Subclinical Hypothyroidism: Your Labs Are “Normal.” You’re Not.
Four words. They land in a patient portal at 6:40 on a Tuesday evening, or they get said in about a second and a half on the way out of an appointment.
Everything came back normal.
Except you are tired in a way that sleep does not touch. Your hands are cold in July. There is more hair in the drain than there used to be. The weight is climbing and you have not changed a thing. And your brain feels like it is running through mud.
Here is what actually happened in that visit. They drew one number. It came back inside the range. And that ended the inquiry.
One number, from a gland that is not even your thyroid. And on the strength of it you get told you are fine, and then quietly handed an antidepressant, or a diet, or your own age as an explanation.
I want to be careful here, because this is a topic where the internet gets loud and unhelpful fast. So let me say what this article is not. It is not "everyone has a thyroid problem." It is not a supplement pitch, and I am not going to tell you to fire your doctor. What it is, is an honest look at three things that got baked into thyroid care a long time ago and never got re-examined: what the word "normal" on a lab report actually means, why the test everyone runs is the wrong test for the question everyone is asking, and why the oldest thyroid medication in the world got quietly pushed out of the conversation.
What "subclinical hypothyroidism" actually means
If your TSH comes back above the top of the range and your free T4 is still normal, you get a label: subclinical hypothyroidism. The standard response to that label is to do nothing. Watch and wait. Recheck in six months, or a year.
Now look at what the word is actually saying. Subclinical means without symptoms.
So consider a case of the kind that shows up constantly in thyroid teaching: a TSH of 7.1, which is above the line and therefore "subclinical," alongside a free T3 of 2.2, which is scraping the floor. The clinical note on that patient reads "very symptomatic." We have just called her "without symptoms" while she is sitting in the room having symptoms.
The label is not a finding. In her case the label is simply wrong, and it is the reason nobody looked any further.
That is the whole problem in one word. And the rest of this article is about why the label survives.
"Normal" is a description of a crowd, not a target for a person
Start with the word itself, because everything downstream depends on it.
A reference range does not mean healthy. It does not mean optimal. It means we measured a big group of people, drew a line around the middle chunk of them, and called that the range. It is an average of a population. It does not tell you whether those people feel their best, function their best, or are their best.
This is not a fringe complaint. Fitzgerald and Falhammar, writing in Frontiers in Endocrinology, put it plainly: there are "no readily identified individually specific optimum levels of thyroid hormones for any individual," and population-based cut-offs for TSH may not be an appropriate treatment target.
Here is where it gets concrete. Take free T3, and hold that name, because it is the hero of this article. Free T3 falls with age. That is measured, not asserted: van den Beld, Visser and colleagues, in the Journal of Clinical Endocrinology and Metabolism, found a clear age-dependent decline in free T3, while free T4 stayed essentially flat.
Now think about what that does to a "normal range" built by averaging a population. The older, and frankly the sicker, that population is, the lower the bar drops, and the more comfortably you can land inside it while feeling nothing like a healthy young adult did.
Sit with that. The "normal" you are being measured against is, in part, the average of people who are older and sicker than you would like to be. Landing inside it is not a clean bill of health. It is a statement that you resemble the crowd.
That is why we say optimal, not normal. Not as a slogan. As a correction to a category error.
TSH is a pituitary hormone, not a thyroid hormone
Now the bigger problem: the test itself.
TSH stands for Thyroid Stimulating Hormone, and here is the thing almost nobody is told. TSH is not made by your thyroid. It is made by your pituitary, a gland in your head. TSH is the message your brain sends to your thyroid. It is the request. It is not the hormone that does anything in your body.
Think about a thermostat. The thermostat is not the heat. It is the thing asking for heat. If you walk into a freezing house and check the thermostat and it reads "set to seventy," you have learned exactly one thing, which is that the thermostat is asking. You still have no idea how cold the room is. For that, you would have to measure the room.
TSH is the thermostat. Free T3 is the room temperature.
I want to be fair to the test, because it does have a real job. The position I hold, and I am flagging it as my clinical position rather than a trial finding, is this: TSH is a good test for diagnosing primary hypothyroidism, and a poor one for assessing how well a thyroid is actually functioning, for monitoring replacement, and for tracking whether symptoms are improving. Those are different questions, and the same number can be genuinely good at one and genuinely bad at the others. The published data below is why I hold it.
And this has been measured. Fitzgerald and colleagues, in the journal Thyroid, went through nearly 1,900 tested associations between thyroid blood tests and actual clinical findings. Free T4 tracked with the clinical picture about half the time. T3, about half the time. TSH: 23 percent. Their conclusion was that no clinical parameter associated better with TSH than with the actual thyroid hormones.
So the number we built the entire system around is the number that tracks worst with how the patient is actually doing. And it is not close.
There is a study that goes one step further. Patients with hypothyroid symptoms, a normal TSH between 1 and 4, and a normal free T4 were measured by mass spectrometry, and were confirmed to be biochemically hypothyroid. Normal TSH. Genuinely low thyroid hormone (Ling, Soldin and colleagues, Annals of Thyroid Research).
TSH has no metabolic activity at the cell level. It does not warm you up. It does not grow your hair. It does not clear the fog. It just asks.
Why the guidelines say "don't treat," and what those trials actually tested
In fairness, the guidance did not come from nowhere. In 2019 the BMJ published a Rapid Recommendation, Bekkering and colleagues, advising against thyroid hormone for adults with subclinical hypothyroidism. It was built on real trials that genuinely did not show people feeling better. That happened, and I am not going to pretend otherwise.
