Bae 2015: desensitising toothpastes versus placebo across the main actives.

Bae, Kim and Myung's 2015 meta-analysis of 31 randomised trials found that toothpastes containing potassium, stannous fluoride, potassium with stannous fluoride, calcium sodium phosphosilicate or arginine each reduced dentine hypersensitivity more than a placebo toothpaste, and that strontium toothpastes did not¹. It cannot tell you which of the five is best, because each active was compared with placebo rather than with the others, each estimate rests on three to eight trials, and the review reported that those trials disagreed far more than chance explains (I² between 86% and 95%)¹. S3 Sensitivity Science™ contains potassium nitrate at a clinically proven 5% dose, a member of the potassium group Bae pooled, together with two forms of hydroxyapatite that the review did not include.

What was checked12 peer-reviewed publications (ten systematic reviews, one randomised trial, one commentary), NHS and Oral Health Foundation pages, product information as published by each brand

What did Bae, Kim and Myung set out to answer, and how?

They asked one question: does a desensitising toothpaste reduce dentine hypersensitivity more than a placebo toothpaste¹? Ji-Hyun Bae and colleagues at Seoul National University Bundang Hospital and the National Cancer Center, Korea, searched PubMed, CENTRAL and Embase on 20 December 2013, identified 626 articles and kept 31 randomised trials for the meta-analysis, publishing the result in the Journal of Clinical Periodontology (doi:10.1111/jcpe.12347)¹. A 2016 commentary in Evidence-Based Dentistry adds that the trials held 2436 patients, that two authors assessed each trial with the Cochrane risk-of-bias tool, and that 23 trials were judged moderate quality, three high and five low².

The abstract does not say what the placebo pastes contained, which trials sit under each active, which stimulus (a probe, a puff of air) was pooled, or at what time point. We could not reach the full text: the publisher's page refused access on 13 September 2026, so each of those details stays unknown on this page rather than guessed¹. The review declares no funder in its PubMed record, and none of its authors' listed affiliations is a toothpaste maker¹.

Table A — the review, as reported

Which actives beat placebo, and by how much?

Five of the six groups beat placebo and one did not: the table below reproduces the six estimates exactly as the abstract gives them, with the number of trials behind each and the heterogeneity the review reported¹. A negative number favours the desensitising paste.

The unit needs a sentence of its own, because it is easy to misread. A standardised mean difference (SMD) takes the gap between the two groups' average pain scores and divides it by how widely those scores were spread, which lets the review pool trials that used different pain scales on the same footing¹. By the usual statistical convention all five significant estimates would be called large, but an SMD of −1.28 is not 1.28 points on any scale a patient fills in, and it does not tell you how much less a cold drink will hurt¹. It says how far apart the groups were, in units of spread.

Strontium's result is the one plain negative in the review: across four trials the estimate was 0.05, and its interval from −0.34 to 0.44 runs either side of no effect¹. That is a finding of no detectable difference from placebo in those four trials, not proof that strontium can never help anyone¹.

Why can the review not tell you which active is best?

Because none of the 31 trials was pooled as a head-to-head contest: every estimate is a separate comparison against placebo, drawn from a different set of trials, run on different people with different stimuli¹. A bigger number from one set of trials and a smaller number from another is not a measured difference between the two actives¹.

Heterogeneity is the second reason. I² is the share of the variation between trials that is larger than chance alone would produce; the review reported values from 86% to 95% for the five significant actives, which means the trials under each heading disagreed with one another so much that a single pooled number averages results that were very far apart¹. At that level the pooled figure answers "on average, across these trials, did the paste do better than placebo?" and very little else¹.

Read the intervals and they say less than the point estimates seem to. Potassium (−2.05 to −0.51) and stannous fluoride (−2.30 to −0.44) overlap almost entirely¹. Arginine's interval (−3.87 to −2.63) does not overlap either of those two, yet it overlaps both calcium sodium phosphosilicate (−3.72 to −1.00) and potassium with stannous fluoride (−4.10 to −0.91)¹. Because those are separate comparisons against placebo in different trials, even the gap between arginine and potassium is a hint, not a head-to-head result¹. An earlier meta-analysis of 14 arginine trials by Yan and colleagues (Quintessence International, 2013) did report arginine ahead of potassium salts at moderate quality, but its abstract does not say who paid for the trials it pooled⁹.

The third reason is time. The search closed on 20 December 2013¹. A 2022 scoping review, a systematic review that counts trials rather than pooling them, catalogued 138 randomised trials and 368 formulations, and found hydroxyapatite as the active in 2% of those formulations against 19% for potassium compounds¹¹. Its search ran to August 2021, and it reports that the number of formulations tested grew fastest after 2010, so part of that literature was published after Bae stopped looking¹¹.

