What happens inside a dentine tubule when you drink something cold.

Cold at the surface of exposed dentine chills the fluid sitting in the open tubules; the fluid contracts, the column is pulled outwards, and the nerve endings at the pulp end of the channel read that movement as a short, sharp pain. Nothing cold ever reaches the nerve — the stimulus that arrives is mechanical, and the temperature is only what sets it going¹. What almost no page about cold sensitivity says is that the numbers printed on the boxes were not measured with a drink: the standard test stimulus is a blast of air from a dental syringe, and when a systematic review and meta-analysis of six four-week randomised trials reported the stimuli separately, nano-hydroxyapatite came out ahead of its comparators on the evaporative and tactile measures and no better than them when the stimulus was cold, at a standardised mean difference of −0.17 (p = 0.61)². S3 Sensitivity Science™ pairs 5% potassium nitrate with 10% nano-hydroxyapatite and 5% biomimetic hydroxyapatite, both as solution, and keeps 1450 ppm fluoride, all of which is readable on the pack.

What was checked21 peer-reviewed studies, the Oral Health Foundation and the NHS

What is inside the tubule before the drink arrives?

Fluid, at roughly the temperature of the rest of you, in a channel that runs from just under the surface down towards the nerve. Dentine is not a solid block: it is threaded with tiny channels linked to the nerve of the tooth, and while enamel and gum cover their outer ends nothing gets to them, which is why most people go decades without noticing they exist⁴. When enamel wears or the gum recedes, the outer ends open onto the mouth, and the Oral Health Foundation's list of how that happens starts with brushing too hard and gums shrinking back and runs through acid wear, gum disease, grinding, cracked teeth and, temporarily, whitening⁴. The Journal's page on dentine tubules and the glossary entry on their size and number take the anatomy further than this page needs to.

Two features of the arrangement matter for what happens next. The exposed part is usually at the gumline, where the enamel thins into the root surface, which is exactly where a mouthful of iced drink washes first. And the pulp sits inside rigid mineralised tissue with nowhere to expand into, which a 2026 review of thermal mechanisms gives as one reason that even mild thermal stimuli, cold in particular, can produce fast and intense pain in a tooth when the same temperature on skin would produce almost nothing¹.

What does cold do to the fluid in the channel?

It shrinks it, and the direction of travel is outwards, away from the nerve. That is the hydrodynamic account, set out by Brännström in the mid-sixties and read line by line on the page about the hydrodynamic theory without jargon; this page starts where that one stops, at the second-by-second detail.

The most detailed picture available is a simulation rather than an observation. A three-dimensional fluid–structure interaction simulation of one tubule and the odontoblast process inside it, dimensioned from eleven cats' teeth, modelled a hot and a cold stimulus and reported three asymmetries⁵. In the simulation cold moved the fluid faster and the modelled stress built about 2.9 times more quickly than under heat; the stress from heat, once established, persisted 71% longer; and the peak stress in both cases fell on the tip of the odontoblast, the cell process that reaches up into the tubule, which the authors take as the likely site of activation⁵. Cat geometry and computed physics, so it is a description of what the theory implies rather than a record of what happened in anyone's tooth — but it does explain the quality of the pain. A stimulus that arrives fast and passes fast produces a jolt; one that arrives slowly and lingers produces an ache.

What the movement does at the far end is being rewritten as this page is published. The account now favoured combines fluid movement with ion channels: mechanical stress opens channels and enzymes in the odontoblast, which releases ATP, glutamate and nitric oxide onto the pulpal nerve fibres, and a 2025 review calls the integrated version the prevailing hypothesis rather than a finding⁶. A 2026 review of thermal mechanisms in teeth names the same thermomechanical coupling as the most widely accepted explanation for thermal pain¹. Neither review claims that any of this has been demonstrated inside a living human tooth, and neither points at anything you can buy.

Is a cold drink the same stimulus as the dentist's puff of air?

No, and the difference is the reason this page exists. At least four different provocations get called "cold", and they act on the tooth in different ways, for different lengths of time, over different amounts of surface.

Start with the one the field runs on. A short blast from a three-way dental syringe is an evaporative stimulus as well as a thermal one: measured in vitro on human teeth, an air blast raised evaporative water loss from the dentine surface between fifteen and thirty times over what the same dentine lost sitting in still air, and the effect grew as the syringe came closer and as the air got warmer³. The paper's own conclusion is the sharpest thing anyone has written about this category's standard test: because water evaporates readily across a smear layer that badly impedes bulk fluid movement, air blasts may overestimate dentine sensitivity³.

