17 Flossing, and what replaced it
In 2016 the Associated Press asked the US federal government for the evidence behind its flossing recommendation, did not get a satisfactory answer, and ran a story. “Medical benefits of dental floss unproven” went around the world. The recommendation was quietly dropped from the US dietary guidelines. Most people who read about it concluded that flossing had been debunked.
That is not what the evidence says, and it is not what the guideline says either. This chapter is the one where almost everybody is wrong, including me.
The advice
Interdental plaque control:
- clean daily between the teeth to below the gum line before toothbrushing
- where there is space for an interdental or single-tufted brush, this should be used
- for small spaces between teeth, use dental floss or tape
Strength of recommendation: Conditional
— Delivering Better Oral Health, chapter 2, table 2b (1)
Note the label before going further. Conditional, not Strong. And note the ordering: interdental brushes first, floss only for gaps too small to take one. Floss has been demoted from the default to the fallback.
What the guideline says its evidence is
Recommendation based on low certainty evidence for an added benefit from flossing, though it is unclear if this benefit is clinically important; EFP recommends floss only where gaps are too small for interdental brushes. Low certainty evidence that use of interdental brushes is beneficial and that they are more effective than floss, but again the clinical importance of the difference is uncertain.
— Delivering Better Oral Health, chapter 13 (1)
I want to record that this is a good evidence statement. It gives the certainty, names the direction of the effect, and twice flags that statistical benefit may not amount to clinical benefit. That last distinction is one most guidelines skip entirely. Nothing in this chapter is a complaint about it.
Where I was wrong
Here is what I wrote in 2023:
While some studies have assessed the sole effect of flossing on interproximal dental caries, an updated Cochrane review in 2019 showed that we still don’t have any trial evaluating the impact of flossing+toothbrushing versus toothbrushing alone.
That is factually wrong, and reading the review properly is what corrected it.
Worthington and colleagues included 35 randomized trials in 3,929 adults (2). Of those, fifteen compared flossing plus toothbrushing against toothbrushing alone. That is precisely the comparison I said did not exist.
What does not exist is something narrower, and the review states it in one sentence:
No trials assessed interproximal caries, and most did not assess periodontitis.
— Worthington and colleagues, 2019 (2)
So the accurate claim is not “nobody has tested flossing.” It is: the review found thirty-five trials of interdental cleaning, fifteen of them testing floss added to brushing, and not one of the thirty-five measured tooth decay between the teeth. They measured gingivitis and plaque, over one to six months. I wrote “thirty-five trials have tested flossing” in an earlier draft, which quietly inflates the floss literature by counting trials of interdental brushes, sticks and irrigators.
The difference matters. My version implied an empty literature, which invites the conclusion that the recommendation is baseless. The true version says the literature is reasonably large and points the right way for gum inflammation, while this review measured nothing about tooth decay. That is a more interesting finding and a more defensible one, and I should not have needed a full text to get it right. I had been relying on the abstract and on my memory of it.
Caries and floss: what was measured, and in whom
I then made the second half of the same mistake, and said the decay outcome “has never been measured”. It has, outside this review and long before it.
Hujoel and colleagues pooled six trials in 808 children aged 4 to 13 (3). What they found depends entirely on who was holding the floss:
| Comparison | Risk ratio | 95% CI |
|---|---|---|
| Professional flossing on school days, 1.7 years, mostly primary teeth | 0.60 | 0.48 to 0.76 |
| Professional flossing every three months, 3 years | 0.93 | 0.73 to 1.19 |
| Self-performed flossing, young adolescents, 2 years | 1.01 | 0.85 to 1.20 |
A hygienist flossing a child’s teeth on every school day for nearly two years cut interproximal caries by about 40%. Children flossing their own teeth showed nothing at all: 1.01 is as close to no effect as a number gets.
The review’s own scope sentence is the one that matters for the recommendation DBOH actually makes: “No flossing trials in adults or under unsupervised conditions could be identified.” The authors add that the professional-flossing finding “should be extrapolated to more typical floss-users with care, since self-flossing has failed to show an effect.”
So the honest position is narrower than either of my drafts. The decay question has been studied, in children, mostly on baby teeth, under low fluoride exposure, and the answer depends on whether a professional does it. For an adult flossing their own teeth at home, which is what the guideline recommends and what you are being told to do, there is still no trial with a decay outcome. The gap is real. It is just a different and much more specific gap than “nobody has tested it”.
What 35 trials found
| Device, added to toothbrushing | Gingivitis | Certainty |
|---|---|---|
| Floss | SMD −0.58 (95% CI −1.12 to −0.04) at 1 month, 8 trials, 585 participants; effects also at 3 and 6 months | low |
| Interdental brushes | MD −0.53 (95% CI −0.83 to −0.23) at 1 month, 1 trial, 62 participants | very low |
| Wooden cleaning sticks | bleeding sites MD −0.25 (−0.37 to −0.13) at 3 months, 1 trial, 24 participants | very low |
| Rubber cleaning sticks | no effect on gingivitis (MD −0.01, −0.19 to 0.21, 1 trial, 12 participants) | very low |
| Oral irrigators | SMD −0.48 (−0.89 to −0.06) at 1 month, 4 trials, 380 participants; not at 3 or 6 months | very low |
And head to head, at low certainty, interdental brushes may reduce gingivitis more than floss at one and three months. That is where DBOH’s ordering comes from, and it is the substantive change since the 2011 version of this review.
