Ice cream and iced drinks: why cold food is the sharpest trigger, and how to get back to it.

Two kinds of toothpaste have a claim on a tooth that hurts at a cold drink: the ones that narrow the open dentine tubule, which is nano-hydroxyapatite, calcium sodium phosphosilicate, stannous fluoride or arginine, and the one that quietens the nerve behind it, potassium nitrate. The caveat belongs in the same breath as the answer, because nearly everything known about all of them was measured with a puff of air on a tooth in a dental chair, and where a systematic review and meta-analysis has separated the stimuli, nano-hydroxyapatite came out ahead of its comparators on air and on touch and level with them on cold, at a standardised mean difference of −0.17 with p = 0.61¹. S3 Sensitivity Science™ pairs potassium nitrate for the nerve with nano-hydroxyapatite inside the tubule and biomimetic hydroxyapatite on the surface, which is checkable against the ingredient list on the pack. What follows is the cold-liquid evidence named trial by trial, the argument the field is currently having about how a tooth senses cold at all, and the half of the question no toothpaste page answers: what is known about giving cold food up, and about getting it back.

What was checked24 peer-reviewed studies, two further papers cited for attribution only, S3 consumer trial (ADSL, 2026), the Oral Health Foundation and the NHS

Which toothpaste helps teeth that are sensitive to cold drinks and ice cream?

One that contains an occluding active, used morning and night and judged after weeks, not over a weekend; and where the twinge is sharp and immediate instead of a slow ache, ideally one that carries potassium too⁷. That is a shortlist and not a name, for a reason set out below: this literature is organised by active ingredient, and the pooled comparisons rank the actives against a plain fluoride paste, not against each other in the mouths of people eating ice cream²⁸.

Reading the pack is simple enough once you know what to look for. The Oral Health Foundation's own advice names potassium citrate, potassium nitrate and stannous fluoride as the ingredients that help, and adds a line most brands would rather you skipped: the effect is maintained only while you keep using the paste⁷. Hydroxyapatite is not on that list, which reflects how recently it arrived on the UK shelf rather than a verdict on it; the pooled evidence for it is further down, including the part that does not flatter it.

There is no route through an ingredient list to a promise. A paste that has narrowed the channel and quietened the nerve has done what a paste can do; whether the ice cream is comfortable at the end of it is a separate question, and the published trials have never asked it.

Why is a cold drink sharper than a cold breath?

Because it lasts longer and it takes more heat out. A breath crosses the tooth and is gone; a mouthful of iced water or a spoonful of ice cream sits against the enamel until you swallow, cooling it the whole time. What happens next is set out in the Journal's page on what dentine tubules are and, at length, on the sibling page about the inside of a tubule during a cold drink: fluid in the exposed channels moves, and the nerve at the far end reads the movement as pain. That has been the working account since Brännström set it out in the mid-sixties⁹. The Journal also covers why cold hurts your teeth and the accommodations people make around it.

The degrees are less dramatic than the sensation. In a non-randomised clinical study that taped thermocouples to the outside of sixteen molars and had people drink to a fixed regimen, the surface of an intact natural tooth bottomed out at about 31.5 °C on average during the cold drink, a handful of degrees below resting mouth temperature rather than anywhere near the temperature of the drink¹⁰. Enamel and dentine are poor conductors, and a tooth is defended by its own thickness. The same clinical study put a molar with a gold inlay in it at 25.0 °C, cooling roughly three times faster, which is worth knowing if the tooth that objects is a restored one¹⁰. No pain was recorded there, and nobody in it is described as having a sensitive tooth, so it tells you what the surface reaches and nothing about what that feels like.

