13 Sugar: how much is too much?
Everything else in this book is about defending teeth from decay. This chapter is about the thing that causes it.
It is also the chapter where I have to be most careful, because the causal claim is not in doubt and the quantitative claim, which is what the recommendation actually makes, rests on much less than people assume.
The advice
Minimise the amount and frequency of consumption of sugar-containing food and drinks
Strength of recommendation: Strong
— Delivering Better Oral Health, chapter 2, table 1f (1)
That is the adult wording. The advice is not phrased identically across the age tables, and the difference matters for the point I am about to make:
| Table | Population | Wording |
|---|---|---|
| 1a | children up to 3 | “Minimise consumption of sugar-containing foods and drinks” |
| 1b, 1d | children 3 to 6, 7 to 18 | “Minimise amount and frequency of consumption of sugar-containing food and drinks” |
| 1f | adults | “Minimise the amount and frequency of consumption of sugar-containing food and drinks” |
Note that the adult and older-child lines each contain two instructions, amount and frequency, with no bullets to separate them. It is a bundle that Chapter 3’s counting method scores as one, which is why that chapter calls its own figures a floor. The under-threes line does not contain that bundle; it says only “minimise consumption,” and an earlier draft of this chapter attributed the two-instruction wording to all four tables, which was wrong.
What the guideline says its evidence is
Recommendation based on moderate-certainty evidence that dental caries is lower when free-sugars intake is <10% and on very low certainty evidence that dental caries is lower when free-sugars intake is <5% energy, and in line with WHO and SCAN [sic] guidelines.
— Delivering Better Oral Health, chapter 13, table 6 (1)
The “[sic]” is mine and it is not a cheap shot. DBOH prints SCAN in all eight places this sentence appears, and the GOV.UK markup expands the abbreviation as “South Central Antimicrobial Network,” which is a real body and has nothing to do with sugar. The footnote the sentence hangs on is the Scientific Advisory Committee on Nutrition, SACN, and its 2015 report Carbohydrates and health.
So it is a typo, and everyone can see it is a typo. I flag it for one reason: an earlier draft of this chapter silently printed “SACN” inside the quotation marks. A book whose method is verbatim quotation does not get to tidy up its source, even helpfully, even when the correction is obvious. If I can quietly improve a guideline’s wording when it suits me, nothing else I quote is trustworthy.
Two thresholds, two very different certainties, in one sentence. The citation is Moynihan and Kelly’s review, written to inform the WHO guideline (2).
Vipeholm, and how not to cite it
Before the modern evidence, a word about the study everybody reaches for.
The Vipeholm study ran in Sweden from 1945 to 1955. Its participants were 436 adults institutionalized at Vipeholm Hospital with intellectual disabilities, and they were fed large quantities of sweets, in various patterns, to see what happened to their teeth. What happened was a great deal of tooth decay, particularly in the groups given sticky sweets between meals.
In my original blog post I wrote that this “showed a causal relationship between sugar and tooth decay.” I have changed that, for two reasons.
The first is scientific. Vipeholm was not randomized in any modern sense, the groups were not comparable, and the exposure was extreme. It is cited so often because its finding is memorable and its direction is right, not because its design would support a causal claim today.
The second reason is that describing it in one clause, as I did, sanitises it. The participants could not consent. The study deliberately caused irreversible harm to people who were in institutional care, and it is a standard case in research ethics teaching for exactly that reason. A book about evidential honesty should not use it as a throwaway citation for a claim that has better support elsewhere. I mention it here because it is part of how we came to believe what we believe, and then set it aside.
The modern evidence
Moynihan and Kelly screened 5,990 records and assessed the body of evidence with GRADE. Their conclusions:
This in-depth systematic review shows consistent evidence of moderate quality supporting a relationship between the amount of sugars consumed and dental caries development. There is evidence of moderate quality to show that dental caries is lower when free-sugars intake is < 10% E.
— Moynihan and Kelly, 2014 (2)
So: moderate certainty that less sugar means less decay, and moderate certainty for the 10%-of-energy threshold specifically.
In grams, the figures most people will have met are the 5% ones, because those are what UK guidance publishes: DBOH’s own chapter 10 gives no more than 19 g of free sugars a day for a 4- to 6-year-old and no more than 30 g from age 11 upwards. WHO illustrates its 10% level as roughly 50 g on a 2,000 kcal diet. An earlier draft of this chapter gave a set of gram figures for the 10% level that I could not source back to either document, and I have replaced them with the published ones rather than leave arithmetic of my own standing in for a citation.
The 5% threshold is different, and DBOH says so: very low certainty. The 2014 review is candid about why. The evidence for going below 5% rested on a handful of Japanese ecological studies from 1959 and 1960, using per-capita sugar availability rather than measured intake, which is why the profile downgrades for risk of bias.
The update I missed
That was true of the evidence as it stood in 2014. It is not true now, and my search should have found out.
Moores, Kelly and Moynihan published a ten-year update in 2022, covering the literature from 2011 to 2020 (3). From 488 new papers they found 23 eligible studies. According to PubMed, its GRADE conclusions are:
- the evidence that caries is lower when free sugars are below 10% of energy remains moderate quality, now confirmed by new cohort data;
- the evidence for below 5% of energy is upgraded from very low to low, on the strength of new cohort studies.
Their summary sentence is that the findings “support and strengthen original evidence underpinning the WHO recommendations for sugars.”
So the 5% figure is better supported than DBOH’s evidence statement says, and better supported than an earlier draft of this chapter said. It is still the weaker of the two thresholds, and low certainty is still low. But “a stretch goal resting on very low certainty evidence,” which is how I put it, has been overtaken by evidence that DBOH could have cited: the update predates the September 2025 edition by more than three years.
