How Long Does a Migraine Last? Three Rules Change the Answer

Four to seventy-two hours, in an adult, untreated. That is the window the diagnostic criteria use, and it is where every honest answer to this question starts.

What the explainer pages leave out is that the same criteria carry three counting rules, and the rules decide which of your attacks fall inside the window at all. They are the difference between a duration a neurologist can work with and a figure you assembled in the waiting room.

The window describes an attack you probably did not have

ICHD-3 is the classification headache specialists actually work from. Its entry for migraine without aura asks for “headache attacks lasting 4-72 hr”, and then adds four words in brackets that change the meaning: untreated or unsuccessfully treated.

That parenthesis is doing a great deal of work. The attack you broke with a triptan ninety minutes in was still a migraine. Its natural length is simply not observable any more, because you intervened, which is the entire point of having intervened.

People discount those attacks anyway. “It only lasted an hour and a half, so it can’t have been a proper one.” That reasoning quietly deletes your best-treated attacks from your own record, and it pushes your migraine-day count downwards at exactly the moment the count is being used to decide what you get offered next. Undercounting is already the common direction of error, and this is one of the mechanisms behind it.

One adjustment for age: in children and adolescents under 18, the same criteria allow attacks from two hours.

If you slept through it, the clock runs to when you woke

This is the rule almost nobody has read, and it is written into the criteria as a note: when the patient falls asleep during a migraine and wakes up without it, the duration is reckoned until the time of awakening.

Work an evening attack through it. Pain starts at nine, you go to bed at eleven because being upright is unbearable, and you wake at seven with it gone. The recorded duration is ten hours. Not two, and not “I slept it off”, which is what most people write if they write anything.

Sleeping through an attack feels like the attack ending, which is why the rule exists. It is the single commonest way a real duration gets recorded as a short one, and it matters most for the people whose attacks reliably arrive in the evening, because nearly all of their attacks are being logged at a fraction of their length.

Past three days it has its own name, and an afternoon off does not reset it

An attack that runs unremitting beyond 72 hours and is debilitating with it is classified separately, as status migrainosus.

The useful detail sits in that entry’s notes: remissions of up to 12 hours, due to medication or sleep, are accepted. So the attack that lifted for an afternoon after you took something and then closed back over you is not necessarily two short attacks. Under twelve hours of relief, it is one long one, and recording it as two is how a three-day event gets filed as a pair of ordinary Tuesdays.

Pain that returns after a longer gap is a different pattern with a different name, and recurrence is worth recording as its own thing: same attack, second act.

If you are actually in the middle of one of these, past 72 hours and not shifting, that is a call to your GP or specialist rather than a diary question.

The number you would give out loud is a reconstruction

Asked “how long do yours last?”, you answer from memory, and the two ends of the memory are not equally reliable.

The start is easy. Something happened, you noticed it, there was a time on a clock. The end is the hard one, because nothing announces it. Pain fades rather than stopping, and what follows it is the postdrome, a phase of the attack in which fog and slowed thinking are the main symptoms. You are asked to timestamp the end of the attack during the period when your ability to timestamp anything is at its worst.

That is not a character flaw and no amount of trying harder fixes it. It is a measurement problem, and measurement problems are solved by writing things down closer to when they happen.

What to record, on paper or otherwise

Four marks per attack, and none of them require you to be functional:

• The time it started. Written when you notice, not later.
• The time you were free of it and stayed free. Check once more an hour on, because pain that fades and returns inside the hour has not ended.
• Whether you treated it, and when. This is what tells a reader that a two-hour attack was a two-hour treated attack.
• Whether you slept, and when you woke. So the sleep rule can be applied rather than guessed at.

On paper, a wall calendar with a start time and an end time in the day’s box is enough, and the arithmetic can wait until the end of the month. Migraine Journal stamps the time on the entry as you log it, which removes the half of this that depends on remembering what time it was when things were bad.

Either way, the sentence you are building towards is one a clinician can act on: “Six attacks last month, typically eight to fourteen hours, two of them cut short by medication inside two hours, one that ran three days.” That is a description of a pattern. “They last quite a long time, usually” is a description of a feeling, and it is the answer the room already expects.