Migraine and the Nervous System

People with migraine forecast their own attacks, and diaries prove it

A logged warning symptom was followed by a headache 72 percent of the time in the one study that locked its diaries against hindsight. The yawning, the stiff neck and the craving are physiology, recorded hours before the pain.

Ask someone who has had migraine for twenty years how they know one is coming. Many will answer without hesitating. They yawned all afternoon. They could not find an ordinary word. The neck went stiff, they wanted something sweet, and by evening they were certain. The headache arrived the next morning.

Doctors heard that story for a century with no way to test it. A team led by the neurologist Nicola Giffin finally tested it, and the design was the clever part. One hundred and twenty people with migraine carried handheld electronic diaries that prompted them at random times through the day and would not let them edit an entry afterwards. That removes the memory problem, because nobody can tidy the record once the headache has arrived. Ninety-seven produced usable data across three months.

In that electronic diary study of warning symptoms, a logged warning was followed by a headache 72 percent of the time. The commonest warnings were tiredness, in 72 percent of forewarned attacks, trouble concentrating in 51 percent, and a stiff neck in 50 percent. One limit is built into the design and the authors state it. Only people who already believed they could forecast their attacks were recruited. Seventy-two percent is the accuracy of self-selected forecasters and not a general figure.

What survives that limit is narrower and more interesting. These patients are reading something real off their own physiology, hours before anything hurts.

The forecasting is not superstition. Something is running for hours before the pain, and it can be recorded.

Four phases, and the recovery stage that doubles them

Migraine is conventionally described in four phases. First the premonitory phase, hours to days of warning. Then aura, a slow spreading disturbance of vision or sensation that some people get and others never do. Then the headache.

Then a fourth phase that a London neurologist named after interviewing 50 patients about how their attacks ended rather than how they began. Joseph Blau found that 47 of the 50 still felt unwell once the headache had gone, and he called that stage the recovery phase. It could double the length of the whole episode.

How a headache is made, how pain fibers wrap the vessels and membranes of the head, and what the spreading electrical wave of aura is, all belong to the headaches page. Migraine itself is the cycle around that machinery: the state between attacks, and the strange truth about triggers. First, the small number of headaches that are not migraine at all.

The headaches that need a doctor today

Thunderclap onset, fever with a stiff neck, and any sudden neurological change mean emergency care now rather than an appointment. The SNNOOP10 red-flag list carries the rest of the warning signs, and it comes before anything else about migraine.

Most headaches are not dangerous. A small number are, and they come first.

An international group of headache specialists reviewed the warning signs that a headache is caused by something other than a primary headache disorder. They assembled them into a checklist, published as the SNNOOP10 list of red and orange flags. Get medical attention for any of the following.

The flags that mean emergency care

Sudden and severe. A headache that reaches full force within about a minute, often called thunderclap, needs emergency assessment the same day.

Fever, or a stiff neck with systemic illness. Systemic symptoms alongside a headache are on the list. Fever with a stiff neck can mean infection of the membranes covering the brain, and that belongs in emergency care now rather than in an appointment.

Any neurological change. Weakness, numbness, trouble speaking, disturbed consciousness, or swelling of the optic disc found on eye examination. Sudden weakness down one side, or sudden loss of speech, is treated as a stroke until proven otherwise. That is emergency care in the same hour, because the treatments for it run on a clock.

A first headache after about 50. A headache pattern that begins for the first time after roughly the age of 50 should be examined rather than watched. The formal SNNOOP10 flag is set higher, at a first onset after 65. Everyday practice starts looking earlier than the list does.

A headache that has changed character. A pattern you have had for years that now behaves differently is a reason to be checked, not a reason to wait.

Onset after head trauma. Also headache that changes with posture, and headache brought on by coughing or exertion.

The rest of the list covers a history of cancer, pregnancy and the period after childbirth, and a painful eye with signs from the automatic nerves such as a drooping lid or a watering eye. It also covers immune system disease such as HIV, a progressive or atypical headache, and painkiller overuse or a new drug at onset.

What the list is, and what it is not

The authors are careful about what this is. Prospective studies of how well each flag performs are lacking, and questions about their sensitivity and predictive value remain unanswered. It is a screening aid rather than a validated instrument. Anything on it is a reason to be seen by a physician. Thunderclap onset, fever with a stiff neck, and any sudden neurological change are reasons to go to emergency care instead of booking one.

The wider discipline holds throughout. Structural, vascular, infectious and malignant causes of head pain are real, and they are treated as such. Nothing here is a reason to skip the examination that rules them out. What follows is about migraine as a primary disorder, meaning the cases where that examination comes back clean and the person still loses two days a month.

The migraine warning phase runs through the hypothalamus

Eight patients scanned during triggered warning phases showed the posterolateral hypothalamus, the midbrain and the dorsal pons active while still pain free. Yawning, thirst and food craving are the brain's housekeeping hub announcing the attack hours early.

Studying the beginning of a migraine ran into an obvious problem. Scanners are booked weeks ahead and attacks arrive when they like, so almost every image of migraine was taken during the pain.

A London headache group found a way around it. They gave an intravenous infusion of glyceryl trinitrate, the nitrate compound used for angina, to 44 people with migraine and 12 people without it. Thirty-three of the 44 went on to develop a genuine attack meeting the international criteria. Twelve patients described the classic warning symptoms first, which had never been documented with this trigger before. Repeat the infusion and almost every patient reproduced the same attack.

It was an open provocation with no placebo arm. A placebo arm is the group given an inactive dummy in place of the drug, which is how a study finds out how much of the result came from the procedure rather than the compound. The 12 non-migraine controls were there to rule out an ordering effect, and they establish only that the drug does not produce a migraine attack in everyone.

Migraine had become something a scanner could be booked for.

The scan that caught the warning phase

The design mattered as much as the result, and the result rests on eight patients. Farooq Husain Maniyar, a headache neurologist, took the trigger into a positron emission tomography scanner. That machine tracks blood flow as a stand-in for how hard a brain region is working. Eight patients were scanned three times each: at rest, during the warning phase while still free of pain, and once the headache had arrived.