What have the reviews since found, and do they agree?

They agree on the first half of Bae's answer and not on the second. Most of the syntheses below found desensitising actives ahead of a non-desensitising paste, with the 2011 oxalate review the exception; none produced an order of the actives that the others confirm, and the order with the most certainty rests on trials that were mostly industry-funded³⁵⁷¹⁰. The table puts them side by side; each reports its effect in its own unit and against its own comparator, and those units are not converted here because converting them would invent a precision none of the reviews claims³⁴⁵⁷.

Table B — the other syntheses, before and since

Hu and colleagues' 2018 review of 53 trials and 4,796 patients is the closest to a repeat of Bae: pastes with an active beat a paste without one, strontium and amorphous calcium phosphate did not, and potassium-only paste was rated moderate-quality evidence, while potassium with stannous fluoride was rated low³. The same group's 2019 network of 30 trials then found potassium paste not significantly different from placebo, and measured a significant placebo effect across the trials⁴. That second finding is the one to hold on to when reading any number in this field: part of the improvement happens in the comparison arm too.

Strontium is where the reviews openly contradict each other. Bae found no effect across four trials and Hu 2018 advised against strontium-only pastes¹³. West and colleagues' 2015 review of 105 trials, by contrast, listed strontium among the pastes that appeared effective⁸. Martins and colleagues' 2020 network reported a large tactile benefit for strontium over fluoride paste at moderate certainty⁵. We report both sides; the reviews used different comparators, and the page does not average them.

Pollard and colleagues' 2023 network of 32 trials concluded that stannous, potassium with or without stannous, and arginine pastes can be recommended; its Bristol authors declare advisor fees, lecturer fees or research grants from GSK, Colgate-Palmolive, Sunstar, Procter & Gamble and Unilever, each a manufacturer of toothpaste⁶. Gormley and colleagues' 2026 network of 93 trials, whose senior author led that Bristol group, rated stannous fluoride's two-week benefit over a benchmark fluoride paste as high confidence on ten studies, and it reported that 96% of stannous studies, 86% of arginine studies and 76% of potassium studies were industry-funded⁷. Its authors still advised choosing between stannous fluoride and arginine by preference, tolerability and availability rather than by an expectation that one works better⁷.

Funding does not prove bias, and the page does not infer it⁷. It does mean that the most certain result in the table, stannous fluoride's, comes from trials almost all funded by industry, and potassium, the active S3 carries, is no exception: 76% of its studies were industry-funded⁷. The Cochrane review of potassium toothpastes is read on its own page, the Cochrane review on potassium toothpastes, and not again here.

S3 pairs potassium nitrate with hydroxyapatite: what do Bae and the later reviews say about that?

Bae says nothing about it, because the review had no hydroxyapatite arm at all¹. The only combination Bae could pool was potassium with stannous fluoride, across three trials; S3 contains potassium nitrate but no stannous fluoride, and hydroxyapatite, the other active ingredient in S3, was not in the review¹.

The later networks say a little, and it is thin. Martins and colleagues' 2020 network includes a potassium-plus-hydroxyapatite node with an SMD of 2.47 against fluoride paste on the tactile stimulus, but that node rests on two trials and 140 patients, it has no cold-stimulus estimate, and the review's appendix does not name the two trials⁵. Hu's 2019 network ranked nano-hydroxyapatite on its own as the most probable best paste at two and four weeks, and Gormley's 2026 network gave nano-hydroxyapatite a moderate-confidence two-week result from two studies, but a ranking probability is not a demonstrated difference, and neither review reports a paste combining hydroxyapatite with potassium nitrate⁴⁷.

The costly sentence has to be said plainly. A 2019 network meta-analysis found potassium toothpastes not significantly different from placebo, and the 2026 network rated potassium's two-week effect small, from one study, at low confidence⁴⁷. No published trial on this page tested S3. What S3 can point to in this literature is that potassium, the class its potassium nitrate belongs to, is one of the five actives Bae found ahead of placebo, and nothing more than that¹.

Where are these actives on the UK shelf?

Every active Bae found ahead of placebo is on the UK shelf except the one combination, and no strontium paste turned up at all¹. This page compares ingredients and stated actions only, not clinical performance, based on what each brand states about its own formula. The rows below come from the category review, read from Boots listings on 9 September 2026 for Sensodyne, Colgate and Oral-B and from the brand product page for S3 on the same day. Declared ingredients, levels and prices may change, so check the ingredient list on the tube you are holding.

Table C — the UK shelf by declared active

Two limits sit under the last row. The 10% nano-hydroxyapatite and 5% biomimetic hydroxyapatite figures are inclusion levels as supplied, and the active hydroxyapatite content in the paste is lower. Of the three potassium products, every one lists potassium nitrate but just one prints a level, so on two of them the dose cannot be read off the Boots listing. The Bae abstract states no concentration for the trials it pooled either¹.

