About That Trip to the Dentist

Sometime in the next two years, someone you pay for medical advice is going to hand you an oral microbiome test. It will cost several hundred dollars. It will come back with a species list, an index of some kind, and a protocol. The protocol may be perfectly sensible. The framing will be that your mouth is a source of bodywide inflammation, and that inflammation is implicated in heart disease, diabetes, cognitive decline, arthritis, and mood.

I can get behind every clause in that sentence. Strung together, though, and it's something else. Marketers are counting on you to translate all that nuance into a simple promise: fix your gums to prevent heart attacks and treat depression. Don't buy it. You are listening to a sales pitch, not facts.

The question is how to tell the difference. A case unfolding in public, in the business pages, gives us the chance.

Witness the machine assembling itself

The Wall Street Journal ran a piece in late July 2026 on the mouth-body connection. It is a nice piece of health journalism. It covers the oral microbiome, biofilms, bacterial translocation across inflamed gum tissue, and the emerging interest in treating periodontal disease as an inflammatory condition rather than a plumbing problem.

If you read this previous Nāhua Fieldnote, you’ll recognize the example. I did not expect to find it in the WSJ quite so soon.

The science makes for a good read. The more revealing story is how much of medicine and business is already reorganizing around it. And here I thought psychedelics were the only industry willing to move this quickly on unsettled science.

Dental schools are revising curricula around the systemic effects of oral disease. Medical educators are calling for oral health to become part of physician training. Harvard's dental school has a standing initiative on oral-health policy and health-system design. PDS Health is putting dental practices onto Epic so dental and medical records live in one place. Periodontists are making the case for Medicare dental benefits partly on downstream medical costs. Two companies are developing immune-targeting treatments for gum disease. And a forty-billion-dollar consumer oral-care industry is repositioning brushing and flossing as a wellness practice, explicitly modeled on skin care.

None of that is a study. All of it is infrastructure.

Inflammation is an organizing principle, a commercial engine, and a scientific hypothesis—in roughly that order. Long before every causal link is proved, the "inflammation explains everything" idea reorganizes which specialists talk to one another, what gets measured, what startups get funded, and what insurers are asked to cover.

Studies can be corrected. Infrastructure is harder to unwind. Once schools rewrite curricula, health systems rebuild records, startups raise money, and insurers are asked to pay, the inflammation story has consequences beyond the evidence. It is no longer merely a hypothesis waiting to be tested. People and institutions now have an interest in keeping it persuasive.

What the article leaves out

The article establishes that gum disease and several systemic illnesses travel together. Good to know. But the practical claim now being built around that association is much stronger: treat the gum disease and you may prevent the heart attack, the stroke, or the dementia.

That requires different evidence. The article moves past the distinction almost without stopping.

The associations are real and reasonably robust. Roughly half of American adults over thirty have periodontitis, and close to sixty percent of those over sixty-five. Periodontitis travels with cardiovascular disease, diabetes, rheumatoid arthritis, and cognitive decline in observational data, consistently enough that the pattern is not a fluke.

Researchers have shown that treating gum disease can improve several things associated with cardiovascular risk: blood pressure, cholesterol, inflammatory markers, and measurements of how well blood vessels respond. That is useful.

But a better laboratory value is not the same thing as fewer heart attacks or strokes. No clinical trial has yet shown that periodontal treatment prevents either.

Researchers call those stand-in measurements "surrogate endpoints." They are useful because they change faster than heart-attack rates. They are dangerous when improvement in the stand-in is sold as improvement in the disease.

The absence of that trial is not a scandal. A study large enough to detect fewer heart attacks would be expensive, slow, and difficult to conduct when the treatment is something as obvious as intensive dental care.

More revealing is what happens when the evidence falls short. In 2020, researchers reviewed 24 randomized trials of periodontal treatment and asked whether the abstracts accurately described the results. None had measured heart attacks or strokes. Yet 64 percent still described the treatment positively even when the study’s main finding was not statistically significant.

Sixty-four percent. In a body of research that has never shown fewer heart attacks or strokes, nearly two-thirds of the abstracts still found a way to sound positive.

Something similar is happening with the pneumonia finding the article cites: a preventive dental visit before hospitalization was associated with a 10 percent lower risk of hospital-acquired pneumonia, and professional gum treatment within six months with a 30 percent lower risk.

Those numbers come from insurance records, not clinical trials. People who get regular dental care differ from people who don't in dozens of ways (income, overall health, mobility, home care) that also protect against pneumonia. The dental care might be helping, or it might just be tracking who has a functional life and support system. Health records cannot untangle the two. The finding is interesting, but it is an association, not proof of what the trip to the dentist accomplished.