The biggest of those trials deserves to be named, because it is the one everything rests on. TRUST, Stott and colleagues, in the New England Journal of Medicine in 2017: 737 community-dwelling adults over 65 with persistent subclinical hypothyroidism, randomized, double-blind, placebo-controlled, across four countries, with mock dose adjustments in the placebo arm so nobody could work out who was on what. It is a well-built trial. I want to say that before I say anything else.
Now look at what they gave. Levothyroxine, starting at 50 micrograms a day, adjusted by computer toward a TSH inside the reference range. And here is the number almost nobody quotes: the median dose at one year was still 50 micrograms. It started at 50 and it finished at 50.
Then they measured. The between-group difference on the two primary outcomes, a hypothyroid symptom score and a tiredness score, was 0.0 on one and 0.4 on the other, against a nine-point threshold for a difference a patient would actually notice. No symptom benefit, no tiredness benefit, no cognitive benefit, no subgroup that benefited. That is a real randomized trial with a real null result, and the authors' own confidence intervals excluded any benefit larger than about two points, which means I do not get to call it underpowered either. It failed cleanly.
So why am I still writing this?
Because of what got tested. There is no T3 anywhere in TRUST. The word liothyronine does not appear in the paper. They tested a small, essentially fixed dose of the storage form of thyroid hormone, and the free T4 went up by a little, and the TSH came down. That was the entire biochemical intervention. Then they measured how people felt.
That is not a surprising result. That is a prediction, and the physiology that predicts it was in print before the first participant was randomized. The next section is that physiology.
One more thing worth having, because it belongs in the same breath. TRUST also reported new atrial fibrillation in 3.0 percent on levothyroxine versus 3.5 percent on placebo, fractures in nine versus eight, and new osteoporosis in three versus four. Randomized and double-blind. Hold onto that, because it comes back later.
T4 is storage. T3 does the work.
Your thyroid makes two hormones that matter here, T4 and T3, and the names are just counts of iodine atoms. Here is the part that changes everything.
T4 is largely storage. T3 is the one that does the work. T4 is a prohormone, inventory sitting in the warehouse. To use it, your body has to pull one iodine off and convert T4 into T3, out in your tissues.
So when you are handed levothyroxine, you are being handed T4. Pure storage. And the entire strategy quietly rests on one assumption, which is that your conversion works fine. For a lot of people it does. For a lot of people it does not.
And now the genuinely counterintuitive part. Giving T4 can, in a way that took decades to work out, lower your T3.
T4 itself inactivates the enzyme that makes T3. The enzyme is called type 2 deiodinase, or D2, and it is destroyed in the act of doing its job. Ettleson and Bianco, in the Journal of Clinical Endocrinology and Metabolism, describe it as exhibiting "a suicidal behavior." The more T4 you put into the system, the less of that enzyme survives to convert it.
And here is the twist. The same enzyme in your hypothalamus, the part of your brain that sets your TSH, is more stable and largely spared. In their words, serum thyrotropin "is normalized at higher serum T4 levels that are insufficient to normalize serum T3."
Read what that means. Your brain keeps sensing plenty of thyroid hormone, so it stops asking, and your TSH normalizes. Meanwhile your muscle, your liver and your actual tissues have lost the enzyme they needed, and they are still running low. The thermostat gets fed first. The rest of you does not.
The lab number and the patient come apart, and the physiology explains exactly why. You give more T4. The T4 on the report climbs. The TSH drops into a beautiful-looking range. And the hormone that actually does the work can be falling the whole time. The chart looks better while the patient gets worse.
Is that measurable in real people? Yes, and I will give you the honest version rather than the tidy one.
Gullo and colleagues, in PLoS ONE, studied 1,811 patients who had no thyroid gland at all, so every molecule of thyroid hormone in them came out of a tablet, against 3,875 people with working thyroids. They cut both groups into five bands by TSH and compared band against band. In all five bands, free T4 was higher and free T3 was lower in the treated patients. 15.2 percent of them had a free T3 below the reference range outright, with a normal TSH and a clinician telling them they were fine. And as the levothyroxine dose went up, the imbalance got worse, not better. That study was retrospective, and its authors are explicit that they did not measure how any of those patients felt, so it is biochemistry and not an outcome.
Ettleson and Bianco summarize the picture the same way: in most levothyroxine-treated patients, normalizing the TSH results in a lower T3 to T4 ratio with relatively lower T3, and in at least 15 percent of cases the T3 falls below normal. Notice who is writing that. Antonio Bianco is a past president of the American Thyroid Association. This is not a complaint from outside the field. It is coming from inside it.
And now the part I owe you, because I would rather hand you the other side's best card than have you find it later.
Jonklaas and colleagues published a study in JAMA in 2008: fifty patients measured before and after their thyroid was removed, concluding that levothyroxine alone restores serum T3 to where it was before surgery. That is the study most often reached for when somebody wants to say the T3 question is settled. So I pulled it and read the whole thing, discussion and limitations included.
