8 How much fluoride should be in your toothpaste?
Every chapter so far has ended by saying the evidence is thinner than the label suggests. This one does not, and the book would be worth very little if no chapter did.
Fluoride toothpaste is the intervention that earns the Strong label in every brushing recommendation in Delivering Better Oral Health. It is the “main component” that chapter 13’s drafting rule refers to. It works, we know it works, and I would defend a Strong recommendation for it against anybody.
What I want to show here is that even this recommendation contains a bundle, and that the second half of it is a great deal weaker than the first.
The advice
- with toothpaste containing 1,350 to 1,500ppm fluoride
Strength of recommendation: Strong
— Delivering Better Oral Health, chapter 2, tables 1d and 1f (1)
For children under six the figure is “at least 1,000ppm,” rising to 1,350 to 1,500ppm for those at higher risk.
Read as a reader would read it, this single bullet makes two different claims:
- use fluoride toothpaste rather than a fluoride-free one;
- use one in the 1,350 to 1,500ppm band rather than a weaker one.
Hold those apart, because the evidence does not treat them alike.
What the guideline says its evidence is
Recommendation based on moderate certainty evidence for toothbrushing with fluoride toothpaste and concentration of 1,000ppm fluoride and above (the evidence for toothpaste 500 to 1,000ppm fluoride is inconclusive).
— Delivering Better Oral Health, chapter 13, table 1 (1)
and for children at higher risk:
Recommendation based on moderate certainty evidence of added benefit over 1,000ppm fluoride.
— Delivering Better Oral Health, chapter 13, table 3 (1)
The citation is the Cochrane review of fluoride toothpaste concentrations (2). Unlike the previous three chapters, this citation is exactly on point: the review’s question is the recommendation’s question.
Following the citation
Walsh and colleagues updated their 2010 review in 2019, including 96 randomized trials published between 1955 and 2014, and analyzed the permanent dentition of children and adolescents in a network meta-analysis of 81 studies covering 21 comparisons across seven concentration bands.
This is a serious piece of work. Searches ran across the Cochrane trials register, CENTRAL, MEDLINE and Embase to August 2018, with two people independently selecting studies, extracting data and assessing risk of bias. It is the kind of review the previous chapters wished they had.
Claim 1: use fluoride toothpaste
| Comparison | Population | Effect | 95% CI | Studies | Certainty |
|---|---|---|---|---|---|
| 1,000 to 1,250ppm vs no fluoride | children, permanent teeth | SMD −0.28 | −0.32 to −0.25 | 55 | high |
| 1,450 to 1,500ppm vs no fluoride | children, permanent teeth | SMD −0.36 | −0.43 to −0.29 | 4 | moderate |
| 1,500ppm vs no fluoride | young children, primary teeth | MD −1.86 dfs | −2.51 to −1.21 | 1 | moderate |
| 1,000 or 1,100ppm vs no fluoride | adults | MD −0.53 DMFS | −1.02 to −0.04 | 3 | moderate |
Fifty-five randomized trials, at high certainty for the 1,000 to 1,250ppm comparison and moderate for the 1,450 to 1,500ppm one. An earlier draft printed both as “high to moderate,” which is not something GRADE does: a certainty rating attaches to one estimate and takes one of four values, not a range. Consistent direction, tight confidence intervals nowhere near zero, and replicated in primary teeth and in adults.
This is what a Strong recommendation is supposed to look like. Note in particular what the previous chapters were missing and this one has: the studies are randomized, they measure tooth decay rather than a chemical proxy, and they compare the thing the advice tells you to do against the thing it tells you not to do.