Meanwhile, in cardiology

While I was drafting this essay, the biggest headline machine of the inflammation hypothesis had a bad week of its own, in the organ where the whole idea started, the heart.

Cardiology has run the test dentistry has not: if you lower inflammation, do fewer people actually have heart attacks or strokes?

The answer has been maddeningly mixed. In 2017, canakinumab (an antibody that blocks one inflammatory pathway) modestly reduced repeat cardiovascular events in more than ten thousand patients. It also increased fatal infections, and the drug was never brought to market for this use.

Two trials of colchicine, an older anti-inflammatory drug, looked more promising. Then came the largest colchicine trial ever conducted, with more than seven thousand patients. The drug lowered C-reactive protein (CRP), a standard blood marker of inflammation, exactly as expected. It did not reduce heart attacks or strokes.

The marker improved. The patients did not. The lead investigator took his own father off the drug the day he saw the data.

Then in July, Novo Nordisk reported the results of ZEUS. The trial was built to answer the most obvious defense of the earlier failures: perhaps anti-inflammatory treatment works only in people who are clearly inflamed.

So researchers enrolled more than 6,300 patients who had heart disease, kidney disease, and a CRP high enough to qualify them as inflamed. These were the people the theory said should benefit. The drug, ziltivekimab, lowered CRP and IL-6 exactly as designed. But it prevented no more heart attacks, strokes, or cardiovascular deaths than placebo. The hazard ratio was 0.99, which is a technical way of saying nothing happened.

The usual defense of a failed trial is that the drug was given to the wrong patients. That defense is much harder to make here, because these patients were chosen to fit the theory before anyone knew the answer. (Granted, the defense isn't gone entirely. Everyone in ZEUS also had kidney disease, and someone will inevitably argue that kidney disease is its own flavor of inflammation. But the escape route is much narrower than it was, and no one is running a trial this large to test it again anytime soon.)

A fair caveat, in the spirit of not getting ahead of the evidence. What exists so far is the company's public announcement, not a peer-reviewed publication. I am analyzing an early signal in real time. The full tables could still contain subtleties, but the primary result is hard to misread: a massive trial designed around the inflammation theory came up empty.

But my first reaction was triumph. Here, I thought, was proof that the inflammation story had been oversold. Yeah, no. That was too easy.

ZEUS does not show that inflammation is irrelevant to heart disease. A drug can fail because it blocks the wrong pathway, arrives too late, or moves a marker that was never driving the disease. Two more trials of the same drug are still running, in heart failure and after heart attack, and both report next year.

What ZEUS takes away is the easiest excuse. The markers fell. The patients did not benefit.

Mechanism, marker, and treatment are three separate claims. Evidence for one does not automatically validate the next.

Next up: the brain

The theory connecting gum disease to Alzheimer’s has already produced its own major drug trial. It also produced exactly the kind of unfinished result that later becomes easy to oversell.

The premise was compelling. Porphyromonas gingivalis (a primary bacterium behind advanced gum disease) produces toxic enzymes called gingipains. Researchers have found those enzymes in the brains of deceased Alzheimer’s patients. In mice, infection with the bacterium produces Alzheimer’s-like damage, and blocking the gingipains appears to prevent it.

That was enough to justify a serious test. Cortexyme, a biotech startup, developed a drug designed to enter the brain and block gingipains, then enrolled 643 people with mild-to-moderate Alzheimer’s in a placebo-controlled trial.

The drug failed its main test. Across the entire trial, it did not significantly slow cognitive decline or the loss of everyday functioning.

But one subgroup produced a striking result. Among the 242 participants who entered the study with detectable P. gingivalis DNA in their saliva, decline was 42 percent slower on the lower dose and 57 percent slower on the higher one. The patients whose bacterial levels fell most also tended to do best.

To a non-scientist, the obvious reaction is: then the drug worked in the people who had the bacterium. Why not give it to them?

Because that is precisely what the trial did not establish. It found a pattern suggesting that the drug worked in those patients. But the trial had not been designed to test that proposition directly, so the result could not tell researchers whether the same pattern would appear again in a new group of patients.

That distinction matters. The next trial would not be bureaucratic repetition of something already known. It would be the first trial designed specifically to test the interesting thing the first one discovered.

The first trial generated the hypothesis. The second had to test it.

Importantly, the researchers had identified this subgroup before looking at the results. They had not rummaged through a failed trial afterward until something appeared positive. The company announced a confirmatory trial specifically in the biomarker-positive patients, the subgroup where the promising result had appeared.

That trial never happened. Facing FDA safety holds over liver toxicity, Cortexyme rebranded as Quince Therapeutics and pivoted to an entirely different disease area.

