Did it work, or did it just feel like it worked? Judging a sensitivity toothpaste honestly.

When sensitive teeth seem to be getting worse, the first suspect should often be the reading rather than the teeth or what you brush them with: you started when your teeth were at their most sensitive, what you feel and what a clinician's instruments measure are different things, and sensitivity rises and falls with acid and mood from one week to the next. An honest verdict on whether a sensitivity toothpaste worked comes from one fixed trigger and a written record read against its own first entry, not from how the last few days felt. S3 Sensitivity Science™'s own evidence about its paste is a consumer trial of 51 adults with sensitive teeth over eight weeks that recorded what people said, which is exactly the kind of number this page asks you to read with care.

What was checked17 peer-reviewed studies, S3 consumer trial (ADSL, 2026), guidance from the Oral Health Foundation and the NHS

Why does a toothpaste often seem to work best in the first fortnight?

Because you almost certainly started it at a bad moment, and a reading taken at its worst tends to come back less bad the next time, whatever happens in between. That tendency is called regression to the mean. It is neither imagination nor the placebo effect: it is what happens to anything that goes up and down when you act on a high reading.

Nobody has put a size on it for sensitive teeth. A PubMed search run for this page on the tenth of September 2026, pairing regression to the mean with dentine hypersensitivity or tooth sensitivity, returned no records, as the planner's identical search had earlier the same day. The nearest figures come from pain research in general. A 2022 modelling study simulated the placebo group of a chronic-pain trial and found that the regression effect grew as the pain score needed to join the trial was raised¹. Across the simulation's scenarios the effect ran from 0.13 to 0.78 points on a ten-point pain scale, largest at the highest entry bar, although that scenario also had a wider spread of scores and a looser link between one reading and the next, so the bar was not the only thing that changed¹.

Trials in this field are built with that in mind. The consensus guidelines they still follow ask for a negative control and a benchmark control alongside the paste being tested⁵, because whatever moves in everyone moves in the control group too, and only the gap between the groups belongs to the active ingredient. One industry-funded double-blind trial, paid for by Procter & Gamble, shows where the entry bar sits: it admitted only adults with at least two teeth scoring above one on a cold-air scale and a probe threshold of ten to twenty grams⁶.

You can watch the early pattern in that trial's control group, whose members brushed for eight weeks with a toothpaste carrying no desensitising active. On the cold-air score their adjusted average went from about 2.45 at the start to 2.24 at week two and then barely moved, reaching 2.22 at week eight, according to the industry-funded trial's own tables⁶. On the probe their threshold rose from about 10.6 grams to 14.2 grams by week two and then slid back, to 13.5 grams at week eight and 12.7 grams three weeks later, in the same manufacturer-funded trial⁶. Most of what happened to them happened in the first fortnight.

Whatever produced that change, whether the response to being treated, regression to the mean or simply joining a trial, it arrived early and then stopped. How large the response to being treated can be in this condition is set out on the page about relief that builds over weeks. The point here is its shape, which is the shape of a paste that worked for a fortnight and then stopped.

A person repeats the pattern on a smaller scale. Few people start a sensitivity toothpaste on a good week. If the week you started was the roughest of the month, the weeks after it will tend to feel easier whatever is in the tube, and when the ordinary ups and downs come back, the paste appears to have given up. It may have been doing its job throughout, or it may never have done much. A fortnight cannot tell you which, and neither can a control group of one.

Why can your own sense of improvement and a clinician's measurement disagree?

Because they measure different things, and neither is the whole truth. An examiner aims a controlled puff of cold air at one tooth, or presses a calibrated probe against it, and scores the response on the day; you notice, or fail to notice, twinges spread across weeks of ordinary eating, drinking and brushing.

The sharpest example is a trial built to test the control pastes themselves. In an industry-funded randomised study of 249 adults, paid for by GSK Consumer Healthcare, six toothpastes with no desensitising active ran for eight weeks beside a 5% potassium nitrate paste and a stannous fluoride paste³. Every one of the eight improved from baseline on every measure, the controls included, in that manufacturer-funded study³. On the examiner's measures the potassium nitrate paste separated from the controls at weeks four and eight, and on the cold-air score the stannous fluoride paste separated from all six controls at week four but from only three of them at week eight, in the same industry-funded study³. The participants' own verbal rating of the air blast did not separate the active pastes from the controls, in that study sponsored by GSK Consumer Healthcare³.

