Weather is near the top of every list of reported migraine triggers, and people are often extremely confident about it. It is also the trigger where the gap between confidence and evidence is widest — not because the reports are dishonest, but because weather has a set of properties that make it unusually good at fooling a diary.
This article is about the reasoning, not about a verdict. Whether pressure affects you is a question your record can inform and probably cannot close.
Four reasons weather is hard
1. It happens to everyone, constantly
An exposure that occurs every single day is the worst possible candidate for this kind of inference. You cannot compare exposed days against unexposed days if there are no unexposed days — only degrees, and choosing where to put the cut-off is a decision that can produce almost any answer you like.
2. Weather is autocorrelated, and so is migraine
A falling barometer today makes a falling barometer tomorrow more likely. An attack today makes an attack tomorrow more likely. Two patterns that both cluster in time will appear to line up far more often than chance would suggest for two independent events, and the usual intuitions about “that has happened too often to be coincidence” are calibrated for independent events.
3. There are dozens of weather variables
Pressure, pressure change, rate of change, temperature, temperature swing, humidity, wind, storms, sunshine hours. Test enough of them against your attack days and one will look convincing. This is the multiple-comparisons problem, and it does not announce itself — the spurious result looks exactly like a real one.
4. The premonitory phase gets in the way
Sensitivity to light and to the environment can be part of the attack's early phase, before the pain. Someone already in that phase is more likely to notice that the sky has gone heavy. The weather then gets recorded as something that preceded the attack, which it did — but so did the attack.
This is the same confound that makes food cravings look like triggers, and it is discussed at greater length in Triggers, prodrome, and the difference between them.
What a personal record can do
It can weaken a strong claim. If pressure drops are following a normal seasonal pattern where you live and your attack rate is flat across them, a large effect in you is unlikely. Negative results are the ones an individual record handles best.
It can stop you attributing. Much of the cost of a believed-in trigger is not the avoidance, it is the explanation: every attack gets filed under “the weather” and nothing else gets examined. A record that shows attacks distributed across all sorts of days reopens the question.
It can give a clinician context. “Clusters in autumn” is a real observation, even if the mechanism behind it is unproven.
What it cannot do
It cannot establish that a weather variable causes your attacks. There is no volume of personal logging that gets around the four problems above, because they are problems of study design rather than of sample size. An individual keeping a diary is running an observational study with one participant, no control, no randomisation and no blinding — and weather is the exposure that punishes each of those weaknesses hardest.
What we do with it here
MigraSens can join local weather to your entries so that the observation is at least accurate rather than remembered. When a factor shows up more often than chance around your attacks, it is flagged as a question, and the wording is deliberate: it is a thing to look at, not a finding.
The practical position: log the attack, let the weather be recorded alongside it, and do not reorganise your life around a signal this weak. If the pattern is strong enough to matter, it will still be there in six months.
Sources
- Headache Classification Committee of the International Headache Society. The International Classification of Headache Disorders, 3rd edition (ICHD-3). Cephalalgia 2018;38(1):1–211. ichd-3.org
Written in house and checked against the ICHD-3. Where this page describes the state of the evidence rather than a specific finding, it says so, and it cites nothing it cannot point at. A named clinical reviewer is to be appointed before launch. Found an error? Tell us and we will correct it on the page.