Their result first, because that is the order that keeps me honest. By the end of the study, on levothyroxine, T3 averaged 127.2 ng/dL. Their own T3 before surgery, with their own thyroid, 129.3. P = 0.64. No difference. It is a well-built study: each patient served as their own control, and they ran the T3 twice, by immunoassay and by mass spectrometry, and the two agreed. That is a clean null result, it is honestly reported, and I am not going to squint at it. Two details worth having alongside it: 34 of the 50 needed a dose adjustment partway through to get there, and 17 of the 50 were thyroid cancer patients deliberately dosed to push their TSH down, which the paper states plainly.
Now here is the scope limit, and it is in their own methods sentence: each thyroid profile consisted of a serum TSH level, free thyroxine, and total T3. The free fraction for T4. The total for T3. Free T3 was never measured, and free T3 is the number this entire article is about. I want to be fair about how much that costs them, because it costs them less than I would like. Total T3 misleads mainly when binding proteins drift, and comparing a person against themselves over four months, with hormone-affecting medications excluded, is the design where that matters least. It is still not the number the argument is about.
Two more things from that paper, and both are their sentences rather than mine. Their discussion concedes that patients on levothyroxine "clearly have higher FT4 / T3 ratios than the ratios that were characteristic of their period of endogenous thyroid function," which is the paper granting that the mix is different, while its conclusion addresses only the level. And they raise, then leave open, the possibility that the high free T4 needed to normalize T3 "is associated by some unappreciated mechanism with an adverse impact on patients."
And nobody wrote down how these people felt. That is theirs too: "our study, however, is limited by the fact that we did not document patients' symptoms." They did examine them, blood pressure and pulse and reflex relaxation time, and report no differences. But fifty people had a thyroid removed, went onto levothyroxine, gave four sets of blood, and the thing the patient would actually tell you is not in the paper.
So my position is that the field overstates that study, not that the study is wrong. Its own conclusion is carefully scoped. It gets quoted as though it said combination therapy is unnecessary and everyone on levothyroxine is fine, and those are not the same sentence. Attack the citation, not the author.
Nine years later, in her 2017 review, Jonklaas revisited her own study and wrote that it "showed maintenance of T3 levels, based on the average T3 levels," and that "However, 8 of the 50 individual patients (16%) did have lower T3 levels when being treated with LT4." Eight out of fifty. And I have to fence that twice. It is a blood test and nothing else, and nobody showed those eight felt worse. And against a null that clean, a comparable number moved the other way, which she does not break out. Her own 2008 paper says as much: preoperative T3 values within individuals "can be observed to be both higher and lower than their postoperative T3 values."
The study that did measure free T3
There is a second study of the same shape that measured the number Jonklaas did not. Ito and colleagues, in the European Journal of Endocrinology in 2012: 135 consecutive patients after total thyroidectomy for papillary cancer, each one compared against their own numbers from before surgery. Free T3 fell, from a median of 3.01 to 2.92 pg/mL, P = 0.029. That is a small move on a reference range of 1.7 to 3.7, and I will say so rather than hide behind the p-value. The free T3 to free T4 ratio fell further, 3.01 to 2.17, with the caveat Ito attaches himself: postoperative bloods were drawn two to four hours after the levothyroxine dose, which lifts the free T4 side of that ratio.
The part that matters is what happened when they split those patients by where the TSH landed. Below 0.03, free T3 came out higher than their own natural level, and in six of those thirty patients it came out above the top of the reference range, which is overshoot and worth saying out loud. Between 0.03 and 0.3, free T3 matched what their own thyroid used to make, 3.06 before and 3.03 after, P = 0.51. And in the people whose TSH landed inside the normal range, which is the range every guideline tells you to aim for, free T3 sat significantly below their own baseline, 3.01 down to 2.77, P < 0.001.
They put a number on the crossover, the point where free T3 on levothyroxine equals what your own thyroid was making. It is a TSH of about 0.1, read off a fairly loose correlation, so treat it as a direction rather than a dial setting. Ito’s own lab printed a TSH reference range starting at 0.3, so that middle band sits at or below his floor. Against a US report it is further below still: Quest Diagnostics prints 0.40 as the bottom of normal for anyone 20 or older, so a TSH of 0.3 is under the range, not at the low end of it. Either way, the doses that restored these patients to their own native T3 were suppressive doses. Which is why the title of that paper reads as it does: TSH-suppressive doses of levothyroxine are required to achieve preoperative native serum T3 levels.
I have described these two papers in the past as a standoff, one this way and one that way and the field not knowing. Having read both, that is lazy, and so is the tidier version where I claim they agree. They asked different questions. Jonklaas asked whether her most-suppressed patients had higher total T3 than her normal-TSH patients, and they did not. Ito asked whether suppression restored each patient's own free T3, and below 0.03 it did. Those are not the same question and they can both be right. Ito's middle band, the one that matched native levels, sits entirely inside the bin Jonklaas called suppressed. The one printed criticism is Ito's own, of the earlier paper: their results "came from compound data from all cases, including cases in various thyroid states," and they "did not demonstrate how the serum T3 level changed from the preoperative native levels in each group." That is one peer-reviewed paper making a specific complaint about another one, in print. I am quoting it, not inventing it.
And the fences on Ito, because they are real and there are four. It is retrospective, single-hospital and entirely thyroid cancer patients, while Jonklaas is prospective. Its free T3 decline stayed inside the normal range, and statistically significant is not the same as abnormal. Ito did not measure how anybody felt either, which is exactly the fence I just put on Jonklaas and I am not going to apply it in one direction only. And the fourth is the trap: Ito's proposed direction is more levothyroxine, not adding T3, and he explicitly declines to recommend even that until somebody correlates T3 with how patients actually feel. Which lands the honest version here: in Ito's patients, restoring an individual to their own prior free T3 on T4 alone took a TSH below the reference range. Their free T3 at a normal TSH still landed inside the normal range, so this is a statement about restoring someone to their own number, not about making them abnormal. And it comes from a cancer population dosed to cancer targets, so I am not going to stretch it into a rule for routine replacement.
