Air pollution and dementia risk: what the evidence shows, and what actually helps

Short answer: Long-term exposure to fine particle pollution is associated with a modestly higher risk of dementia, and the association is consistent enough that the Lancet Commission counts air pollution among the fourteen modifiable risk factors and WHO added a recommendation about it in 2026. But the effect per person is small, it is not established as causal, and the pollutants do not all behave the same way — several show nothing at all. The most useful things you can do about it are unglamorous: know your local air quality, treat wildfire-smoke days as different from ordinary days, and keep your effort proportional to one factor out of fourteen — and the one the Lancet Commission itself calls the most shaped by policy rather than personal choice.

This is educational content, not medical advice. It describes associations found in population research, not predictions about any individual.

How big is the association, really?

The most careful synthesis to date pooled 32 studies in The Lancet Planetary Health in July 2025. Its headline findings:

• Fine particles (PM2.5) — 21 studies, more than 24 million people: a hazard ratio of 1.08 (95% CI 1.02–1.14) for every 5 µg/m³ increase in long-term exposure.
• Nitrogen dioxide — 16 studies, over 17 million people: 1.03 (1.01–1.05) per 10 µg/m³.
• Black carbon (the sooty fraction, largely from traffic and burning) — six studies: 1.13 (1.01–1.27) per 1 µg/m³.

An 8% higher risk is not nothing across a whole population. It is also not the kind of number that should dominate anyone's personal planning. For comparison, that is smaller than the associations reported for untreated hearing loss, midlife high blood pressure or physical inactivity — all of which you have far more direct control over.

The review's authors rated their own overall certainty as moderate, which in this field is a respectable grade and an honest one.

The part almost nobody prints: three pollutants showed nothing

The same meta-analysis tested six pollutants. Three of them came back without a significant association:

Look at the PM10 row for a moment. The point estimate is 1.52 — a 52% higher risk, a number that would make a terrifying headline. The confidence interval runs from 0.80 to 2.87. It includes no effect at all, it includes a 20% reduction, and it includes a near-tripling. Four studies, wildly inconsistent. The correct reading of 1.52 here is "we don't know."

That is worth carrying around, because it is the most common way a real research finding turns into a misleading article. We wrote a whole guide to it: seven ways a dementia statistic can mislead you.

How strong is the evidence — not the effect, the evidence?

There are two defensible answers, and they are both correct, because they grade different things.

Answer one: "highly probable." An umbrella review published in Age and Ageing in 2026 pulled together 54 systematic reviews, 66 meta-analysed outcomes and 270 exposure–outcome associations covering the whole urban environment — noise, light, green space, social contact, traffic, pollution. Air pollution came out with the strongest evidence of anything examined: PM2.5 showed significant detrimental effects in 10 of its 11 meta-analysed outcomes. Among all the things about where you live that might affect your brain, this is the one with the most support behind it.

Answer two: "two stars out of five." Researchers applying the Burden of Proof framework — a deliberately conservative method built to stop over-confident conclusions, which relaxes the usual statistical assumptions and penalises unexplained disagreement between studies — analysed 28 longitudinal cohorts in Nature Aging in

1. Their verdict: a real and statistically significant association, with a minimum 14% excess risk across the typical range of exposure, rated two stars, meaning "weak and/or inconsistent evidence." Alzheimer's disease specifically scored three stars.

Those two answers are not in conflict. Consistent direction across dozens of studies, small effect sizes, and substantial disagreement about magnitude — that is exactly a "highly probable but weakly quantified" finding.

And the Nature Aging analysis contained one genuine surprise: no significant association with vascular dementia. The mechanism everyone assumes — particles inflame blood vessels, blood vessels supply the brain — is the one the data did not confirm. If you want to know why nobody should be quoting this factor with the confidence they quote blood pressure, that is why.

Wildfire smoke is a different question

If you live anywhere that gets smoke seasons, this is the part that matters most.

A study of 1,223,107 Kaiser Permanente members aged 60 and over in Southern California, published in JAMA Neurology in 2025, separated wildfire-generated fine particles from everything else. Over the study period 80,993 people received a dementia diagnosis. Per 1 µg/m³ of three-year average exposure:

• Wildfire PM2.5: odds ratio 1.18 (1.03–1.34)
• Non-wildfire PM2.5: odds ratio 1.01 (1.01–1.02)

Microgram for microgram, the wildfire particles carried an association roughly eighteen times larger. Smoke particles are smaller, chemically different, and arrive in short, intense episodes rather than as a steady background — and the association was strongest in people under 75, in racially minoritised groups, and in higher-poverty neighbourhoods, which is a finding about who is able to stay indoors, filter their air and leave, not about anyone's biology.