The review is well conducted; these are limits in the underlying trials, and the review names every one of them.
Nobody can be blinded. You know whether you have been flossing. Every trial is at high risk of performance bias for that reason, and it is not fixable. Only two of 35 were otherwise at low risk of bias.
A floor effect. “Many participants had a low level of baseline gingival inflammation.” If your gums are already close to healthy, there is very little room for an intervention to improve them, and the average effect across such a population understates what might happen in people with inflamed gums. This cuts in the intervention’s favor, and it is the strongest single argument that these trials underestimate the benefit.
Short follow-up. One to six months. Periodontitis develops over years and interproximal caries over years. A six-month trial cannot see either, which is the mechanical reason the caries outcome is missing rather than an oversight by trialists.
Tiny trials. Look at the participant counts in the table: 62, 24, 12, 380. Several of the device comparisons rest on a single small trial. The flossing estimate, at 8 trials and 585 participants, is by some way the best-supported line in the table, which is ironic given the public story.
The review’s own summing up is the sentence to keep: “the effect sizes observed may not be clinically important.” Gingivitis was measured with several different indices, most commonly Löe-Silness on a 0 to 3 scale scored by an examiner, which is why the pooled figure is a standardized mean difference rather than a number of points. So −0.58 means roughly half a standard deviation across mixed scales, not 0.58 points on a 0-to-3 index, and Cochrane did not back-translate it to a common index. It is a real statistical signal. Whether it corresponds to anything you would notice in your mouth is genuinely unknown.
PubMed, 21 August 2026, for an update to CD012018. None found; the 2019 version (.pub2) remains current. The review’s own search ran to 16 January 2019.
Unlike the earlier chapters, I have the full text for this review, supplied during writing, including the characteristics of included studies and the risk-of-bias tables. Statements above about blinding, baseline inflammation and trial sizes are from the review body, not from its abstract.
Full record in appraisals/searches/ch17-interdental.md.
Did the newspapers get it wrong?
Partly, and in an instructive way.
The 2016 story was accurate about one thing: the evidence for flossing is weak, short-term, and measured on surrogate outcomes. It was wrong to present that as a discovery, since Cochrane had been saying so since 2011, and wrong to let readers infer that flossing had been shown not to work.
“No trial has measured whether flossing prevents cavities between your teeth” and “flossing does not prevent cavities between your teeth” are different sentences. The first is true. The second has never been tested. This is the same distinction Chapter 19 turns the other way: there, a precise estimate centered on zero really does mean the intervention does not work. Here there is no estimate at all.
It is worth noticing what happened next. A recommendation was withdrawn because journalists asked for its evidence and found the answer thin. Whatever you think of the outcome, that is the process this book is doing, at scale and in public. The lesson I take is not that the journalists were wrong to ask. It is that a guideline which had labeled the recommendation honestly in the first place would have had a much better answer ready.
Verdict
- Certainty of evidence
- Low that flossing added to toothbrushing reduces gingivitis in the short term. Very low that interdental brushes added to toothbrushing do so, from a single 62-person trial, and separately low that interdental brushes may reduce gingivitis more than floss at one and three months. Those are two different comparisons at two different certainties, and an earlier draft ran them together at low. No estimate at all on interproximal caries: of the 35 trials in this review, none measured it.
- Directness to the advice as worded
- Good on population and intervention, poor on outcome and duration. The trials are in adults doing the thing the advice describes. But they measure gingival indices and plaque scores over one to six months, and the advice is ultimately aimed at preventing periodontitis and decay over decades.
- Is the strength label defensible?
-
Yes. Conditional is the right label for low-certainty evidence of an effect of unclear clinical importance, and DBOH says so in those words. The ordering of devices, interdental brushes before floss, follows the head-to-head evidence rather than tradition.
This is the third chapter running where the guideline comes out well, and I would rather say so than manufacture a complaint. The criticism in this chapter is directed at my own earlier writing and at the press coverage, not at Delivering Better Oral Health.
- What would change my mind
- A trial long enough to measure interproximal caries, which means two years minimum and realistically longer, with radiographic scoring. No such trial was among the 35 the review found to January 2019, and I have not searched the primary literature published since, so read that as the review’s absence rather than a current one. Also useful: a trial recruiting people with gingival inflammation rather than the near-healthy participants who dominate the current literature, since the floor effect plausibly hides a real benefit.
What this means for you
Clean between your teeth. A toothbrush does not reach roughly a third of each tooth’s surface, and the trials of these devices point towards less gum inflammation, even if they disagree about how much and the certainty is low.
Use an interdental brush if one fits. That is what the head-to-head evidence supports and what the guideline now recommends first. Floss where the gaps are too tight for a brush.
What nobody can tell you is whether any of this prevents cavities between your teeth, because in the 35 trials Cochrane assembled, nobody measured it. If a dentist tells you flossing prevents interproximal decay, they are stating something plausible that this body of evidence has not measured. And if someone tells you flossing was debunked, they have confused an unasked question with an answered one.