That same 2017 review also concludes against me, and you get that half too: trials of combination therapy with liothyronine "have failed to show clear superiority." Meanwhile the review that gives me the mechanism says, in its own sentence, that the vast majority of patients on levothyroxine with a normal TSH are asymptomatic.
So here is the whole thing in one sentence, and it is the honest one. T4 alone leaves a measurably lower T3 than you would have at the same TSH; most people appear to be fine anyway; and nobody has yet shown you which ones are not.
Which is exactly why, when the paper and the person disagree, I am going to believe the person.
The panel that should actually get drawn
Here I am going to disappoint the internet, because the answer is shorter than the panel you have probably seen sold online.
Three things. TSH. Free T4. And free T3, the one that is almost never on the order. "Free," by the way, just means unbound: the fraction actually available to your cells, as opposed to the portion locked to a carrier protein.
Since somebody is going to ask what the number should be: in my practice, optimal free T3 starts above 4.0. I want to be precise about what that is. It is a floor, not a target, and it is my own clinical practice, not a cited trial endpoint. Where you personally land above it depends on you and on how you actually feel, and that varies a lot between people. The published direction that supports looking at all: Cui and colleagues, in the International Journal of Endocrinology, found that relatively lower free T3 concentrations, even within the normal reference range, were related to impaired quality of life in levothyroxine-treated patients.
Now the two I am going to tell you not to bother with, which is probably not what you expected from me.
Reverse T3. Very popular test, and I never order it. It costs money, and it goes up simply because you are ill or under-eating. Whether that rise is protective or pathological is still genuinely argued in the literature, and Hannoush and Weiss call the adaptation-versus-pathology question "a question of great controversy." Honestly, that argument does not matter to me, because either way it has never once changed what I do. The free T3 to reverse T3 ratio: I do not use that either.
Thyroid antibodies. You can run them. They will tell you whether the cause is Hashimoto's. But they do not change what we do, there is no target to chase, and they do not reliably fall with treatment. Hashimoto's is the most common thyroid disorder in the world and the treatment is the same either way. Chasing that titer is chasing a ghost.
Natural desiccated thyroid: the oldest thyroid medicine in the world
This is the part I most wanted to write, because I think it is the clearest example of dogma we have.
Long before there was a synthetic thyroid pill, there was desiccated thyroid extract: dried, standardized porcine thyroid. You will see it today as NP Thyroid, as Armour Thyroid, and as compounded preparations. And I do not have to editorialize about its history, because Bianco's group laid it out in the Annals of Internal Medicine. Thyroid hormone replacement has been used for more than a century, and from the early 1890s through the mid-1970s, desiccated thyroid was the preferred form of therapy.
Here is the number from that paper that tells the whole story. In 1965, roughly four out of every five thyroid prescriptions in this country were for natural thyroid. Four out of five. That is not a fringe treatment somebody dug up on the internet. That was the treatment, and then, over about a generation, it became the thing you get side-eyed for asking about.
So what happened? Not a trial showing it failed. Two pieces of genuine progress, per that same paper: we invented a good TSH blood test, and we worked out that the body converts T4 into T3 in the tissues, which seemed to mean you could give T4 alone and let the body sort it out.
And then look at what the TSH test did to dosing. Before it existed, doses were adjusted to how people were doing, by symptoms and metabolic rate. Then the test arrived, everyone's TSH looked low, and the conclusion drawn was that everyone had been overtreated. Maintenance doses of levothyroxine, in that paper's own numbers, ranged from 200 to 500 micrograms a day before the TSH assay, and became typically closer to 100 to 150 micrograms a day after it. Depending on which end of each range you take, the new dose is somewhere between about a fifth and about three quarters of the old one. In one generation. Not because patients got worse on the old doses, but because a new number said they should not be on them.
Did they get worse? We know something better than that, from the same paper. Even on the new doses, with a normal TSH, basal metabolic rate remained at about 10 percent below normal controls after three months of therapy. And across 18 studies of replacement in overt hypothyroidism, every one lowered total cholesterol, but in 14 of the 18 the mean post-treatment cholesterol remained above the normal range.
The TSH is normal and the metabolism is still running 10 percent slow.
What is actually in desiccated thyroid
T4 and T3, both, together, in one tablet. Plus T1 and T2, two smaller thyroid hormones that a synthetic T4-plus-T3 combination simply does not contain.
And here is the part that matters most, and it is the same point I make in every episode of this series. Desiccated thyroid comes from a pig thyroid gland, and the hormones in it are structurally identical to the hormones your own thyroid makes. Bioidentical T4 and T3. Not a patented near-miss. Not a synthetic stand-in built to be similar enough to your hormone while being different enough to profit from.
Levothyroxine is one molecule. Desiccated thyroid is the whole symphony.
Is it actually better, or is that just nostalgia?
Fair question. Here is what the published data says, with its limits attached.