Comparing the pain-free warning scans against baseline showed activation in the posterolateral hypothalamus, in the midbrain and the dorsal pons, and in parts of the occipital, temporal and prefrontal cortex. The midbrain and the pons are two levels of the brainstem, the stalk at the base of the brain that every signal from the face has to pass through. One of those brainstem sites is the periaqueductal gray, which is the brain's own volume control for pain.

The hypothalamus is roughly the size of an almond and sits low in the middle of the brain. It runs the housekeeping. Hunger, thirst, temperature, the switch between sleep and waking, the hormonal cascade, and yawning. Now read the warning symptoms again with that in front of you. Yawning. Thirst. Food craving. Heaviness. The warning phase looks like housekeeping gone strange because that is where the activity is.

Eight patients, and the attacks were brought on with a drug rather than arriving on their own. The study shows what the brain does in a triggered warning phase. It does not show that the hypothalamus starts a spontaneous attack.

The neurologists Peter Goadsby and Nazia Karsan, working in a headache group in London, gathered the warning-phase evidence into one argument about what those symptoms reveal. Symptoms can begin hours to days before any head pain. They include lethargy, yawning, sensitivity to light and sound, thirst and food craving, and in some people they forecast the attack reliably.

It is a narrative review, and the authors are principal investigators on much of the work they review. Their claim is that this earliest phase is the most informative window into migraine biology, because it shows the disorder starting somewhere other than the pain system.

Most named migraine triggers are the attack introducing itself

Chocolate provoked no more headaches than carob under double-blind testing, and patients' named triggers kept matching their own warning symptoms. The craving did not cause the attack. The attack asked for the chocolate.

Chocolate is the most famous migraine trigger in the world. It appears on printed avoidance lists handed out in clinics.

Chocolate against carob, double blind

Dawn Marcus, a pain physician in Pittsburgh, put it on trial. The comparison was carob, which looks and behaves like chocolate and lacks the compounds under suspicion. Sixty-three women with chronic headache first spent two weeks on a diet stripped of those compounds. Then they received two chocolate samples and two carob samples in random order, double blind.

Neither the women nor the person handing out the samples knew which was which, so expectation could not decide the answer. Chocolate was no more likely to provoke a headache than carob in any diagnostic group. The result held among the women who were certain chocolate was their trigger.

Sixty-three women, only half of whom had migraine, with headache recorded by diary. A negative provocation study cannot exclude a trigger effect in a small subgroup. This one also used a fixed dose on a fixed day rather than testing when the system was already primed.

Light and exercise did little better. A Copenhagen headache group led by the neurologist Anders Hougaard took 27 people with migraine with aura who named bright or flickering light, or strenuous exercise, as reliable triggers. Then they gave them exactly that in the laboratory. Three of the 27 produced an aura attack.

Three more produced migraine without aura. Flickering light on its own produced nothing at all, while four of the twelve exercise patients did get a migraine. Each patient was exposed once, which may miss a trigger that only works when the system is already close.

The trigger list and the warning list are one list

Then the study that turns the question inside out. The neurologist Nazia Karsan and colleagues interviewed 53 people in detail about what they believed set their attacks off and what warning symptoms they got. Then attacks were brought on with a nitrate infusion so the beginning could be watched.

Where the two lists could be compared, they kept landing on the same item. People who named light as a trigger had premonitory light sensitivity. People who named noise had premonitory noise sensitivity. People who named food or a skipped meal had premonitory food craving. Stress lined up with premonitory mood change.

The craving was the attack introducing itself. The chocolate arrived because the attack had already started, and then took the blame for it.

Why did nobody notice for so long? Because a trigger and a warning symptom are observed in the same window, and from the inside the person cannot tell them apart. A methods review led by the neurologist Richard Lipton works through the consequences.

A trigger is an exposure that raises the chance of an attack. A warning feature is a symptom that announces one. Retrospective questionnaires cannot separate them, and the review calls establishing that an antecedent event caused a headache a formidable challenge. Prospective electronic diaries are the method of choice.

The agreement between the two lists was significant for light, sound, food and skipped meals, and it was not a blanket finding for every named trigger. Some triggers are genuine. What the evidence will not carry is the confident printed food list, and an enormous amount of daily avoidance has been built on it.

A system reading its own early output as an external cause produces exactly this. The tone had already begun to shift, and the only visible part of that shift was a wish for chocolate. What the person then avoided was their own physiology. The name for a nervous system acting on its own forecast is prediction, and in migraine it is one of the signature aspects of tone.

Daily scanning shows the migraine cycle turning before the pain

One volunteer scanned on 30 consecutive mornings, then nine patients scanned daily for a month, showed hypothalamic responses climbing within the last 48 hours before each attack. Migraine is visible before it hurts.

Laura Schulte and Arne May, neurologists in Hamburg, solved the timing problem by brute force. They scanned a single volunteer every single morning for 30 days, covering three complete untreated attacks. On each visit the same standard painful stimulus was delivered to the face and the brain response recorded.

Two things moved across the month. The hypothalamic response to that stimulus climbed as each attack approached. And in the last 24 hours before pain, the hypothalamus changed how it was coupled to the dorsal pons and to the first brainstem relay for the face, the spinal trigeminal nuclei.

Coupling deserves a plain definition, because the finding turns on it. Two regions are coupled when their activity rises and falls together, so that knowing one tells you something about the other. The strength of that relationship can change while neither region changes its own volume. Coupling is the conversation, not the voices.

The paper is titled the migraine generator revisited, and the argument in it is a correction of the field by one of the people who built the field. The celebrated brainstem generator may be downstream. What drives the cycle may be the shifting conversation between hypothalamus and brainstem.

The argument about which site is the generator is the wrong argument. Regulation here is distributed across several sites, and no one of them runs the others. What moves in the days before an attack is the relationship between them, which is why the coupling is the finding and not the location.

