The Cochrane review on potassium toothpastes: what it found, what it did not, and what has happened in the twenty years since.

The 2006 Cochrane review pooled six randomised trials of potassium toothpastes and found that the two instrument tests improved at six to eight weeks while the patients' own assessment of their sensitivity did not, and its authors closed by writing that no clear evidence was available to support these toothpastes¹. That version was published on 19 July 2006 and it is still the current one: the Cochrane record, read again on 2026-09-10, lists two versions of this review and no third, with no withdrawal notice and no replacement². S3 Sensitivity Science™ carries 5% potassium nitrate, the active those six trials tested. So the most-quoted sentence about the ingredient in our own tube says the evidence for it is not clear, and this page is not going to pretend the twenty years since have settled it.

What was checked17 peer-reviewed studies, the Cochrane Library record, the Oral Health Foundation and the NHS

What did the Cochrane review actually measure, and what did it find?

It measured four things and found that three of them moved. The reviewers looked for randomised trials in which a potassium toothpaste was tested against the same toothpaste without the potassium salt, in adults with sensitivity from exposed root surfaces, and they took the readings at six to eight weeks. Only trials that published a mean and a standard deviation could be used, because only those can be pooled¹. Six trials qualified. All six used a paste carrying 5% potassium nitrate, all six ran parallel groups, four lasted eight weeks and two lasted twelve, and between them they involved 390 participants¹.

Sensitivity was provoked in four ways: a fine probe pressed against the exposed dentine, a blast of air, a hot or cold stimulus, and the patient's own account of everyday life. The probe and the air blast are what the review is quoted for. Both moved, and clearly. The pooled probe estimate across five of the trials was a standardised mean difference of 1.19, with an interval of 0.79 to 1.59; the sixth trial had scored the probe on an ordinal scale and could not be pooled with the others, but on its own it also favoured the potassium paste¹. Air blast, across all six pooled trials, gave −1.25 with an interval of −1.65 to −0.85¹. One trial applied a thermal stimulus, and it too favoured the potassium paste. Three trials asked patients to rate their own sensitivity, and pooling those three gave −0.67 with an interval that crossed zero at 0.10 — a result the review describes as failing to detect a difference¹.

The earlier version of the same review had found the same shape. Its search closed in April 2000, it identified eight randomised trials, pooled four of them, and reported a significant air-blast effect at a standardised mean difference of −1.51, a significant tactile effect, and a subjective assessment at six to eight weeks that was not significant³. Three of the four trials it could not pool showed no effect at all³. Its conclusion was stated more strongly than the 2006 one: no strong evidence supports the efficacy of potassium nitrate toothpaste for this condition³.

One footnote on dates, because it will confuse anyone who checks. The Cochrane record dates that earlier version 24 January 2000, while PubMed indexes the same pooled review as the second issue of 2001³. Both are printed here rather than one of them being quietly chosen.

What does "no clear evidence" mean, and what does it not mean?

It means the reviewers did not think six small pooled trials, measured in ways that disagreed with each other, were enough to settle the question¹. It does not mean the toothpastes were shown not to work. The review says so itself, in the two sentences that come before the famous one: the evidence rests on a small number of individuals, and the effect varies with the method used to assess sensitivity¹. Both are statements about the size and the quality of the evidence, not about the ingredient.

Three details sharpen that. The subjective analysis, the one that failed, was the smallest of the three: three trials and 108 patients, against 390 for the air blast¹. All six trials were described as randomised and double blind, but the review found the randomisation procedure and the blinding measures unclear in every one of them, which left allocation concealment unclear as well, and only one trial described what happened to the people who dropped out¹. And the review's closing request is for more large, well designed and well conducted randomised controlled trials — the language of a question still open, not of a case closed¹.

The declared interest belongs in the same paragraph, because it is on the record and because declaring it is to the review's credit. One of the four authors had been awarded a prize by a Danish toothpaste company in 1997, and the record prints that under the authors' declarations where anyone can read it². The review's conclusion runs against the interest of the industry that sells these pastes, so the declaration cuts the wrong way for the usual suspicion — but the point of disclosure is that the reader decides, not the writer.

