6  Brush last thing at night

This is the recommendation that started the whole project, so it gets the fullest treatment. It is also the one where I got the argument wrong the first time, in an instructive way: I led with the weakest objection and buried the fatal one.

The advice

  • last thing at night (or before bedtime) and on at least one other occasion

Strength of recommendation: Strong

— Delivering Better Oral Health, chapter 2, tables 1b, 1d and 1f (1)

The instruction appears as a bullet inside a larger brushing recommendation for every age group. The wording above is that of tables 1b, 1d and 1f; table 1a, for children under three, says “and on one other occasion” without the “at least”. Table 4a, on tooth wear, carries the same advice as Good practice rather than Strong, which is worth noticing: the identical instruction holds two different strength labels in the same document, depending on which disease the table is about.

Note what it actually asks. It is not “brush twice a day,” which is Chapter 5. It is a claim about when: that one of your two brushings should be the last thing you do before sleeping, rather than, say, at nine in the evening, or after dinner.

What the guideline says its evidence is

Low to very low certainty evidence around initiation stage, frequency and timing.

— Delivering Better Oral Health, chapter 13, table 1 (1)

and, for adults:

Low-certainty evidence from children and adolescents for frequency and timing.

— Delivering Better Oral Health, chapter 13, table 6 (1)

The citation, again, is SIGN 138 (2).

Following the citation

SIGN 138 §5.7.2 is short enough to quote entirely:

Brushing last thing at night before bedtime allows fluoride concentration levels to remain high during the night as salivary flow rates are lower during sleep. An observational study found that fluoride concentrations in saliva 12 hours after brushing last thing at night were comparable with those found one to four hours after brushing during the day.

Children’s teeth should be brushed last thing at night before bedtime and on at least one other occasion.

— SIGN 138, §5.7.2 (2)

Two sentences, then the recommendation. Two references: reference 95, a British Dental Journal review, for the physiological rationale that salivary flow falls during sleep (3); and reference 101 for the observational study.

Note what precedes the recommendation in the printed guideline: the tick symbol that SIGN’s own key, on page ii, defines as a Good Practice Point, “recommended best practice based on the clinical experience of the guideline development group.” The recommendation directly above it, on rinsing, carries a grade A. SIGN classified this one as clinical experience, deliberately and in its own notation.

That observational study is reference 101, and it is:

Duckworth RM, Moore MS. Salivary fluoride concentrations after overnight use of toothpastes. Caries Res 2001;35(4):285.

One page. Volume 35, issue 4, page 285, single page, which is the abstracts supplement of the 48th congress of the European Organization for Caries Research (4). There is no full paper. Both authors were at Unilever Dental Research in Bebington.

That is the only empirical study cited for telling every person in the United Kingdom to brush their teeth immediately before bed. The other reference supports the physiology, not the recommendation.

SIGN did not claim more than it had. It described its source accurately and classified the recommendation as clinical experience. The Strong label appears later, downstream, when this becomes a bullet in a DBOH table alongside fluoride toothpaste.

Does it answer the question?

No, and the reason is not that the study is small or old or industry-funded. The reason is that it is about something else.

Population. Twelve adults, in a cross-over. The recommendation covers everyone from erupting first teeth onwards.

Intervention and comparator. This is the fatal one. To find out whether brushing last thing at night beats brushing earlier, you need one group brushing last thing at night and another brushing earlier. In this study everyone brushed at bedtime. There was no early-brushing group. What was compared was two toothpaste formulations, a chalk-based and a silica-based paste with the same 1,500ppm sodium monofluorophosphate content. The study cannot compare timings because it did not vary timing.

Outcome. Fluoride concentration in saliva. Not tooth decay, not pain, not tooth loss.

A surrogate outcome is one that stands in for the thing you actually care about. Salivary fluoride is several steps removed from a patient-important outcome: more fluoride in saliva is thought to mean more fluoride available at the enamel surface, which is thought to mean more remineralization, which is thought to mean fewer lesions, which is thought to mean less pain and fewer extractions.

Each arrow in that chain is plausible. None was measured here.

In the original blog post I called this outcome “too abstract to form the basis for such a crucial recommendation.” I would put it more precisely now: the outcome is entirely reasonable for the study Duckworth and Moore were doing, which was a pharmacokinetic comparison of two toothpaste bases. It is the use of that study, as evidence for a behavioral recommendation about timing, that is indirect. Blaming the study for not answering a question it never asked is unfair to the authors. The mismatch was introduced by whoever cited it.

The argument I got wrong

In the blog version I devoted a long section to the sample size, arguing that twelve participants is too few to conclude anything, with a digression on statistical power and research waste.

That argument is weak and I have dropped it.

Twelve participants may be entirely adequate for a cross-over study of salivary fluoride kinetics. Each person acts as their own control, the within-person correlation is high, and the measurement is a chemical assay rather than a subjective rating. Pharmacokinetic studies with a dozen people are routine and often perfectly well powered. If I had gone after a badly powered trial of a drug on a dozen people, that would be a criticism; here it is not.

The reason this matters is more than pedantry. Leading with the weak objection invites the reply “twelve is fine for a cross-over,” which is correct, and which then makes the real objection look like more of the same. The real objection is that the study did not compare brushing times. No sample size fixes that. A version of this study with twelve thousand participants would tell us about overnight fluoride kinetics with great precision, and would still not tell us whether brushing at eleven beats brushing at nine.

