The Bleeding Test

The mouth is the only organ I maintain by hand, twice a day, for life. Nothing else in the body gets that treatment. I don’t scrub my liver, and the skin I do wash replaces its own outer layer on a schedule that forgives me. Enamel doesn’t turn over. Periodontal attachment, once it’s gone, doesn’t come back.

My shelf is unremarkable. A Suri sonic brush, morning and night. Floss picks once a day, swapped for wire when I travel because the picks are bulky and a spool packs flat. A Waterpik a few times a week. A tongue scraper on roughly the same schedule. Mouthwash occasionally, and almost entirely before seeing people rather than as a hygiene step.

My dental record is clean. No caries, no periodontal findings, no orthodontics, no restorative work. Every appointment I’ve had has been maintenance.

For the one tissue I clean with motorized bristles every single day, I hold zero numbers. I track sleep staging, hydration, continuous glucose, VO2 max, grip strength, quarterly blood panels, epigenetic age, DEXA, a whole-body MRI. The mouth gets the most maintenance and the least measurement. I’ve never once asked my hygienist for a figure.

Observational epidemiology links bad gums to bad arteries, consistently and for a long time. Adults with severe periodontitisperiodontitisPeriodontitis: chronic inflammatory destruction of the tissue and bone anchoring the teeth, driven by bacterial biofilm below the gumline. Distinct from gingivitis, which is reversible inflammation of the gums without attachment loss. show higher rates of heart attack, stroke, and cardiovascular death. The mechanism sounds right: ulcerated pockets leak bacteria into the blood, chronic inflammation follows, plaque builds.

Genetic epidemiology takes that chain apart. Bell et al. tested it with Mendelian randomizationMendelian randomizationMendelian randomization: an analytic method that uses inherited genetic variants as natural instruments. Because alleles are assigned at conception and don’t change with lifestyle, they approximate the randomization of a trial and sidestep much of the confounding in observational data., which uses inherited genetic variants as a stand-in for random assignment. Across hundreds of thousands of people, genetic liability to gum disease moved coronary risk essentially not at all. Every interval sat on top of no effect. Their reading is that the observational link may be confounding rather than cause.

Zhao et al. then did this across the whole literature, pooling dozens of Mendelian randomization studies. Two outcomes survived: cardioembolic stroke and depression. Coronary atherosclerosis didn’t. Neither did Alzheimer’s, Parkinson’s, rheumatoid arthritis, or the various cancers that get attached to this story.

The American Heart Association updated its position in early 2026, its first statement on gum disease and heart disease since 2012. It treats periodontitis as a risk marker, notes that treating it improves surrogate measures like inflammatory markers, and declines to say that treating gums prevents heart attacks. The organisation best placed to make that claim looked at the same evidence and didn’t.

Tooth loss behaves the same way, and it’s the cleanest illustration of the trap. Losing teeth predicts dying, reliably, across eighteen prospective cohorts. Losing teeth does not reliably predict dying of heart disease, which is the mechanism everyone assumes. The authors call it a risk marker and describe even the mortality finding as inconclusive. Teeth fall out for the same reasons people die earlier: smoking, poverty, diabetes, twenty years of not going to a dentist.

Dementia is the one place the mechanism looks mechanical rather than microbial. Tooth loss tracks cognitive decline, and among people who replaced their teeth with dentures the association lost significance. Put the teeth back and the signal fades. That points at chewing and nutrition rather than bacteria travelling from gum to brain, though these are observational cohorts and denture wearers differ from non-wearers in income and access as well as in teeth.

The bacterial story for Alzheimer’s did get its shot. Researchers found *Porphyromonas gingivalis* and its gingipain proteases in Alzheimer’s brain tissue, a company built a drug to block them, and it reached a late-stage trial. No approved medicine came out of it, and the genetic work found no causal link to begin with.

Diabetes is the one place where treating the mouth demonstrably moves something systemic. A Cochrane review of thirty-five randomized trials found that cleaning under the gumline lowers HbA1c by 0.43 percentage points at three to four months, on moderate-certainty evidence.

That is roughly half what a second oral glucose-lowering drug delivers, obtained with a dental instrument. It is also the only claim in this entire article backed by randomized trials pointed at an outcome a physician would act on. Everything else here is a correlation wearing a lab coat.

Notice how narrow the exception is. One condition, out of the dozen the oral-systemic literature has claimed over four decades.

The trial evidence for cleaning between your teeth is much weaker than the advice implies. Cochrane’s review of interdental cleaning rates the whole body of evidence low to very low certainty, and notes the effects it did find may not be clinically important. Interdental brushes probably beat floss. Water flossers sit in the category the authors called limited and inconsistent.

The deeper problem is what nobody measured. Not one trial in that review tested whether flossing prevents cavities between teeth. Most didn’t test periodontitis either. The literature runs on plaque scores and bleeding indices over a few weeks, in people whose gums were already healthy. We have been recommending a daily habit for decades on the strength of surrogate endpoints.