What did the review not cover?

A great deal, and most of what people quote it about. The comparator was always the identical paste with the potassium salt taken out, so the review never asked how potassium compares with arginine, stannous fluoride or a bioactive glass¹. Its title says "potassium containing", but every one of the six pooled trials used a five per cent potassium nitrate paste, so it is not a reading of potassium citrate or potassium chloride products, whatever the title implies¹. It covered toothpastes only, not rinses. It required a mean and a standard deviation, so trials that reported their results another way were left out, and the reviewers say plainly that they cannot estimate whether that biased the answer¹.

Time is the other gap. The pooled readings are at six to eight weeks and the longest trial ran twelve, so the review has nothing to say about a year of use, about what happens when you stop, or about whether the effect keeps growing¹. Nothing published after September 2005 is in it. And no trial in it combined potassium with anything else, so a paste that carries potassium alongside a mineral or another active is outside the review's scope entirely — the page on potassium nitrate and hydroxyapatite in one tube takes that question on its own evidence, and the page on what happens after you stop takes the other one.

When was it last updated, and how would you check for yourself?

The record answers this in one line, and you can read it without a subscription. At the top of the review, under the author list, it says "Version published: 19 July 2006", followed by a link to the version history². Follow that link and you get two rows: the 2006 version, and the earlier one under its old title. There is no third row, no withdrawal notice and no banner saying the review has been replaced². The plain-language summary carries the same date and the same conclusion in plainer words: this review of trials found there was not enough evidence to show that potassium is effective in desensitising teeth².

Whether an update is coming is a separate question, and it has a separate answer. Cochrane registers a protocol before it writes a review, so a planned update becomes visible in the library before the review itself exists. A search of the Cochrane Library on 2026-09-10, in the September 2026 issue, returned two Cochrane reviews on dentine hypersensitivity and no protocols at all: this one, and a 2021 review of laser therapy². On that evidence, nobody is currently rewriting it.

None of which makes the review withdrawn, superseded or void. Those are specific words with specific meanings in this library, and none of them applies here. What is true is narrower and more useful: the current version was published on 19 July 2006, its searches closed in September 2005, and everything published since sits outside it². Anyone can check that in about thirty seconds at the Cochrane Library record, which is why the link is here rather than a summary of it.

What have the syntheses published since 2006 found?

Seven of them, and they do not line up. Two reach a conclusion close to Cochrane's, four reach a different one, and one sits between. Setting them side by side is more honest than averaging them, and averaging them would in any case be meaningless, since they use different comparators, different stimuli and different statistics.

Four other syntheses do not carry a pooled potassium figure but are worth a line each. A 2013 systematic review that found only seven usable studies concluded that the reported efficacy of both strontium and potassium pastes is questionable and the evidence for either is minimal¹¹. A 2019 systematic review of 74 randomised trials grouped the actives by when their effect showed up, and placed potassium nitrate in the long-term window only, never the immediate or medium-term ones¹². A 2015 systematic review of 105 randomised trials across eleven agents gave up on pooling altogether, because the trials varied too much among themselves¹³. And the strongest pre-Cochrane statement is a narrative review from 2000, which found sixteen double-blind randomised trials of potassium toothpastes and reported that every one of them showed a significant reduction — a narrative review, which is why it sits below the pooled ones rather than above them, and which also states that the nerve mechanism has never been confirmed in an intact human tooth¹⁴.

Now the arithmetic, which is the part almost nobody quotes. Cochrane pooled six trials¹. Fourteen years later a network meta-analysis drew on 125 randomised trials and 12,541 patients⁸. Six years after that, the newest network meta-analysis could find one contributing study for its potassium comparison at two weeks¹⁰. The literature did not shrink; the potassium literature did, because the trials moved to other actives. In that same network meta-analysis the class rated at high confidence was stannous fluoride, at −0.85 on the cold-air score from ten contributing studies, and S3 does not contain stannous fluoride¹⁰. That is the honest ranking of the current evidence, and it is not the one that suits us.