Peterson and colleagues surveyed more than 12,000 hypothyroid patients in the journal Thyroid, and the people taking desiccated thyroid reported higher median treatment satisfaction, a median of 7, than those on T4 alone or on T4 plus T3, both of which came in at 5. That was a self-selected online survey, so it measures preference and satisfaction, not head-to-head efficacy.
Pepper and Casanova-Romero, in the Journal of Endocrinology, Diabetes and Obesity, looked specifically at hypothyroid patients who still had persistent complaints on T4 alone. 78 percent of them preferred desiccated, with improvement in symptoms and no serious adverse events. That is a conversion study, not a blinded trial.
And now the part that I think explains the whole thing. You have probably heard that the trials showed adding T3 does not help. Go look at how those trials were built. The T4-to-T3 ratios they used ranged from 20 to 1 all the way down to 1 to 1. Hold that against the real number: a human thyroid gland secretes T4 and T3 at a ratio of about 13 to 1 by weight. So those trials were scattered on both sides of physiology and wildly different from each other, using T3 doses that in practice are close to worthless.
Duntas and Wartofsky, in Metabolism, looked at that pile of studies and concluded it would be, their word, "tenuous" to draw any conclusion about combination therapy from them at all. That is a very polite way of saying the experiment did not test the thing everyone quotes it for.
My own read, and I am marking this as opinion rather than a sourced finding: the denigration of desiccated thyroid has been driven more by economics and politics than by data.
The honest caveats
It is porcine, which is a real objection for some people on religious grounds, and there is a synthetic compounded equivalent for exactly that reason.
The old reliability complaint, that the potency varies, was fair, and it was fixed 40 years ago. In 1985 the US Pharmacopeia changed the standard from measuring iodine content to measuring actual T3 and T4 content, and the result was stable potency (McAninch and Bianco, Annals of Internal Medicine; the modern HPLC assay under the USP standard is described in Hoang and colleagues, JCEM, 2013). But the reputation never recovered, and I still spend appointments undoing that.
The "you cannot use it in Hashimoto's, the pig has antibodies" objection: I would gently point out that Hashimoto's is a human autoimmune condition. And more to the point, I have never seen a study showing that desiccated thyroid worsens Hashimoto's or drives antibodies up, and I have looked. If somebody has that paper, I would genuinely like to read it.
And it is not automatically the right answer for everyone. This is a conversation to have with a clinician who knows the landscape, not a thing to demand by brand name.
I will also give you the line from that same Annals paper that cuts against me, because it belongs here: neither desiccated thyroid nor levothyroxine monotherapy recreates a biochemical state of euthyroidism as defined by the serum T4 to T3 ratio. Neither one perfectly reproduces what your gland does. I say desiccated gets closer. I do not say it is a copy, because the author I am citing says it is not.
What "optimal" looks like on paper, and why it can look alarming
If this gets addressed and nobody warns you, your next lab report is going to frighten you for no reason. So here is the warning.
Your free T3 comes up, and it may come up past the top of the printed range. This next part is the pattern I see in my own practice rather than a published trial result, and I am flagging it as such: on desiccated thyroid, circulating T3 rises and may transiently exceed the upper limit of normal, while the average T4 settles below the bottom of its range, and the TSH falls. So you can get a report with an H next to your free T3 and an L next to your free T4, and both of those can be exactly what we were aiming for. You still want T4 in reserve. It is the gas tank. It just is not the number we are steering by.
And your TSH is going to drop. Possibly a lot. Go back to the thermostat. TSH is the request. Once there is finally enough thyroid hormone on board, your pituitary stops asking. A low TSH on replacement is not a sign you are toxic. It is a thermostat that finally got what it ordered.
There is one situation where a very low TSH genuinely matters, and it is the mirror image of this one: someone who is not on any thyroid therapy, whose TSH is bottomed out and whose free T4 is climbing. That is a screen for Graves' disease. It is roughly one in 500, and it is essentially the reason we still draw a TSH at all.
The two fears attached to a low TSH
Atrial fibrillation and osteoporosis. I am going to be direct, because this one does not deserve a soft answer.
Those risks are real in endogenous subclinical hyperthyroidism, meaning a low TSH driven by your own gland rather than by a tablet. That distinction is not something I am asserting. It is what the study found.
Belaya and colleagues, in Hormones (Athens) in 2007, measured bone density in 66 postmenopausal women with a low TSH, sorted by cause, against 22 healthy controls. Two of those groups had a low TSH from their own thyroid. The women with Graves' disease, who had been on antithyroid drugs for a mean of about three years and whose free T3 and free T4 had been normal for at least six months, had lower bone density than controls at the total hip, P = 0.013, and at the radius, P = 0.0003. And the women with a toxic multinodular goiter, which is not autoimmune at all, had the worst bones in the study, significantly lower than controls at every region of the skeleton measured. So the mechanism is not "autoimmunity does the damage," and I am not going to claim it is. Belaya's own conclusion names both: endogenous subclinical hyperthyroidism, "both of autoimmune and non autoimmune etiology," might be considered an additional risk factor for osteoporosis.
Now the group in that same study whose TSH was suppressed because they were taking levothyroxine after thyroid surgery, at a median TSH of 0.06: bone density no different from the controls, at any site. The paper's own conclusion, in its own words, is that "exogenous subclinical hyperthyroidism has no effect on BMD." That is her term, not mine. A suppressed TSH on thyroid hormone is not hyperthyroidism, and her own data is the reason: that group's bones were fine.
