How potassium nitrate is thought to calm a sensitive tooth, and why the mechanism is still unproven.

Potassium nitrate is thought to work by raising the potassium concentration around the nerve endings inside a tooth until those endings stop passing signals on, a working model that dates from the 1970s and has never been confirmed in an intact human tooth¹. Whether the explanation is right and whether the toothpaste helps are two separate questions, and this page keeps them apart: the trials are what they are regardless of which mechanism turns out to be doing the work. S3 Sensitivity Science™ carries 5% potassium nitrate, the concentration the published toothpaste trials used.

What was checked15 peer-reviewed studies, plus NHS and Oral Health Foundation guidance

What is actually happening when a cold drink hurts?

Dentine is not solid. It is threaded with tubules running from its outer surface towards the pulp, each one holding fluid, and when enamel wears away or the gum edge moves back, the outer ends of those tubules open onto the mouth. A mouthful of cold shrinks the fluid nearest the surface, the column moves, and sensory endings at the inner end of the tubule respond to that movement. The pain is short because the movement is short.

That is the trigger side of sensitivity, and the companion page on what happens inside a dentine tubule when you drink something cold sets it out properly. This page begins one step later, at the far end of the tubule, and asks what a potassium salt is supposed to be doing there. The Journal's page on what dentine tubules are covers the anatomy.

What is potassium nitrate supposed to do about it?

The proposal is about voltage. A nerve at rest holds far more potassium inside its membrane than outside, and that difference is most of what sets its resting voltage. Raise the potassium concentration outside and the difference narrows, so the membrane sits closer to the level at which it fires. Keep it there and the effect inverts: the channels that generate a nerve impulse cannot reset while the membrane stays depolarised, and the ending falls silent. The name for that last step is accommodation. Applied to a tooth, the argument runs that potassium from a toothpaste dissolves at the tubule mouth, works its way inward, builds up around the nerve endings, and leaves them unable to respond to a fluid movement they would otherwise report.

It is a good hypothesis. It explains why potassium nitrate, potassium chloride and potassium citrate have all been used for the same job¹, and why relief in the pooled trials is measured at six to eight weeks rather than in minutes⁹. Potassium nitrate has been in desensitising toothpaste since the 1970s and in placebo-controlled trials since the 1990s, so the model has had half a century in which it could have been demonstrated.

Has anyone shown that it happens?

Not in a person. The 2000 review that the ingredient's reputation mostly rests on identified 27 clinical trials, including 16 double-blind randomised trials of potassium toothpastes, all of which reported significant reductions in tactile, air and subjective sensitivity, and it then says that the proposed nerve-inactivation mechanism has never been confirmed in intact human teeth and remains uncertain¹. That sentence has sat in the literature for a quarter of a century, and it is not one you will find on a pack.

What has been shown is each step separately, in preparations that are not a human tooth. The dose-response was measured in 1995 on isolated rat spinal nerve: attenuation of the compound action potential rose with extracellular potassium across the 8 to 64 mmol/l range, half-maximal attenuation came at 17.4 mmol/l for the fast fibres, and a full conduction block generally needed at least 32 mmol/l². Potassium chloride and potassium nitrate did exactly the same thing at every concentration tested, which is the cleanest available evidence that the working part is the potassium and not the anion beside it².

Recordings from nerves inside a tooth exist, and they are feline. In an animal study in cats, potassium chloride placed in a deep cut cavity produced a burst of firing followed by a long silence during which a normally reliable stimulus evoked nothing, and the same biphasic pattern followed an injection into the artery, which points at the potassium ion rather than the route³. A later animal study, also in cats, complicates the picture: potassium carried into the dentine by inward fluid flow made the receptors more sensitive to inward flow and less sensitive to outward flow, so what potassium does to an intradental receptor depends on which way the fluid is going⁴.

That is the whole of it. To establish that nothing better exists, three PubMed searches were run on the day this page was written, and they are set out below so that anyone can repeat them or overturn the claim.

No recording has been made from a nerve inside a living human tooth before and after a potassium dentifrice. That is a large gap for an ingredient this old, and it is the reason every honest sentence about this mechanism has to begin with "thought to".

Read the table downwards and the shape of the problem is clear. The step with the best evidence is the last one, which is the one that does not require the mechanism to be true at all.

What argues against the simple version?

In 1996 McCormack and Davies took the model apart from the inside, in a review and hypothesis paper in the journal Pain⁶. Its authors accepted that potassium dentifrices work; what they doubted was the explanation. Measurements of solution velocity inside individual dentinal tubules, made by scanning electrochemical microscopy, suggested that the outward movement of tubular fluid is a far greater barrier to potassium diffusing inward than the earlier estimates from hydraulic conductance across bulk dentine had implied⁶. A tooth, on this reading, is not a passive sponge waiting to soak up potassium. It is pushing outward, gently and continuously, and anything trying to get in is swimming upstream.

