocrinologist, with a GP continuing the prescription after a 3-month specialist review under local shared-care arrangements. That is national guidance and not a legal restriction, it is applied through local formularies so the practical picture varies by area, and Scotland, Wales and Northern Ireland run their own arrangements.

Desiccated thyroid extract, sold as natural thyroid, Armour or ERFA, is unlicensed here, so the regulator has not assessed its quality, safety or effectiveness. The 2 randomised trials against levothyroxine found no advantage. The practical objection is that the T3 in it is short-acting, which makes levels swing and doses hard to settle.

If you take liothyronine, timing matters more than for any other thyroid marker. T3 peaks about 2 to 2 and a half hours after a dose, so the same person on the same dose reads high or low depending on when the blood was taken. The knock-on effects run on different clocks again, with heart rate lifting by about 5 hours and TSH bottoming out at 12 hours and staying suppressed for 2 or 3 days. Be consistent, write down what you did, and make sure whoever reads the result knows you are on T3.

Somewhere between 5 and 15% of people whose TSH has been brought into range on levothyroxine still have symptoms, depending on which study you take. That gap is why this argument will not go away. The counterweight has to be stated alongside it. Symptoms that could be attributed to hypothyroidism are common in people with entirely normal thyroid function, so these symptoms do not distinguish a thyroid problem from ordinary life, and a normal TSH with fatigue is not evidence that the thyroid is the answer.

Reverse T3

Reverse T3 is produced by the enzyme that inactivates thyroid hormone, and it climbs with energy deficit, illness, injury, inflammation and high cortisol. A raised reverse T3 is a genuine readout of the body shifting into conservation. The Body Guide sets out the enzyme behind it.

The mainstream dismissal is narrower than it looks, and it is worth being precise about what has and has not been tested. Every negative source tests one of 2 things, whether reverse T3 diagnoses an underactive thyroid, or whether it should guide a levothyroxine or liothyronine dose. The American Thyroid Association's hypothyroidism treatment guideline does not mention it as a recommendation at all. None of that evidence tests whether the number reflects metabolic state, and on that question the physiology is on its side.

Where it goes wrong is what people do next. A high reverse T3 gets read as a conversion problem and answered with more thyroid hormone, which is the one response the evidence does rule out and which risks pushing someone into over-treatment. A 2025 study measured reverse T3 by mass spectrometry in nearly 1,000 treated, fatigued patients and found the pattern runs the wrong way for that story. A raised reverse T3 was most common in people on levothyroxine alone and least common in people taking T3, and it rose alongside free T4 and free T3 instead of against them. It is behaving as a marker of how much T4 is around, not as something blocking T3 from working. The thing actually driving a high reverse T3 in a well person is usually energy, training load, sleep and recovery, and those are what move it.

Free T3 moves earlier than reverse T3 and at shallower deficits, so it warns you sooner. Reverse T3 assays are not standardised, and one comparison found immunoassay reading about 2 and a half times higher than mass spectrometry on the same samples, though not by a constant factor across the range. No reference change value has been published for it and it does not appear in the European biological variation database, so unlike free T4 and free T3 there is no figure for how far it has to move before the change is real. The only way to track it usefully over time is the same lab every time.

One correction to the usual telling. In fasting, most of the rise is the body clearing reverse T3 more slowly instead of making more of it. The number still climbs and still means what it appears to mean. The route there is different from the one in genuine illness, where the inactivating enzyme really is switched on.

Units

The UK reports free T3 in pmol/L and the US in pg/mL. Multiply pg/mL by about 1.54 to get pmol/L, so 3.0 pg/mL is about 4.6 pmol/L.

Thyroid Antibodies

TPO and Tg antibodies

Antibodies the immune system makes against the thyroid. Thyroid peroxidase antibodies are aimed at the enzyme that builds thyroid hormone, thyroglobulin antibodies at the protein the gland stores hormone on. Why the immune system is the usual reason a thyroid fails, and how often antibodies turn up in people whose thyroid works perfectly, is in Thyroid, in the Body Guide.

TSH, free T4 and free T3 describe how the gland is performing now. Antibodies describe which way it is likely to go.

Guidelines key on the TPO antibody. Thyroglobulin antibodies appear nowhere in NICE's recommendations and come bundled into private panels because they point at the same process on the same sample.

A negative result is not permanent

People who test negative can turn positive later, and it happens often enough to matter. Among antibody-negative women with a family history of autoimmune thyroid disease, 14.5% developed TPO antibodies within 5 years and 20.1% developed one antibody or the other. The only thing measured at the start that predicted who would was where their TSH already sat. One thing that happened during follow-up predicted it too, which was stopping smoking.