Here is where the scientific story and the public story part company. The full trial failed. The promising subgroup was never tested again. The company moved on. Yet the sentence that survived was not: “A failed Alzheimer’s trial produced an intriguing subgroup result that remains unconfirmed.” It was: gum-disease bacteria may travel to the brain and contribute to Alzheimer’s.

That is how an unfinished scientific result becomes a durable public "fact." The failed main result disappears. The missing follow-up disappears.

The possibility remains free to replicate in the marketing wild.

Four questions before you buy the test

By now the problem should be clear. Inflammation science is not fake. It is just unusually easy to enlarge. That word "inflammation" can refer to a blood marker, an immune pathway, a symptom pattern, or a cause of disease. The commercial story benefits when those meanings blur together. Here are four questions that help us keep them separate.

1. Do the conditions merely travel together, or does treating one change the other?

Inflammation is where many roads meet. Obesity, smoking, poor sleep, chronic stress, inactivity, poverty, gum disease, and aging can all raise CRP. So when CRP travels with depression (or gum disease travels with heart disease) the overlap may reflect causes the two conditions share. It does not necessarily mean that one is producing the other.

That is why studies of gum disease and depression keep changing their answer. The apparent relationship shifts depending on which overlapping factors researchers can measure and adjust for. Smoking, obesity, sleep, stress, healthcare access, and social disadvantage all matter, and no dataset can subtract an entire life.

The useful question is not merely whether two conditions are associated. Ask whether treating one changes the other.

2. What did the test actually measure?

CRP is an inflammatory marker produced largely by the liver in response to signals such as IL-6. It is a useful blood test. It is not a window into the brain.

The immune system can affect the brain through several routes, but those routes are difficult to measure and do not rise and fall in lockstep with a blood test. A high CRP may tell you that inflammation is present somewhere in the body. It cannot tell you what is happening in the brain, whether inflammation is causing a person’s depression, or what treatment would reverse it.

That sounds obvious once stated. Much of the consumer inflammation market depends on skipping the sentence.

So when a newspaper says “inflammation in depression,” it can collapse several different scientific questions into one phrase. A blood marker is not a brain state. Immune signaling is not a diagnosis. Merging those distinctions is how a blood test becomes a brain claim.

A test that measures peripheral inflammation is telling you about peripheral inflammation.

3. Which patients are we talking about?

This is the one with the most clinical weight. Immune activation can produce a recognizable cluster: low energy, loss of pleasure, slowed movement, extra sleep, appetite change, and social withdrawal. Researchers call this sickness behavior. It overlaps with depression, but it is not the whole of depression. It does not explain features such as guilt, self-blame, rumination, or suicidality.

Major depressive disorder is defined by a symptom count. Two people can receive the same diagnosis with very different underlying problems. Put both into the same drug trial, along with hundreds of others, and any immune-linked subgroup may disappear inside the average.

That is what happened in a trial of infliximab, an immune-blocking drug, for treatment-resistant depression. The drug did not help the group as a whole. It appeared to help the patients who entered the trial with markedly elevated CRP.

That result did not prove that infliximab works for depression. It suggested that the diagnosis “treatment-resistant depression” may contain an immune-linked subgroup that needs to be studied separately.

GAIN and the infliximab trial point to the same possibility: the average patient may be the wrong unit of analysis. A mechanism may matter greatly in one subgroup and not at all in the rest.

That is a reason to run a more targeted trial. It is not yet a treatment recommendation.

4. Was the promising result ever repeated?

A promising subgroup result is not worthless. It is a lead. The next step is straightforward: enroll patients who fit that subgroup from the beginning and see whether the result appears again. Until that happens, researchers have an interesting possibility, not a validated treatment.

GAIN never got that second trial. Neither did the infliximab signal. ZEUS did something close: researchers selected in advance the patients the hypothesis said should benefit, lowered the intended inflammatory markers, and still saw no clinical benefit.

“Is there a plausible mechanism?” isn't good enough here. Almost every good sales pitch has one.

The question is: “Has anyone tested this exact claim again, on purpose, and gotten the same answer?”

A good policy does not need a glamorous rationale

None of this is an argument against better dental care. Medicare should cover it. Gum disease destroys teeth, causes pain, and makes eating harder. Physicians and dentists should share records because the mouth is part of the body. Dental students should understand systemic medicine.

The problem is not the policies. It is the borrowed rationale.

These good ideas are increasingly being justified with the promise that dental care may prevent heart disease, dementia, and other systemic illness. If that promise weakens, opponents can pretend the policy weakened with it.

Dental coverage never needed an Alzheimer’s claim. Teeth were reason enough.

Borrowing a glamorous rationale for a boring good idea is a bad trade. The glamour gets attention. It also becomes the weak point.

