Wildfire Smoke and Migraine: What Air Quality Does to Attacks

The sky turns the color of weak tea, the air smells like a campfire nobody put out, and a couple of days later you are in a dark room waiting for your head to stop. Smoke season raises a reasonable question: is the air doing this, or is the timing a coincidence? The research points to a real but modest effect, weaker than the smell of the air would suggest, and the measured signal tends to arrive on a delay rather than the same afternoon.

That delay is the useful part. The studies that find a link mostly do not measure the day the smoke arrives; they measure exposure windows of two to seven days. So the attack you want to pin on this morning's haze may trace back to last week's. Below is what the evidence supports, what it does not, and what to write down so your own pattern becomes visible instead of guessed at.

What a smoke day puts in the air

Wildfire smoke is dominated by fine particulate matter, written as PM2.5. The EPA defines it as particles with diameters of 2.5 micrometers and smaller. For scale, the agency points out that an average human hair is roughly 70 micrometers across, about 30 times wider than the largest fine particle.

Size is what makes PM2.5 the pollutant health agencies watch most closely. Particles under 10 micrometers reach deep into the lungs, and the EPA notes that some may pass into the bloodstream, with fine particles posing the greatest risk to health of the regulated sizes.

The number you see on a smoke day is usually the Air Quality Index rather than a raw concentration. AirNow, the EPA's reporting service, splits the AQI into six color-coded bands: 0 to 50 is Good, 51 to 100 is Moderate, and a value of 100 corresponds roughly to the short-term national standard for a given pollutant. Above 100, air quality is described as unhealthy for sensitive groups first, then for everyone as the number climbs.

What the wildfire research found

The most direct study on this question came out in Headache in 2023. Elser and colleagues ran a case-crossover analysis on adult Californians in a large claims database from 2006 to 2020, linking daily wildfire-specific PM2.5 to residential zip codes. They identified 13,623 emergency department encounters for primary headache disorders among 9,898 people. Migraine was the most common diagnosis at 47.6% of visits.

For all primary headaches combined, a 10 microgram per cubic meter rise in 7-day average wildfire PM2.5 carried an odds ratio of 1.17, with a confidence interval of 0.95 to 1.44. By subtype the point estimates were higher still for tension-type headache at 1.42 and cluster headache at 1.29. Every one of those intervals crossed 1.0, which means none of them cleared the usual bar for statistical significance.

One result did clear it, and it is not the one you would expect. Overall PM2.5, from all sources rather than fires specifically, was associated with tension-type headache at an odds ratio of 1.29 (1.03 to 1.62), and was not associated with migraine. The authors called their findings imprecise and framed them as a suggestion that wildfire PM2.5 may be linked to headache visits, not as proof that it is.

Anyone selling you certainty about smoke and migraine is ahead of the evidence. The direction of the estimates is consistent. Precision is what this field does not have yet.

The broader air pollution picture

Wildfire smoke is one slice of a larger question, and the larger question has more data behind it. A systematic review published in Headache in 2023 by Portt and colleagues at the University of Toronto pulled together 12 studies covering more than four million participants, using designs ranging from case-crossover to time series.

Rather than pooling incompatible exposures into a single meta-analysis, the reviewers tabulated how often each pollutant's point estimate landed above 1.00, the value that marks increased risk. For PM2.5 it was 13 of 15 estimates, or 87%. Nitrogen dioxide came in at 10 of 13, ozone at 7 of 8, and carbon monoxide at 5 of 5. Their conclusion was that results were most consistent for nitrogen dioxide, a traffic-linked gas rather than a smoke particle.

Two caveats from that review deserve to travel with the headline. More than 75% of the two-pollutant confidence intervals still included the null value. And the reviewers observed that the stronger-quality studies generally reported weaker associations, which is the pattern you see when a real effect is smaller than early work implied.

The lag that trips people up

Where the evidence gets practical is timing. In Edmonton, Szyszkowicz, Stieb and Rowe analyzed 56,241 emergency department visits for migraine between 1992 and 2002, with women accounting for 78.5% of them. An interquartile increase of 6.2 micrograms per cubic meter in daily PM2.5 was associated with a 3.3% rise in migraine visits among women in the cold season, confidence interval 0.6 to 6.0, lagged two days.