The group study dates the warning phase to 48 hours

It is one patient. That has to be said before anything is built on it, and the same group said it by doing the group study. Nine patients were scanned daily for at least a month, seven produced analyzable data, and 27 spontaneous attacks were captured. Hypothalamic activation to pain appeared within the last 48 hours before headache onset and not before. The warning phase now has a measured duration rather than an estimated one, defined by one marker in one paradigm.

The brainstem relay climbs toward each attack

A second instrument watches the same cycle with a different signal. The systems neuroscientist Anne Stankewitz, working with Arne May, delivered a standard painful stimulus to the face inside an MRI scanner and watched the first relay station in the brainstem answer. That relay is where sensation from the face and the head arrives before anything else happens to it.

It carries the trigeminal name because the nerve feeding it splits three ways across the face. Its response was not fixed. Between attacks it sat lower than in healthy people. It then climbed steadily as the next attack approached, matching healthy levels shortly before one arrived. The relationship was tight enough that the distance to the next headache could be read off the scan. During the attack itself it dropped back down.

Different patients were scanned at different points in the cycle, instead of the same patients being followed through one. Reading a distance off a scan is a pattern across a sample, and not a forecast made for one person.

Is the premonitory phase universal? The dissent

The dissent is serious. The neurologists Cedric Gollion, David Dodick and Hakan Ashina, with a colleague, wrote up a formal debate held in Copenhagen. They accept that hypothalamic transmitters modulate trigeminal pain signaling, orexin and dopamine among them. Orexin is the chemical that holds a brain awake. Dopamine sets what a system will bother to pursue.

They accept the imaging too. Then they argue the field moved too fast, because the studies are small, inconsistently reported, and often do not use the official definition of a premonitory symptom. They go further and question whether a distinct premonitory phase exists as a universal feature at all.

The reading here does not need the hypothalamus to be the generator. It needs the observation that both sides accept, which is that something changes in the days before the pain. Even the word generator is disputed by the people who use it, and that is a healthy sign in a field.

The migraine brain braces for the attack a day early

The contingent negative variation is a scalp voltage of anticipation. Between migraine attacks it runs larger, refuses to settle with repetition, and peaks the day before the pain arrives.

A nervous system does not wait for the world and then answer it. It runs a forecast of the next few seconds and acts on the forecast, correcting itself when the world disagrees. That is what prediction means as a property of a nervous system, and the prediction page follows it through the whole nervous system.

The cost of running a forecast is that the forecast can be wrong. A system can start preparing for something that is not coming, and it can start preparing before there is anything to prepare against.

In migraine, the thing being prepared for is the attack. That preparation can be recorded, and it was recorded first with a cheap instrument.

There is a slow electrical wave the brain produces while it is waiting for a signal it expects, recordable from scalp electrodes, called the contingent negative variation. Give someone a warning tone followed a few seconds later by a target they must respond to, and the voltage drifts negative in the gap. That drift is anticipation, made visible. In a healthy brain it shrinks as the task repeats, because a signal that keeps arriving on schedule stops being worth bracing for.

The wave that would not settle between attacks

Peter Kropp and Wolf-Dieter Gerber, two researchers in medical psychology in Kiel, recorded it in 16 people with migraine without aura and 22 healthy people, across 32 trials each. Between attacks the migraine wave was significantly larger than normal and did not shrink with repetition. During an attack it fell back to the healthy pattern and habituated normally. The brain looked most abnormal precisely when the person felt fine.

Three years later the same pair followed the wave across the cycle. Its amplitude, and particularly its early component, rose toward maximum negativity in the day before an attack. In most cases the attack arrived when the wave reached its peak. For two to three days afterwards it sat at ordinary values.

Anticipation has a waveform. In migraine it peaks the day before the pain, and settles once the pain has been and gone.

Small samples, no blinding, and no formal forecasting statistics, so this is a demonstrated pattern rather than a device anyone could buy. What the pattern establishes is the shape of the thing. A brain that is braced has already committed resources to something, and here the commitment is heaviest on the day when nothing has happened yet. A person in that state is being asked to live inside a forecast their own physiology is making.

A migraine brain keeps the volume up on well days

Shown the same flashing checkerboard in repeated blocks, healthy visual responses shrank and migraine responses grew, in 36 patients tested between attacks against 16 controls. The abnormality sat in brains that felt fine.

Every patient in the studies that follow was well on the day of testing. Hold on to that while the numbers arrive, because it is the whole of the argument.

Show a healthy brain the same flashing pattern over and over and the electrical answer recorded from the back of the head gets smaller. That shrinking is habituation, a nervous system declaring a repeated signal unimportant, and the gain page treats it as a foundation of tone in any nervous system.

The neurologist Jean Schoenen, in Liege, asked whether migraine brains do it. A flipping checkerboard went to 16 healthy volunteers and 36 migraine patients between attacks, in five blocks of 50 flashes. The healthy response shrank across the run and was deepest by the third and fourth blocks. The migraine response grew instead, peaking in the second to fourth block. Attention could not explain the difference, and aura made none.

Fifteen minutes without a break, same answer

Two minutes is a short recording, and that was the obvious objection. So the group ran it for fifteen minutes without a break, in 25 healthy controls, 25 patients with migraine without aura and 15 with aura, all tested between attacks.

The neurologist Judit Afra, in the same Liege department, led that one. Over the quarter hour the controls declined steadily, as expected. Both patient groups stayed flat and never came down. The size of the deficit had no relation to how often the patients had attacks or how long they had been ill.

Remember what was being recorded and when. Nobody in either patient group had a headache. The abnormality sat in what a nervous system did with a signal it had already seen forty times, on a day when nothing hurt. That is where tone is read, in the answer to a repeated demand and not in any single value taken at rest. A system that will not turn its own answer down is a system running with the volume up.

Blinded recordings challenge the migraine habituation deficit

Once the analyst was blinded, 27 migraine patients and 34 controls in Trondheim habituated identically, and no blinded-recording study has found the missing habituation since. The strongest challenge to the between-attack claim is methodological, and it is serious.

The habituation deficit is the single most quoted neurophysiological fact about migraine, and it is directly disputed.