Two other numbers set the scale of the personal part. In a non-randomised clinical study that monitored twenty-nine adults across three weeks, the cold temperature that produced moderate to strong pain in a sensitive tooth held steady within each person and varied hugely between them, averaging about −13.7 °C with a standard deviation near ten degrees¹¹. Your threshold is yours, it is not drifting, and it is not anyone else's. And in a non-randomised clinical study of eighty adults, split between an air group and a touch group, the interval a sensitive tooth needs before it will answer the same way again came out at roughly two minutes for cold air, and longer for touch¹². A tub of ice cream is therefore not one long injury. It is a run of separate provocations, each landing on a tooth that has largely reset.

The cold-air version of the question is a different stimulus with its own trial literature, and it is answered on the sibling page about breathing cold air.

How does a tooth sense cold, and why is that still being argued about?

Because the obvious answer was tested and failed, and what replaced it has not been confirmed in a person. This belongs on a page about ice cream because it is the reason nobody can yet design a toothpaste aimed at cold specifically.

The obvious candidates were the two cold-activated channels that carry cold and irritant pain elsewhere in the body. The animal work that tested them in teeth came back negative: deleting TRPM8 or blocking TRPA1 did not stop a tooth responding to noxious cold, and the expression of both channels turned out to be much the same in nerves supplying the pulp as in nerves supplying the cheek or the mucosa³. That animal paper carries its result in its title: TRPM8 and TRPA1 do not contribute to dental pulp sensitivity to cold³.

A large international group then proposed a different answer. Their animal work put cold transduction in the odontoblast, the cell that lines the pulp and sends a process up into the dentine tubule, with a channel called TRPC5 acting as the cold sensor inside it; blocking TRPC5 pharmacologically cut cold responses by about three-fifths rather than abolishing them⁴. The same study, in the human teeth it examined alongside the animal work, identified TRPC5 in the odontoblast layer of healthy extracted adult teeth and in nerve fibres running inside the dentine tubules; in teeth with pulpitis the proportion of TRPC5-positive fibres rose while TRPM8-positive fibres fell⁴. The commentary that greeted it in Cell Calcium — a short review piece rather than new data — was titled "I scream for ice cream — TRPC5 as cold sensor in teeth", which is a fair indication of how surprising the field thought it was¹³.

Then the complication. A single-channel in vitro study recorded human TRPC5 in a kidney-derived cell line and confirmed that it is intrinsically cold-gated, with its steepest response between about 16 °C and 11 °C; but the temperature at which it began to open shifted substantially when the cell's own signalling state was altered⁵. The cold behaviour of the channel is therefore a property of the channel and its surroundings together, so a threshold measured in one cell type cannot simply be read across to an odontoblast. Three papers, one disagreement, no resolution. What it changes for a person with a sensitive tooth today is nothing at all, which is the honest answer: none of this work says anything about any ingredient, and none of it has been tried as a treatment.

What have the trials actually measured with a cold liquid?

Less than the packaging would lead you to expect. The table below holds the two searches this page was built on, written out in full with the day they were run, so that the count can be checked and argued with rather than taken on trust.

Two things follow. One is that eleven toothpaste trials is a thin base for the commonest complaint in this category, and the search undercounts even that: the eight-week nano-hydroxyapatite trial discussed below wrote "ice-cold" without the word "water", so it is not among the thirty-six at all. The other is flatter. No published trial has used ice cream, or any other food, as a test stimulus. The stimuli here are a calibrated blast of air, a probe pressed against the tooth at a known force, and iced water from a syringe. Everything anyone tells you about ice cream is an inference from those three.

Here is the cold-liquid evidence, source by source.

The pooled result is the one that costs the occluding ingredients most. A systematic review and meta-analysis of six four-week randomised trials rated nano-hydroxyapatite's overall effect high quality on GRADE and put it clearly ahead of its comparators on evaporative stimuli, at a standardised mean difference of −1.09, and on tactile stimuli at −0.93; on cold, pooled across the same trials, the figure was −0.17 at p = 0.61¹. A second meta-analysis, of nine randomised trials in people whose sensitivity followed a deep clean, put desensitising agents as a class at −0.78 on a water stimulus and −1.21 on an air stimulus, rated the water result very low certainty, and found the control group doing better on the mechanical stimulus¹⁴.

