The Fluoride Decision

Fluoride is the most-swallowed drug never prescribed to anyone. This episode follows it from a brown stain on children's teeth in 1901 Colorado, through a trial whose control city fluoridated itself halfway through, to the Cochrane review that put the modern benefit at a quarter of one tooth — and to the federal court that found an unreasonable risk of lost IQ, and the appeals court that vacated the finding without touching the science.

Show notes

Colorado Springs, 1901. A young dentist named Frederick McKay arrives from Philadelphia, opens a practice, and finds something he cannot explain. A large proportion of the people born and raised in that town have teeth covered in brown stains. Not decay — stains. Chocolate-coloured patches, sometimes chalky white flecks, sometimes an enamel so mottled it looks like someone spattered it with paint. The locals call it Colorado brown stain and treat it as a fact of life, the way you might treat hard water.

McKay spends the next thirty years on it. He recruits G.V. Black, the most famous dentist in America, and together they establish two things. The first is that the stain is in the enamel itself, laid down while the tooth was forming under the gum, which means it happens in childhood and is permanent by the time the tooth erupts. The second is stranger, and it is the reason we are still having this conversation a century later.

The stained teeth had less decay.

That observation — a disfigurement that came with a benefit — is the seed of every argument about fluoride that has happened since. It took until 1931 for anyone to name the cause, and the person who named it was not a dentist. He was H.V. Churchill, the chief chemist at the Aluminum Company of America, and he was investigating because aluminium smelting produces fluoride waste and his employer was being blamed for something. He ran the water from the affected towns and found fluoride at concentrations far above what anyone expected.

Hold onto that detail. The first person to connect fluoride to human teeth was working for a company with a fluoride disposal problem. That is not a scandal and it does not make the chemistry wrong. It is simply the first instance of something that recurs in this story for the next ninety-five years: almost everybody who has funded a study of fluoride has had a stake in the answer, on both sides.

Let us establish what fluoride actually is, because almost nobody is told, and the chemistry explains most of the argument.

Fluorine is an element — one of the entries on the periodic table, the chart of everything the universe is made of. It sits in the column chemists call the halogens, which is a family of elements that all share one feature: each of them is exactly one electron short of a full outer shell. An atom with a full outer shell is content. An atom one electron short is not, and it will take that electron from almost anything.

Fluorine is the most extreme case in the entire periodic table. It grabs electrons harder than any other element in nature. Chemists measure this with a property called electronegativity — literally, how greedy an atom is for electrons — and fluorine holds the record. Nothing beats it. Which is why you will never find pure fluorine lying around: it reacted with the first thing it touched, billions of years ago, and it is still holding on.

Fluoride is fluorine after it has won. It has taken its electron, it now carries a single negative charge, and it has calmed down. It is a small, stable, negatively charged particle — an ion — and it is found naturally in rock, in soil, in groundwater, in tea leaves, and in the ocean. Every human being who has ever lived has had fluoride in their body. The question was never whether to be exposed to it. The question has always been how much, from where, and who decides.

Now the part that matters for teeth. The hard outer shell of your tooth is enamel, and enamel is the hardest substance your body makes. It is a crystal, and the crystal has a name: hydroxyapatite. Break that word in half and it stops being intimidating. Apatite is a family of calcium phosphate minerals — calcium and phosphate are the two things bones and teeth are built from. Hydroxy means it carries a hydroxide group, which is one oxygen and one hydrogen bonded together, tucked into the crystal at regular intervals.

Here is the trick. That hydroxide group sits in a slot in the crystal. And fluoride — small, negatively charged, and desperate to be somewhere stable — fits that slot. When fluoride is present in the fluid around your teeth, it swaps in. Hydroxyapatite becomes fluorapatite.

Why does that matter? Because tooth decay is, chemically, a dissolving problem. Bacteria in your mouth eat sugar and excrete acid. Acid dissolves the enamel crystal. Saliva then tries to put the minerals back. Decay happens when the dissolving outruns the rebuilding. And fluorapatite dissolves at a lower pH than hydroxyapatite does — it holds on in acid that would have taken the original crystal apart. Fluoride does not make the tooth harder in the sense of tougher. It makes the tooth less soluble.

Notice what that mechanism requires. It requires fluoride to be present in the fluid at the surface of the tooth, at the moment the acid is there. It is a surface chemistry event, happening in your mouth, in a film of saliva a fraction of a millimetre thick.

Keep that in mind for the next forty minutes, because the entire public health apparatus of fluoridation is built on delivering fluoride through the bloodstream to achieve an effect that happens on a surface. Philippe Grandjean — a Harvard environmental health professor we will meet properly later — put the question plainly in a 2019 review: topical application in the mouth appears to be a more direct and appropriate way of preventing decay, so why is the delivery system systemic? 1 The dental literature has answers. They are not as strong as you would expect for a question this obvious.

Back to the history, in order, because the order is the argument.