That does not make the participants wrong, and it does not make the examiner right. It says that, in that trial, a question put to the person was too blunt to tell a desensitising paste from an ordinary one, while the examiner's scales were not. Pooled mouthwash trials show the split running the other way: across seven placebo-controlled trials, five of them testing potassium nitrate, a 2017 meta-analysis found that patients reported significantly less day-to-day pain at eight weeks while the touch and air tests showed no significant difference⁴.

Even the plainest question, whether someone has sensitive teeth at all, gets a different answer depending on who asks it. In a 1992 cross-sectional clinical study of 635 patients at a dental clinic in Rio de Janeiro, 25% said they had sensitive teeth and 17% were diagnosed on examination with air and a probe⁷. In a 2019 cross-sectional clinical study of 380 patients at a Brazilian university dental clinic the gap ran the other way: 41.7% said so, and 88.7% were diagnosed⁸. Neither figure is a prevalence for anyone reading this; side by side, they show that asking and examining count different things.

So "is it working?" has at least two honest answers, and they need not agree. If your teeth feel no different but a dentist finds less response to the probe, or the other way round, nobody has got it wrong.

What has never been done, in sensitive teeth, is the piece of work that would say how closely the two agree. A PubMed search on the tenth of September 2026 pairing the Schiff, probe or air-blast scores with self-rated scales and the words correlation, agreement or concordance turned up seventeen records; each was a paper that listed both kinds of score, a paper about an instrument or a survey, and none of them set out to test agreement. The way those scores are produced is explained on the page about how desensitising toothpastes are tested. For what the actives in these pastes are meant to do, the Journal's article on how sensitive toothpastes work and why yours might not be working is a shorter read.

Why can the same teeth feel worse this week?

Because what sets sensitivity off changes from week to week even when the tooth does not. Acid, low mood and the small ways people protect a sore tooth all move, and each has been tied to how much sensitive teeth hurt.

Timing mattered in the one study of people found to have looked at it: across 350 adults under 35 seen in hospital and practice clinics in south-east England, a cross-sectional clinical study found sensitivity on the biting and cutting edges was more severe the less time had passed since their last acidic food or drink⁹. That study saw each person once, so it cannot show that the drink caused the twinge, and its authors read the association as a sign that sensitivity comes and goes with active acid wear⁹. Tracing where the acid comes from, and what else to rule out if the record really does show a decline, is the job of the page on what to rule out when sensitivity keeps climbing.

Mood and habit move the reading too. In a two-week daily diary study of 101 people at a UK university who had diagnosed their own sensitivity, an observational record rather than a trial, funded by GlaxoSmithKline Consumer Healthcare, the number of twinges each person logged swung widely from one day to the next¹⁰. In the same industry-funded diary study, doing more to protect sore teeth on one day, such as keeping food away from them, predicted more pain on the days that followed, and a lower mood went with pain that came more often and bothered people more¹⁰. The authors are careful about the direction, and so is this page: someone already hurting protects their teeth more, and pain and mood can each feed the other.

One more finding from that manufacturer-funded study needs the same caution: people who used a sensitivity toothpaste more often logged pain more often, which says more about when people reach for the tube than about what the tube does¹⁰.

Held in the same conditions, the tooth itself is steadier than the week around it. In a three-week clinical study of 29 adults, the cold-air temperature at which a sensitive tooth began to hurt moderately to strongly barely changed within each person from one session to the next, though it varied a great deal between people¹¹. A tooth's pain threshold holds still when the trigger is held still. Your week does not hold still, and that is the reason the record at the end of this page fixes the trigger. A bad week is a real bad week; on its own, it is just not news about the paste.

Nobody has followed untreated sensitivity week by week to learn how far it drifts on its own. A re-run search on the tenth of September 2026 for the natural history, natural course or spontaneous remission of dentine hypersensitivity turned up five records, and not one of them followed people over time; the planner's broader version the same morning turned up fifteen, with the same outcome.

Before blaming the tube, check whether it was put down. A lapse can let sensitivity creep back, although in one older follow-up the creep was slight. Twelve weeks after 40 adults stopped eight weeks of supervised brushing with one of two strontium chloride toothpastes, their sensitivity had drifted back a little and was still significantly below where it started, in a follow-up of a randomised trial whose funding is not stated¹².

Do people change the yardstick as they go?

Yes, and it has been counted, even if what the direction means is less clear than the count. In a secondary analysis of an eight-week trial of 75 adults with sensitive teeth, 36% of participants were classed as having recalibrated their internal standard for their own symptoms downward over the trial, and 14.7% upward¹³.