Blinding the recording erased the difference

A group of clinical neurophysiologists at a university in Trondheim, led by Petter Omland, set out to work out which stimulus settings explained the disagreement in the literature. They recorded 27 migraine patients between attacks and 34 healthy controls, using small and large checkerboard squares in random order. Critically, the analyst was blinded to which block was which.

Blinding means the person reading the trace does not know what they are reading, so expectation cannot lean on the measurement. Both groups habituated. None of the habituation measures differed between patients and controls, and check size made no difference either. They recommended that all future habituation studies be blinded.

They tried again with a fifth check size, in 41 patients and 30 controls, and again found nothing. Their conclusion is the sharpest sentence in the whole dispute. No study that blinded the recording has ever found the missing habituation, so it cannot be treated as a reliable hallmark of migraine.

Blinding the analysis kept the deficit

The blinded record is not uniform, and the distinction it turns on is worth having. Omland's charge is about blinding while the recording is being made. An Italian group at a headache unit in Pozzilli, working with Jean Schoenen in Liege, tested the other half of it and blinded the analysis.

Recordings from 25 healthy volunteers and 78 patients with episodic migraine, 52 between attacks and 26 during one, were read twice. One investigator knew the diagnosis and the phase and the other did not. The two readings agreed with each other. Habituation was normal in the healthy volunteers and normal during attacks, and deficient between them. Schoenen was senior author, so the group with the most to lose ran the test.

The cycle confounds the between-attack average

There is also a complication, and it came from the same group. They recorded visual and auditory responses at four points relative to an attack: the day before, during, one day after, and two days after. The comparison group was recorded at least three days from any attack, 66 patients for the auditory measure and 39 for the visual one.

On the day before the attack, habituation had returned to normal. During the attack it was normal too. Two days after, it swung back past normal into the migraine pattern. Different patients contributed to different time points, and the group recorded before an attack was only 11 people.

Follow the consequence. Classing a patient as between attacks by history alone pools someone two days past an attack with someone one day short of the next, and those two states point in opposite directions. Averaging them returns something close to normal. A blinded protocol fixes the observer and does nothing whatever about that.

Blinding corrects the person reading the trace. It cannot correct a measurement whose true value depends on which day of the cycle the subject walked in on.

So the Trondheim group went and controlled it. Forty-nine migraine patients were recorded four times each, with every session sorted from a diary kept as it happened into between attacks, just before one, during one, or just after. The analysts were blind to the diagnosis and to the phase.

The stimulus was a laser pulse to the back of the hand rather than a flashing checkerboard, so this is the pain measure and not the visual one. Habituation was present and similar in the 29 patients recorded between attacks and in 30 controls, and it barely moved across the phases. The cycle-timing argument is still open on the visual measure. On the pain measure it has been tried and it failed.

The same test, the same protocol, two different underlying organizations, two different answers. That is input meeting tone inside a laboratory instead of a clinic. The work from Trondheim is careful work on a moving target, and its recommendation to blind is correct. The abnormality between attacks is well supported on several measures. On this one, the blinded studies disagree with each other.

Structure comes out mostly normal in 296 scans

The same caution belongs on the structural side. Small studies had reported all kinds of cortical differences in migraine. The neurologist Rune Christensen and colleagues at a headache center in Copenhagen scanned 296 people with migraine and 155 matched controls on the same machine and compared the whole cortex.

Almost everything came out the same. Reduced surface area in the left insula survived, and the insula is the patch of cortex folded away inside the side of the brain that reads the body's own internal state.

The chronic subgroup also showed increased surface area in the right caudal anterior cingulate, a strip along the midline that weighs how much something matters. Aura made no difference and having a headache during the scan made no difference. How little survived a well-controlled comparison is a warning against reading small structural studies confidently, including the ones that would have suited this argument.

Migraine behaves as a sensory threshold disease

Fifty-two people with migraine, headache free on the day of testing, reached discomfort from rising light and sound at significantly lower levels than 48 controls. The excitability question stayed open, and the threshold question did not.

A magnetic pulse held over the back of the head can make a person see a flash of light that is not there, called a phosphene. Nothing enters the eye. The pulse fires the visual cortex directly, and the brain reports light because that is what this tissue is for. How strong the pulse has to be is a rough reading of how excitable that cortex is.

The headache researcher Sheena Aurora and colleagues in Detroit tested 11 patients who had migraine with aura against 11 matched controls. The coil sat seven centimeters above the bump at the base of the skull. Every single patient produced phosphenes. Only three of the 11 controls did. Patients needed 44 percent of maximum stimulator output on average against 69 percent for controls, and the patient with the lowest threshold went on to have an aura.

Afra and Schoenen found the aura brain under-excitable between attacks, opposite to Aurora

The same year, using the same technique, a larger study reached the opposite verdict. Judit Afra and Jean Schoenen stimulated 27 controls, 33 patients with migraine without aura and 25 with aura, over both the motor cortex and the visual cortex, between attacks.

Patients with aura needed more stimulation to fire a hand muscle, not less. They produced phosphenes less often than controls. Resting thresholds, response sizes and every other measure they took came out the same across the groups. The authors concluded that the aura brain is under-excitable between attacks.

Neither study was blinded and phosphene threshold is highly technique-dependent. The question is genuinely open, and a single number labeled excitability was probably never going to survive it.

A threshold that moves with the cycle

Kuan-Po Peng and Arne May, in the same Hamburg department, proposed a frame that does more work. Pulling together studies of light, sound, touch, heat and pain thresholds, they describe migraine as a sensory threshold disease. People with migraine sit lower than healthy people in general, and the threshold moves on a schedule.

It reaches its lowest point as the headache begins, rises after the headache ends, then drifts back down toward the next attack. It is a review, and the authors are explicit that studies disagree because designs and phase definitions differ.

One measurement makes that concrete for a reader. Fifty-two people with migraine, all headache free on the day of testing, and 48 matched healthy people were exposed to steadily rising light and steadily rising sound until they called it uncomfortable. The migraine group reached discomfort at significantly lower levels for both. Discomfort is a subjective endpoint and headache-free was established by report on the day. Even so, the finding is the one patients describe constantly. The volume was still up when the pain was gone.