Lagged two days. Not that afternoon. The California wildfire study used a 7-day average for the same reason: the effects researchers can detect are spread across a window, not concentrated on a single hour of bad air.

This has a blunt implication for self-diagnosis. Same-day blame is the least reliable version of the question you are asking. If you only notice air quality on days you already have a headache, you will confirm whatever you already believed, and you will miss the weeks where the smoke rolled through and nothing happened. The same reasoning applies to barometric pressure, where the drop often precedes the attack by a day or more.

Fewer hospital visits does not mean fewer headaches

A 2026 study in Headache adds a wrinkle worth understanding before you read any air quality statistic. Shapiro and colleagues at UCSF examined 206,230 pediatric emergency department encounters for primary headache in California between 2006 and 2019, ages 5 to 19.

The odds of an emergency visit went down after smoke exposure, not up. Using a binary threshold of wildfire PM2.5 at or above 15 micrograms per cubic meter, the adjusted odds ratio fell from 0.99 on the day itself to 0.84 by day six, with a confidence interval of 0.78 to 0.92. The authors' reading is that care-seeking behavior changes during smoke events rather than headaches becoming rarer, and they recommend that young people with headache disorders be equipped to manage attacks at home on smoke days.

Keep that in mind whenever emergency department counts are used as a proxy for how many people are in pain. When the air is bad, people stay inside, and the same instinct that keeps you off the roads keeps you out of the data.

What people do when the air turns

There is no treatment for smoke exposure to describe here, and nothing below replaces what your own doctor tells you. What follows is what public health guidance and people managing attacks in smoke-prone regions tend to converge on.

Check the number, not the sky. Haze is a poor gauge of concentration. AirNow publishes a Fire and Smoke Map alongside the daily AQI, and the 100 threshold is the point where guidance shifts to advising sensitive groups to cut back on outdoor exertion.

Reduce what you breathe indoors. Standard smoke-event advice is to close windows during the worst hours, filter indoor air where you can, and avoid adding more particles inside. None of this is migraine-specific, which is rather the point: the exposure is the thing you can change.

Protect the routine the smoke disrupts. Smoke events cancel plans, wreck sleep, and often land during hot stretches, and those are established trigger territory in their own right. Our piece on heat and migraine covers the overlap, and a smoke week is a bad week to also let meals and sleep timing drift.

Bring a changed pattern to your doctor. If your attack frequency shifts during smoke season, or you have breathing symptoms alongside the headaches, that is a specific and useful thing to describe at an appointment rather than something to sit with until the air clears.

Separating smoke from everything else in your log

Here is the hard part, and it is the reason the published estimates stay imprecise. Smoke days rarely arrive alone. They come with heat, with pressure changes, with the poor sleep that comes from a closed-up house, and with whatever stress the fire itself is causing. Any one of those can carry the blame that you are trying to assign to air quality.

I built this site and the app behind it for my mom and my wife, who both get migraines. I do not get them, so everything I understand about how these patterns surface, I learned from reading their logs rather than from my own head, and the lesson that keeps repeating is that single-cause explanations fall apart once you have enough entries to check them against.

What makes a smoke week readable later is logging the exposure days themselves, not only the attack days. The windows that showed effects in the research ran from two to seven days, so a record that marks when the air was bad is what lets you look backward across the right span.

Some tracking apps let you extend the trigger list past the defaults. MigrAid, for example, ships with 12 built-in triggers and lets you add custom ones, so a smoke tag can sit alongside the weather it records automatically, and the confounders stay visible next to the thing you are testing.

Our guide to tracking migraine triggers goes deeper on how to avoid fooling yourself with a short run of data.

The bottom line

Fine particles from wildfire smoke are associated with more headache-related emergency visits, with an odds ratio of 1.17 per 10 micrograms per cubic meter of 7-day average wildfire PM2.5, and a confidence interval that does not rule out no effect at all. Across the wider pollution literature, 87% of PM2.5 estimates point toward increased migraine risk, and the better-designed studies report the smaller effects. Measured lags run two days and longer.

So treat smoke as a plausible contributor rather than a verdict. Reduce exposure on the days the number is high, and record when the air was bad instead of trusting your memory of it later. Your own log, built across several smoke seasons, will answer this question for you more convincingly than the population data currently can.