A negative describes the day it was taken. If circumstances change, particularly after a pregnancy or if TSH starts drifting up, the question is open again.

What a positive result predicts

A positive antibody is a risk marker and not a diagnosis. It raises the odds that the gland fails eventually and says nothing about whether it has failed yet.

The clearest figure for a younger population comes from UK women aged 16 to 40 who were antibody positive with normal thyroid function, followed for about a year before and through pregnancy. Around 7% became hypothyroid, almost all of it subclinical, and 83% crossed over before conceiving instead of during pregnancy. Two things about that figure. Half the women in it were taking levothyroxine as part of a trial, and among those who were not the rate was about 9%. And they had all had a miscarriage or fertility trouble, so this is not a general population number.

The best long-term picture comes from a population followed for 18 years, and it is more reassuring than the progression figures alone suggest. Antibody levels followed 4 patterns. About 81% stayed low throughout, about 13% stayed high, about 2% started high and fell, and about 3% started low and climbed. So most people do not move, a small number turn positive, and a small number go the other way and become negative again. A TSH of 5 or above roughly tripled the chance of turning positive.

Antibodies alone and a raised TSH alone carry similar weight. Together the risk is several times higher than either on its own, and that combination is the one that matters.

Higher titres do progress faster, but the level at which that starts is not settled and, for the reason below, a number derived on one analyser would not carry across to another.

The number does not transfer between labs

Thyroid antibody assays are not harmonised, and the international reference preparations have not fixed it. Run the same healthy samples across different platforms and the upper reference limit for TPO antibodies varies by about 4 and a half times, and for thyroglobulin antibodies by about 18 times. Modern TPO assays agree with each other on positive or negative far better, with the newest generation matching other platforms 94 to 98% of the time against 89 to 91% for the one before it, but the values themselves still do not compare.

There is a second thing hidden in that. For the newest TPO assay, the 95th centile of healthy people sits below 8 while the cut-off used to call a result positive is 20. The normal range and the diagnostic threshold are not the same number, and a result between them is neither clearly normal nor positive.

So the cut-off printed on your own report is the only one that applies to your number, and a change of lab invalidates any comparison over time. If your lab changed assay, your antibody number can move without your immune system moving.

There is more than calibration going wrong here. The 2026 European guideline on interference lists what else these assays react to, and for TPO antibodies that includes intravenous immunoglobulin, heterophile antibodies, rheumatoid factor and cross-reactivity with myeloperoxidase, an enzyme from white blood cells that is structurally close to thyroid peroxidase. Thyroglobulin antibody assays can be thrown by a very high thyroglobulin in the same sample. The same guideline states plainly that the sensitivity and specificity of TPO antibodies for autoimmune thyroid disease are relatively poor, which is not how the test is usually sold.

Biotin. Both antibodies are markers biotin affects, and on these tests they read falsely high. Biotin, at the top of this guide, has the detail, and the TSH entry sets out why the whole panel then looks contradictory instead of obviously wrong.

What to repeat, and what not to

NICE's position is one sentence. Consider measuring TPO antibodies in an adult whose TSH is above the reference range, and do not repeat the test. The committee's reasoning was that changes in the level do not guide treatment decisions.

What antibody status changes is the monitoring interval and not the treatment threshold, and that distinction is easy to lose. NICE's levothyroxine thresholds for subclinical hypothyroidism contain no antibody criterion. A positive result buys yearly TSH and free T4, a negative one every 2 to 3 years. The older UK lab guidance puts the negative interval at approximately every 3 years, and that guidance is now 20 years old.

So the thing to repeat is TSH. The titre is not a progress bar.

Pregnancy and fertility

Antibody-positive women with normal thyroid function deliver early more often. Across 47,000 pregnancies the rate was 6.6% against 4.9%, an absolute difference of about 1.6 percentage points. They also miscarry more often, with roughly double to quadruple the odds in pooled analyses, though nobody has reliably established how big that difference is in absolute terms, so the relative figures are doing more work than they should.

Treating it does not help. Three randomised trials have given levothyroxine to antibody-positive women with normal thyroid function, covering previous miscarriage, recurrent loss and fertility treatment, and none improved live birth rates. The 2026 American guideline is unusually firm for this field. For antibody-positive women with normal thyroid function and infertility it says levothyroxine should not be offered, regardless of TSH level or miscarriage history, and it grades that as a strong recommendation on high-quality evidence. A recommendation of that strength against an intervention is rare here.

The same guideline says TSH and free T4 may be rechecked every 3 to 6 months while trying to conceive, which is where the progression rate above earns its keep. That one is a conditional suggestion on low-quality evidence, so it is worth doing and it is not firmly evidenced.