Claim 2: use 1,350 to 1,500ppm specifically
| Comparison | Effect | 95% CI | Studies | Certainty |
|---|---|---|---|---|
| 1,450 to 1,500 vs 1,000 to 1,250ppm | SMD −0.08 | −0.14 to −0.01 | 10 | moderate |
| 1,055 vs 550ppm (primary teeth) | MD −0.05 dmfs | −0.38 to 0.28 | 2 | moderate |
| 1,450 vs 440ppm (primary teeth) | MD −0.34 dmft | −0.59 to −0.09 | 1 | moderate |
| 1,700 to 2,200 vs 1,450 to 1,500ppm | SMD 0.04 | −0.07 to 0.15 | indirect only | moderate |
| 2,400 to 2,800 vs 1,450 to 1,500ppm | SMD −0.05 | −0.14 to 0.05 | 2 | moderate |
The step from ordinary 1,000ppm toothpaste to the 1,450 to 1,500ppm band is SMD −0.08, 95% CI −0.14 to −0.01. The interval excludes zero, so the effect is probably real, and it is about a quarter the size of the effect of using fluoride toothpaste at all. The upper bound of the interval, −0.01, is an effect of essentially nothing.
The review’s own summary of this half of its work is worth quoting, because it is notably more cautious than the recommendation it supports:
Evidence for the effects of different fluoride concentrations is more limited, but a dose-response effect was observed for D(M)FS in children and adolescents. For many comparisons of different concentrations the caries-preventive effects and our confidence in these effect estimates are uncertain and could be challenged by further research.
— Walsh and colleagues, 2019 (2)
And note the review’s own statement about the network: “The certainty of the evidence was judged to be low” apart from the handful of comparisons listed above. Most of the 21 comparisons in that network are low certainty. Five are not, and the guideline uses three of them, which is the correct way round and worth crediting. The other two are the higher-concentration comparisons in the second table, both moderate. I said four in an earlier draft.
What those two comparisons show needs stating carefully, because I first wrote that they show “no added benefit” and that is not what a confidence interval spanning zero says. SMD −0.05 (95% CI −0.14 to 0.05) is compatible with a small benefit, with nothing, and with a small harm. The honest reading is that a benefit of the size these trials could have detected was not found, not that the benefit is absent.
A standardized mean difference of 0.28 means the groups differ by 0.28 of a standard deviation of the caries increment. That is not a quantity anyone has intuitions about, so the honest thing is to translate it.
Caries increment distributions in these trials are heavily skewed, with standard deviations typically somewhere between 2 and 4 D(M)FS over three years. Taking that range, an SMD of 0.28 corresponds very roughly to 0.6 to 1.1 fewer decayed, missing or filled permanent surfaces per child over three years, and the SMD of 0.08 for the concentration step to something like 0.15 to 0.3 surfaces.
Two cautions. Back-transforming an SMD this way is approximate, and it is approximate in a direction that matters: with skewed data the standard deviation is a poor summary, and the average conceals that most of the benefit accrues to the minority of children who would have developed most of the lesions. The review reports the pooled result as an SMD precisely because the included trials used different indices, and my arithmetic here does not recover information the review did not have.
I give it anyway, with the caveat attached, because “0.28 of a standard deviation” invites the reader to nod without understanding, and a rough number with a stated error is more honest than a precise number nobody can picture.
Where I overstated this before
The blog post said:
Among different doses, 1450–1500 ppm F toothpaste was the only one which led to lower tooth decay in children and adolescents compared to all the lower doses.
That is not what the review found, and it is the kind of sentence this book exists to object to.
The review reports one head-to-head comparison in which 1,450 to 1,500ppm beat a lower band: against 1,000 to 1,250ppm, at SMD −0.08. It did not find that band superior to “all the lower doses.” In primary teeth the comparison of 1,055 against 550ppm found essentially nothing (MD −0.05 dmfs, 95% CI −0.38 to 0.28). And I gave no confidence interval anywhere, which turned an estimate whose upper bound is nearly zero into a flat statement of fact.
I have also dropped the accompanying claim that “amounts over 1500 ppmF didn’t perform better.” The two relevant comparisons are SMD 0.04 (95% CI −0.07 to 0.15), from indirect evidence only, and SMD −0.05 (95% CI −0.14 to 0.05) from two studies. Those intervals include worthwhile benefit as well as none. “No evidence of a difference” is right; “didn’t perform better” implies a demonstrated equivalence that these data cannot support. That is the absence-of-evidence error, committed by me, in a post complaining about it.
The Cochrane review’s own search ran to 15 August 2018 across the Cochrane Oral Health trials register, CENTRAL, MEDLINE and Embase, with two trial registries for ongoing studies. I have not attempted to reproduce that; a single person with PubMed cannot improve on it.