The individual cold-liquid trials are few enough to name. An eight-week double-blind randomised trial funded by a nano-hydroxyapatite manufacturer, completed by eighty-five adults, took visual analogue scores against ice-cold water and against air every two weeks: all four pastes improved from baseline at every visit, and the nano-hydroxyapatite pastes did not differ from a calcium sodium phosphosilicate paste at any point¹⁵. In that manufacturer-funded trial's full text, the paste pairing nano-hydroxyapatite with potassium nitrate reduced cold-stimulus sensitivity more than the same paste without potassium at weeks two, four and six, though not at week eight¹⁵. A four-week double-blind randomised trial of thirty adults put a 5% potassium nitrate paste ahead of a non-desensitising paste on air and on cold water at four weeks, and behind a calcium sodium phosphosilicate paste on both¹⁶. In that trial's companion electron microscope work, the potassium paste had closed no tubules at all, which is the mechanism stated plainly: a quieter nerve behind an unchanged channel¹⁶. A UK examiner-blind randomised trial scored teeth with iced water and put a paste combining potassium nitrate, hydroxyapatite and aluminium lactate ahead of a potassium-nitrate-only paste at every time point, with relative risk reductions on the Schiff score of 55% immediately after brushing, 81% at seven days and 88.6% at a fortnight¹⁷. In the same randomised trial the whole-mouth rating and the quality-of-life questionnaire showed nothing between the two pastes¹⁷.

The two big network meta-analyses treat cold differently from each other. The 2020 network meta-analysis of 125 randomised trials and 12,541 patients covered tactile, cold and air stimuli and named calcium sodium phosphosilicate the most beneficial formulation across all three, while placing potassium on its own in the tactile ranking rather than the air one⁸. The 2026 network meta-analysis pooled only two-week outcomes and only the cold-air and tactile scores, so it carries no cold-liquid node at all; its authors conclude that stannous fluoride and arginine dentifrices should be considered first-line self-care options, chosen for preference, tolerability and availability rather than for any expectation of superior efficacy, and they record that ninety-six per cent of the stannous trials, eighty-six per cent of the arginine trials and seventy-six per cent of the potassium trials in it were industry funded². That conclusion is printed here on a page published by a company selling neither stannous fluoride nor arginine. The strongest single trial of nano-hydroxyapatite, a double-blind randomised trial in which a fluoride-free 15% paste reduced sensitivity more than both a fluoride paste and a placebo at two and four weeks in 105 adults, used cold air rather than cold water¹⁸. And the Cochrane review of potassium toothpastes, six randomised trials, found an effect on air blast and touch at six to eight weeks, no significant movement in what patients themselves reported, and concluded that clear evidence for potassium toothpastes was lacking¹⁹.

The empty cell in the last row is the most useful one in the table. Each of those three actives has a trial literature behind it; the finished paste has not been through a randomised trial of its own, and this page will not imply otherwise.

How long does it take, and how do you get cold food back?

Weeks for the paste, and longer for the habit, which is the part almost every page leaves out.

The onset figures exist and they agree with each other. In an industry-funded double-blind randomised trial of 120 adults, paid for by a maker of stannous fluoride toothpaste, all three test pastes beat a plain fluoride control on the cold-air score by day three; by week eight the improvements over control were 57% for stannous fluoride, 47% for an experimental oxalate paste and 44% for potassium nitrate, and the gaps between the three actives at that point were not significant²⁰. A separate manufacturer-funded randomised trial of 215 adults put a calcium sodium phosphosilicate paste significantly ahead of a fluoride reference from day three, improving progressively out to day fifty-six²¹. Neither trial used a cold liquid. Read together they describe a curve rather than a switch: something measurable inside the first week, most of the movement across the first two months.