Then the only attempt to put the different provocations on one scale. In an in vitro study on extracted human crown segments, acid-etched to behave like hypersensitive dentine, five stimuli were applied in turn — hot water at 56 °C, cold water at two degrees Celsius, an air blast, an osmotic stimulus and a tactile one — and the fluid each one shifted was converted into the hydrostatic pressure that would have produced the same movement⁷. The ranking, largest to smallest, ran hot water, cold water, air blast, osmotic, tactile, and it held both before and after conversion⁷. The authors end by saying the approach should be verified in a living person, which as far as the searches below found has not happened in the thirty years since.

The last two rows are thin, and they are thin because nothing has been put in them. A consensus guideline that still shapes this literature asked, back in 1997, for tactile, cold and evaporative stimuli together, for blinded parallel-group work lasting eight weeks, with negative and benchmark controls, and for two independent replications before a product is approved¹⁰. Applying more than one stimulus is common; reporting them apart is not, which is why the single pooled analysis that did carries so much weight below. Meanwhile a scoping review of 71 studies found the visual analogue scale and the Schiff cold-air scale to be the field's two standard measures, and its authors note that the scales are not validated for the condition they are used on¹¹. That caution applies to every number on this page, including the ones that flatter the actives in the tube whose maker publishes it.

One number deserves rescuing from the general vagueness. In a non-randomised clinical study, thermocouples taped to the outside of sixteen molars recorded what happens during ordinary drinking: an intact natural tooth bottomed out at about 31.5 °C during a cold drink, a few degrees under mouth temperature rather than anywhere near the temperature of the drink, while a molar carrying a gold inlay reached 25.0 °C and cooled roughly three times faster⁸. Enamel and dentine are poor conductors and a tooth defends itself largely by being thick. That measurement was taken on the outside of the tooth. What the inside reaches during a drink is the number this page would most like to give, and it is the first row of the search table below: nothing published measures it in a living person. So the inside of a tooth during an iced coffee has been modelled and never measured, and anybody who quotes you a figure for it is quoting a model.

Three of the absences on this page are absences rather than opinions, and the searches that found them are printed here so they can be re-run and argued with.

What does the evidence actually say about cold?

Less than the shelf implies, and what there is splits by stimulus.

The pooled result is the one the occluding minerals like least, and it is the reason this page was written. The 2019 systematic review and meta-analysis behind it rated its overall pooled effect high quality on GRADE, and it separated the three stimuli instead of averaging them: on the evaporative measure nano-hydroxyapatite came out at a standardised mean difference of −1.09 against its comparators, on the tactile measure at −0.93, and on the cold measure at −0.17, which carried a p value of 0.61². That is one review, of six trials, at four weeks, and it does not separate concentrations — but it is the one pooled analysis behind this page that reported the three stimuli apart, and where it looked, the choice of stimulus decided the answer².

Only a handful of published measurements used a cold liquid at all. Thirty adults, four weeks, double-blind and randomised: by the four-week visit a 5% potassium nitrate paste had cut both the air score and the cold-water score further than a non-desensitising control, while a calcium sodium phosphosilicate paste had cut them further still, at two weeks and at four¹². The companion electron-microscope work in the same paper found the potassium paste had occluded no tubules at all, which is the mechanism stated without decoration: a quieter nerve behind a channel that is exactly as open as it was¹². In a 12-week double-blind randomised trial of 120 adults with no placebo arm, co-authored by the developer of the calcium sodium phosphosilicate technology it tested, cold-water reductions ran 34% at two weeks and 79% at twelve for 5% potassium nitrate, against 49% and 91% for the phosphosilicate paste and 26% and 85% for a stannous fluoride gel¹³. And in an eight-week double-blind randomised trial funded by the company that made the three test pastes, in which 85 of 105 recruited adults completed, ice-cold water and air scores were taken every two weeks: every arm improved from baseline at every visit, the nano-hydroxyapatite arms were no different from a calcium sodium phosphosilicate comparator, and the arm pairing 10% nano-hydroxyapatite with potassium nitrate did better on the cold measure than the same paste without potassium at weeks two, four and six¹⁴.