I searched PubMed on 15 August 2026 for any update or successor to CD007868.pub3. None was found; the 2019 version remains current. Full record in appraisals/searches/.
The gap is worth stating plainly: the newest trial in the review was published in 2014. The evidence base for the concentration in your toothpaste is now more than a decade old, and the review’s authors say themselves that these estimates “could be challenged by further research.”
The bundle, one layer down
So the single bullet “with toothpaste containing 1,350 to 1,500ppm fluoride” contains a very strong claim and a weak one, welded together.
Use fluoride toothpaste is supported by 55 randomized trials at high to moderate certainty with an effect four times larger. Use this particular band rather than an ordinary 1,000ppm one is supported by ten studies at moderate certainty with an effect whose confidence interval nearly touches zero.
Both get the same three words of typography and the same Strong label.
This is exactly the pattern of Chapter 3, and finding it inside the best-evidenced component of the guideline is, I think, the strongest form of the argument. It is not that DBOH picked bad evidence. It picked excellent evidence, described it accurately in chapter 13, and then wrote a summary table that cannot transmit the distinction.
Verdict
- Certainty of evidence
- For “use fluoride toothpaste”: high against no fluoride at 1,000 to 1,250ppm, from 55 randomized trials, consistent, precise and direct; moderate for the same comparison at 1,450 to 1,500ppm, from four. For “use 1,350 to 1,500ppm rather than 1,000ppm”: moderate, for a small effect (SMD −0.08, 95% CI −0.14 to −0.01), and from children and adolescents rather than adults.
- Directness to the advice as worded
-
Excellent for children and adolescents, indirect for adults. The review’s question is the recommendation’s question: right comparison, right outcome, randomized, followed for at least a year.
But the adult recommendation names the 1,350 to 1,500ppm band, and the only adult evidence in this review is 1,000 or 1,100ppm against no fluoride (MD −0.53 DMFS, 95% CI −1.02 to −0.04, 3 studies). The band itself is evidenced in immature permanent dentition in children and adolescents, and DBOH says so: “Evidence for the concentration is based on studies on immature permanent dentition in children and adolescents.”
The other indirectness worth noting is the age of the evidence: the newest included trial is from 2014, and background fluoride exposure has changed.
- Is the strength label defensible?
- Yes for the first claim, and emphatically so. This is what a Strong recommendation should rest on, and I would defend it against anyone. Arguable for the second. An effect of SMD −0.08 whose interval reaches −0.01 is thin ground for a Strong recommendation to buy one product rather than another. Conditional would fit it better. But because the two claims share a bullet, no reader can tell that they are differently supported.
- What would change my mind
- For the first claim, essentially nothing short of a large contrary trial, and I do not expect one. For the second, an updated review including trials published since 2014, or any head-to-head trial of 1,450ppm against 1,000ppm powered for a patient-important outcome rather than a surface count, would move the estimate either way. I would also revise this if someone showed that the SMD back-transformation above is misleading in a direction I have not accounted for.
What this means for you
Use fluoride toothpaste. This is the single best-supported thing in this book and it is not close: dozens of randomized trials, measuring actual tooth decay, consistently finding less of it. If you take one instruction away from these pages, take this one.
The specific number on the tube matters much less. Ordinary adult toothpaste in the UK is already 1,350 to 1,500ppm, so for most people this is a decision they never have to make. If you find yourself choosing between a 1,000ppm paste and a 1,450ppm one, the higher one is probably very slightly better, and “very slightly” is doing real work in that sentence. What is not worth doing is paying a premium for anything above 1,500ppm on your own initiative. For preventing decay on the crowns of teeth, the higher concentrations have not been shown to add anything, and have not been shown not to.
Root caries is the exception, and I left it out of an earlier draft. DBOH’s evidence statement for the 2,800 and 5,000ppm row does not stop at the sentence I quoted. It ends: “Moderate-certainty evidence for effectiveness of 5,000ppm fluoride for root caries.” Root caries is decay on the exposed root surface, which is largely an older-adult problem and a different disease process from decay on the crown. Trimming the statement before that clause made the guideline look as though it were prescribing on no evidence when it had named the one indication where the evidence is strongest.