The line that holds, then, is that where the low TSH came from matters more than how low it is. And I want to be square about this one study's limits rather than leave them for somebody else's rebuttal. It is cross-sectional and observational, and it is 25 treated women against 22 controls, so read it as a null rather than a proof of safety. But two things people reach for as caveats here actually turn the other way once you read Belaya's own discussion. She names the two older meta-analyses, from 1994 and 1996, that did find bone loss on suppressive levothyroxine, and then in her very next sentence she sets against them the two more recent systematic reviews, which found the bone-density changes reflected thyroid hormone's normal physiological influence on bone rather than pathological loss. And the treated group's bone turnover markers were up, osteocalcin 28 percent above controls at P = 0.005 and CTx 36.2 percent above at P = 0.050, while their density was not. That is not a mark against the treated women. Turnover was up in all three low-TSH groups; bone was lost only in the two whose low TSH came from their own gland. The same signal in every group, bone loss only where there was underlying disease, is the control that makes the point rather than a hole in it.
The atrial fibrillation warning traces back to a single study: Sawin and colleagues, in the New England Journal of Medicine in 1994, a community cohort of people over 60 who had a low TSH from their own glands. Here is the move worth watching. When the thyroid-cancer guidelines wanted to warn that a suppressed TSH on thyroid hormone is dangerous, the study they reached for was this one, in which the subjects were not on thyroid hormone at all. That is the citation Kelly audited in 2015: a study of an overactive gland, used to warn about a tablet. And the number runs with that reading, not against it. Take out the small share of Sawin's people who were on thyroid hormone and the atrial-fibrillation signal gets stronger, because the signal belongs to the endogenous disease. For the record the proportion not on hormone is 94.3 percent, not the ">96 percent" you will sometimes see, because 115 of the 2,007 were taking a preparation. The decimal is not the point. A warning about prescriptions was built on a study of people who were not taking one. The strongest version of this point sidesteps the argument entirely: the two cardiac-risk meta-analyses Biondi rated most credible excluded anyone on thyroid medication by design, so they describe the endogenous disease and say nothing about people on a prescription.
What matters more than that percentage is what the study was and what it actually reported. It is a community cohort of people aged 60 and over, with one TSH drawn once between 1978 and 1980 and ten years of follow-up. In it, the single total thyroxine measurement did not predict atrial fibrillation: P = 0.71 adjusted for age, P = 0.60 adjusted for age and risk factors. Of the 192 people who developed atrial fibrillation, only 2 also had spontaneous hyperthyroidism. Sawin attributes that thyroxine null to the wide spread of thyroxine values in a general population and to variation in cardiac sensitivity, so it is a statement about what those two tests predicted in that cohort, not a demonstration that thyroid hormone is irrelevant to the heart.
And Sawin's own closing clinical sentence belongs here, because it is more useful than the warning built on top of it. Among people receiving thyroid hormone who have low serum thyrotropin concentrations, he writes, "the risk of atrial fibrillation can be lessened by avoiding excessively high doses." That is not "a suppressed TSH is nothing." It is dose discipline, which is exactly what I am going to say two paragraphs from now.
Something similar happens with bone. Toft 2001, which gets cited in support of the osteoporosis warning, contains this conclusion of its own: "The evidence that exogenous thyroid is a risk factor for osteoporosis is therefore inconclusive." The source says inconclusive, and it keeps getting cited for a warning. (That quotation reaches me through Kelly's audit; I have not read Toft in the original, and I am flagging that rather than dressing it up.)
And the largest dataset anyone has agrees. 162,369 hypothyroid patients in Britain, over 863,000 TSH measurements. At the lowest TSH levels: no increase in atrial fibrillation, no increase in fractures, and heart failure actually came out lower, at a hazard ratio of 0.79. What they did find harm from was the other direction entirely, the undertreated thyroid, TSH above 10, which is where the fragility fractures showed up (Thayakaran and colleagues, BMJ, 2019).
There is one number in that paper that does not go my way, and I would rather put it on the table myself than let somebody else find it for me. In the same cohort, all-cause mortality was higher below a TSH of 0.1, at a hazard ratio of 1.18. So there it is. And now let's actually look at it, because the moment you do, it stops holding together.
Same patients. Same TSH. Fewer heart failures. No excess atrial fibrillation. No excess fractures. Every specific way a low TSH is supposed to hurt you that this database could capture came back null, or came back in the other direction. Bone density itself was never measured, and I will say that for them because they said it themselves: osteoporosis is poorly recorded in primary care, so fragility fracture stood in for it. What they could count, they counted, and it came back clean. And then the death count went up. From what? The paper cannot tell you. Cause of death is not recorded in the database they used, and again, the authors say so themselves. So what we actually have is an eighteen percent higher hazard of death, cause unknown, in a group whose hearts and fracture rates were measurably fine.
That is not a safety signal. That is a loose end.
And the authors leave the likeliest explanation sitting in their own limitations section: "the possibility of reverse causality also exists, whereby the last TSH measurement close to death may have been aberrant." That sentence sits in their general limitations, directly after one about the deaths at the high end, so reading it onto the low end is my inference rather than their claim, and I want to be square about that. But we are not guessing at the mechanism, because it is one of the better-documented findings in cardiology. When people get seriously ill, the thyroid axis drops. Iervasi and colleagues, in Circulation in 2003, followed 573 consecutive cardiac patients for a year: the ones with a low free T3 died at 14.4 percent versus 3 percent, and free T3 was the strongest independent predictor of all-cause death in their model. And here is the part that matters for our purposes: in that same population, a low TSH did not separate the survivors from the rest. Survival was 86.6 percent against 90.4 percent, not statistically different. That comparison rests on 57 patients and their threshold for "low" was 0.3 rather than 0.1, so take it as a weak null rather than a proof. The direction is still the direction.