After Churchill identifies fluoride in 1931, the US Public Health Service hands the problem to a dental officer named H. Trendley Dean. Dean does something genuinely impressive: he builds a numerical scale for the severity of the mottling, then travels the country measuring both the mottling and the decay rate in city after city. By the end of the 1930s he has data from twenty-one cities, and the shape of it is unmistakable. As fluoride in the water goes up, decay goes down. And as fluoride goes up, mottling goes up too.

Two curves, moving in the same direction, one good and one bad. Dean's job became finding the crossing point — the concentration at which you get most of the benefit and, in his judgement, an acceptable amount of the damage. He landed on roughly one part per million.

Let me be precise about what one part per million means, because it sounds like nothing. One part per million is one milligram of fluoride in one litre of water. If you drink two litres a day, that is two milligrams a day from water alone. Over a childhood, that is grams. Fluoride accumulates in bone. It is not passing through.

And be precise about something else. Dean's 1 ppm was not the concentration at which mottling stopped. It was the concentration at which Dean judged the mottling acceptable. At that level he was still finding visible enamel changes in roughly one child in ten. That was the deal from the beginning: a cosmetic cost, deliberately accepted, in exchange for fewer cavities. Everything since has been an argument about whether the cost is bigger than Dean thought and whether the benefit is smaller.

One more thing about the chemistry before we leave it, because it changes how you should hear every number that follows.

Fluoride you swallow does not pass through. It is absorbed rapidly and almost completely from the gut, and then it leaves the blood by two routes: the kidneys excrete some, and calcified tissue takes up the rest. The standard reference on this — written by a dental researcher, in a dental journal — puts it starkly: about 99% of the body burden of fluoride is in calcified tissue, and most of it is not exchangeable.2 Bone is built from the same apatite crystal as enamel, fluoride fits the same slot, and once it is in, it largely stays.

Three consequences follow, and all three matter later.

Children accumulate more. Clearance of fluoride from plasma into the skeleton is inversely related to how far skeletal development has progressed — a growing skeleton is an open account.

Kidney function decides your dose. Renal clearance of fluoride falls and tissue levels rise when the filtration rate is chronically depressed.2 Which means the people least able to handle the exposure accumulate the most, and nobody adjusts a water supply per person. You cannot write a lower prescription for a patient with stage three kidney disease when the prescription is the tap.

And the number everyone argues about may not measure what they think. That same review, in 1994, warned that the concentration of fluoride in drinking water "may not be a reliable indicator of previous intake", because of dental products, bottled water, home filtration, and the halo effect — foods and drinks manufactured in fluoridated cities and sold everywhere.2 Hold onto that. Thirty years later it becomes the hinge of the entire IQ argument.

January 25, 1945. Grand Rapids, Michigan becomes the first city in the world to deliberately add fluoride to its public water supply. It is designed as a fifteen-year experiment. Grand Rapids gets fluoride; Muskegon, another Michigan city, serves as the control and does not.

Here is what happened to that experiment, and it is the single most underreported fact in this entire field. The control city fluoridated itself in 1951, six years into a fifteen-year trial. There was no longer a control. And the Public Health Service endorsed fluoridation nationally in 1950 — five years into the trial, one year before the control disappeared, and a full decade before the study was meant to report.

I want to be careful here, because this is exactly the kind of fact that gets weaponised. It does not mean fluoridation is a fraud. It means the foundational evidence for the largest sustained population intervention in American history was an uncontrolled before-and-after comparison in one city, endorsed before it finished. That is a real limitation, it is a matter of record, and if a pharmaceutical company proposed it today the FDA would not accept it.

The policy hardened anyway. In 1962 the Public Health Service set a recommended range of 0.7 to 1.2 parts per million, varying by climate — warmer places got less, on the reasoning that people there drink more water. That range stood for fifty-three years. In 2015 it was replaced with a single national figure of 0.7 parts per million, and the stated reason for lowering it was that too many American children were developing dental fluorosis, because fluoride was now arriving from toothpaste, mouthwash, processed food and beverages as well as the tap.

Sit with that for a second. The reason the recommendation went down in 2015 was an acknowledgement that the 1962 dose had become too much. Not because the water changed — because everything else did.

There is one more reason this argument has gone so badly for so long, and it is not a scientific reason.

Fluoridation arrived in 1945 and became a national programme in the early 1950s — which is to say, it became a mass government intervention in the domestic water supply at the precise historical moment when mass government intervention was the most politically loaded idea in America. The opposition that formed around it was not primarily a scientific opposition. It was a Cold War opposition, and some of it was genuinely unhinged.

The effect of that has been permanent and it has damaged both sides. Public health officials learned, correctly, that fluoridation opponents included people who believed it was a communist plot, and they generalised — so that for seventy years any question about fluoride was heard as a signal of who you were rather than as a question. And opponents learned that they would be dismissed regardless of the quality of their evidence, which selected hard for people willing to be dismissed.

