If you wake up with a migraine, your log already started late.

3:40 a.m., and the clock is already wrong

You wake at 3:40 a.m. and the pain is already there. It's settled behind one eye, well past the point where a dark room and a cold pack would have done anything. Nausea's close behind. You find the phone, open your tracker, tap the start field. It offers the current time.

So you accept it. Most people would.

That timestamp is a guess wearing the costume of a measurement. The attack began earlier, sometime while you were asleep, and the phase that comes before the pain (the yawning, the stiff neck, the irritability you'd catch on a weekday afternoon) played out with nobody awake to see it. Your log opens partway through the attack, and every chart built on top of it inherits the gap.

The wake-up anchor

We call it the wake-up anchor. Migraine tracking, as a category, pins an attack's start to the moment you were conscious enough to log it. For a daytime attack that's close enough: you notice the aura or the first throb and log within minutes.

For an attack that starts in sleep, the anchor lands late by an amount nobody can measure. Worse, the error always points the same way. Every sleep-onset attack gets logged after its true start, never before, so the bias piles up across months instead of washing out. Averaging won't surface it.

Three things disappear when the anchor drops at wakeup. The prodrome. Any aura that happened overnight. And the trigger window, which for a 3 a.m. attack sits mostly in the evening before.

A log that starts at wakeup opens partway through the attack.

Why "log it the moment you notice" breaks overnight

The standard advice for migraine diaries is sound: log in real time, because memory fades fast and a recalled attack comes back distorted. We agree with that advice for every attack you're awake for.

But the advice quietly assumes you're present at the onset. Asleep, you aren't. "Log the moment you notice" becomes "log the moment you wake up," and those are two different moments.

The logging screen tends to compound it. A start field that defaults to now invites you to keep it, and moving it earlier feels like making data up, so people leave it alone. A trigger prompt that looks back over the last few hours is asking about hours you spent unconscious. The tool does exactly what it was designed to do. It's just pointed at the wrong slice of the night.

Why do I wake up with a migraine?

The clinical picture backs treating these attacks as a pattern of their own. Studies of attack timing find that onset clusters in the early morning hours. A systematic review of migraine chronobiology found most studies reported morning hours (6 am–12 pm) as the peak time of onset for migraine attacks. ([source](https://pmc.ncbi.nlm.nih.gov/articles/PMC8287677/)). Sleep shows up as a trigger in both directions. Too little, too much and an irregular schedule all appear on standard trigger lists. The American Migraine Foundation's sleep resource is titled "Guide to Healthy Sleep" at https://americanmigrainefoundation.org/wp-content/uploads/2024/09/2203_AMF_Guide_Healthy_Sleep_V4_Digital.pdf ([source](https://americanmigrainefoundation.org/wp-content/uploads/2024/09/2203_AMF_Guide_Healthy_Sleep_V4_Digital.pdf)).

The overnight mechanics are less settled. Older sleep-lab work linked attacks that wake people to particular sleep stages, including REM. Dexter JD and Weitzman ED published "The relationship of nocturnal headaches to sleep stage patterns" in Neurology, 1970 May;20(5):513-8. ([source](https://pubmed.ncbi.nlm.nih.gov/4314825/)). The hypothalamus also helps regulate the sleep-wake cycle, and it is a region researchers keep returning to in migraine. The research hasn't cleanly separated those threads yet.

For your log, the point is simpler. ICHD-3 describes premonitory symptoms that can start hours to a day or two before the headache. ICHD-3 defines premonitory symptoms as symptoms that may start up to 48 hours before the headache. ([source](http://e-hpr.org/upload/pdf/hpr-2024-0031.pdf)). When that window overlaps sleep, the log can't see it.

Morning headaches have other causes too. Sleep apnea and medication overuse are both on the list a clinician will want to rule out.

One hypothetical month, one wrong conclusion

Picture someone whose attacks mostly arrive overnight. This is a constructed example, but the shape will be familiar to a lot of readers.

Over a month they log four attacks, with start times of 4:15, 5:50, 6:05 and 6:20 a.m. The tracker's pattern view does what it was built to do. It reports morning attacks, scans its lookback window and finds the waking routine: coffee, the commute, the first screen of the day. Those are the only entries anywhere near the logged starts.

The real candidates sit outside the window. A late dinner. A bedtime that slipped past 1 a.m. twice. A weekend lie-in. None of it lands close enough to the logged start to register, so none of it surfaces.

Durations come out wrong as well. An attack logged from 6:05 a.m. to 3 p.m. reads as nine hours. If it actually began around 3 a.m., it ran closer to twelve, and the shorter number is the one a neurologist sees.

What a late start time costs you

If the wake-up anchor is real, three things follow.

First, trigger analysis drifts toward morning activities, which can send you after the wrong suspects for months. Second, attack durations come out short, and duration matters clinically. ICHD-3 defines a migraine attack as lasting 4 to 72 hours untreated ICHD-3 criterion C for migraine without aura (1.1) requires headache attacks lasting 4-72 hours (untreated or unsuccessfully treated), in patients with at least five attacks. ([source](https://ichd-3.org/1-migraine/1-1-migraine-without-aura/)), so a log that trims the front off every overnight attack understates what you're living with.

Third, treatment timing gets harder to discuss. When you wake into established pain, the chance to treat at the very first sign is already gone, and that is worth saying out loud to your doctor. A log that clearly flags "woke with it" gives your neurologist a reason to ask about sleep, about prevention, or about formulations that don't depend on keeping a pill down through nausea.

The fix starts with recording the uncertainty honestly.

How to log an attack you slept through

Stop treating the start time as a single number. For a sleep-onset attack, a useful record looks more like this:

1. Last known fine: when you fell asleep without symptoms. 2. Woke with it: a note that separates these attacks from the ones you watched begin. 3. Estimated onset: a range, like "between 1 a.m. and 3:40 a.m." 4. Sleep context: bedtime, wake time, how many times you surfaced, how rested you felt. 5. The evening before: food, alcohol, screens, stress, missed doses. Filling it in the next day is fine.

Whatever you use to keep this record, make sure it lets you describe an attack that began before you noticed it. A single timestamp can't do that, and a range can. Keep tracking through the after-phase too, long after the pain stops. Postdrome is built around that fuller picture, and your data stays on your phone, so you can export it when it's time to show your neurologist.

What we're tracking next

We're watching whether consumer sleep data can narrow the onset range. Wrist-based sleep staging is an estimate, so we'd treat it as a hint and keep the range as the record. We're also watching for research that studies nocturnal onset as its own pattern, since acute-treatment advice is usually written for attacks you're awake to catch.

Until then, log a range. "Between 1 and 3:40" tells your neurologist more truth than a confident 3:40 ever did.