The best-known trials in this area used cold air rather than cold liquid, which is exactly the confusion the page is trying to clear. Take the trial most often quoted for the mineral: 105 adults, four weeks, double-blind and randomised, a fluoride-free 15% nano-hydroxyapatite paste against a fluoride paste and a placebo, with the nano-hydroxyapatite arm lower on the cold-air and tactile measures at both the two-week and the four-week visit¹⁵. Or the trial that reads the clock most closely: 120 adults across four arms, double-blind and randomised, seven of its eight authors employees of the company whose stannous paste was one of the arms, with the Schiff cold-air score read at day three and again at weeks two, four and eight¹⁶. Set against a plain fluoride control at week eight, that industry-funded trial put the stannous arm 57% better, the oxalate paste 47% and potassium nitrate 44%, with no significant distance between the three actives by then¹⁶. And in an eight-week double-blind randomised trial of 67 adults where the only difference between the arms was the potassium — both pastes carried the same monofluorophosphate — the cold-air score fell further in the potassium arm at four weeks, 1.12 against 0.32 on a visual analogue scale¹⁷.

Two network meta-analyses set the frame, and they treat cold differently from each other. The older of the two, a network meta-analysis pooling 125 trials and 12,541 patients, is the one with a cold node: it put calcium sodium phosphosilicate at the top across tactile, cold and air together, and it entered potassium on its own under the tactile stimulus only, not the air or the cold one; its node for potassium combined with hydroxyapatite showed large effects against fluoride toothpaste on tactile, at 2.47, and on air, at 2.44¹⁸. The 2026 network meta-analysis of 93 randomised trials and 9,548 participants pooled only two-week outcomes and only the cold-air and tactile scores, so it has no cold-liquid node at all; it put stannous fluoride at −0.85 on the cold-air score with high confidence, nano-hydroxyapatite at −0.96 from two studies with moderate confidence, arginine at −0.78 and potassium with or without fluoride at −0.42 both with low confidence, concluded that stannous fluoride and arginine should be first-line self-care options chosen on preference and tolerability rather than on any expectation of superior efficacy, and recorded that most of the trials it pooled were paid for by industry — 96% of the stannous ones, 86% of the arginine ones and 76% of the potassium ones¹⁹. S3 carries neither of the two actives that review puts first, and that sentence stays on the page.

The mineral's own headline number does not help with cold either way. Reductions of 39.5% against placebo and 23% against fluoride toothpastes come from a 2023 meta-analysis pooling 44 clinical trials, in which the gap against other desensitising agents was 10.2% and not statistically significant; two of its three authors are scientists employed by Dr Wolff, a maker of hydroxyapatite toothpastes, and the review declares no external funding²⁰. It pools formats, concentrations and forms of hydroxyapatite together and does not report the stimuli apart, so it cannot settle the question this page is asking.

Which leaves an honest summary. The second row of the search table above went looking for anything that sets a cold-liquid outcome beside an evaporative one and asks whether the two agree; nothing that came back does it. The one pooled analysis that reported them apart found they behaved differently. Everything else in this category is measured on the stimulus that overestimates.

What happens with the second mouthful?

Not what most people assume, and the reason anyone knows is a piece of methodology rather than an interest in ice cream. A trial that provokes the same tooth twice needs to know how long to wait in between, and until this one nobody had measured it²¹.

The design was simple: 80 adults in a non-randomised clinical study, split evenly between a cold-air arm and a touch arm, each tooth provoked again at ten minutes, at five, at two, and once more the moment the first pain had faded²¹. About two minutes was enough for a cold air-blast; touch needed longer, with a mean visual analogue change from immediate to two minutes of 8.0 for the tactile stimulus against 0.8 for the air in that clinical study²¹. And the examiner's mean Schiff score did not climb as the gap between provocations shortened. In the same clinical study it drifted down, from 2.38 at ten minutes to 2.15 immediately after the pain had gone²¹. A second spoonful, on that evidence, does not land harder than the first.

The other thing worth knowing about your own tooth is that its threshold is a fixed property rather than a mood. In a non-randomised clinical study that monitored 29 adults across three weeks with a purpose-built cold-air device, the temperature that produced moderate to strong pain held steady within each person, with an intra-class correlation of 0.83, while ranging enormously between people, around a mean of −13.7 °C with a standard deviation near ten degrees⁹. Neither of those studies tested a toothpaste, and neither used a drink. They describe the tooth, not the treatment, which on a page like this is worth more than another effect size.

Which toothpaste ingredients have evidence against a cold drink, and where does S3 sit?

The ones with cold-specific numbers behind them are potassium nitrate and calcium sodium phosphosilicate, with nano-hydroxyapatite supported on cold air and unsupported on cold liquid once the trials are pooled²¹³¹⁹. That is a narrower answer than any pack gives, and the table below is why.

The last row is the useful one, and its emptiness is the point. The pooled analyses in this literature are built by ingredient class rather than by product, and the finished S3 paste has not been through a randomised trial of its own; nothing on this page should be read as saying otherwise¹⁸.