A threshold is not a symptom. It is what the system does when a rising signal is put to it, measured on a day the person feels well. That answer is a direct reading of tone. Tone is the organization behind the answer, and never any single number the answer happens to produce.

Light reaches the migraine pain system by an image-free route

Photophobia in migraine runs through retinal cells that build no picture, and green light drives the circuit least at every level tested: retina, thalamus and cortex.

Photophobia is usually explained as the eye being sensitive. The route runs somewhere else. Light reaches the pain system through retinal cells that build no picture at all. They report how much light is present and they run the body clock. That is why blind people who have lost image-forming vision, yet still have those cells, report that light makes their migraine worse, and why blind people who have lost the eye itself do not. The senses page traces that route and that observation.

The circuit was traced in the rat. Single-cell recordings found neurons in the posterior thalamus that answered stimulation of the covering of the brain, the membrane where the head's pain fibers run, and light changed how those neurons fired.

They also received direct input from retinal ganglion cells, mostly the light-reporting kind, and sent their own axons out across sensory, visual and association cortex. The thalamus is the switchboard almost every sense passes through on its way to the cortex, which is why a circuit that reaches it can color everything downstream.

The human observation is a survey of a small and unusual group, and the circuit tracing is rodent work. The bridge between them is an inference, though a well constrained one.

Green, tested at three levels of one circuit

If light reaches the pain system by a specific retinal route, then some colors should hurt more than others, and that is testable in an afternoon. Patients with normal sight rated their headache under white, blue, amber, red and green light. Green made it worse by far the least.

Then the neuroscientists Rodrigo Noseda and Rami Burstein, in Boston, put the same question to the tissue. Recordings were taken from the retina and the scalp in patients, and from light-sensitive and membrane-sensitive thalamic cells in rats. Every level lined up with the ratings. Green drove the cone pathways less than white, blue or red. The thalamic cells answered blue most and green least. Cortical responses to green were the smallest of all. Three levels of one circuit agreed about a single color.

None of that is a treatment claim. What it explains is a preference many patients report and have been told was imagination. The route is the same in everyone. Light is delivered into a pain circuit in every human head. What decides whether it hurts is where the threshold on that circuit is already sitting, on a day with no headache in it at all.

CGRP, from a jugular vein in 1990 to a pill you can swallow

Calcitonin gene-related peptide was the one messenger raised in blood leaving the head during migraine attacks. Infusing it produced attacks, and blocking its receptor became a drug class. Twenty-seven years separate the jugular vein from the tablet.

The headaches page pins the pain of an attack to a chemical messenger. The story of how that messenger was run down is the best told in headache medicine.

In 1990 the reasoning was simple. A firing trigeminal nerve should be dumping its chemical messengers into the blood leaving the head. Peter Goadsby, then a neurologist at a hospital in Sydney, drew blood from the external jugular vein on the painful side and from an arm vein at the same moment, during real attacks.

The laboratory of the vascular physiologist Lars Edvinsson, in Lund in Sweden, measured the peptides, which is why four of them could be assayed at all. Only one was raised, calcitonin gene-related peptide, and only in the blood draining the head rather than the arm. The rise was larger in migraine with aura.

A neuropeptide is a chemical a nerve releases to change how the tissue around it behaves. This one widens blood vessels and raises the sensitivity of the nerve endings near them. The 1990 study was small, used the antibody-based assay of the day, had no placebo condition, and established an association with the attack rather than a cause of it.

Infusing the peptide produced the attack

Twelve years later a Copenhagen group settled the direction by infusing the peptide itself. Twelve people with migraine without aura received either human alpha-CGRP or placebo intravenously, double blind and crossed over. Every patient developed headache after the peptide against one after placebo, and in three of them the delayed headache met full migraine criteria. Three were excluded, so the analysis rests on nine, and three of nine is a modest hit rate. The senior author was the neurologist Jes Olesen.

Blocking the receptor worked, and became a class

Then block the receptor and see whether attacks stop. The international team that ran the first blocker trial gave 126 patients placebo or one of six intravenous doses of a highly selective blocker, the first of the class now called gepants. At the dose selected, 66 percent responded against 27 percent on placebo.

Nausea, light sensitivity, sound sensitivity and function all improved. The effect showed inside thirty minutes, and no serious adverse events occurred. It was a phase 2 dose-finding study given by drip in a clinic. It is also the trial that made an entire drug class possible.

Twenty-seven years after the jugular blood, the first monoclonal antibody against that receptor reached a large trial. An antibody is the large protein an immune system builds to lock onto one target and nothing else. A monoclonal antibody is a single one of those proteins, manufactured in bulk and given as a drug.

The international group that tested this one gave 955 patients monthly injections or placebo for six months. Migraine days per month fell by 3.2 and 3.7 on the two doses against 1.8 on placebo, from a baseline of 8.3. Half the patients on the higher dose halved their migraine days, against 26.6 percent on placebo, and side-effect rates matched placebo.

Then a pill. In a trial of 873 adults, a once-daily oral gepant cut migraine days by 3.7 to 4.2 across three doses against 2.5 on placebo, from a baseline near 7.7. Both trials were industry funded and both report a gain over placebo of roughly one to two migraine days a month.

A messenger was identified in human blood, shown to produce the symptom when infused, then blocked, and the block worked in thousands of people. That is medicine doing exactly what it is supposed to do, and the relief it gives is worth having.

Raised between attacks, and active in the brain

Two facts complicate the tidy version. The messenger is raised between attacks, not only during them. A neurology group in northern Spain, led by Eva Cernuda-Morollon and Julio Pascual, drew ordinary arm blood from 103 women with chronic migraine, well away from any attack.

The average came out at 74.9 picograms per milliliter, against 33.7 in healthy women and 46.4 in episodic migraine. Age, painkiller overuse, depression and preventive medication did not move the numbers. Measuring this peptide in ordinary peripheral blood has produced conflicting results across laboratories, so it is not a clinical test.

A messenger sitting high on a day with no attack is not a fact about attacks. It is a fact about the organization the system is holding in between them, which has already moved before anything asks it a question.