The result is a field where almost nobody is a neutral party, where the Chinese literature sat untranslated for two decades because nobody serious wanted to be caught reading it, and where a 2024 federal court finding was reported as vindication by people who had not read the ruling and as a fringe event by people who had not read the evidence.

I am telling you this because it is the reason I have given you every number with its confidence interval and every paper with its address. There is no trusted referee here. You are going to have to do the arithmetic yourself, so I am going to show you all of it.

So: does it work?

In October 2024, Cochrane published the most thorough review ever conducted of community water fluoridation. 3 Cochrane is a global network of researchers that exists to do one thing — pool all the evidence on a question according to rules set in advance, so that nobody can cherry-pick. When Cochrane speaks on a clinical question, it is as close to a neutral referee as medicine has.

They found 157 studies. Every single one was non-randomised, which is not anyone's fault — you cannot randomly assign families to water supplies. But it means every study was downgraded for risk of bias, and it means the strongest claim available in either direction is association, not proof.

Then they did something the coverage almost entirely missed: they split the evidence at 1975, the year fluoride toothpaste became widespread. Because the question is not "does fluoride prevent decay" — it plainly does. The question is whether putting it in the water adds anything on top of brushing with it twice a day.

Here is the contemporary answer, the post-1975 answer, the answer that applies to a child alive now. Starting community water fluoridation changed decayed, missing and filled baby teeth by 0.24 teeth. Roughly a quarter of one tooth. And the confidence interval — the range of values the data is compatible with — runs from minus 0.03 to plus 0.52. 3

A confidence interval is the honest version of a result. It says: given what we measured, the true answer is probably somewhere in this range. When that range crosses zero, it means the data cannot rule out that the effect is nothing at all. This one crosses zero.

The proportion of children with no decay at all: four percentage points better in baby teeth, three points better in adult teeth. Both confidence intervals cross zero. Cochrane graded all of it low certainty.

Two more findings from that review, both of which deserve more attention than they got.

There is no data on adults. Not weak data. None. Cochrane writes that they found no eligible study reporting decay outcomes in adults. Seventy-nine years of the largest public health intervention in the country, and the outcome literature stops at childhood.

And nobody knows what happens when a city stops. One study, on cessation, graded very low certainty. Which is awkward, because as of 2026 a lot of cities have stopped.

Now the cost side, from the same review. At 0.7 parts per million, roughly 12% of people have dental fluorosis of aesthetic concern, with a confidence interval of 8 to 17%, drawn from 40 studies and 59,630 people. And roughly 40% have dental fluorosis of some degree — 90 studies, 180,530 people. 3

Dental fluorosis is what Frederick McKay found in Colorado Springs in 1901. It is fluoride disrupting the cells that lay down enamel while the tooth is still forming, before it ever erupts. Mild cases are faint white flecks that only a dentist notices. Moderate cases are visible white or brown patches. Severe cases are pitted, crumbling enamel. It is permanent, and it is the only harm from fluoridation that literally everyone in this argument agrees on.

So the trade, stated as plainly as the evidence allows: a quarter of a tooth, low certainty, confidence interval including zero — against a two-in-five chance of marking a child's enamel and a one-in-eight chance of marking it visibly enough to matter to them.

That may still be a good trade for a population with no dental access and a high decay burden. It is a harder trade for a suburban child who brushes twice a day with fluoride toothpaste and sees a dentist every six months. And notice what kind of question that is. It is not a scientific question. It is a distributional one — who bears the cost, who gets the benefit — and it is being answered by people who mostly talk about it as if it were settled science.

Topical or swallowed — the question the whole thing rests on

I want to come back to the mechanism, because there is a version of this argument that has nothing to do with IQ and it is the one I find hardest to answer.

Fluoride's anti-decay action happens at the enamel surface. Everyone agrees on this — it is in the pro-fluoridation reviews as plainly as anywhere else. A 2025 synthesis of the whole field describes the mechanisms as enamel remineralisation, inhibition of demineralisation, and direct antibacterial activity against the organisms that make the acid.4 All three of those happen in the mouth.

The same review lists the delivery forms with clinical evidence behind them: toothpaste, varnish applied by a dentist, mouth rinse, supplements, and silver diamine fluoride painted directly onto a cavity to arrest it. Four of those five are topical. The one that is not — supplements — has the thinnest evidence of the set.

So what is the water doing?

The historical answer was that fluoride swallowed during childhood gets incorporated into the enamel of teeth that have not erupted yet, making them inherently more resistant. That is called the pre-eruptive effect, and it is real — it is exactly what causes dental fluorosis, which is the same process overshooting.

The modern answer, which the dental literature itself now gives, is different: the main benefit of fluoridated water is topical after all. You drink it, it bathes your teeth on the way past, and more importantly it raises the fluoride concentration in your saliva slightly and continuously, so the remineralisation side of the equation runs a little faster all day.

Which is a real mechanism. And it means the case for fluoridated water is a case for a low-dose, all-day topical delivery system that happens to also pass through your kidneys, your thyroid, your bones and your placenta on the way.

