No trial has tested whether flossing prevents cavities. Cochrane says so twice.

Flossing has never been tested against the outcome it is recommended for. Cochrane's 2011 review of flossing found that none of its 12 included trials reported data on dental caries [s2]; its 2019 successor, covering 35 trials of every interdental cleaning device, reported that no trial assessed interproximal caries at all, and that most did not assess periodontitis either [s1]. What both reviews did find is a small reduction in gum inflammation, on evidence they rate low to very low certainty [s1] [s2].

That is a narrower claim than "flossing does nothing", and a much narrower one than "flossing prevents cavities". The gap between those two is where most writing on this subject goes wrong.

What the trials measured instead

The 2019 Cochrane review pooled 35 randomised controlled trials with 3,929 adult participants, comparing toothbrushing plus a home-use interdental device against toothbrushing alone, with a minimum trial duration of four weeks [s1]. Fifteen of those trials tested floss; the rest tested interdental brushes (2), wooden cleaning sticks (2), rubber or elastomeric cleaning sticks (2) and oral irrigators (5) [s1].

Because participants cannot be blinded to whether they are flossing, every study carried a high risk of performance bias, and only two were otherwise at low risk of bias [s1]. Many participants also started with a low level of gingival inflammation — leaving little room for improvement [s1]. Outcomes were gingivitis, measured by indices such as the Löe-Silness scale (0 to 3) or by the proportion of bleeding sites, and plaque, most often by the Quigley-Hein index (0 to 5) [s1].

For floss specifically, low-certainty evidence suggested that adding it to brushing may reduce gingivitis measured by gingival index at one month — standardised mean difference −0.58 (95% confidence interval −1.12 to −0.04; 8 trials, 585 participants) — and at three and six months [s1]. Results for the proportion of bleeding sites and for plaque were inconsistent, on very low-certainty evidence [s1].

The size of the effect, in the review's own units

The 2011 review is the more useful of the two on magnitude, because it converted its pooled effects back onto the clinical scale. It included 12 trials, with 582 participants flossing and brushing against 501 brushing alone; seven trials were at unclear risk of bias and five at high risk [s2].

Flossing plus brushing produced statistically significant reductions in gingivitis at all three time points: SMD −0.36 (95% CI −0.66 to −0.05) at one month, −0.41 (95% CI −0.68 to −0.14) at three months, and −0.72 (95% CI −1.09 to −0.35) at six months [s2]. Re-expressed on the Löe-Silness index, which runs from 0 to 3, those are reductions of 0.13, 0.20 and 0.09 points respectively [s2].

A tenth of a point on a three-point scale is the honest picture of the benefit measured. The review also found weak, very unreliable evidence that flossing may be associated with a small reduction in plaque at one or three months, and reported that none of the included trials provided data on caries, calculus, clinical attachment loss or quality of life [s2].

Floss is not the best-performing device in these reviews

The 2019 review's secondary objective — comparing devices with each other — produced a result that receives almost no attention outside dentistry. Low-certainty evidence suggested that interdental brushes may reduce gingivitis more than floss at one and three months, though they did not differ for periodontitis measured by probing pocket depth, and the plaque evidence was inconsistent [s1].

Against toothbrushing alone, interdental brushes reduced the gingival index by 0.53 points (95% CI −0.83 to −0.23) in a single trial of 62 participants, on very low-certainty evidence, with no clear difference in bleeding sites (mean difference −0.05, 95% CI −0.13 to 0.03; one trial, 31 participants) [s1]. They reduced plaque more than brushing alone (SMD −1.07, 95% CI −1.51 to −0.63; 2 trials, 93 participants), on low-certainty evidence [s1].

Oral irrigators fared worse. Very low-certainty evidence suggested they may reduce gingivitis at one month (SMD −0.48, 95% CI −0.89 to −0.06; 4 trials, 380 participants) but not at three or six months, and low-certainty evidence showed no reduction in bleeding sites at one month (MD −0.00, 95% CI −0.07 to 0.06; 2 trials, 126 participants) or in plaque at any time point [s1].

Why the evidence is this thin, and what it does not license

The absence of caries data is not evidence that flossing fails to prevent caries. It is an absence: a trial capable of detecting a difference in interproximal decay would need to run for years, enrol large numbers, and radiograph everyone repeatedly, and nobody has funded one. The 2019 review's list of trial durations — a minimum of four weeks — explains the whole problem. Gingivitis can be moved in a month. Cavities cannot be counted in one.

Nor does low certainty mean the effect is imaginary. Both reviews found the direction of effect consistently favouring flossing on gingivitis across time points [s1] [s2]. What "low certainty" records is that the trials were small, unblindable, short, and conducted mostly in people whose gums were already fairly healthy [s1].

The defensible summary is this. Adding floss to brushing produces a small measured reduction in gum inflammation, on weak evidence [s1] [s2]. Interdental brushes may do it slightly better, on evidence that is weaker still [s1]. Whether any of it prevents cavities between the teeth is unstudied [s1] [s2]. And no review can say what any of this means for a particular mouth, which is what a dentist or hygienist examining it is for.