Interdental brushes outranking floss is the part I can’t act on. My contacts are tight enough that a brush doesn’t pass without being forced, and forcing a wire between teeth to chase a ranking is a good way to injure a gum that was fine. Picks go in daily and reach what they can reach. A tool that ranks higher in a trial and won’t fit in your mouth ranks nowhere in your bathroom.

The Waterpik sits where its evidence sits, which is as a supplement rather than the base layer. Picks do the daily work. The irrigator is a deeper pass a few times a week, and I’d be uncomfortable if it were the only thing between my teeth, because the trials don’t support that job.

Powered brushing holds up better than either, and the long cohorts suggest it compounds. Over eleven years in a German population study, powered-brush users lost less attachment and kept more teeth than manual brushers. That isn’t randomized and powered-brush users skew younger, so treat it as a nudge rather than a proof.

Within powered brushes, oscillating-rotating heads outrank sonic ones for plaque and gingivitis. My Suri is sonic, so on that axis I’m a tier down. It’s worth being precise about what the axis measures: plaque removed per session, in a trial, with a charged device and a supervised participant.

Two things I care about aren’t on that axis. The first is what happens when the battery dies. A sonic brush with a flat battery is still a brush, and you finish the job by hand. An oscillating-rotating head with a flat battery is a short plastic stick that cannot be used properly, because the cleaning motion is the motor. On the road, that’s the difference between a worse brush and no brush. The second is whether I pick it up at all. The Suri is quiet and slim and the case is not embarrassing, and a brush that travels well gets used twice a day in hotel rooms where a bulkier one would stay in the bag.

Neither of those shows up in a plaque score, and both change the number of days a year the thing gets used. That’s the tradeoff I made knowingly: a few percent of per-session efficacy against resilience and adherence. I’d make it again, and someone who never travels and always charges should probably buy the oscillating-rotating one.

The device I did abandon was a U-shaped mouthpiece that brushes every tooth at once. I bought it because thirty seconds of standing still beats two minutes of technique. Tested head to head, it removed no more plaque than not brushing at all, and a second trial found that even forty-five sloppy seconds with an ordinary brush beat the whole mouthpiece category comfortably.

The principle underneath is contact time, not motor design. A tray that touches every surface for an instant does worse than one small head that visits each surface in turn. Every shortcut in this category sells you the removal of the only variable that matters, which is how long something abrasive is actually in contact with the tooth.

The tongue scraper is the honest gap in my routine. I use it a few times a week because I’m inconsistent about it, not because I decided that was the right frequency. The reason I’ve never fixed that is that the evidence only supports it for one thing, which is breath, and the certainty behind even that is very low. Nothing suggests it prevents decay, slows gum disease, or touches a systemic marker. Inconsistency on a step that only affects how you smell at dinner costs close to nothing, so it stays unfixed.

Mouthwash is the only habit on my shelf with a credible route to net harm. The mouth isn’t a chamber that wants sterilizing. Dietary nitrate concentrates in saliva, bacteria on the back of the tongue convert it to nitrite, and once swallowed that becomes nitric oxide, which relaxes blood vessels. Kill the bacteria and you break a loop your circulation is using.

That isn’t theoretical. A week of chlorhexidine rinsing collapses oral nitrite production and raises blood pressure by a few millimetres of mercury, with the rise tracking the fall in nitrite. The effect shows up within a day.

The sharpest demonstration involves exercise. Blood pressure drops for hours after a workout, which is part of why exercise works. When people rinsed with an antibacterial mouthwash straight after a treadmill session, most of that drop disappeared, and by two hours it was gone entirely. Rinsing after a run cancels a good part of what the run just did to your vasculature.

Then the trials pool and the effect evaporates. A meta-analysis of the controlled crossover studies found a systolic rise too small to distinguish from zero, and a sequential analysis that couldn’t settle the question either way. Five small trials is not a literature.

So: a demonstrated mechanism, a consistent direction, a vivid acute result, and a null pooled estimate. That is an unresolved small harm, and it is genuinely small. The honest position is not that mouthwash is dangerous. It’s that a habit with an unproven downside needs a proven upside, and for a healthy mouth the upside is thin.

My own bottle complicates that slightly, because it’s an anticavity fluoride rinse rather than a pure antiseptic. It carries a caries-prevention rationale a germ-killing rinse doesn’t, while still being antiseptic enough to hit the nitrate pathway. What I’d be trading, twice a day, is a small unquantified caries benefit in a mouth that has never had a cavity against a small unresolved vascular cost. Occasionally, before seeing people, that trade doesn’t arise. I landed on that pattern by accident and the evidence happens to endorse it.

Tooth colour is dentin seen through enamel. Enamel is mostly mineral, semi-translucent, and scatters light. The dentin underneath is softer, wetter, and runs yellow through amber. What you see is the core, filtered.

Both layers move with age, in the same direction. Chewing and acid thin the enamel so it transmits more and scatters less, while the pulp keeps laying down new dentin inward, thickening and darkening the core. A healthy tooth gets yellower over a lifetime because that is what healthy teeth do. Brilliant white in an adult mouth means bleach or porcelain, not vitality. Whiteness and dental health are uncorrelated, and the cosmetic market has spent decades selling the first as evidence of the second.