Grandjean's framing of this is the sharpest I have read: topical fluoride appears to be a more direct and appropriate means of preventing caries, and systemic uptake is the part suspected of causing harm.1 You do not have to accept his conclusions about neurotoxicity to notice that if the benefit is topical, the systemic exposure is a side effect rather than the point.

The other question

Everything so far has been about teeth. The argument that has actually moved governments in the last two years is not about teeth.

It starts, as these things often do, in a place nobody was looking. Through the 1980s and 1990s, researchers in rural China — in regions where groundwater naturally carries fluoride at four, six, eight parts per million, many times anything any government has ever added — began publishing small studies comparing children's intelligence test scores between high-fluoride and low-fluoride villages. The studies were published in Chinese, in Chinese journals, and for twenty years almost nobody outside China read them.

In 2008 a Chinese team pooled sixteen of them. Children in high-fluoride areas scored about five IQ points lower.5 The paper appeared in a small journal and made no impression whatsoever.

Then in 2012, Anna Choi and Philippe Grandjean at the Harvard School of Public Health did the thing that changed everything: they searched the Chinese literature properly, found twenty-seven eligible studies, translated them, and ran a formal meta-analysis in a mainstream Western journal.6

A meta-analysis is not a new study. It is a statistical method for combining the results of studies that already exist, weighting each one by how much information it carries, to produce a single pooled estimate. Its great strength is that it sees patterns no single study is large enough to show. Its great weakness is that it inherits every flaw of every study it swallows.

Their pooled result was a standardised mean difference of minus 0.45.

That number needs unpacking. A standardised mean difference expresses a gap between two groups in units of the spread of the data itself, so that results measured on different tests can be compared. Minus 0.45 means the high-fluoride children scored, on average, a bit under half of one standard deviation below the others. On a standard IQ scale, where the spread is fifteen points, that is roughly six or seven points.

Six or seven IQ points in an individual child is nothing much — it is within the noise of having a cold on test day. Six or seven points shifted across an entire population is an enormous public health event. Move the whole distribution left by half a standard deviation and you roughly double the number of people below any given threshold, and you lose a large fraction of the people at the top. That asymmetry — trivial for a person, serious for a population — is the single most misunderstood idea in this argument, and both sides abuse it.

Now the honest problems with that 2012 paper, which Choi and Grandjean themselves named and almost none of the people who shared it did.

These were villages in endemic fluorosis regions being compared with other villages. Not fluoridated cities compared with unfluoridated cities. The exposures were four to eight parts per million and sometimes higher — many times the 0.7 that American water carries. The studies were cross-sectional, which means everything was measured at one moment in time, so you cannot tell which came first. And nobody had adequately accounted for the thousand other ways a high-fluoride village in rural China differs from a low-fluoride one: iodine in the soil, arsenic in the water, lead, coal smoke, nutrition, family income, how far the school is.

The word for that problem is confounding — when the thing you measured travels with something else you did not measure, and you cannot tell which one did the damage. It is the oldest problem in epidemiology and it has never been solved by finding more studies with the same flaw.

Which is why what happened next mattered more than the meta-analysis.

Three prospective cohort studies were run — two in Mexico, one in Canada — and they were a different animal entirely. A prospective cohort follows people forward in time. You measure the exposure first, before anybody knows the outcome, and then you wait. In these studies the measurement was maternal urinary fluoride during pregnancy — fluoride in the mother's urine, which reflects what actually got into her body from all sources, rather than what the tap was rated at.

That matters enormously, because it removes two of the biggest objections at once. It is individual, not ecological. And it cannot be explained by the child scoring badly causing the family to live somewhere, because the exposure was recorded years before the test.

All three found the same direction: higher maternal fluoride, lower childhood cognitive scores. Grandjean's 2019 review pulled these together and did a benchmark dose calculation — a method regulators use to estimate the exposure at which a defined amount of harm begins — and concluded that safe exposures are "likely to be below currently accepted or recommended fluoride concentrations in drinking water."1

In 2023 an Italian group led by Federica Veneri and Marco Vinceti did the most sophisticated analysis anyone has attempted, and it is the one I would point a sceptic at first. Instead of splitting children into high and low, they modelled the shape of the relationship across the whole exposure range using a technique called a cubic spline, which lets the curve bend where the data says it bends rather than forcing it to be a straight line.7

What they found: comparing the highest exposure category with the lowest, a difference of 4.68 IQ points — and for drinking water specifically, 5.60 points. The curve is substantially linear above 1 milligram per litre, falling about 3.05 IQ points for every additional milligram per litre up to 2 mg/L, and steepening above that. For urinary fluoride the slope is weaker and appears to begin around 0.28 mg/L, which they note corresponds roughly to drinking water at 0.7.