The period after birth matters as much. Up to half of women who are antibody positive in early pregnancy develop postpartum thyroiditis, against a background rate of around 5%. The textbook course, an overactive phase followed by an underactive one, is the minority pattern and happens in about a quarter of cases. About 4 in 10 have an underactive phase only and about 3 in 10 an overactive phase only. Where it is biphasic the overactive phase usually runs from about 1 to 4 months and the underactive one from about 4 to 8. Most settle within a year, and between a fifth and two fifths stay hypothyroid for good. Symptoms are more common in the underactive phase and any of it gets mistaken for ordinary exhaustion.

Symptoms with a normal TSH

Antibody-positive people with normal thyroid function do report more fatigue, low mood and general symptom burden on average. The effect is small, nearly all of the evidence is cross-sectional, and the populations studied were already seeking thyroid care.

The idea has been tested another way. The largest trial of selenium in autoimmune thyroid disease lowered TPO antibodies over a year and produced no improvement on any thyroid quality of life measure against placebo. If circulating antibodies were driving the symptoms, that should have worked. The people in that trial were already on levothyroxine and not euthyroid, so it is not a direct test of the antibody-positive, normal-function case, but it is the closest thing there is.

A positive result does not explain an individual's symptoms, and treating it as the answer tends to stop the search early.

What has been tried

Selenium lowers TPO antibodies and modestly lowers TSH in people not taking levothyroxine. It does not move free T4, free T3, thyroglobulin antibodies or the size of the gland. In the largest placebo-controlled trial the antibodies fell and neither quality of life nor levothyroxine dose changed. No trial has used progression to hypothyroidism as a prespecified endpoint, so the claim that it protects the gland has not been tested rather than disproved.

The picture has also turned less favourable. Two studies published in 2026 found antibodies going up on supplementation and not down, one of them also reporting more autoimmune diagnoses and higher mortality in the supplemented group. Both are retrospective and neither measured selenium status at the start, so they are a reason for caution and not evidence of harm. They sit against a meta-analysis of 35 randomised trials pointing the other way.

Dose is the other half of it. Trials use 200 µg a day. Europe cut its tolerable upper intake for selenium to 255 µg a day in 2023, and once ordinary UK food is counted a 200 µg supplement puts most people at or above that. The UK's own guidance figure for supplements is more permissive at 350 µg. Long-term supplementation at 200 µg in people who were not deficient raised the rate of type 2 diabetes by about half, and the risk was concentrated in those who already had the highest selenium levels to begin with. Most UK diets are adequate without it.

Myo-inositol combined with selenium is sold specifically for this. The positive results come almost entirely from one Italian group testing a branded product, and the largest of those studies had no control group. When the supplements were pooled against each other, selenium alone significantly reduced both antibodies while the combination did not reach significance, though its estimates were larger and rested on far fewer trials, so that is a difference in certainty and not in direction. A second pooled analysis of the same field ranked them the other way round, and a 2026 one found the combination lowered TSH more than selenium alone without touching TPO antibodies. Nobody should read a ranking out of this.

Low dose naltrexone has no randomised trial in any thyroid condition, and nothing registered either. The one national prescription-registry study compared how much thyroid hormone 898 people took in the year before and the year after starting it and found no difference, if anything a slight drift upward. That measures dispensed medication and not symptoms, so it cannot detect a benefit that did not change anyone's dose.

A gluten-free diet in people without coeliac disease has now been tested. Three small randomised trials covering 110 people between them were pooled in 2025, and the result is not the one the internet expects. Thyroglobulin antibodies fell slightly and TPO antibodies rose, with no change in thyroid function either way. The evidence was rated as having serious methodological problems and very serious imprecision, so this is not a finding to lean on hard in either direction. The study usually quoted in favour is a separate one, and it enrolled only women who already had positive coeliac antibodies and allocated them by whether they complied and not at random, so it does not answer the question asked of it. Testing for coeliac disease is the useful move, for the reason in the last section below.

Vitamin D trials agree that TPO antibodies fall and disagree about everything else, with one pooled analysis finding thyroglobulin antibodies and thyroid function improved and another finding no effect on either. And women who went on to develop thyroid antibodies did not have lower vitamin D beforehand than those who did not.

Nothing has been shown to stop a positive antibody turning into an underactive thyroid. Lowering the number is not the same as changing what happens, and the one trial that separated those 2 things found exactly that.

What raises the risk

Iodine is the clearest modifiable factor and it cuts both ways, which is why the advice around it is such a mess.