That is not the same as saying the study is worthless. It supports the biological plausibility of the recommendation, and I return to that below. It cannot estimate the comparative clinical effect the recommendation asserts.

The industry question

Both authors worked for Unilever, which sells toothpaste. This is worth noting and worth not overplaying.

The empirical basis for caring is a Cochrane methodology review, which finds that industry-sponsored studies more often report results favorable to the sponsor than independently funded studies do, and that this is not explained by differences in risk of bias (5). That is a reason for extra care with the body of literature, not a reason to dismiss a particular paper.

Here it barely signifies. The study is not being used to promote a product; it is being used, by third parties, to support a behavioral recommendation the authors did not make. The problem with this citation is not who paid for it.

The problem is that it is a conference abstract. There is no methods section to appraise, no full results to check for selective reporting, and no way to know what else was measured. When the entire evidence base for a Strong recommendation is one page, “we cannot tell” is the appraisal.

Is there better evidence?

PubMed, 15 August 2026, for trials comparing toothbrushing at different times of day with a caries or patient-important outcome. The query returned 249 records, of which the top 20 by relevance were screened on title and abstract. Nothing directly relevant was found. The top hit was the brushing-frequency review from Chapter 5; the rest concerned frequency, supervised brushing programs, or fluoride kinetics.

The limits matter and I would rather state them than imply a completeness the search does not have: one database, one screener, and 20 of 249 records read. A trial ranked below 20 with an uninformative title would have been missed. This supports “I did not find one,” not “there is none.” Full record in appraisals/searches/ch06-brushing-timing.md.

The recommendation’s own citation could not have been found this way in any case. A one-page congress abstract with no DOI and no PubMed record is invisible to a database search, and had to be traced by hand through SIGN 138’s printed reference list.

In the blog post I discussed a cluster-randomized evaluation of an oral health education program called “Brush Day & Night,” and argued from one of its figures that the program had not worked.

I have removed that argument, and I have not replaced it with a corrected version. Two reasons. The claim I made came from reading a bar chart for a secondary outcome in one of the participating countries, which is the same reading-off-a-figure move I criticize elsewhere in this book, and a cluster trial needs an analysis that accounts for clustering before anything can be said about it. And the trial does not bear on this chapter’s question in any case: everyone in both arms was encouraged to brush at night, so once again no comparison of timings exists. Rather than restate someone else’s results at second hand from memory, I leave the point out.

The rationale, taken seriously

There is a mechanistic argument here and it deserves better than dismissal.

Salivary flow falls substantially during sleep. Saliva clears fluoride from the mouth, and it also buffers acid and supplies calcium and phosphate for remineralization. So the hours of sleep are both the period when the mouth is least protected and the period when anything left on the teeth persists longest. Putting fluoride there immediately beforehand, and not eating afterwards, is a coherent piece of reasoning, and the Duckworth abstract is consistent with it: fluoride does hang around overnight.

I think this rationale is probably right. I follow it myself.

But “probably right and untested” is a different thing from “Strong,” and the distinction is the whole subject of this book. In GRADE’s vocabulary this is indirect evidence supporting a mechanism, which is exactly the sort of thing that justifies a Conditional recommendation and a note that the reasoning is biological rather than experimental.

Verdict

Certainty of evidence
Very low. A single conference abstract reporting a surrogate outcome, with no full report available to appraise, supporting a mechanism rather than the recommendation. DBOH’s own rating for this component is low to very low, and I agree with the lower end.
Directness to the advice as worded
Very poor. The cited study did not vary the timing of brushing, which is the only thing the advice is about. It compared two toothpaste bases, in twelve adults, measuring salivary fluoride.
Is the strength label defensible?

Not on the published reasoning. The recommendation is supported by plausible physiology and by one indirect measurement, with no study of the comparison it asserts. On DBOH’s own definitions that is the profile of a Good practice point: advice resting on extrapolation from related research and on clinical consensus. SIGN 138 classified it exactly that way, in its own notation, before DBOH bundled it into a Strong recommendation.

Two qualifications. Calling it a good practice point is not the same as saying the evidence is unrateable; I have rated it very low above, and both statements can stand because DBOH’s category is defined by the kind of evidence available, not by the impossibility of grading it. And as everywhere in this book, I have not run a full evidence-to-decision assessment (Chapter 4), so this is a judgment about the published justification, not a claim that no panel could reach Strong.

What would change my mind
A randomized trial allocating people to brush last thing at night versus ninety minutes to two hours earlier, everything else held constant, with caries increment or self-reported tooth pain at two years or more. A cross-over with an intermediate outcome would not do it; the point of this chapter is that intermediate outcomes have already been measured and cannot settle it. Even a well-conducted prospective cohort with objectively recorded brushing times would move me from very low to low.

What this means for you

Brush before bed. Given that you were going to brush twice anyway, choosing when costs you nothing extra that I can see, and the physiological reasoning behind it is coherent even though the comparison it implies has not been tested.

What has not been demonstrated is how much it matters. If you brush at nine and go to bed at eleven, the searches described here found nothing that can tell you what that costs you. And if you are given the impression that this specific instruction is backed by strong evidence, the summary table is the reason: it does not disclose that the component-level certainty behind that bullet is low, or that the guideline it came from classified it as clinical experience.