And then, in the same abstract, this: the inverse association was particularly strong in the studies at high risk of bias, and no adverse effect emerged in the only study judged at low risk of bias. They could not rule out residual confounding, and they said so.7

I am going to read that twice because it is the most important sentence in this episode, and it appears in a paper that is usually cited as proving harm. The worse the study, the bigger the effect. The one good study found nothing.

The monograph, the meta-analysis, and the number everybody misquotes

In 2024 the National Toxicology Program — a US federal programme run out of the National Institute of Environmental Health Sciences, whose entire job is evaluating whether substances are toxic — published a monograph concluding, with moderate confidence, that fluoride exposure above 1.5 milligrams per litre is associated with lower IQ in children. In March 2025 the same team published the underlying meta-analysis in JAMA Pediatrics.8

Seventy-four studies. Sixty-four cross-sectional, ten cohorts. Fifty-two rated high risk of bias, twenty-two low. Pooled standardised mean difference: minus 0.45.

The same figure Harvard got in 2012 with twenty-seven studies. Two independent teams, thirteen years apart, nearly three times the evidence base, identical answer. That kind of stability is not nothing.

Where there were individual-level measurements, they found IQ falling 1.63 points for every 1 milligram per litre of urinary fluoride — and restricting to the low-risk-of-bias studies only, 1.14 points.

Now the part that decides the entire policy question, and I want to give it to you exactly as the paper states it, because it is misquoted by both camps with equal enthusiasm.

When exposure was measured by fluoride in drinking water, the inverse association held when the exposed groups were restricted to under 4 milligrams per litre, and under 2 — and was null below 1.5.

When exposure was measured by fluoride in urine, the inverse association held below 4, below 2, and below 1.5.

The authors' own conclusion names the gap: there were "limited data and uncertainty in the dose-response association between fluoride exposure and children's IQ when fluoride exposure was estimated by drinking water alone at concentrations less than 1.5 mg/L."8

American water is fluoridated at 0.7.

So here is the honest position, and it satisfies nobody.

Above 1.5 milligrams per litre, the association between fluoride and lower childhood IQ is consistent across thirty years, multiple countries, multiple study designs, and two independent federal-grade reviews, and it is dose-dependent. That is a serious body of evidence and the people dismissing it as Chinese village data stopped reading in 2012.

At 0.7 milligrams per litre — the actual American exposure — the drinking-water evidence does not resolve the question, and the urinary evidence suggests the curve does not stop. Which of those two you find more persuasive depends on whether you think water concentration or body burden is the better measure of what a person actually received. There is a real argument that urine is better, because it captures tea and toothpaste and processed food. There is a real argument that a single spot urine sample is noisy and reflects the last few hours rather than a pregnancy.

In December 2025 the NTP authors published a direct response to their critics and went further than the meta-analysis had: they wrote that among the high-quality evidence, inverse associations were still observed below 1.5 milligrams per litre based on both urinary and drinking-water measurements.9 That is a stronger claim than the abstract of their own paper makes. It may well be correct. It is also a research team defending its own finding, and you should file it accordingly.

One more recent data point, because this episode is supposed to be current. In September 2025, a systematic review of eleven Indian studies covering over six thousand children found that above 2.0 parts per million, IQ was lower by a pooled 6.5 points, with the effect concentrated in six-to-ten-year-olds. The authors reported substantial heterogeneity and some publication bias in their own data.10 It is consistent with everything else. It is also, again, above 1.5.

The case against, given properly

This show does not do the thing where the opposing view gets one dismissive sentence. So here is the other side, at full strength, from the people who have made it best.

In 2023, Jayanth Kumar and colleagues published a meta-analysis restricted to studies from non-endemic areas — places with fluoride levels relevant to actual community water fluoridation, rather than rural regions with naturally extreme groundwater. Eight studies. Standardised mean difference 0.07, confidence interval minus 0.02 to 0.17, and an I-squared of zero.11

I-squared measures how much the studies disagree with each other. Zero means they are all telling the same story. That is a genuinely clean result, and it says: at the levels we actually use, no signal.

Eight studies is a thin evidence base on which to clear a nationwide intervention, and to their credit the authors say so — their own closing line is that the association at higher levels "requires further investigation."

Then in 2026, Kumar with Steven Levy and John Warren published a review in the Journal of the American Dental Association concluding flatly that there is no association between fluoride at fluoridation-relevant concentrations and IQ, that the endemic-area studies suffer from weak designs, poor exposure measurement, inadequate handling of confounders and inappropriate statistics — and that "there is no established biological mechanism to explain fluoride's effect on IQ."12

Two things about that paper, and both are true at the same time.

The methodological critique is largely correct. The Chinese cross-sectional literature is weak. Exposure measurement in it is poor. Confounding is inadequately handled. Anyone who tells you otherwise has not read the studies. And the "no established mechanism" point is fair as far as it goes — we will get to what the proposed mechanism is in a moment, and to the experiment that recently tested it.