The paper does not translate those directions into everyday terms, and this page will not guess at them. What survives is plainer: by the end of the trial, half of its participants were no longer judging their teeth by the standard they had used at the start, and a person judging their own paste from memory is in the same position. A number written down on day one is a version of that standard that cannot drift.

One threshold for a change that matters has been published in this condition, and it is not for pain. In a validation study that pooled 311 participants from three earlier trials, funded by GlaxoSmithKline Consumer Healthcare with two company scientists among its authors, the smallest important change on the Dentine Hypersensitivity Experience Questionnaire, which asks how sensitivity affects daily life, was put at between 22 and 39 points¹⁴. Nothing equivalent exists for the cold-air score, the probe or a pain scale: a search for a minimal important difference alongside any of them, re-run for this page on the tenth of September 2026, turned up two records, that questionnaire paper and one on quality of life in periodontal patients. How far the clinical scores and that questionnaire parted company in a real trial is covered on the page linked in the first section.

So there is no published number that tells you whether your change counts. Settle, on day one, the change that would make you buy the tube again, and put that line in the record next to the first score; then the ruler you judge by at the end is the one you picked at the start.

Laid side by side, the distortions push a verdict in different directions, and each comes with a gap in the evidence and something you can do about it at home.

How should you read S3's own consumer-trial results?

As a record of what people said, from a consumer trial whose full readout has not been published. S3's evidence about its own paste comes from an independent third-party consumer trial run by ADSL in Devon, in which 51 adults with sensitive teeth used it for eight weeks and answered questions about their teeth along the way.

Those figures come from asking 51 people what they noticed, and by the reasoning on this page they cannot tell you how much of what was noticed belonged to the toothpaste. The trial report describes a single panel, all using the same paste, and tests each answer against a fixed pass mark rather than against a group using another toothpaste or a placebo.

Run the three distortions over it. Everyone on the panel started from teeth sensitive enough to take part, so some improvement was likely whatever they brushed with, and the report describes no comparison group to absorb it. Every answer was what a panellist said, which is the kind of reading that could not tell active pastes from plain ones in the manufacturer-funded control-paste study above³. And eight weeks of anyone's life hold good weeks and bad ones, which averaging across 51 people softens but cannot remove.

What a figure of that kind can tell you is what people reported about their own teeth after using the paste, which is worth knowing when you choose what to try. What it cannot tell you is how much of that report belongs to the paste, and how much to starting on a bad week, to being asked, and to the ordinary run of the weeks. That is why no percentage from the trial appears anywhere on this page, including the ones S3 publishes elsewhere.

S3 also asks to be judged at two to four weeks, which is earlier than the eight weeks the field's consensus trial guidelines expect most trials to run⁵.

None of the distortions on this page takes sides between brands. Each applies to our own paste exactly as it applies to the tube already in your bathroom, which is why the table below grades a brand's consumer trial, ours included, on the same columns as a randomised trial and a record kept at home. In its last row, S3's consumer trial of 51 adults over eight weeks is graded like everything else.

So how do you tell whether it worked?

By reading a record rather than a mood. Pick one trigger you meet anyway, score it on the same scale each time, date every entry, and compare the latest entries with the first, not with the last few days. The method itself, with its dates, its triggers and the gap to leave between tests, is laid out on the page about how long to give a new sensitivity toothpaste; what follows is how to read what it gives you. S3 is sold as the toothpaste you use every day rather than a fortnight's course, and a record kept over weeks is the fairest way to judge any paste used like that.

Read against the first entry and the fixed trigger. A cold drink from the same fridge, scored on the same line, is a reading. An ice cream on a hot afternoon, remembered the next morning, is an anecdote. When the two disagree, go with the reading: not because feelings do not count, but because the reading was taken the same way both times and the memory was not.

Look at runs, not days. A single bad day is the kind of swing the industry-funded diary study above recorded from one day to the next¹⁰. Three or four weeks of entries worse than the first is a trend, and a trend is what the record is for.

Write down what else happened. Something acidic shortly before a test, a stressful week, a missed evening brush, a new whitening paste in the rotation: a few words beside the score turn a puzzling entry into an explicable one. The acid and mood notes are there because of the cross-sectional clinical study and the industry-funded diary study described above⁹¹⁰.

Allow for the record itself. Keeping a record changes what you notice. In that same industry-funded diary study, forty-three of the ninety-nine people who answered its closing questions said taking part had changed their experience of sensitivity: some noticed twinges they had not registered before, others found the twinges came less often than they had assumed, and the authors warn that watching a symptom can alter it¹⁰. So a change in the first days of keeping a record may belong to the record rather than the tooth. Whether being watched, or expecting a paste to let you down, moves sensitivity scores has never been examined in its own right: a search on the tenth of September 2026 for nocebo or Hawthorne effects in dentine hypersensitivity turned up one record, a fourteen-day toothpaste evaluation that looked at neither.