Why did the patients' own ratings not move when the instruments did?

Nobody knows, and the candidates are all plausible. The pooled subjective analysis was the smallest of the three, so it had the least chance of detecting anything¹. Pain is hard to score, and the review says the methods for scoring it disagree with each other¹. And there is the placebo response, which in this condition is large: a network meta-analysis of 30 randomised trials measured a significant placebo effect and found potassium not significantly different from placebo within its own network⁷. If everyone improves, including the people on the plain paste, a patient rating is a blunt instrument for telling the two apart.

The pattern has not gone away with better trials. In a manufacturer-funded randomised trial published in 2026, with four arms and 118 completers, the examiner-measured cold-air and probe readings separated the pastes clearly, while the patients' own rating and the sensitivity questionnaire did not, which the authors put down to differences between the groups at the start¹⁷. Twenty years on, the same gap between what the instruments record and what the person says.

What nobody has shown is that the gap is an artefact of measurement. It could be that the instruments are picking up a real change too small for a person to notice. It could be that they are picking up something that is not sensitivity at all. Both readings are live, and a page that told you which one to believe would be inventing the answer. The page on how desensitising toothpastes are tested sets out what each instrument actually does.

If the review is unconvinced, why does S3 use potassium nitrate?

Because of what the review found, not because of what it concluded. The pooled objective measures improved in both versions of it, and improved by a lot: −1.25 on the air blast in 2006, −1.51 in the earlier one¹³. That is a real signal in a small evidence base, and the reviewers' caution is about the base, not about the direction.

Because potassium is the one class of home-use active that addresses the nerve at all. A 2013 review published in a journal supplement, written by authors employed by Johnson & Johnson Consumer and closing by introducing that company's own potassium oxalate rinse, sets out the classification the field works with: potassium salts act on the nerve, while strontium salts, bioglasses, arginine with calcium carbonate and oxalates work by occluding the tubule¹⁵. If you want the nerve half of a sensitive tooth addressed, potassium is what the shelf offers. The page on the nerve mechanism is honest about how much of that mechanism is still a proposal, and the page on the five per cent dose is honest about the fact that no published comparison of one concentration against another exists.

And because S3 does not rest on it. A sensitive tooth is failing in two places at once: the nerve has become too easy to set off, and the tubules that lead to it are no longer sealed. Most sensitivity toothpastes are built to deal with one of those two, and not the other. The potassium is there for the nerve, the nano-hydroxyapatite for the inside of the tubule and the biomimetic hydroxyapatite for the surface, and no one of the three stands in for another. That is an argument about formulation, and it is not an argument that the Cochrane reviewers were wrong. They read the trials that existed in 2005 and reported what they saw¹. No review on this page tested this formula, and none of them is evidence about it.

What should someone with a potassium toothpaste in the bathroom do?

Give it the time the trials gave it. The pooled readings in the Cochrane review are at six to eight weeks of unbroken twice-daily brushing¹. The 2019 systematic review that grouped 74 trials by follow-up window puts potassium nitrate among the long-term actives, not the quick ones¹². A fortnight is not a test. A week is not even the beginning of one, and the page on how long potassium nitrate takes has the trial-by-trial timings. The Journal's article on how sensitive toothpastes work covers the everyday version of the same point.

Keep using it, and treat it as your ordinary toothpaste. The Oral Health Foundation still names potassium citrate, potassium nitrate and stannous fluoride as the actives in toothpastes that help with sensitivity, advises brushing gently with a soft brush, advises spitting rather than rinsing afterwards, and says the effect has to be maintained by continued use¹⁸. Its explanation of how sensitive toothpastes work is that they block the tiny channels in dentine, which is a fair description of the mineral and stannous pastes and not of what potassium is thought to do. A UK charity naming potassium as an active it recommends, next to a Cochrane review that will not vouch for the class, is exactly the sort of disagreement this subject is full of, and it does not resolve neatly. The Journal's piece on the two causes of sensitive teeth explains why the two mechanisms are not interchangeable.