Read those two papers next to each other and the picture is not subtle. A suppressed TSH in a dying patient and a suppressed TSH in an optimized patient look identical on a printout. A database cannot tell them apart. A doctor sitting across from the person can.
The authors' own numbers point the same way, incidentally. That low-TSH mortality signal did hold in both men and women, so I am not going to tell you it evaporated. But when they split their cohort by age, it stopped being statistically significant, while the signal at high TSH held up in every subgroup they looked at.
Which brings us to the finding in that paper that is not ambiguous at all, and is a great deal bigger. Above a TSH of 10, undertreated, the hazard ratio for death was 2.21. Not eighteen percent higher. A hundred and twenty-one percent. Between 4 and 10 it was still 1.29. And in that same dataset, 32.4 percent of all the annual TSH measurements sat above 4. Nearly a third of the readings, across 162,000 treated patients, were above the range the guidelines they cite recommend. That is the emergency in this paper. Nobody is holding a press conference about it.
The rule underneath all of it is the same rule as this whole article, pointed the other way. If your labs look "abnormal" and you feel well, that is usually not a problem to correct. If your labs look "normal" and you feel terrible, that is.
Common questions
What is a normal TSH range for a woman?
Ranges vary by lab and by assay. Quest Diagnostics prints 0.40 to 4.50 mIU/L for anyone aged 20 or over, and other labs print something close but not identical, so read the range on your own report rather than one you found online. It matters more than it sounds: a TSH of 0.3 is below a range that starts at 0.40, not at the low end of it, and that distinction is the difference between a suppressed number and a normal one. The more useful point is the one this whole article makes: that range is a description of the population that got tested, not a target for you, and in the analysis of roughly 1,900 associations described above, TSH tracked with actual clinical findings 23 percent of the time. A TSH inside the range does not tell you your thyroid hormone level is adequate for you.
What does "TSH with reflex to FT4" mean on my lab slip?
A reflex order means the lab runs your TSH first and only adds the free T4 if the TSH falls outside its range. The consequence worth knowing is that if your TSH lands inside the range, the reflex never fires, so no thyroid hormone gets measured at all. And free T3 is not part of that reflex. If you want free T3, it has to be ordered by name.
What are normal thyroid levels?
Every lab prints its own reference ranges for TSH, free T4 and free T3, and they differ by assay, so the numbers on your report are the ones that apply to your report. What I would push back on is the question. "Normal" is where the middle of a tested population landed. It is not a health target, and free T3 in particular declines with age, which drags the bottom of the range down with it.
Is a high TSH with a normal T4 a problem?
That combination is what gets labeled subclinical hypothyroidism, and the standard answer is to watch and wait. My answer is that the label describes two numbers and says nothing about the third one, free T3, which is usually not on the order. If you have symptoms that fit and nobody has measured the hormone that does the work, the picture is incomplete rather than reassuring.
Is NP Thyroid the same as Armour Thyroid?
Both are desiccated thyroid extract, made from porcine thyroid, containing T4 and T3 together. They are different products from different manufacturers, and whether one can be substituted for another is a conversation to have with your own clinician and pharmacist rather than a swap to make on your own.
Can you take desiccated thyroid if you have Hashimoto's?
Hashimoto's is a human autoimmune condition, and the concern that porcine tissue drives antibodies up is a claim I have never seen a study support, and I have looked. Hashimoto's is the most common thyroid disorder in the world, and the treatment approach is the same either way. Treating desiccated thyroid as off-limits in Hashimoto's means treating a myth instead of a patient.
Should I ask for free T3?
Yes, by name, because it will not appear otherwise. That is the single most useful change most people can make to a thyroid workup: TSH, free T4 and free T3, ordered together.
Watch and listen
The full breakdown is Episode 5 of Exposing Outdated Dogma, and it runs across two formats this week.
Watch: "Your Thyroid Labs Are 'Normal.' You're Not." on YouTube, the complete visual walkthrough of everything above, including the physiology.
Listen: "The Hormone Your Doctor Is Afraid Of," a three-part conversation on This Isn't Your Grandma's Health Podcast, which goes considerably deeper than the video does.
Part One takes the safety question apart: the atrial fibrillation and bone-loss warnings, traced back to the studies the guidelines actually cite.
Part Two covers how the dose collapsed in a single generation when the TSH assay arrived, what your position inside a "normal" range does to your metabolism and your brain, and the TRUST trial in full.
Part Three covers the mechanism underneath all of it: mitochondria, the brain's own T3 supply, the psychiatric presentations of low thyroid, PMOS (formerly PCOS), what I actually see in practice, and what to do when your own provider is unhappy about your number.
Between the video and the three podcast parts, everything in this article gets discussed at length.
The bottom line
Come back to those four words at the top. Nothing about that situation is mysterious. Symptoms with a recognizable shape, and one number drawn, from the pituitary and not the thyroid, landing inside a range built from a crowd of other people. And that was the end of the inquiry.
