11 Fissure sealants
A back tooth is not a smooth dome. Its biting surface is a landscape of pits and narrow grooves, some of them narrower than a toothbrush bristle. Those grooves are where decay in permanent molars usually starts, for the straightforward reason that nothing can clean them.
A sealant is a resin painted into the grooves to fill them in. The logic is almost too obvious to need testing, which is exactly the situation this book keeps warning about. So it is a relief to report that it was tested, thoroughly, and it works.
The interesting problem in this chapter is when.
The advice
Apply resin sealant to permanent teeth on eruption
Strength of recommendation: Strong
— Delivering Better Oral Health, chapter 2, table 1e (1)
This sits in table 1e, for children from 7 years and young people up to 18 giving concern because of dental caries risk. It is not universal advice.
What the guideline says its evidence is
Recommendation based on moderate certainty evidence of a benefit of resin-based sealant maintained up to at least 48 months of follow-up in both low risk and high risk populations.
— Delivering Better Oral Health, chapter 13 (1)
The citation is the Cochrane review of pit and fissure sealants (2).
Following the citation
Thirty-eight trials, 7,924 children aged 5 to 16. Fifteen compared resin-based sealant against no sealant; three compared glass ionomer against no sealant; twenty-four compared one sealant against another.
The headline estimate below is narrower than that whole set: it is the seven trials in children aged 5 to 10 at 24 months. The review says the effect was maintained at longer follow-up, but that “evidence quality and quantity was reduced,” so the 48-to-54-month figure should be read as supporting rather than as separately certified.
The headline result is one of the largest in this book:
| Comparison | Effect | 95% CI | Trials | Certainty |
|---|---|---|---|---|
| Resin sealant vs none, 24 months, children aged 5 to 10 | OR 0.12 | 0.08 to 0.19 | 7 | moderate |
| Resin sealant vs none, 48 to 54 months | OR 0.21 | 0.16 to 0.28 | 4 | not separately rated |
| Glass ionomer vs none, 24 months | inconclusive | — | 3 | very low |
| One sealant type vs another | unknown | — | 24 | very low |
An odds ratio of 0.12 is an enormous effect. The review translates it into absolute terms, which more reviews should do. Quoting it exactly:
If we were to assume that 16% of the control tooth surfaces were decayed during 24 months of follow-up (160 carious teeth per 1000), then applying a resin-based sealant would reduce the proportion of carious surfaces to 5.2% (95% CI 3.13% to 7.37%). Similarly, assuming that 40% of control tooth surfaces were decayed (400 carious teeth per 1000), then applying a resin-based sealant would reduce the proportion of carious surfaces to 6.25% (95% CI 3.84% to 9.63%). If 70% of control tooth surfaces were decayed, there would be 19% decayed surfaces in the sealant group (95% CI 12.3% to 27.2%).
— Ahovuo-Saloranta and colleagues, 2017 (2)
An earlier draft reproduced those three lines as the chapter’s own illustration. Read them again: a baseline of 16% and a baseline of 40% give almost the same sealed risk, 5.2% against 6.25%. One odds ratio cannot do that. Two and a half times the baseline risk cannot leave the treated risk essentially unchanged.
The standard conversion from an odds ratio to an absolute risk, given a control risk \(p_0\), is
\[p_1 = \frac{\text{OR} \times p_0}{1 - p_0 + \text{OR} \times p_0}\]
Applying it to the odds ratio the review prints in the same sentence:
| Assumed control risk | The review says | The odds ratio gives | Absolute reduction |
|---|---|---|---|
| 16% | 5.2% (3.13 to 7.37) | 2.2% (1.5 to 3.5) | 13.8 points |
| 40% | 6.25% (3.84 to 9.63) | 7.4% (5.1 to 11.2) | 32.6 points |
| 70% | 19% (12.3 to 27.2) | 21.9% (15.7 to 30.7) | 48.1 points |
I am not accusing the review of inventing numbers, and I do not know how the published figures were derived; the second column is what its abstract says. What I can say is that the three printed values are not transformations of the odds ratio printed beside them, and that the middle row is the one that gives it away.
The lesson is the book’s own, turned on the book. I copied an absolute-risk translation out of a Cochrane abstract because it was helpfully provided, and did not spend the two minutes needed to check it against the relative effect in the same paragraph. A book about following citations to their source has no business reproducing arithmetic it has not done.
The qualitative point survives either way, and is in fact stronger under the correct transformation: the higher the baseline risk, the more decay sealing prevents in absolute terms, rising from about 14 percentage points to about 48. This is why the recommendation targets children at higher risk, and it is the guideline reading its evidence correctly.
The problem with the seven trials
Now the caveat, and it is a substantial one that DBOH does not mention.
The 24-month estimate, the moderate-certainty headline, comes from seven trials. Five of them were published in the 1970s. Two are from the 2010s.
That matters for the same reason it matters in Chapter 24: caries rates were far higher then, so the amount of decay a sealant could prevent was correspondingly greater. A child in 1976 with unsealed molars was at much higher risk than a child today.
The review adds a detail that makes this harder to resolve:
Trials rarely reported background exposure to fluoride of trial participants or baseline caries prevalence.
— Ahovuo-Saloranta and colleagues, 2017 (2)
So the one variable that would let you adjust for the era mostly was not recorded. I should be precise about what that does and does not license me to say. An earlier draft called these a “pre-fluoride-toothpaste population,” which the review does not support: of the seven trials, three were conducted in fluoridated-water areas, two stated the water was not fluoridated, and two did not report it. The trials date from 1976 to 1979, by which point fluoride toothpaste was already widely used in several of the countries involved. The honest version is that background fluoride exposure was mostly unrecorded, not that it was absent.