And: this is the dental profession, in the dental profession's own journal, reviewing the safety of the dental profession's flagship public health measure. That is not an accusation of dishonesty. It is a statement about where a paper sits, and it belongs next to the sentence about the ALCOA chemist in 1931. In this field, everybody has an address. The correct response is not to dismiss either side but to notice that the referee position is vacant — which is exactly why the Cochrane review and the NTP monograph carry the weight they do.

The iodine question

"No established biological mechanism" is a strong claim, and there is a proposed mechanism. It is elegant, it is widely repeated, and the best direct test of it came back negative. All three of those facts belong in the same segment.

Remember the halogens — the column of the periodic table where every element is one electron short. Fluorine is the first and smallest and most aggressive. Iodine is much further down the column: large, heavy, and comparatively relaxed.

Iodine matters because your thyroid gland runs on it. Thyroid hormone is, structurally, a small molecule with iodine atoms bolted onto it — four of them in the storage form, three in the active form, which is why they are called T4 and T3. Your thyroid is essentially an iodine concentrating machine, and thyroid hormone sets the metabolic speed of every cell in your body. In a developing fetus it also directs brain development. Iodine deficiency in pregnancy is, worldwide, the leading preventable cause of intellectual disability.

So the question writes itself: two halogens, one of them essential for brain development, the other one the pushiest element on the table. Does one get in the way of the other?

Let me kill the popular version first, because it is wrong and repeating it costs you the argument. Fluoride does not knock iodine off the thyroid hormone molecule. That is not a reaction that happens under the conditions inside a human being. If you have heard "fluoride displaces iodine" and pictured a swap at the atomic level, that picture is not right.

The serious hypothesis is about a pump.

The thyroid does not wait for iodine to wander in. It has a protein in the cell membrane called the sodium-iodide symporter — symporter meaning it carries two things in the same direction at once. It drags sodium down its concentration gradient and uses that energy to haul iodide up its own gradient, concentrating iodide inside the thyroid cell to perhaps twenty or thirty times its level in blood. It is a genuine molecular pump, and if it stops, the thyroid starves regardless of how much iodine you eat.

In 2019 Declan Waugh published the fullest version of the fluoride argument: that fluoride suppresses this pump — not by competing with iodide for the binding site, but indirectly, by raising inflammatory signalling molecules. Specifically TNF-alpha, interleukin-6, interleukin-1-beta, interferon-gamma and TGF-beta-1, each of which is independently known to turn the symporter down.13

That is a coherent hypothesis. It is also a single-author narrative review by an environmental consultant, assembling other people's findings into a proposed chain. It is where you go for the hypothesis. It is not where you find out whether it is true.

In 2026, somebody finally tested it directly.

Sodium fluoride was run against the sodium-iodide symporter itself in a cell assay, at concentrations up to 300 micromolar. Result: no change in iodide uptake. It did not inhibit thyroid peroxidase — the enzyme that attaches iodine to the hormone precursor. It did not bind thyroid hormone receptors. At up to 10 micromolar it did not bind oestrogen, androgen, progesterone, glucocorticoid, PPAR-alpha, PPAR-gamma, AhR, CAR, PXR or RAR-alpha receptors either. At up to 316 micromolar it did not alter oestrogen or testosterone synthesis.14

To put 300 micromolar in perspective: plasma fluoride in a person drinking fluoridated water sits in the low single-digit micromolar range at most. This assay went two orders of magnitude past realistic exposure and the pump did not budge.

That is a clean negative on the most specific version of the mechanism, and I am not going to soften it.

And now the part that must be said in the same breath, because leaving it out would be exactly the kind of one-sided reporting this show exists to refuse.

All three authors of that study work for Procter & Gamble, which sells fluoride toothpaste.14

Say the result. Say the address. Then say why neither one cancels the other.

A conflict of interest is not a refutation. Industry scientists run good assays; the receptor panels they used are standard, the concentrations are generous, and nobody has published a contradicting direct measurement. If you dismiss this study because of who paid for it, you are doing the thing you object to when it is done to the NTP.

And: an isolated cell assay, at fixed concentrations, for a fixed duration, is not the same experiment as a human thyroid drinking fluoridated water for thirty years while also being short of iodine, pregnant, and inflamed. A negative in a dish constrains the mechanism. It does not close the question.

The population evidence points the other way, and it has its own problem.

In 2015, Stephen Peckham and colleagues compared hypothyroidism rates across English general practices and found that practices in the fully fluoridated West Midlands were nearly twice as likely to report high hypothyroidism prevalence as practices in unfluoridated Greater Manchester.15

Public Health England replied in the same issue of the same journal, and their objection is a good one: the study never measured iodine status — the actual dominant cause of thyroid disease worldwide — and Birmingham differs from Manchester in a hundred ways that were not adjusted for.16

This is an ecological study: it compares groups, not people. It cannot tell you that any individual with hypothyroidism drank fluoridated water. That is a real and serious limitation, and Peckham's reply did not overcome it. The exchange ran on into 2017 with neither side moving.