Two findings keep that curve modest. When the same industry-funded trial of 120 adults moved everyone onto the control paste for three more weeks, all three actives held most of their benefit instead of losing it, so a tube that runs out does not reset the tooth overnight²⁰. And in an industry-funded randomised trial that followed seventy-six adults for twenty-four weeks, the examiner-measured cold-air score improved a little at every visit while the participants' own reported sensitivity barely shifted across the whole half-year²². The long view is thinner still: a meta-analysis restricted to randomised trials with at least six months of follow-up found significant reductions for several agents, the largest for glutaraldehyde and low-level laser therapy, and no significant long-term benefit for the class it labels calcium-phosphate-based, which its abstract does not break down any further²³.

Now the recovery half of the question. The nearest thing to evidence that treating sensitivity gives people their food back is a systematic review and meta-analysis of six clinical trials, in which quality-of-life scores improved alongside the sensitivity itself, with two of the six at high risk of bias, high heterogeneity, and certainty graded very low to moderate⁶. The instrument those trials lean on was built by asking patients: the Dentine Hypersensitivity Experience Questionnaire came out of in-depth interviews and focus groups before it was ever tested as a questionnaire, and was validated in a general population sample and a clinical sample, though its funding is not declared on the record and two of its authors are recorded elsewhere as employees of a sensitivity-toothpaste manufacturer²⁴.

What the qualitative work adds is that avoidance outlives the pain. In an interview study of eighteen Australian adults with confirmed dentine hypersensitivity, funded by a manufacturer of sensitivity toothpaste, seventeen had changed what they ate or how they ate it, and nine described the condition as having no daily impact, because the adaptation had stopped registering as a limitation²⁵. In that same manufacturer-funded study, eighty-nine per cent were using a desensitising toothpaste, every one of them chosen without professional advice, and not one had been asked about sensitivity by a dentist²⁵. A UK study of one hundred and one people with self-diagnosed sensitivity, funded by GlaxoSmithKline Consumer Healthcare and co-authored by two of its employees, found that the impact a month later was predicted by how often the teeth twinged and also by whether people made sense of the condition and whether they coped with it passively, by working around it²⁶.

None of which is a study of reintroducing ice cream, because no such study exists. What follows is sensible practice, not a finding. Give a paste eight weeks of uninterrupted use, morning and night, before you judge it. Spit out and do not rinse, so the actives stay on the tooth. Then test the thing you have been avoiding on purpose and in steps, a cold drink without ice, then with, then something genuinely frozen, instead of waiting to feel brave and finding out at a party. Keep using the paste afterwards, because the Oral Health Foundation is explicit that the effect lasts only as long as the habit⁷. If the twinge comes back at a particular step, that step is your marker, and it is a far more useful thing to tell a dentist than "cold hurts".

Most people never get that far. In a cross-sectional survey rather than a trial, of 815 Australian adults, funded by a manufacturer of sensitivity toothpaste, 67.2% reported symptoms consistent with dentine hypersensitivity, 74.3% of them had never sought professional care and 26.3% had never used any treatment at all²⁷. Managing this by not ordering the dessert is the normal response, not an unusual one.

Where does S3 sit if cold food is the trigger?

On the two-mechanism side of the argument, and with the same hole in its evidence as everything else on the shelf. Calm the nerve but leave the tubules open and the triggers keep arriving; seal the tubules but leave the nerve over-reactive and it still complains, which is why single-active pastes do half the job by design. A mineral cannot quieten a nerve and a potassium salt cannot narrow a channel, so anything meant for the cold-liquid trigger has to bring both jobs with it. That is the specification S3 was written to: potassium at the nerve end, nano-hydroxyapatite down inside the tubule, biomimetic hydroxyapatite across the surface.