The second complication is stranger. The antibody is a large molecule thought not to cross into the brain. A preregistered placebo-controlled imaging study scanned 40 patients before treatment and four weeks after, and found reduced activity in the thalamus, the putamen and the operculum in response to face pain. The putamen sits deep in the brain and helps shape movement and habit. The operculum is the lid of cortex folded down over the great fissure in the side of the brain.

The systems neuroscientist Hauke Basedau and colleagues, again in Hamburg, say plainly that they cannot yet tell whether that is a direct central action or the downstream consequence of a peripheral one. Only seven of 21 on the drug and four of 19 on placebo met the response threshold, which left the planned subgroup analysis underpowered. Because the study was preregistered, that shortfall is visible instead of buried.

Skin sensitivity decides when a migraine drug can work

A triptan taken before scalp tenderness developed left patients pain free in 25 of 27 migraine attacks. Taken after, it worked in 5 of 34. The state the dose lands on decides the response, not the hour on the clock.

Why does the same drug work brilliantly in one attack and do nothing in the next? Patients ask that constantly and are rarely given an answer. There is one, and it turns on something that happens to the skin partway through.

Partway into a bad attack, the scalp itself can become painful to ordinary touch. Pain from something that should not hurt is called allodynia, and the field's name for the change behind it is central sensitization. The headaches page teaches the machinery, and the gain page treats it as a foundation of tone. The neuroscientist Rami Burstein measured when it arrives.

His team tested the pressure and the temperature at which skin around the eye and on the forearm first became painful, both away from attacks and during them. In 79 percent of patients ordinary touch became painful during the attack. In 21 percent it never did, and those people matter as much as the 79.

Which side of the tenderness the dose lands on

Then the timing. The drugs in question are the triptans, the tablets taken at the start of an attack. The team studied 31 patients three times each and sorted 61 attacks by whether the skin had already become tender at the moment the drug was taken.

Where tenderness had set in, the drug left patients pain free in 5 of 34 attacks. Where it had not, the same drug worked in 25 of 27. Treating at one hour or at four hours made no difference. What mattered was which side of the tenderness the dose landed on.

That is a mechanism and nothing else. It describes an association between a sensory state and a drug response, in 31 patients, with open treatment and no placebo arm. It is not a dosing instruction. Decisions about what to take and when belong with the physician who knows the whole case.

Allodynia forecasts the course of the disorder

The forward edge is the part with consequences. A Dutch group in Leiden, led by the neurologist Mark Louter, followed 2331 well-characterized migraine patients for a median of two years. Seventy percent reported allodynia.

After correcting for early age at onset, high baseline headache load and depression, allodynia independently predicted an increase in migraine days over the follow-up. It was a web-based cohort with self-reported allodynia and 86 percent women, and predicting more headache days is not the same as predicting the formal transition to chronic migraine.

A sensitivity that appears inside an attack and forecasts more attacks later is a system moving along a time course rather than repeating an event. Each attack is an input that meets a tone and leaves the range a little narrower than it found it. That is what the word chronification points at, and it is why the interval between attacks is where the interesting measurements are.

That raises a frightening question, and the answer is kinder than most patients have been told. A population cohort was rescanned nine years after the first study that alarmed everybody. Among women, 77 percent of the migraine group showed progression of deep white-matter spots against 60 percent of controls. There was no relation to how many or how frequent the attacks had been, and no association with any decline in thinking or memory across those nine years. In men there was no association at all.

Menstrual migraine tracks the fall in estrogen, not the level

Peak and average estrogen were identical in 114 women with migraine and 223 controls. What differed was the slope of the drop after the peak: 40 percent in two days against 30. The migraine system fails to absorb the fall.

Menstrual migraine is the clearest natural experiment available in this disorder, because the input arrives on a schedule the body publishes in advance. Hormonal rhythm across the whole female lifespan is worked out on the women's health page. The narrow question of why attacks keep time with the cycle is migraine's own.

A physician working in Australia published the governing idea in 1972, in a paper on the role of estradiol withdrawal. The trigger is the withdrawal of estrogen rather than a high level or a low one. Three years later the same physician reported the exposure time it takes.

Giving long-acting or short-acting estradiol to six women showed that several days of high estrogen exposure were needed before withdrawal would produce a migraine at all. Supplementing estrogen premenstrually in four more women did not prevent the attacks.

The withdrawal hypothesis, tested with daily urine

Thirty-four years later a London group tested the hypothesis properly. Forty women were recruited, each with regular cycles, one to four attacks a month, and at least one attack reliably around the first days of bleeding. They used a fertility monitor and gave a urine sample every morning for three cycles while keeping a headache diary. Thirty-eight were analyzed.

Attacks clustered in the late luteal and early follicular phase, which is the second half of the cycle and the days just after it starts again, when estrogen is falling. Attacks were fewer than expected while estrogen was rising. Rising estrogen looked protective. The study took only women with regular cycles and none on hormonal contraception, so the result does not extend to women on hormonal treatment or through the menopause transition.

Same hormones, faster fall

The sharpest version came from a large women's health cohort with daily hormone measurements. The neurologist Jelena Pavlovic and colleagues compared 114 women with a migraine history against 223 controls. Peak levels were the same in both groups. Average daily levels were the same.

What differed was the rate of the fall. In the two days after the peak in the second half of the cycle, estrogen dropped 40 percent in the migraine group against 30 percent in controls. The faster decline was present whether or not a headache occurred in the cycle studied.

The migraine brain did not have different hormones. It had the same hormones falling faster, and it was the slope that marked it out.

Nothing here is an abnormal level, so nothing here would appear on a hormone panel. What differs is a rate of change the system has to absorb. The study is observational, and the faster fall looks more like a trait than like an attack trigger, in pre-menopausal and early perimenopausal women only.

A system whose range has narrowed absorbs a steep change less well than a system whose range is wide. A steep fall is an input, and whether it lands as an attack depends on the tone it arrives at. The disorder lives in the regulation and not in the value.