What the table does support is an argument about design. Two faults sit in the same tooth at the same time: the channel is open, and the nerve behind it is provoked too easily, so a paste with one active can only reach one of them. The two faults need different chemistry, because nothing in the hydroxyapatite family quietens a nerve and nothing in the potassium family narrows a channel. Three actives answer that in one tube, each sized and chosen for a different place: potassium nitrate for the nerve ending, nano-hydroxyapatite small enough to work inside the tubule, biomimetic hydroxyapatite for the surface around its mouth. It is also, on the cold question, an argument from mechanism rather than from a cold-stimulus result, and the difference between those two things is what most of this page has been about.

Two caveats travel with it. Read the two hydroxyapatite percentages on the pack as what they are, the level of the ingredient as it arrives from the supplier rather than the weight of mineral in the tube, which is lower; S3 publishes both numbers. And the clock runs in weeks rather than days, because both mechanisms accumulate — potassium around the nerve, mineral on and in the surface — so S3 starts working from the first brush and is judged at two to four weeks, an argument taken further on the page about why relief that builds over weeks is a feature. If cold is your only trigger and you want the single active with the strongest two-week cold-air number, the pooled two-week evidence in the 2026 network meta-analysis points at stannous fluoride¹⁹, and that page is the comparison of what occlusion and nerve-calming can each do.

There is a reason nobody has built a toothpaste aimed at cold in particular. The 2025 review of ion channels in dentine hypersensitivity concludes that the channels involved do too many other jobs around the body to be targeted safely with a drug, and that closing tubules and desensitising dentine remain the safest and most effective approaches available⁶. The two levers a toothpaste can pull are still the two worth pulling, then, even though the mechanism underneath them is more complicated than any pack has room to say⁶.

When is cold pain not a toothpaste's problem?

When it outlasts the cold. Dentine pain arrives with the stimulus and leaves with it: the twinge is over by the time the tooth is back at mouth temperature. A dull ache that turns up after the cold rather than during it and then sits there for an hour has been described as something else entirely — a hypothesis paper on dental pain following a swing in ambient temperature sets it apart from dentine hypersensitivity, and the page on sensitive teeth in winter carries that argument in full²².

The thresholds for booking an appointment are not subtle. The Oral Health Foundation says to see a dentist when the pain is severe, when it has run on for more than a few weeks, when only one tooth is involved, or when it started suddenly, any of which can point to decay, a crack, a gum problem or an infection⁴. The NHS puts its own line at toothache lasting more than two days, and adds pain that painkillers do not touch, a swollen cheek or jaw, and the combination of a high temperature, pain on biting, red gums or a bad taste; it sends swelling that reaches the eye or the neck, or any difficulty breathing, swallowing or speaking, straight to A&E²³. Neither list has a toothpaste in it, and neither should.

Frequently asked questions

Does S3 help with cold drinks specifically?

S3 carries potassium nitrate and both forms of hydroxyapatite. Of those, potassium nitrate is one of the two actives with cold-liquid numbers behind it and hydroxyapatite is not, so the honest answer is narrower than a yes: the pooled evidence on cold is weaker than the pooled evidence on air and touch, and for nano-hydroxyapatite the pooled cold result was no different from its comparators². The trials in this field test ingredient classes rather than finished pastes, so none of them is a trial of S3, and relief of any kind is judged at two to four weeks rather than on the first cold drink.

Why does ice cream seem worse than a cold drink?

Contact time and temperature, on the mechanism as it is understood — a spoonful sits against the tooth and keeps drawing heat out, where a mouthful of drink is swallowed. That is an inference, not a measurement: nothing published has used a food, or any drink, as a test stimulus, and the searches behind that statement are set out above with their terms and the day they were run. The sibling page on ice cream and iced drinks works through what that gap means for anyone deciding what to order.

Why does the pain stop the moment I swallow?

Because the tooth warms back up, the fluid in the tubules stops moving and the nerve stops firing. The mechanism has no memory: once the temperature settles, the provocation is over. Pain that carries on after the tooth is warm again is not following this pattern, and that is the version to take to a dentist.

Why do my teeth hurt from cold but not from hot?

Not, on the bench evidence, because heat is a gentler provocation. When five stimuli were put on one scale in vitro, hot water at 56 °C shifted more fluid than cold water at two degrees Celsius⁷, and the simulation of a single tubule found the stress from heat lasting 71% longer than the stress from cold, though it built more slowly⁵. Cold is sharper because it arrives quickly; heat, when it hurts, tends to ache instead, and a tooth that reacts to heat and then holds the pain is the pattern discussed on the page about being sensitive to hot and cold at once. The Journal's page on why cold hurts your teeth covers the everyday version of the question.