Being treated for this usually helps, even outside a trial¹⁶. In a prospective cohort of patients seen in ordinary American dental practices, people rated their own pain across eight weeks, and most of them ended the period better than they began it — with no control group, so nobody can say how much of that belonged to the treatment¹⁶. Hold that next to the placebo finding above: both are true at the same time, and neither cancels the other.

And know when the answer is not a toothpaste. The Oral Health Foundation lists severe pain, pain in a single tooth, pain that starts suddenly and sensitivity lasting more than a few weeks as reasons to see a dentist, because those can point to decay, a crack, gum problems or infection¹⁸. The NHS is blunter about toothache: if it has lasted more than two days, book an appointment, and go to A&E if there is swelling around the eye or the neck¹⁹. No reading of any review changes that.

Frequently asked questions

Has the Cochrane review been updated?

No. The version on the Cochrane Library in September 2026 was published on 19 July 2006, the version history shows it and one predecessor and nothing else, and the searches behind it closed in September 2005². A search of the library on the same day found no protocol registered for a new review of this condition, which is the signal that would appear first if an update were being written².

That is a statement about the record, not a verdict on the review. "Not updated since 2006" is not the same as withdrawn, and it is not the same as wrong. It does mean that twenty years of trials sit outside it, and that anyone quoting its conclusion without its date is quoting half of it.

Does the Cochrane review mean S3's potassium nitrate does not work?

No, and it does not mean the opposite either. The review found no clear evidence to support potassium toothpastes as a class, on six pooled trials, in 2006¹. It has not been updated. The syntheses published since disagree with each other: two land where Cochrane landed, four find in potassium's favour to varying degrees, and one is mixed⁵⁷⁸¹⁰.

No review anywhere has tested this formula, so none of them is evidence about this product in particular. What is checkable is what is in the tube: 5% potassium nitrate, the concentration used in all six of the trials the review pooled, alongside two forms of hydroxyapatite and full adult-strength fluoride.

Why did the patients' own ratings not improve when the instrument readings did?

Nobody has established why. The pooled patient rating came from three trials and 108 patients, the smallest of the review's three analyses, and its interval ran from −1.44 to 0.10, so it was also the least able to detect anything¹. Sensitivity is a subjective pain and the ways of scoring it disagree, which the review says in its own discussion¹.

The placebo response is the other candidate, and it is measurable: a network meta-analysis of 30 randomised trials found a significant placebo effect in this condition and no significant separation between potassium and placebo⁷. What has not been shown is that the objective tests are measuring something the person cannot feel. That would be a convenient conclusion and there is no evidence for it.

Which of these reviews should I believe?

All of them, as evidence of how unsettled this is. The useful way to read the set is by comparator: reviews that test potassium against an identical paste without it tend to be cautious, and reviews that test it against a plain fluoride paste in a network of many actives tend to find a modest benefit. The 2026 network meta-analysis is the largest and the newest, and its potassium estimate rests on a single contributing study at low confidence, which tells you how thin the recent evidence for this particular active is¹⁰.

The one reading that is not available is an average. These reviews use different comparators, stimuli, statistics and years, and a number produced by splitting the difference between them would correspond to nothing that was ever measured.

Is potassium nitrate still recommended in the UK?

Yes, by the Oral Health Foundation, which lists toothpastes containing potassium citrate, potassium nitrate or stannous fluoride among the things that help with sensitivity and advises keeping up their use to hold the effect¹⁸. That advice and the Cochrane conclusion have coexisted for twenty years without either being retracted.

If sensitivity is severe, sudden, confined to one tooth or lasting more than a few weeks, the recommendation changes: that is a reason to see a dentist rather than to change toothpaste¹⁸. The NHS says the same about any toothache that has lasted more than two days¹⁹.