Sleep ends migraine attacks, and short nights failed as a trigger

Across 4406 diary days and 870 headaches, sleeping six and a half hours or less was not followed by more migraine the next day. Yet 85 percent of clinic patients choose sleep to end an attack.

Sleep occupies a strange double position in this illness. It is the thing patients most often blame for the attack, and it is the thing they reach for to end one. Both beliefs are old, and only one of them survives measurement.

The headache specialists Leslie Kelman and Jeanetta Rains put numbers on the anecdote. They interviewed 1283 migraine patients, drawn from 1480 consecutive headache patients at a specialist clinic, about their sleep. Half said disturbed sleep set off their attacks. Seventy-one percent had headaches that woke them.

Eighty-five percent said they chose to sleep because of a headache, and 75 percent said they were forced to sleep by one. Thirty-eight percent averaged six hours a night or less, and those short sleepers had more frequent and more severe headaches. It is cross-sectional interview data from a tertiary clinic, so it captures the severe end and cannot separate cause from effect.

Sleep is also the way out. Of the 50 patients interviewed about how their attacks resolve, 14 could cut one short by lying down during the day, averaging two and a half hours.

Sleep instructions as treatment, one small trial

The restoration side has one small trial worth knowing. Anne Calhoun and Sutapa Ford, working in neurology, randomized 43 women with transformed migraine to real behavioral sleep instructions or to placebo behavioral instructions, on top of their usual care. The real instructions produced significantly lower headache frequency and intensity at six weeks, and none of the control group had reverted to episodic migraine at that point.

After everyone received the real instructions, 48.5 percent had reverted. It is a 43-woman pilot at a single center with a six-week blinded phase, behavioral instructions cannot be fully blinded, and it has not been replicated at scale. The regulation sleep performs is worked out on the sleep page.

The trigger belief fails prospective measurement

Then the prospective test, and it is a null. Suzanne Bertisch, a sleep medicine researcher in Boston, and her colleagues had 98 adults with episodic migraine keep twice-daily electronic diaries and wear wrist actigraphs for six weeks. An actigraph is a wrist device that infers sleep from movement, so it records what the body did rather than what the person remembers.

The study produced 4406 days of data and 870 headaches. Sleeping six and a half hours or less was not followed by more migraine the next day. Poor self-rated sleep quality was not either. Only fragmentation showed anything at all: diary-reported low sleep efficiency raised the odds of headache the following day by 39 percent, while actigraphic fragmentation was associated with slightly lower odds on the same day.

The most universally believed sleep trigger in the disorder did not survive prospective measurement, and the two instruments pointed in opposite directions. The model expects exactly that shape from a trigger result. Averaging matched and mismatched nights across an unstratified sample returns a modest number that describes nobody in it.

Why the same input lands differently on different people is the subject of the recovery page. A short night is an input. What it produces depends on the tone it arrives at, and that is the one variable this literature has never stratified on.

The migraine attack that arrives as dizziness

Vestibular migraine affects just under one percent of the population and has had formal criteria since 2012. Its entire prevention literature is three trials with 209 participants, which is absence of evidence rather than evidence of absence.

For decades, people who came in with recurring vertigo and a history of migraine were sorted by whichever specialty saw them first. Some were given a balance diagnosis. Some were given a headache diagnosis. Balance itself is assembled from a running comparison between the inner ear, the eyes and the joints, and the vertigo page works that assembly out. The two fields eventually sat down together and wrote shared criteria for vestibular migraine, with the neurologist Thomas Lempert in Berlin as first author.

Diagnosis requires recurrent vestibular symptoms, a history of migraine, a timed association between the two, and the exclusion of other causes. Episodes must be moderate or severe and last between five minutes and 72 hours, which is a far wider window than a headache. It is a consensus definition and not a study, and it was placed in an appendix of the classification as a first step, which tells you how the field rated the evidence at the time.

An attack whose most disabling feature is dizziness rather than pain is exactly what a model organized around the state of a system, and not around the site of a lesion, expects to find. One disorder, several outputs, surfacing wherever a particular body has least capacity to absorb it.

A prevention literature of 209 participants

What the evidence amounts to for prevention is worth stating exactly. A Cochrane review team in Oxford searched the entire controlled-trial literature and found three studies with 209 participants in total, one on a beta blocker and two on a calcium channel blocker.

There were no trials at all for antiepileptics, antidepressants, diuretics, CGRP antibodies, botulinum toxin or hormonal approaches. All the evidence found was rated low or very low certainty, and the reviewers declined to draw conclusions from the numbers. The condition affects just under one percent of the population and many people go undiagnosed.

That is not a treatment that failed. It is a literature that barely exists, and absence of evidence is a different thing from evidence of absence. The distinction matters because the first is routinely reported to patients as the second.

Medication-overuse headache is real, disputed, and best read as learning

In 720 patients with chronic migraine and medication overuse, randomized across 34 clinics, keeping the overused medication proved not inferior to switching away from it. The category is real, and the confident version of it is weaker than the handouts imply.

Medication-overuse headache is defined as headache on 15 or more days a month for more than three months, in someone with an existing headache disorder, caused by overuse of the medication taken to relieve it.

A review by the neurologist Hans-Christoph Diener and colleagues in Essen puts it at around one percent of the general population and far commoner in people who already have headaches, especially chronic migraine. The resulting headache takes its character from both the original disorder and the drug being overused. Success rates for the standard approach run around 50 to 70 percent, with higher relapse where opioids are involved.

The dissent, and the trial that tested the rule

The dissent is substantial and comes from serious people. The epidemiologist Ann Scher and the neurologists Paul Rizzoli and Elizabeth Loder argue that the observational studies are confounded by indication. The people taking the most painkillers are the people whose headaches were worst to begin with.

The withdrawal studies were mostly uncontrolled with high dropout. Their conclusion is that only a minority of patients told to limit medication actually benefit, and that frequent use is better read as a marker of poorly controlled headache than automatically as its cause.