A 2025 review of everything published on fluoride and the thyroid since 1976 reports reduced T3 and T4, raised TSH, oxidative stress, and thyroid follicular cell damage on histology.17 Read the methods and most of that weight is animal work at fluoride doses far above 0.7 milligrams per litre. It is real evidence that fluoride can damage a thyroid. It is not evidence that fluoridated water does.

Which brings me to the finding that should embarrass this entire field, and it is the one I would put on the poster.

In 2023 a group at the University of Kansas went looking for studies that had measured both iodine and fluoride in pregnancy. That is the only design that could answer the question that actually matters — not "does fluoride hurt everyone", but "does fluoride matter more in a woman whose iodine is already low". The hypothesis is specific, plausible, and testable: a pump under pressure is more vulnerable when the substrate is scarce.

They found one study. Their conclusion is four words long: more studies are needed.18

Eighty-one years after Grand Rapids. One study.

What the courts did

In 2016, a coalition of groups petitioned the Environmental Protection Agency under section 21 of the Toxic Substances Control Act — a provision that lets citizens ask the EPA to regulate a chemical and, if refused, sue. They asked for a ban on adding fluoride to drinking water on the grounds of neurotoxicity.

EPA denied the petition in 2017. The plaintiffs went to federal court in the Northern District of California.

On September 24, 2024, Judge Edward Chen ruled for the plaintiffs. The finding: fluoridation of water at 0.7 milligrams per litre — "the level presently considered 'optimal' in the United States" — "poses an unreasonable risk of reduced IQ in children." He ordered EPA to begin rulemaking under section 6(a).19

And he was careful in a way the headlines were not. He wrote explicitly that the finding "does not conclude with certainty that fluoridated water is injurious to public health" — only that there is an unreasonable risk of such injury, sufficient to require a regulatory response.19

That is a legal standard, not a scientific one, and the difference is the whole ballgame. TSCA asks whether a risk is unreasonable, not whether harm is proven. A judge can find an unreasonable risk on evidence that would not get a paper past peer review, and that is by design — regulation is supposed to act before certainty arrives.

EPA appealed in January 2025.

On May 21, 2026, the Ninth Circuit vacated the ruling and sent it back.20

Read what they vacated it on, because it matters enormously and almost nobody reported it correctly. The appeals court did not decide whether fluoride at 0.7 milligrams per litre presents an unreasonable risk. It did not resolve the scientific question at all. It vacated on a procedural ground: the district court had relied on the NTP monograph after the parties had agreed not to rely on it at the first trial, which the panel held transformed the case beyond what either side had presented. The instruction on remand was to rule on the first trial record alone.

So as of today, the legal position is: the ruling that fluoridation poses an unreasonable risk has been vacated, the district court has been told to decide again on a narrower record, and no court has ever held that fluoridation at 0.7 is safe. Nothing has been decided. Something has been un-decided.

The politics did not wait for any of it.

Utah passed a statewide ban on fluoride in public water in March 2025; it took effect on May 7.22 Florida followed with Senate Bill 700, signed the same day.23 Hawaii had never fluoridated outside military facilities. Bills to prohibit fluoridation were introduced in eighteen states that year. And by April 2025, sixty-two communities serving more than nine million people had ended, suspended or blocked it — that last count is the Fluoride Action Network's, which campaigns against fluoridation, and I am giving you their number with their name attached rather than laundering it.

Federally, the Department of Health and Human Services directed the CDC to reconvene an expert panel and produce new recommendations, and EPA was directed to review the health risks. About 63% of the US population currently receives fluoridated water, on the CDC's own figure.

And then — this is the part that cuts against the momentum narrative — as of 2026, no additional state has enacted a ban. Two states, and then a plateau.

What I actually think

I am an internal medicine physician. I did not come to this expecting to find what I found, and what I found is not a clean story in either direction.

Start with what is not in dispute. Fluoride prevents tooth decay — that chemistry is real and I am not arguing with it. Dental fluorosis is real, permanent, affects roughly two in five people at 0.7 parts per million, and is visible enough to matter in about one in eight. Nobody contests that either; it was the accepted price from 1945.

What is in dispute is everything else, and here is where I land.

The benefit is real and it is much smaller than the public has been told. A quarter of a tooth, low certainty, confidence interval crossing zero, no adult data at all. That is not what "one of the ten great public health achievements of the twentieth century" sounds like. It may well have been true in 1945, when most Americans had no fluoride toothpaste and no dentist. The honest question is whether it is still true in 2026, and the honest answer from Cochrane is: barely, and we are not certain.

The cognition evidence above 1.5 milligrams per litre is strong. Not definitive, but strong — consistent across three decades, two continents, multiple designs, and a dose-response curve. Anyone still describing it as Chinese village data is thirteen years out of date.

At 0.7 it is genuinely unresolved, and the two best measures disagree. By drinking water the signal disappears below 1.5; by urine it does not. I do not think you can honestly claim it is settled either way, and I notice that the people most certain in both directions are the ones with something at stake.