My own work does not get an exemption

In the Signal Loss Model, Neuroimmune Dysregulation is one of three possible mechanisms of treatment resistance, alongside Untethered Cognition and Pursuit-Reward Decoupling. One of three. It is not an explanation for every case.

A high CRP does not diagnose Neuroimmune Dysregulation. A sickness-behavior pattern is a clue, not proof. And no inflammatory finding currently tells me which intervention will help a particular person.

Which is why Nāhua does not measure it. We do not order CRP or IL-6 panels at intake, we do not sort guests by inflammatory status, and we do not claim that anything we do lowers anyone's inflammation. We would change that the day a marker is shown to predict who responds to what. Until then we would be charging you for a number we cannot act on, which is nothing more than performative medical theater.

Treatment-resistant depression is unlikely to be one thing. Immune involvement may matter greatly for some people and not at all for others. The subgroup evidence makes that plausible, but it does not yet yield a validated test.

A panel and a protocol are easy to sell. The data, for now, offer neither.

You really should call your dentist

Floss. Treat your gum disease. See a dentist twice a year. Do it because losing teeth is bad, and gum disease is painful and expensive.

But if you are told to care for your teeth to protect your heart or your brain, ask why. Did treating gum disease prevent a heart attack or dementia, or did it merely improve a measurement associated with risk?

When someone wants to charge you several hundred dollars for a panel that ends with instructions to floss, you'll know they were trying to sell certainty the evidence has not earned.

And now they won't have your money.

Essays on treatment resistance, altered states, and the conditions under which change becomes possible.

Subscribe

References

Alzforum. 2023. “Atuzaginstat.” Last updated February 6, 2023.

Baker, Dian, Karen K. Giuliano, Madhuli Thakkar-Samtani, et al. 2023. “The Association between Accessing Dental Services and Nonventilator Hospital-Acquired Pneumonia among 2019 Medicaid Beneficiaries.” Infection Control & Hospital Epidemiology 44 (6): 959–61.

Dominy, Stephen S., Casey Lynch, Florian Ermini, et al. 2019. “Porphyromonas gingivalis in Alzheimer’s Disease Brains: Evidence for Disease Causation and Treatment with Small-Molecule Inhibitors.” Science Advances 5 (1): eaau3333.

Eke, Paul I., Gina O. Thornton-Evans, Liang Wei, Wenche S. Borgnakke, Bruce A. Dye, and Robert J. Genco. 2018. “Periodontitis in US Adults: National Health and Nutrition Examination Survey 2009–2014.” Journal of the American Dental Association 149 (7): 576–88.e6.

Jolly, Sanjit S., Marie-Astrid d’Entremont, Shun Fu Lee, et al. 2025. “Colchicine in Acute Myocardial Infarction.” New England Journal of Medicine 392 (7): 633–42.

Kolata, Gina. 2026. “A Promising Heart Drug Fails, Challenging a Long-Held Theory of Disease.” New York Times, August 6.

Landro, Laura. 2026. “Your Mouth Could Hold the Secrets to a Longer and Healthier Life.” Wall Street Journal, July 26.

Novo Nordisk. 2026. "Novo Nordisk Provides Update on the ZEUS Phase 3 Trial in People with ASCVD, CKD and Inflammation." Company announcement, July 31.

Nidorf, Stefan M., Aernoud T. L. Fiolet, Arend Mosterd, et al. 2020. “Colchicine in Patients with Chronic Coronary Disease.” New England Journal of Medicine 383 (19): 1838–47.

Raison, Charles L., Robin E. Rutherford, Bobbi J. Woolwine, et al. 2013. “A Randomized Controlled Trial of the Tumor Necrosis Factor Antagonist Infliximab for Treatment-Resistant Depression: The Role of Baseline Inflammatory Biomarkers.” JAMA Psychiatry 70 (1): 31–41.

Ridker, Paul M., Brendan M. Everett, Tom Thuren, et al. 2017. “Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease.” New England Journal of Medicine 377 (12): 1119–31.

Sabbagh, Marwan N. 2023. “The Gingipain Inhibitor Atuzaginstat and Results of the GAIN Trial.” Alzheimer’s & Dementia 19 (S12): e079337.

Shaqman, Murad, Khadijeh Al-Abedalla, Julie Wagner, Helen Swede, John Caton Gunsolley, and Effie Ioannidou. 2020. “Reporting Quality and Spin in Abstracts of Randomized Clinical Trials of Periodontal Therapy and Cardiovascular Disease Outcomes.” PLOS ONE 15 (4): e0230843.

Tardif, Jean-Claude, Simon Kouz, David D. Waters, et al. 2019. “Efficacy and Safety of Low-Dose Colchicine after Myocardial Infarction.” New England Journal of Medicine 381 (26): 2497–2505.