It is one cohort, in one state, and it is observational. But it is the single most actionable finding in this whole literature, because smoke days are visible, forecastable and temporary.

(A note on sourcing: this paper carries a published correction. The original version misstated the non-wildfire figure; the corrected value — the one above — is 1%.)

Does reducing exposure actually help?

Nobody has run, or could easily run, a randomised trial of clean air and dementia. The closest evidence is a 2022 analysis in PNAS that followed 2,239 women aged 74 to 92, all dementia-free at the start, 398 of whom developed dementia over about six years.

Instead of measuring pollution levels, it measured change. Women whose local air improved more over the preceding decade had lower dementia risk: HR 0.80 (0.71–0.91) per 1.78 µg/m³ reduction in PM2.5, and 0.80 (0.71–0.90) per 3.91 ppb reduction in nitrogen dioxide. The authors put the size of that in human terms: it was "equivalent to the lower risk observed in women 2.4 years younger at baseline."

Two caveats belong with it. The cohort is small and entirely female. And in the same study, the association with baseline pollution levels was 1.16 (0.98–1.38) — an interval that crosses no-effect. The improvement signal was cleaner than the exposure signal, which is unusual, and worth being a little curious about rather than fully convinced by.

WHO's second-edition dementia guidelines, published in July 2026, added a new recommendation to reduce exposure to air pollution — the first time the organisation has included it. Worth keeping in view alongside everything above: WHO grades its recommendations using the GRADE system, and on a topic where the underlying studies are all observational, a recommendation is a statement that something is worth doing, not a claim that the effect size is settled. You can read what else changed in the WHO's 2026 guidance.

So should you buy an air purifier?

You will find a lot of pages answering this question. A striking number of them are published by companies that sell air purifiers and air filters. Almost none of them mentions the only randomised trial that has looked at filtration and thinking.

That trial ran HEPA purifiers against identical-looking sham units in 119 people's homes — one month on each, in random order, with a washout month in between — and measured Trail Making Test times, a standard pencil-and-paper measure of attention and mental flexibility. The results, published in Scientific Reports in 2026:

• No significant overall difference between HEPA and sham.
• Age changed the picture (interaction p = 0.02). Participants aged 40 and over completed Part B about 12% faster after a month of real filtration — 54.0 seconds versus 61.4 seconds.
• No effect at all in participants under 40.

Take that for exactly what it is: a secondary analysis, of a subgroup, showing a seven-second difference on a cognitive test, in 119 people, over one month. It is a reasonable hint that cleaner indoor air does something measurable. It is not evidence that a purifier lowers anyone's dementia risk, and nobody should sell it to you as such.

A HEPA filter is a sensible purchase if you have smoke seasons, a busy road outside, asthma in the house, or an indoor source you can't remove. It is not a brain-health product.

What is actually worth doing

1. Know your number, and treat the bad days differently. In the US, AirNow gives a free local reading. Most days it will not change your plans. Smoke days should: keep windows shut, run filtration if you have it, and move exercise indoors. The wildfire data says the episodes are where the signal is.

2. Don't stop exercising because of the air — move it. Keeping up physical activity is one of the best-supported things you can do for your brain, and the advice here is to shift the route and timing, not to cut the activity. Away from heavy traffic, earlier in the day, indoors when the reading is bad. (Where you move matters in other ways too — see whether physical work counts as exercise for your brain.)

3. Deal with the indoor sources you control. Cooking without ventilation, wood or coal burning, and smoking indoors are the big ones. Household air pollution is part of what WHO's recommendation covers, and unlike the air over your city, it is yours to change. If you smoke, that dwarfs everything else here — here is what quitting does.

4. Spend some time in green space. The same umbrella review found green and blue spaces associated with lower dementia risk in three of six meta-analysed outcomes. Weaker evidence than the pollution finding, and tangled up with exercise and social contact, but it costs nothing and it is pleasant.

5. Keep it in proportion. Air pollution is one of fourteen modifiable factors on the Lancet Commission's list, and the Commission itself describes it as the one most shaped by policy rather than personal choice. Blood pressure, hearing, blood sugar, activity, smoking, alcohol, social contact, cholesterol and vision are almost entirely yours to act on. If you have bought a purifier and never had your hearing checked, you have optimised the wrong variable. Here is what all fourteen are, in plain English.

How we research and review our content →

An independent physician audited the Solenna app. We published all nine findings →

Educational content only. It does not diagnose any condition and is not a substitute for professional medical advice. If you have breathing problems, heart disease or any condition made worse by poor air quality, discuss your local exposure with your clinician.