Pick one scale and keep it. None of them is a precision instrument. Across 71 studies, a scoping review found the visual analogue scale to be the commonest pain measure in this field and the Schiff cold-air score the next, and found the references papers gave for choosing either to be inconsistent¹⁵. In a cross-sectional clinical study of 72 adults, five pain scales, the Schiff score among them, identified sensitive teeth with an accuracy of between 0.729 and 0.750¹⁶. Each gives an estimate on its own terms, so switching scales halfway through leaves two records that cannot be compared.

When the record, not the mood, shows a decline. Say the fixed trigger has read worse against its first entry for several weeks running. At that point, stop troubleshooting the paste. The accepted definition of dentine hypersensitivity requires other causes of tooth pain to be ruled out, and a 2026 scoping review of 72 papers found that few studies apply that part of the definition in full¹⁷; a record kept at home rules out nothing. The page on what to rule out, linked above, works through the candidates, from acid to a cracked tooth.

Some patterns should not wait for a record at all. The Oral Health Foundation tells people to see a dentist when the pain is severe, when sensitivity has lasted longer than a few weeks, when only one tooth is affected or when the pain arrives suddenly, since each can be a sign of tooth decay, a cracked tooth, gum problems or infection¹⁸. The NHS puts a two-day limit on toothache before a dentist should see it, says the same of toothache that painkillers do not ease, and treats swelling around the eye or neck, or swelling that makes breathing, swallowing or speaking difficult, as a reason for A&E¹⁹.

Frequently asked questions

Are S3's own trial results the kind of evidence this page warns about?

Yes. They are what 51 adults with sensitive teeth said about their own teeth during an independent eight-week consumer trial, and a figure of that kind cannot separate the paste from starting on a bad week, from the response to being treated or from the ordinary run of good and bad weeks. The trial report describes no comparison group and its readout is not published, so none of its percentages appears on this page.

Judge S3 the way this page suggests judging any tube: from your own record, read against its first entry, at the two to four weeks S3 gives for itself and again at eight weeks, where the trials take their verdict⁵.

If it helped for two weeks and then stopped, did it ever work?

Possibly, and those two weeks cannot settle it. In an industry-funded double-blind trial of 120 adults, the control group, brushing with a paste that had no desensitising active in it, did most of its improving in the first fortnight and then levelled off, and on the probe it drifted back⁶. That is the same shape as "it worked, then it stopped", drawn by a tube with nothing in it meant to relieve sensitivity.

Look at your record rather than the fortnight. If the trigger still scores below its day-one entry, the paste may be doing what it did all along against noisier weeks; if it is back at its day-one level, the explanations on the page about why a sensitivity toothpaste stops working are the next place to look.

How much better does it need to be to count?

Nobody has published an answer for pain. There is no agreed smallest worthwhile change for the cold-air score, the probe or a pain scale, and the single threshold this field has, set in a validation study that pooled three earlier trials, funded by GlaxoSmithKline Consumer Healthcare, belongs to a quality-of-life questionnaire, at between 22 and 39 points¹⁴.

So draw your own line before you start and write it next to your first score. The point is that the line was drawn before the result arrived, not after it.

My teeth are definitely worse. Is it the toothpaste?

Take the pain as real and the cause as open; the record is what tells the two apart. If the fixed trigger has read worse against its first entry for several weeks, something has changed. It may well be something no tube reaches: a new acid habit, harder brushing, a gum line that has moved, or a pain that is not sensitivity at all. The page on what to rule out, linked in the section on why teeth can feel worse this week, takes those one at a time.

Do not wait for weeks of entries if the trouble is one tooth, arrived suddenly, hurts when you bite or comes with swelling. The Oral Health Foundation and the NHS both send those patterns to a dentist rather than to the toothpaste shelf¹⁸¹⁹.

Can I trust the percentages on toothpaste packs?

Trust them for what they are. A percentage followed by "said" or "reported" counts answers to a question put to the people who used the product; it is not a measurement of their teeth. Before giving one weight, ask how many people were asked, whether anyone was using a different paste or a placebo at the same time, and whether the full results have been published. If nobody was using anything else, the figure cannot separate the paste from the distortions on this page, and that holds for our own figures as much as for anyone else's.