Their alternative was that potassium confined to the outer part of the tubule could still quieten deeper endings indirectly, through a second-messenger pathway involving nitric oxide, whose release can be provoked by potassium concentrations under 1 mmol/l⁶. That is a hypothesis and it is thirty years old, and this page will not present it as the mechanism. It is on the page because it is the only published alternative anyone has put forward, and because a reader who has been told that potassium travels to the nerve deserves to know that the people who study it have argued about whether it can.

The same year, and pointing the other way, Stead, Orchardson and Warren built a mathematical model of a single dentinal tubule — arithmetic, not a laboratory bench — and asked what a one-minute application of 500 mmol/l potassium to the dentine surface could achieve at the far end: about 30 mmol/l at the inner end, if the barrier between tubule and pulp were closed and the outward flow slow, decaying over 20 to 30 minutes and falling whenever the outflow or the barrier permeability rose⁵. Set that against the figures from isolated rat nerve and the arithmetic is uncomfortably tight: 30 mmol/l sits above the level that halved conduction in vitro and below the level that generally blocked it². The model's own authors made the point that the rise is transient and that the flow velocity is what decides it. Orchardson is a name on both papers, and on the review that says the mechanism is unconfirmed: the people who know this ingredient best are the ones least willing to state the mechanism flatly.

For scale, 5% potassium nitrate comes to roughly half a mole per litre if all of it dissolves: 50 g in a litre, at a molar mass near 101 g per mole, is about 0.5 mol, which is close to what the model held against bare dentine for a minute⁵. Brushing is not that. Brushing is a diluted slurry, two minutes, and then you spit.

There is industry evidence on the other side of the argument, and it should be read as what it is. A 1994 in vitro paper written by scientists employed by the manufacturer of the paste reported that a silica-based formula carrying 5% potassium nitrate allowed rapid penetration through the dentine matrix, in a bench model its authors described as simulating conditions in the mouth⁷. No person took part, and in the manufacturer's own account the surface film that reduced dentine permeability was made of copolymer and silica, which are parts of the base rather than the salt⁷.

How much potassium actually gets in, and does the paste around it matter?

Somebody measured it. In 2015, 50 extracted human molars were divided between five groups and given a single 30-minute treatment, after which nitrate in the pulp cavity was read spectrophotometrically by the Griess assay and the materials' viscosity was measured on a rheometer⁸. The results differed significantly across all five groups, and the highest reading came from the desensitising toothpaste⁸.

Two things follow, and only two. Some potassium nitrate crosses enamel and dentine into the pulp cavity of an extracted tooth within half an hour, and how much arrives depends on the formulation and not on the label alone⁸. Everything else the table might seem to say, it does not. These were extracted teeth in vitro, with no pulpal pressure and therefore no outward tubular flow, which is the barrier the whole argument turns on; half an hour of standing contact is not brushing; nobody reported anything, because nobody was there; and the untreated control already read 3.41 micromolar, so part of the signal was never the treatment⁸. What a paste delivers into a beaker has never been matched against what a person feels⁸. The companion page on the concentration the trials used takes the dose question further.

If the mechanism is unproven, does the toothpaste work?

Better than the mechanism does, and the syntheses still disagree with each other. Pooling six randomised trials, the Cochrane review of 2006 separated the instruments from the people: air-blast and tactile sensitivity improved significantly at six to eight weeks, what the patients themselves said did not, and the reviewers' own conclusion was that clear evidence for potassium toothpastes is lacking⁹. A meta-analysis of 31 randomised trials published in 2015 put potassium toothpastes at a standardised mean difference of -1.28 against placebo, with heterogeneity between the trials so high that the pooled figure has to be quoted with the caveat attached¹⁰. A 2019 network meta-analysis of 30 randomised trials found no significant difference between potassium and placebo at all¹¹.

The honest summary is the one the evidence forces: the instrument readings usually improve, the patients' own ratings often do not, and the reviews reach different verdicts from the same literature. Part of the gap is that sensitivity trials carry a large placebo response. In a six-week double-blind randomised trial of 120 adults, potassium nitrate, strontium acetate and a plain fluoride toothpaste all reduced sensitivity with no significant difference between them, and the fluoride-only group improved significantly from its own baseline¹². A tooth that feels better in week three is not proof that an active has started working¹². The full reading of the Cochrane review sets out what it did and did not test.

One detail is worth carrying away, because it shows how much depends on which instrument is used. In an industry-funded double-blind randomised trial of 120 adults run by a manufacturer's own scientists, potassium nitrate beat the control on the cold-air score at every time point but did not differ from it on the tactile probe at day three, a non-significant 10.7%¹³. Two stimuli, one paste, different answers on the same day.

