Almost every migraine information page carries a list of triggers. Chocolate, cheese, red wine, bright light, missed meals, lie-ins, stress, the let-down after stress. The lists are broadly the same everywhere, and people reorganise years of their lives around them.
The lists are not wrong, exactly. They are a summary of what large numbers of people report. What they leave out is that a thing reported before an attack is not necessarily a thing that caused the attack — and for at least some of the classic entries, there is a strong case that the order of events runs the other way.
The attack starts before the pain does
The ICHD-3 describes migraine as having phases, and the one that matters here is the premonitory phase: symptoms that can begin hours to a day or two before the headache itself. Recognised premonitory features include fatigue, difficulty concentrating, neck stiffness, yawning, mood change, increased thirst or urination, sensitivity to light and sound, and — this is the important one — food cravings.
Read that list again with the trigger list beside it. Craving something sweet is on both. So is neck stiffness. So is noticing that the light is too bright.
If the craving is part of the attack, then eating the chocolate is not the cause of what follows. It is the first thing you did after the attack had already begun. Avoiding chocolate for the next decade will not change the frequency of anything, and you will have given up chocolate.
Why this is hard to settle
The honest position is that for any individual, separating the two is genuinely difficult, and a personal diary alone usually cannot do it. Three reasons:
- Base rates. If you eat chocolate most weeks and have attacks most weeks, they will coincide often, whether or not one causes the other.
- You only notice the hits. The times you ate chocolate and nothing happened are not memorable, and they do not get logged.
- The confound runs the right way to fool you. If a premonitory craving reliably precedes an attack, then chocolate will appear to predict attacks beautifully — because it does predict them. It just does not cause them.
Note what that third point means: a trigger and an early symptom produce the same pattern in a diary. No amount of careful logging separates them on its own, which is why the lists persist.
What your own record can still do
A diary will not prove causation, but it is good at three narrower jobs.
Ruling a candidate out. A negative is much easier than a positive. If you have logged forty exposures and attacks follow at the same rate as on the days without, that candidate is not doing much to you. That is worth knowing, and it is the result people rarely go looking for.
Timing the sequence. Recording when in the run-up something happened — four hours before the pain, or four days — starts to distinguish a plausible external trigger from something happening inside an attack that has already started.
Generating one hypothesis at a time. A candidate you have identified from your own record, tested deliberately, with your clinician aware of it, is a much better use of a diary than avoiding fifteen things at once and never learning anything about any of them.
What not to do with this
Do not use a diary to decide on your own to stop a prescribed medication, to eliminate a food group, or to restrict a child's diet. This article is about how to read a record, not about what to change. Elimination has costs of its own — nutritional, social, and the quiet cost of arranging your life around a rule that turns out to be doing nothing — and whether those costs are worth paying is a decision to make with a clinician who knows the rest of your history.
The useful summary: some of what is on the trigger lists is probably the attack announcing itself. Treat your list as a set of questions, not a set of rules.
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.