Too much is the better-documented direction. In a large Chinese population followed for 5 years across 3 regions with different iodine supply, autoimmune thyroiditis rose from 0.2% to 1.3% and subclinical hypothyroidism from 0.2% to 2.9% as intake went from mildly deficient to excessive. Overt hypothyroidism did not follow the same gradient. The people it caught were those who already carried antibodies. The same work found the lowest subsequent rate of any thyroid abnormality in people whose starting TSH sat between 1.0 and 1.9.

Too little has quietly become a UK problem again. The national diet survey covering 2019 to 2023, published in 2025, found urinary iodine had fallen by about a quarter to a third over a decade and classified girls aged 11 to 18 and women aged 16 to 49 as iodine insufficient. Those groups were classified sufficient in earlier rounds, so this is a reversal. The drivers are less dairy and less white fish, plant milks that are mostly not fortified, and the fact that the UK never adopted iodised salt. The UK reference intake is 140 µg a day. Ordinary diet, iodised salt where you can find it, and a standard 150 µg prenatal are not the problem.

What you cannot do is test your way out of this. There is no reliable blood or urine test of iodine status in an individual, because iodine excretion swings with what you ate yesterday. Urinary iodine works for populations and not for people. Anything sold as a personal iodine check will not answer the question, and this is one of the few places where the honest answer is that the test does not exist rather than that we do not offer it.

At the other end, kelp, bladderwrack and seaweed supplements are not standardised and can deliver thousands of micrograms in a capsule, and high-dose iodine drops sold as thyroid support run many times the daily requirement by design. The European upper limit is 600 µg a day and the UK guidance figure for supplements is 500. Iodine-induced thyroid dysfunction from these products is documented and it is the most common way people do themselves harm here.

Stress is the thing people reach for first and the evidence is split in a specific way. The one prospective study tracked stressful events, daily hassles and mood in women at raised risk and found no difference between those who developed antibodies and those who did not, and none for going on to overt disease either. Retrospective studies consistently find the opposite, reporting more difficult life events in the year before diagnosis. The most likely explanation for that gap is that people who are already ill remember differently, and that early illness itself makes life harder. A 2026 review of the whole field rated the evidence as very low certainty. So the honest position is that the prospective data do not support it and the retrospective data cannot settle it.

Smoking points in opposite directions for the 2 thyroid diseases, which is why it gets quoted both ways, and the 2 halves need keeping apart. For autoimmune hypothyroidism, smokers carry fewer thyroid antibodies, 11% against 18%, and among those who do carry them smokers have about 40% lower odds of a raised TSH. The risk then jumps in the first 2 years after quitting, by 6 or 7 times, before settling back to no excess. For Graves' disease it runs the other way entirely. Current smoking roughly triples the risk and about quadruples the risk of thyroid eye disease, and quitting lowers both. Smoking's relationship with Hashimoto's specifically is inconsistent across study designs, so nobody should take any of this as a reason to keep smoking.

TRAb is a separate test

TSH receptor antibodies are ordered separately and are not included when a panel says thyroid antibodies. They confirm Graves' disease and separate genuine overactivity from a transient thyroiditis that will pass on its own.

They are also the exception to the rule above about repeating, though less of one than it is often made out to be. European guidance asks for them at diagnosis, before stopping antithyroid drug treatment, and in pregnancy to judge risk to the baby, with a repeat a year later if they are still high. It does not ask for routine serial measurement to steer the drug dose, which is decided on how the person is doing. A TPO result tells you nothing about Graves'.

It travels with other autoimmune conditions

Roughly 1 in 7 people with Hashimoto's has another autoimmune condition, and rheumatoid arthritis is the one most often found alongside it. Pernicious anaemia, coeliac disease, vitiligo, lupus and Addison's all carry more than 10 times the background risk, though several are rare enough that the absolute risk stays low. Those figures come from people attending UK hospital thyroid clinics and the other conditions were self-reported, so treat them as indicative.

Coeliac is the practical one, because it is common enough to turn up and it is treatable. Pooling across all ages, biopsy-confirmed coeliac disease runs at about 1.6% in autoimmune thyroid disease against roughly 1% in the general population. In adults specifically it is about 2.7%, in children about 6.2%, which is where the much higher figures in circulation come from, and in hypothyroidism about 1.4%. The pooled figure sits below both age subgroups because it is pulled down by the studies that mixed adults and children together, which came out at about 1%. The pernicious anaemia link is why B12, in the Body Guide covers intrinsic factor antibodies.

Units

IU/mL and kIU/L are the same number, so a result of 35 on one scale is 35 on the other. Some analysers report arbitrary units instead, and those do not convert. None of them mean anything without the reference limit printed on the same report. And matching units do not make matching results. Two labs can both report in IU/mL and still give numbers for the same blood that differ many times over.