Nobody said out loud that the hormone doing the actual work was never measured. That is not a failure of anyone's intelligence. It is what happens when a shortcut gets taught for long enough that everyone forgets it was ever a shortcut.
So if you have been told your thyroid is normal and you do not feel normal, you are not imagining it and you are not being difficult. You may simply have been measured with one test that was never designed to answer the question you were asking.
Go get the full picture. Ask for free T3 by name.
Because "normal" is a description of a crowd. You are not a crowd.
Optimal, not normal.
Want to talk about your own labs with someone who will run the whole panel? Start here: withinyou.health/links
Luke Swift, DNP · Within You Therapeutics. Educational content, not individualized medical advice. Nothing here diagnoses, treats or replaces care from your own clinician.
References
Bekkering GE, Agoritsas T, Lytvyn L, et al. Thyroid hormones treatment for subclinical hypothyroidism: a clinical practice guideline. BMJ. 2019;365:l2006.
Belaya ZE, Melnichenko GA, Rozhinskaya LY, et al. Subclinical hyperthyroidism of variable etiology and its influence on bone in postmenopausal women. Hormones (Athens). 2007;6(1):62-70.
Cui, et al. Int J Endocrinol. 2022;2022:1918674.
Duntas LH, Wartofsky L. Metabolism. 2016;65(4):428-431.
Ettleson MD, Bianco AC. Individualized therapy for hypothyroidism: is T4 enough for everyone? J Clin Endocrinol Metab. 2020;105(9):e3090-e3104.
Fitzgerald SP, Bean NG, Falhammar H, Tuke J. Clinical parameters are more likely to be associated with thyroid hormone levels than with thyrotropin levels: a systematic review and meta-analysis. Thyroid. 2020;30(12):1683-1695.
Fitzgerald SP, Falhammar H. Redefinition of successful treatment of patients with hypothyroidism. Is TSH the best biomarker of euthyroidism? Front Endocrinol. 2022;13:920854.
Gullo D, Latina A, Frasca F, Le Moli R, Pellegriti G, Vigneri R. Levothyroxine monotherapy cannot guarantee euthyroidism in all athyreotic patients. PLoS ONE. 2011;6(8):e22552.
Hannoush ZC, Weiss RE. Rambam Maimonides Med J. 2016;7(1):e0002.
Hoang TD, Olsen CH, Mai VQ, Clyde PW, Shakir MKM. Desiccated thyroid extract compared with levothyroxine in the treatment of hypothyroidism: a randomized, double-blind, crossover study. J Clin Endocrinol Metab. 2013;98(5):1982-1990.
Iervasi G, Pingitore A, Landi P, et al. Low-T3 syndrome: a strong prognostic predictor of death in patients with heart disease. Circulation. 2003;107(5):708-713.
Jonklaas J. Persistent hypothyroid symptoms in a patient with a normal thyroid stimulating hormone level. Curr Opin Endocrinol Diabetes Obes. 2017;24(5):356-363.
Ito M, Miyauchi A, Morita S, et al. TSH-suppressive doses of levothyroxine are required to achieve preoperative native serum triiodothyronine levels in patients who have undergone total thyroidectomy. Eur J Endocrinol. 2012;167(3):373-378.
Jonklaas J, Davidson B, Bhagat S, Soldin SJ. Triiodothyronine levels in athyreotic individuals during levothyroxine therapy. JAMA. 2008;299(7):769-777. (The standard rebuttal, quoted above from the original. Its T3 measure is total T3; free T3 was not measured.)
Kelly T. An examination of myth: a favorable cardiovascular risk-benefit analysis of high-dose thyroid for affective disorders. J Affect Disord. 2015;177:49-58. (Carries the Toft 2001 quotation above; that primary is not independently held.)
Ling C, … Soldin SJ. Ann Thyroid Res. 2018;4(1):122-125.
McAninch EA, Bianco AC. The history and future of treatment of hypothyroidism. Ann Intern Med. 2016;164(1):50-56.
Pepper GM, Casanova-Romero PY. Conversion to Armour Thyroid from levothyroxine improved patient satisfaction in the treatment of hypothyroidism. J Endocrinol Diabetes Obes. 2014;2(3):1055.
Peterson SJ, Cappola AR, Castro MR, et al. An online survey of hypothyroid patients demonstrates prominent dissatisfaction. Thyroid. 2018;28(6):707-721.
Sawin CT, Geller A, Wolf PA, et al. Low serum thyrotropin concentrations as a risk factor for atrial fibrillation in older persons. N Engl J Med. 1994;331(19):1249-1252.
Stott DJ, Rodondi N, Kearney PM, et al; TRUST Study Group. Thyroid hormone therapy for older adults with subclinical hypothyroidism. N Engl J Med. 2017;376(26):2534-2544.
Thayakaran R, Adderley NJ, Sainsbury C, et al. Thyroid replacement therapy, thyroid stimulating hormone concentrations, and long term health outcomes in patients with hypothyroidism: longitudinal study. BMJ. 2019;366:l4892.
van den Beld AW, Visser TJ, Feelders RA, Grobbee DE, Lamberts SWJ. Thyroid hormone concentrations, disease, physical function, and mortality in elderly men. J Clin Endocrinol Metab. 2005;90(12):6403-6409.
Werneck de Castro JP, Fonseca TL, Ueta CB, et al. J Clin Invest. 2015;125(2):769-781. (Rodent and cell work; it establishes the D2 mechanism, not a human clinical outcome.)