This is the fourth time in this book that an update search by known author and title has failed to find an update that existed, and it is the same failure Chapter 32 is about. It is also a case my own tooling would not catch, because that tool reads version numbers out of Cochrane DOIs and this is a journal review with no version number to read.
Two related cautions in the review are worth separating from that, because I ran them together in an earlier draft. Moynihan and Kelly note that at intakes under about 10% of energy the average DMFT of 12-year-olds was under 3, the WHO’s oral health goal for the year 2000, and add that “by modern-day standards, this is not considered to be ‘low’.” They also note that in most studies caries was diagnosed at the cavitation level, a late stage. Both remarks are about the 10% evidence and the literature at large; neither is the reason the 5% evidence is rated very low.
Almost none of this is randomized, and it could not be. You cannot allocate children to a decade of high sugar intake. The review therefore pools intervention, cohort, population and cross-sectional studies, and grades the body accordingly.
That is the correct approach, and moderate certainty from observational evidence is a higher rating than the default, which means a GRADE upgrade criterion from Chapter 2 was engaged. The supplementary GRADE profiles say which one, and it is worth being exact because I was not in an earlier draft.
The upgrade is for large effect size, and for that alone: an SMD for DMFT of 0.82 (95% CI 0.67 to 0.97) comparing higher with lower sugars intake, and a risk ratio for caries prevalence of 7.15 (95% CI 2.82 to 8.14). The profile then adds: “Additionally, dose-response effect noted from Rugg-Gunn cohort study also supported by population studies but not further upgraded.” So the dose-response was considered and deliberately not used for a second step up.
I had also listed consistency among the upgrade criteria. It is not one. Consistency across studies is a reason not to rate down; the three upgrade criteria are large effect, dose-response gradient, and confounding that would work against the observed effect. Getting that wrong in a chapter praising someone else’s GRADE application is not a good look, so it is corrected here and recorded in Chapter 4.
This remains what upgrading observational evidence properly looks like, and it is the contrast case for Chapter 5, where I looked for the same thing and could not find it.
Two limits worth naming. Measuring what people eat is notoriously unreliable, and misreporting of sugar in particular is systematic rather than random. And sugar intake is entangled with everything else that predicts decay: income, education, dental attendance, fluoride exposure. The review’s moderate rating is a judgment that the consistency of the finding survives both problems. I think that judgment is defensible.
The frequency half. The recommendation says “amount and frequency”. The review’s graded conclusion is about amount. Frequency has a long-standing mechanistic rationale, each sugar exposure dropping the pH for a period, so more exposures mean more time demineralized, and it is the part of the Vipeholm finding that has best survived. But the moderate-certainty statement DBOH cites is about quantity, and the two are being carried by one label.
PubMed, 21 August 2026. Moynihan and Kelly’s own search covered multiple databases from 1950 through November 2011; the paper was published in 2014, and I had previously given the search end-date as 2013. I checked for a more recent systematic review of sugars and caries intended to inform guidelines and recorded that I found none superseding it.
That search failed. Moores, Kelly and Moynihan’s ten-year update was published in the Journal of Dental Research in 2022 (3) and I did not retrieve it. The query was built around the 2014 review’s authors and title, which is a known-item search: it finds the paper you already know about and is structurally incapable of telling you a successor exists. The same mistake appears in three other chapters and is the subject of Chapter 32. The chapter above has been rewritten around the update.
I have the full text and all six supplementary files, supplied during writing, including the GRADE profile tables.
Age of the evidence: this review is now twelve years old, its literature search closed fifteen years ago, and it remains the basis of the WHO’s current sugars guidance.
Full record in appraisals/searches/ch13-sugar.md.
Verdict
- Certainty of evidence
-
Moderate that lower free-sugars intake means less dental caries, and for the <10% of energy threshold. The moderate rating represents a legitimate upgrade of observational evidence rather than a downgrade of trials.
For the <5% threshold, low, not the very low that DBOH states. The 2022 update upgraded it a level on new cohort evidence (3), three years before the edition audited here. DBOH’s rating was right for the review it cites and is now a level too pessimistic.
- Directness to the advice as worded
- Good on amount, weaker on frequency. The graded evidence concerns how much; the recommendation says how much and how often. The frequency claim is mechanistically strong and separately less well graded.
- Is the strength label defensible?
-
Yes. This is the clearest Strong recommendation in Part II after fluoride toothpaste. The evidence is consistent, the mechanism is understood, the direction is not in dispute, and the harms of eating less sugar are nil.
The drafting point stands: “amount and frequency” is two instructions in one line, and the <10% and <5% thresholds carry moderate and very low certainty respectively while appearing in the same guidance.
- What would change my mind
- On the direction, essentially nothing. On the 5% threshold, this has already partly happened: the 2022 update moved it from very low to low, and a further review using modern caries diagnostic criteria and measured rather than per-capita intake would settle whether it deserves more. On frequency, a study that separates the number of sugar exposures from the total quantity, which is hard but not impossible with modern dietary monitoring.
What this means for you
Eat less sugar, and eat it less often. Of everything in this book, this is the advice with the least argument against it and the fewest caveats attached.
If you want a number, 10% of your energy intake is the one with the strongest evidence behind it, roughly 55 g of free sugars a day for an adult. The 5% figure in UK guidance is the more demanding target and the evidence for it is weaker, but it is no longer as weak as the guideline says: a 2022 update moved it up a level. Treat it as a good aim rather than a threshold anyone has pinned down precisely.
And the frequency really does matter, or at least the mechanism says it should: a single sweet dessert is easier on your teeth than the same amount of sugar spread across six moments of the day.