If the mechanism is unproven, why does S3 use potassium nitrate?

Here is the sentence that costs us something. The mechanism behind the nerve-calming active in the S3 tube has never been confirmed in an intact human tooth, and the one review that looked directly at the physical barrier argues that potassium probably does not reach the nerve in the way the story on the pack implies¹⁶. That is not followed by a "but". It is followed by the trials, which are a separate fact, and by the reasoning about what to do when a mechanism is uncertain.

If you cannot be sure that calming the nerve is doing the work, the sensible response is not to rely on it alone. A sensitive tooth has two things wrong at once, an ending that responds too readily and tubules that are open, and most sensitivity toothpastes address one of them. Hydroxyapatite occludes and potassium desensitises, and neither does the other's job. S3 carries both, which means this page does not have to be right about potassium for the formula to have an answer.

Two limits belong in the same paragraph. Five per cent is the concentration the published trials used, not a concentration shown to beat another one: no dose-ranging trial of potassium nitrate in a toothpaste has been published, so 3%, 5% and 8% have never been compared in a single study. And whichever mechanism is operating, it is slow. Relief from this class of active is judged at two to four weeks, not at two to four days, which the companion page on how long potassium nitrate takes to work covers in detail.

What does potassium nitrate not do?

It does not seal the tubule. In the companion laboratory arm of a four-week double-blind randomised trial in 30 adults, the electron microscope found that the 5% potassium nitrate paste had left the tubules open while the comparator paste had not¹⁴. That is the cleanest published separation of the nerve route from the tubule route, and the page on whether potassium nitrate blocks tubules works through the electron-microscope evidence. Where a potassium paste does close tubules in the laboratory, including in the manufacturer's own in vitro work above, the credit belongs to the silica and the copolymer in the base⁷.

It does not rebuild a receded gum, replace worn enamel, or treat what is causing the exposure in the first place. In the standard account of how desensitising treatments divide, potassium salts are the class aimed at the nerve while strontium salts, bioglasses, arginine with calcium carbonate and oxalates are the class aimed at the opening, and that account is a review written by scientists employed by a manufacturer, closing with an introduction to that company's own product¹⁵. Reading an ingredient list by what each active is aimed at is still the most useful habit a shopper can have.

It is not a diagnosis. The Oral Health Foundation, on a page produced with an educational grant from an oral-care manufacturer, describes sensitive toothpastes as working by blocking the tiny channels in dentine, which is a description of occlusion rather than of anything potassium does, and tells readers to see a dentist if the pain is severe, if it lasts more than a few weeks, if only one tooth is affected or if it came on suddenly¹⁶. The NHS puts the threshold at toothache lasting more than two days¹⁷. A single tooth that has started to hurt is a question for a dentist, not for a tube.

Frequently asked questions

Does S3 numb the tooth?

No. Nothing in a toothpaste anaesthetises a tooth, and the difference matters. A local anaesthetic blocks conduction outright for an hour or two and you can feel that it has. On the proposed model, potassium works the other way round: it is thought to sit around the nerve ending and raise the level of movement needed before that ending reports anything, so what changes is a threshold rather than sensation itself. You keep normal feeling in the tooth. If a tooth genuinely goes numb, that is a reason to see a dentist, not a sign that a toothpaste is working.

Is it the potassium or the nitrate that does the work?

The potassium, as far as anyone has been able to test it. In vitro, on isolated rat spinal nerve, potassium chloride and potassium nitrate attenuated conduction identically at every concentration between 8 mmol/l and 64 mmol/l, and control experiments ruled out osmolarity and ionic strength as the cause². The nitrate is the counter-ion that makes a stable, soluble salt. No head-to-head comparison of the two salts as toothpastes has ever been published, so "the salts are interchangeable in a tube" remains an inference from a nerve in a dish, not a finding in a mouth.

If nobody has proved the mechanism, why is it printed on every box?

Because it is the best available explanation and it has never been replaced. Regulators ask a toothpaste to substantiate what it claims about people, not to prove the biology behind it, and the review that gathered this literature counted 16 double-blind randomised trials reporting reductions in sensitivity, which is a stronger card than any mechanism paper¹. The problem is not that the model is on the box. It is that it is printed in the present indicative, as if it were anatomy, when the review it comes from calls it unconfirmed.

How long before a potassium nitrate toothpaste makes a difference?

Give it a fortnight before you form a view and eight weeks before you decide. The pooled Cochrane outcomes sit at six to eight weeks⁹, and in a manufacturer-funded randomised trial potassium nitrate had not separated from the control on the tactile probe by day three¹³. Judging this class of active on a weekend is judging it on the placebo response.