The percentages on the pack need their basis attached, and S3 gives both figures: 10% nano-hydroxyapatite and 5% biomimetic hydroxyapatite are inclusion levels of the ingredient as supplied, and the active hydroxyapatite content is lower than either number reads. So does the timescale. Neither of these actives arrives all at once, which is the whole reason for the eight-week rule above: S3 starts working from the first brush and is judged at two to four weeks. It is a daily toothpaste rather than a fortnight's treatment, which is the same point the Oral Health Foundation makes about the category as a whole.

Three things this page will not say. One: that a toothpaste will let you eat ice cream again, because no trial of any paste has ever used a food as a stimulus, so nobody knows. Two: that the two hydroxyapatites in S3 are proven against a cold liquid, because the pooled evidence for that ingredient class is weakest exactly there, and the numbers are printed above¹. Three: that S3 outranks the two actives the newest network meta-analysis puts first, because it contains neither of them².

When is cold pain a dentist's question?

At the point where it stops behaving like sensitivity at all. Dentine pain leaves with the stimulus that caused it: it arrives with the spoonful and has gone by the time the tooth is warm again. Pain that lingers after the cold has cleared, that sits in one tooth, that wakes you at night, or that turns up when you bite down is a different signal, and no toothpaste is the answer to it.

The Oral Health Foundation gives four thresholds for booking: severe pain, sensitivity running on for more than a few weeks, only one tooth affected, or a sudden start, any of which can mean decay, a crack, a gum problem or an infection⁷. The NHS puts its own threshold at toothache lasting more than two days, and adds three things that should not wait for it: painkillers that fail to help, a swollen cheek or jaw, and the combination of a high temperature, pain on biting, red gums or a bad taste. It is blunt about where to go, which is a dentist rather than a GP, and it sends anyone whose swelling reaches the eye or the neck, or who is struggling to breathe, swallow or speak, to A&E instead²⁸. The sibling page on being sensitive to hot and cold at the same time sets out the timing test in more detail, because a tooth that reacts to heat as well as cold and then holds the pain is the pattern that most often turns out not to be sensitivity at all.

Frequently asked questions

Will S3 let me eat ice cream again?

Nobody can promise that, and no toothpaste has been tested against a food, because none of these trials has ever used one as a stimulus. What S3 can say is what its own panellists reported: in an unpublished consumer trial of 51 adults over eight weeks, run for S3 by ADSL, an accredited laboratory in Devon, 86% said by the end that they no longer avoid certain foods or drinks, up from 59% at 24 hours. That is a self-reported change in behaviour, from a trial S3 paid for and has not published in a journal, and it is weaker evidence than the pooled analyses on this page.

Why does a cold drink hurt more than cold air?

Contact time, mostly. A breath is gone in a second; a mouthful of iced drink keeps drawing heat out of the tooth until you swallow it, so the fluid inside the exposed tubules moves for longer. In a non-randomised clinical study the outside of an intact molar fell to about 31.5 °C on average during a cold drink, which is a few degrees rather than the temperature of the drink itself, because the tooth's own thickness does most of the defending¹⁰.

How long before cold food stops hurting?

Give it eight weeks of twice-daily use before you judge it. Across the onset trials something is measurable by day three and most of the improvement accumulates over the first two months; in one industry-funded randomised trial the advantage over a control paste was still growing at week eight²⁰. Note that all of those measurements were taken with cold air, not with a cold drink.

Is it bad to avoid cold food?

It is not harmful, but it hides the problem from you and from your dentist. In a manufacturer-funded interview study, seventeen of eighteen participants had changed how they ate, and half of them no longer experienced those changes as a restriction at all²⁵. If you have quietly stopped ordering anything with ice in it, that is worth saying out loud at your next appointment, because a symptom you have built your habits around still counts as a symptom.

Does drinking through a straw help?

Probably, by keeping the cold liquid away from the exposed surfaces near the gumline, and no trial has tested it. The mechanism is contact time again: less of the tooth in contact with the cold for less of the time. Treat it as a workaround rather than a treatment, and note that it does nothing whatever for ice cream.