On the iodine mechanism specifically, I have to tell you that the premise did not survive. I went into this expecting the displacement story to hold up, because the chemistry is so intuitive. It does not, in the form it is usually told. The direct assay on the sodium-iodide symporter is negative, and the fact that Procter & Gamble ran it does not make the pump move. What survives is narrower and more interesting: nobody has tested whether fluoride matters more when iodine is scarce, because after eighty-one years exactly one study has measured both.

And here is the thing that decides it for me, which is not a scientific argument at all.

The delivery route makes no sense. Fluoride works on the surface of the tooth. We deliver it through the bloodstream, to every tissue in the body, for a lifetime, at a dose nobody can control because it depends on how much water you drink — which depends on your age, the climate, whether you are an athlete, and whether you have diabetes. Infant formula reconstituted with fluoridated tap water delivers a dose per kilogram far above what the same water gives an adult. There is no other medicine in the world administered this way. We do not put statins in the water.

You can defend that as a public health trade-off for populations without dental access, and that defence has real force. What you cannot do is describe a whole-body dose, uncontrolled and lifelong, as a low-risk intervention and then demand certainty from anyone who asks what else fluoride touches.

What to do on Monday

Most of what gets said about fluoride is addressed to legislatures. Almost none of it is addressed to a person standing in a pharmacy aisle. So let me be specific, and let me be clear that this is general education and not advice about you — I do not know your history, your water, your children or your risk.

Find out what is in your water. Every public water system in the United States publishes an annual Consumer Confidence Report, and it will tell you the fluoride concentration in milligrams per litre. If you are on a private well, it is not tested for fluoride unless you ask, and naturally occurring fluoride above 2 and even above 4 milligrams per litre exists in parts of this country. That number is the one that decides how much of this episode applies to you. Under 0.7, most of the cognition literature is arguing about a range you are not in. Over 1.5, you are in the range where the evidence is strong, and it is worth a conversation.

Know where else it is coming from. This is what actually changed the federal recommendation in 2015, and it is the part nobody tracks. Black and green tea concentrate fluoride from soil — tea is one of the largest dietary sources in tea-drinking populations. Foods and drinks made with fluoridated municipal water carry it. Toothpaste is roughly 1,000 to 1,500 parts per million, and small children reliably swallow a fraction of what goes in their mouth. Nobody is adding these up for you, and nobody is adding them up at a population level either.

If you have an infant. Powdered formula reconstituted with fluoridated tap water delivers a dose per kilogram of body weight far higher than the same water gives an adult, because the baby is small and the water is essentially the entire diet. This is not a fringe concern — it is the straightforward reason infant fluorosis rates rose. Ready-to-feed formula, or formula made with low-fluoride water, is a reasonable choice, and it is worth raising with your pediatrician rather than deciding alone.

If you are pregnant. The three prospective cohorts that moved this field measured maternal urinary fluoride, and the window they implicate is prenatal. I am not going to tell you the evidence is strong enough at 0.7 to change what you drink, because I do not think it is. I will tell you that iodine is the part of this you can act on with confidence: adequate iodine in pregnancy is unambiguously important for fetal brain development, iodine deficiency is the leading preventable cause of intellectual disability worldwide, and a meaningful fraction of pregnant women in developed countries are mildly deficient. If the fluoride-iodine interaction turns out to be real, the protective variable is the one you can already fix.

If you want to remove it. Understand what works. Standard carbon filter jugs and most refrigerator filters do not remove fluoride. Reverse osmosis, distillation and activated alumina do. And if you remove fluoride from your water you should be more, not less, diligent about brushing with fluoride toothpaste or using a well-evidenced alternative — because the one thing everybody in this argument agrees on is that topical fluoride prevents decay, and tooth decay in a child is not a small thing.

And if your dentist and this episode disagree, listen to your dentist about your teeth. They can see your enamel and I cannot. What I am disputing is not whether fluoride works on a tooth. It is whether the case for putting it in everybody's water, forever, at a dose nobody controls, is as settled as the word "settled" implies. Those are different questions and conflating them is how this argument has gone badly for seventy years in both directions.

Where this goes next

There is more material here than an hour holds, and the parts I have left out are not footnotes. A second episode should take up: what fluoride does to bone, where it accumulates for decades and where skeletal fluorosis is a real and badly-monitored disease; the pineal gland, which calcifies with age and concentrates fluoride, and the animal work suggesting that matters for melatonin; the total dose question, which is the one that actually changed the 2015 recommendation — tea, processed food, toothpaste that children swallow, and the fact that nobody is measuring anyone's total intake; and the alternatives, chiefly nano-hydroxyapatite, the mineral your enamel is already made of, whose safety review is far more reassuring than its critics claim and whose efficacy evidence is far thinner than its advocates claim.21

That is a second hour and it deserves to be one. Compressing it into this one would mean dropping the caveats, and the caveats are the show.

Sources

Every paper referenced on air, in the order it comes up. Links go to the publisher via DOI.