There is one risk-of-bias problem in this literature that no amount of care can fix, and the review is admirably direct about it:
Although studies were generally well conducted, we assessed blinding of outcome assessment for caries at high risk of bias for all trials (blinding of outcome assessment is not possible in sealant studies because outcome assessors can see and identify sealant).
An examiner scoring a tooth can see whether it has been sealed. There is no way to hide it. So every trial in this literature is at high risk of detection bias, permanently.
Now the part I got wrong in an earlier draft, and it matters because this book spends a lot of its time complaining about GRADE domains being assigned to the wrong thing. I wrote that the moderate rating “already reflects a downgrade for” that detection bias. It does not. The review’s own summary-of-findings footnote says the opposite, twice over:
Downgraded by one level due to indirectness of evidence (5 trials were published between 1976 and 1979; it is not unreasonable to anticipate that a new study could change this estimate, even if it found sealants to be effective). … Nor did we downgrade evidence on the basis of overall risk of bias classification of the studies (all studies were assessed at high risk of bias because blinding of outcome assessment is not possible in sealant studies).
— Ahovuo-Saloranta and colleagues, 2017 (2)
So the single downgrade from high to moderate is for the age of the evidence, which is the caveat this chapter is about, and the unblindable detection bias was noted and deliberately not penalised.
An earlier draft of mine went on to defend that as harmless, on the grounds that a bias every trial shares equally cannot explain a difference between arms. That argument is wrong and I want to retract it explicitly, because it is the kind of error this book exists to catch.
Sharing a bias across studies is not the same as a bias that cancels within each study. If an examiner who can see a sealant classifies borderline surfaces differently because of it, the misclassification runs in the same direction in every trial, and pooling twenty trials averages away random error while leaving that direction untouched. Whether such an effect is large enough to matter here is a separate question, and the review’s judgment not to downgrade is defensible on its own terms. What is not defensible is my claim that shared bias is automatically safe.
The practical consequence is unchanged: the moderate badge is not doing the work I first said it was, and a reader who wants to discount for detection bias has to do it on top of moderate rather than assume it is already inside.
This is worth contrasting with Chapter 17, where blinding is also impossible but the effect is small and the certainty low. Here the effect is so large that even a generous allowance for detection bias leaves a real benefit. Effect size and bias have to be weighed together, not separately.
What the review could not answer. Which sealant material is best is unknown at very low certainty, across 24 trials. Glass ionomer against no sealant is inconclusive. So the recommendation’s specification of resin is doing real work and is supported, while the profession’s choice between resin products is not.
PubMed, 21 August 2026, for an update to CD001830. None found; the 2017 version (.pub5) remains current.
I have the full text of this review, supplied during writing. The statements about the publication dates of the seven trials, the absence of recorded background fluoride exposure, and the universal high risk of detection bias are from the review body, not its abstract.
Full record in appraisals/searches/ch11-sealants.md.
Verdict
- Certainty of evidence
- Moderate that resin-based sealants substantially reduce caries in first permanent molars, sustained to at least 48 months. I agree with DBOH. Very low for glass ionomer and for any comparison between sealant materials.
- Directness to the advice as worded
-
Good on population and outcome, questionable on era. Right children, right teeth, caries measured over years. But five of the seven trials behind the headline estimate are from the 1970s, in a pre-fluoride-toothpaste population, and background fluoride exposure was rarely recorded.
DBOH’s claim that the benefit holds “in both low risk and high risk populations” turns out to be better supported than I first credited, and not only by the arithmetic projection from the pooled odds ratio. The review reports the two modern trials directly: in Liu 2012, carious first molars were found in 7.4% of the sealant group against 17% of the placebo group; in Liu 2014b, 3% against 15%. Both are Chinese cohorts in a country the review classifies as very low caries (DMFT under 1.2 at age 12), against control incidences of 37% to 69% in the five 1976 to 1979 split-mouth trials. The effect survives into a modern low-caries setting; the absolute benefit is much smaller, exactly as you would expect.
- Is the strength label defensible?
-
Yes. The effect is large enough to survive its caveats, and the recommendation is correctly aimed at children at higher risk, where the absolute benefit is greatest. Of everything in Part II, this and fluoride toothpaste are the two I would defend most confidently.
The improvement I would want is transparency about the age of the evidence, which affects how much benefit a child today should expect.
- What would change my mind
- More trials in contemporary populations with recorded fluoride exposure. The two modern trials the review does report point the same way as the old ones on relative effect while showing far lower absolute risk, which is the pattern I would predict and not one that unsettles the recommendation. What would unsettle it is a modern trial finding no benefit. I would also want the harm side filled in: sealants can fail, and a partially lost sealant over a carious lesion is a scenario the trials do not describe.
What this means for you
If your child is offered fissure sealants, and particularly if a dentist has said they are at higher risk of decay, take them. Along with fluoride toothpaste, this is one of the two strongest recommendations in this book.
Ask for resin. That is what the evidence supports; glass ionomer is inconclusive and no one knows which brand of resin is best.
The one thing to hold lightly is the size of the benefit. The figures come mostly from children in the 1970s who had far more decay than children do now. Sealants do still work in modern populations; the review’s two recent trials show that. But where those 1970s trials had 37% to 69% of control teeth decaying, the modern ones had 15% to 17%, so the number of children who need sealing for one to benefit is much higher today than an odds ratio of 0.12 makes it sound.