I notice the dentin directly, because my teeth are sensitive. A handful of blueberries is enough to register, which is a useful reminder that the enamel on top is a finite layer with a nerve supply behind it. Desensitizing toothpaste is one of the few consumer oral products with a serious evidence base behind it, and the actives differ enough that the label is worth reading.

The advice that follows sensitivity around is mostly wrong, though. Waiting half an hour after something acidic before brushing gets repeated everywhere. When researchers pooled the studies, delayed brushing protected bovine enamel and did nothing measurable for human enamel. The rule that reached every dental column was extrapolated from cow teeth. What did help human enamel was using a fluoride toothpaste at all.

Bleaching itself is close to harmless as far as anyone can measure, producing small and inconsistent changes in surface hardness in laboratory conditions. The irreversible damage comes from the drill. A minimally invasive veneer removes a modest slice of the tooth. A full-coverage crown removes roughly two thirds of it.

That gap is the whole risk in cosmetic dentistry, because a tooth sold as a veneer can be prepared as a crown. Enamel is acellular. It doesn’t heal or regrow. Grind a tooth to a peg and it’s in a restorative cycle for the rest of your life, with a meaningful minority eventually needing root canal treatment. Whitening strips are reversible and cheap. The appointment worth scrutinising is the one where someone offers to fix your smile permanently.

What my own file actually holds is a panoramic radiograph: a picture, not a number. It shows a full set of teeth and no restorations, which is reassuring and completely untrackable. You cannot plot it against last year.

Dentistry already has a fast, free, objective test that predicts periodontal stability, and almost nobody tracks it. Lang et al. published it in 1990, following maintenance patients for two and a half years and scoring whether each site bled when probed.

The result is lopsided in a way that’s easy to misread. A site that bleeds has roughly a one in sixteen chance of actually losing attachment, which is why “your gums bled a bit” is such weak information. A site that doesn’t bleed stays stable about 98% of the time.

That asymmetry makes it a negative test: the information is in the absence. It also explains why nobody markets it. There is no product in a result that says nothing happened, and a number that high partly reflects how rarely these patients were deteriorating in the first place. A negative test always looks strong when the base rate is low. It’s still the most predictive thing in the room, it’s generated at every cleaning, and people who track their resting heart rate and lipid subfractions almost universally don’t know theirs.

The six-month cleaning interval doesn’t survive contact with a trial either. When British researchers randomized adults to six-month, two-year, or risk-based recall and followed them for four years, gum health came out the same across all three. Six-month visits only win the economic case once you count how much people like going, which is a preference rather than a clinical result.

Better tests are arriving slowly. Point-of-care tests for an enzyme released during active gum breakdown now detect periodontitis at accuracies that would be useful if anyone standardised them. Direct-to-consumer saliva microbiome kits are a different category: they sample free-floating saliva that shifts with every meal, while the disease lives in the organised biofilm below the gumline, and no trial shows the report changes an outcome.

Bite force tracks grip strength, and it’s the cleaner instrument of the two. Both decline with age, but body mass confounds grip while leaving bite force alone. Grip strength took two decades to become a standard frailty marker. The jaw version is less confounded and almost nobody measures it.

Weak chewing also tracks frailty and death. Older adults who chew poorly are likelier to be frail, and in a four-year cohort of care-home residents, poor measured chewing predicted mortality while self-reported chewing difficulty did not. People don’t know how well they chew, which is an argument for measuring it rather than asking.

The same skepticism I applied to gums applies here. Most of this evidence is cross-sectional, nobody has run the Mendelian randomization, and nobody has randomized anyone to a stronger bite. Reverse causation is the obvious reading: illness arrives, chewing weakens as a symptom, and the gauge catches it. Bite force looks like a good instrument, not a thing to train.

Structural claims about adult jaws mostly don’t survive the anatomy. The mid-palatal suture stiffens through adolescence and into adulthood, so widening an adult upper jaw takes bone-anchored appliances or surgery, not tongue posture. Chewing gum raises blood flow in the prefrontal cortex without improving sustained attention in a rested adult. Masseter botulinum toxin was suspected of thinning jawbone, and the best-designed trial so far found no change.

Teeth may become regrowable, for a narrow group of people. Blocking a protein called USAG-1 with an antibody grows teeth in mice, and that work became a drug now in early human trials out of Kyoto, with a paediatric study cleared in 2026. The indication is congenital tooth agenesis, meaning children born without teeth, not adults who lost them to decay. Popular coverage keeps dropping that distinction.

Hydroxyapatite is a real fluoride alternative with a thin, industry-adjacent evidence base. An eighteen-month trial in adults found it non-inferior to standard fluoride toothpaste, and a supporting meta-analysis agrees while several authors across this literature are affiliated with a manufacturer. Both things are true at once, and that trial is still the only head-to-head of its length.

My shelf looks the same tomorrow. Everything on it is cheap enough that low-certainty evidence still justifies it, and the mouthwash stays in the cabinet for dinners.

What changes is one question. My hygienist already probes around every tooth and records which sites bleed. The quiet sites are the ones carrying that 98%. It costs nothing, it’s already being written down, and I’ve never asked to see it. At my next cleaning, I’m asking for it.