To check a row yourself, turn the tube over: the active appears by its ingredient name (potassium nitrate, stannous fluoride, arginine, calcium sodium phosphosilicate), and a level, where one is printed, sits in the declaration near the fluoride content. For how these actives are meant to work, see the Journal's explainer on how sensitive toothpastes work, and for the four actives side by side, the page comparing potassium nitrate, arginine, strontium and stannous fluoride.

When is a different toothpaste, a dentist, or no toothpaste the better answer?

This page is written by S3 Science, the company behind S3 Sensitivity Science™, a toothpaste that pairs potassium nitrate with two hydroxyapatites and fluoride. Read it with that in mind, because several honest answers to the reader's question do not end at S3.

If you want the active with the most certain evidence in the newest network, that is stannous fluoride: ten studies and a high-confidence two-week result against a benchmark fluoride paste, from trials 96% of which were paid for by industry⁷. S3 does not contain stannous fluoride, and four of the pastes in Table C do.

If you want an active with trials behind it in both the oldest and the newest pooled analyses, arginine has eight trials in Bae and seven in the 2026 network's two-week cold-air comparison¹⁷. Calcium sodium phosphosilicate was the most beneficial formulation across stimuli in Martins' 2020 network of 125 trials⁵. Each is declared by a paste in Table C.

If the symptom is not dentine sensitivity, no toothpaste is the answer. The Oral Health Foundation page, in the copy saved on 2026-09-10, advises a dental visit when the pain is severe, when sensitivity continues beyond a few weeks, when a single tooth is involved or when it starts suddenly, because those can point to decay, a crack, gum problems or infection¹⁴. The NHS toothache page lists sensitive teeth as only one of several causes of tooth pain, alongside decay, abscesses, cracked teeth and gum disease, and it tells readers with toothache that persists past a couple of days, or does not settle with painkillers, to book a dentist rather than their GP¹³.

A dentist can also do things a tube cannot. In a practice-based randomised trial of 75 patients in general dental practice, published by Gibson and colleagues in the Journal of Dentistry in 2013, a desensitising toothpaste and a dentine bonding agent applied in the chair both reduced air-blast sensitivity, and the bonding agent gave the greatest reduction at two weeks and at six months¹². The abstract does not name the active in that toothpaste, so the trial shows what the chair can add, not which paste to buy¹². What a dentist looks for is set out in what a dentist checks for sensitive teeth.

Frequently asked questions

Which sensitivity toothpaste ingredient has the best clinical evidence?

No review has settled it. Bae's 2015 meta-analysis showed five actives ahead of placebo without ranking them¹, the 2019 network put nano-hydroxyapatite and arginine at the top of its probability rankings while finding potassium no different from placebo⁴, and the 2026 network gave stannous fluoride the only high-confidence two-week result, from trials mostly funded by industry⁷. The fair summary is that several actives work better than a plain paste, and which one suits you depends on tolerance, taste and what your dentist finds¹⁷. The Journal's guide to choosing a toothpaste for sensitive teeth covers the same question from the shelf side.

Does strontium toothpaste work?

The reviews disagree. Bae found no effect across four trials (SMD 0.05, interval −0.34 to 0.44) and Hu 2018 found strontium pastes did not beat a negative control¹³; West 2015 listed strontium among the agents that appeared effective, and Martins 2020 reported a tactile benefit over fluoride paste at moderate certainty⁸⁵. No strontium paste appeared in the recorded rows of the UK category review searched for this page.

Is S3 covered by the Bae meta-analysis?

Only in part. S3's potassium nitrate belongs to the potassium group Bae pooled across eight trials, but its two hydroxyapatites were not in the review, and no trial in Bae or in any review on this page tested S3 itself¹.

What does "placebo" mean in a toothpaste trial?

It means a toothpaste without the desensitising active, not no toothpaste. The Bae abstract does not describe what its placebo pastes contained, and we could not read the full text¹. What the later work shows is that the comparison arm improves too: the 2019 network measured a significant placebo effect and found fluoride paste no different from placebo on sensitivity⁴.

Would an NHS dentist recommend a particular sensitivity toothpaste?

Neither public page read for this article names a brand or ranks one active above another. The Oral Health Foundation page suggests pastes containing potassium citrate, potassium nitrate or stannous fluoride, and does not mention arginine or calcium sodium phosphosilicate¹⁴; the NHS toothache page does not discuss sensitivity toothpaste at all and points readers to the foundation¹³. Individual dentists recommend by what they see in your mouth, which is set out in what toothpaste UK dentists recommend for sensitive teeth.