Then somebody tested the rule. The neurologist Todd Schwedt and colleagues randomized 720 adults with chronic migraine and medication overuse, across 34 clinics. One arm received a preventive drug plus a switch away from the overused medication, capped at two days a week. The other arm received a preventive drug with no restriction at all.

At twelve weeks, moderate-to-severe headache days were 9.3 in the switching group and 9.1 in the group that changed nothing. Not switching was not inferior. The trial was open label and pragmatic by design, which is why it carries a lower evidence grade than a blinded trial. Its outcome was headache days rather than total medication burden or long-term relapse.

Normal at rest, abnormal when challenged

The mechanism, where there is one, comes from animals. A pharmacology group in Arizona gave rats triptans repeatedly over six days. The animals developed reversible skin tenderness that faded once the drug stopped. Their dural nerve fibers carried more CGRP than before, and that change persisted long after the drug was gone.

Two weeks later, with sensory thresholds back to normal, a challenge with a nitrate compound of the kind that triggers attacks in people produced exaggerated tenderness and a bigger CGRP surge than it should have. Rats do not get migraine, so this is a shape rather than a proof.

The system looked normal at rest and answered abnormally when it was asked a question. Any resting measurement would have called it healthy.

A nervous system that meets the same input often enough reorganizes around expecting it. The repeated input has been absorbed into the tone, and the tone now assumes it. That is learning working exactly as designed, in a direction nobody chose. Nothing in it is a moral failure, and nothing here is a reason for any reader to change a medication. Anyone using acute medication frequently should have that conversation with their own physician, who can see the whole picture and weigh it.

The Unified Model of Tone reads migraine as one organization

Hypothalamic warning activity, an anticipation wave peaking a day early, low thresholds on good days, and a faster hormonal fall are established science. As they stand they are a list. The model reads the list as one thing.

Tone is the integrated organization of the body's interacting state transitions. It is the mechanical, electrical, chemical, fluid and neural processes taken as one bound state rather than as a list of parts. The notes are the body's rhythms: the cardiac and respiratory cycles, the sleep and hormonal cycles, the coordinated firing of whole populations of neurons. Tone is none of them on its own.

It is how they are held in relation to one another, and the chord is what the ear reads. A healthy nervous system runs a wide and flexible range. It can raise sensitivity sharply when something genuinely matters and drop it again once the moment has passed. The width of that range is what health is, and illness is the narrowing of it.

Every condition expresses all of tone. In migraine the signature runs through gain, oscillation and prediction, and the evidence delivers each one.

Gain: the volume is up between attacks

Light and sound reach discomfort at lower levels in people who are headache free that day. Thresholds across every modality tested sit lower and move on a schedule. The messenger is raised in chronic migraine away from any attack. Ordinary touch becomes painful during attacks in four patients out of five. Whatever is wrong is not confined to the hours of pain, which is why the disorder cannot be a headache that happens to be severe.

Oscillation: the rhythm is not free

Habituation crosses through normal before an attack and swings past normal after it. Brainstem relay activity climbs across the interval. Hypothalamic response rises inside the last 48 hours. The cycle is coupled to the menstrual rhythm through the steepness of an estrogen fall.

It is coupled to sleep in the direction that survived measurement, since sleep reliably ends an attack, and not in the direction everybody assumed, since short and poor nights were not followed by more migraine. A cycle that keeps returning to the same state is a system with too little slack in it.

Prediction: the brain is already braced

The anticipation wave is enlarged between attacks and does not settle with repetition. It climbs toward its peak in the day before the pain, which is preparation for something that has not happened yet. The person reads their own premonitory output as an external cause and builds a life of avoidance around it. Rats given a drug over and over looked normal at rest and answered abnormally when challenged, which is an expectation held quietly until a demand exposes it.

Gain says the volume is up on the quiet days. Oscillation says the volume is on a schedule. Prediction says the system has started planning around the schedule. Those are three readings of one organization.

Load runs underneath all three. Each attack leaves the range a little narrower than it found it, and what that costs over years is worked out on the time course page.

Allostatic load reached toward the same reading

One earlier framework reached toward the same reading. A review written with the neuroscientist Bruce McEwen, who named allostatic load, describes migraine as a feedforward cascade. The mediators that meet a demand become the problem when demands come often enough, so that repeated attacks change the brain state that produces attacks.

That is a conceptual review rather than data, and what holding a state costs is worked out on the allostatic load page. It points at a missing variable. What this model adds is the name, the claim that the readings are one variable, and the tests that could break it.

Inherited structure is real, and it is the other half

The discipline that keeps this model honest has to be stated plainly. Not every disease reduces to poor regulation. Migraine has real inherited structure. A meta-analysis of 375,000 people found 44 variants across 38 distinct locations in the genome, 28 of them never reported before, including the first on the X chromosome. The implicated genes were enriched for vascular and smooth-muscle tissue.

Each variant contributes very little, so it is not a predictive test and cannot tell an individual anything. The responsible formulation is that every disease has a tonal expression, and that many are initiated, maintained or amplified by failures of tonal regulation. Migraine is a strong case of the second half of that sentence, and inherited susceptibility is the first half.

Blocking a messenger and restoring a threshold are two different acts

The CGRP drugs cut migraine days by one to two a month against placebo, which is what quieting one voice produces. Restoring means widening the range, so that the thresholds stand down on their own.

This distinction decides what counts as help.

The CGRP drugs are aimed intelligently at a mechanism that was found the hard way and proved step by step. They reduce attacks in large trials with side-effect rates close to placebo. For a person losing eight days a month, two of those days back is not a small thing, and nobody should read this page as a reason to decline them.

The distinction is about what act is being performed. Blocking a messenger with precision is one act. Restoring a threshold is a different act. Both are worth having, and neither is a criticism of the other. The evidence that they are different sits in the trial numbers themselves.

Half the patients on the highest antibody dose did not halve their migraine days. The oral blocker beats placebo by roughly one to two migraine days a month. Those are ordinary, respectable numbers for a preventive drug, and they are what a hand held firmly on one voice in a chord produces while the organization that sets the thresholds carries on underneath.

Restoring means moving tone itself, so that the range widens. The threshold comes back up because the regulator recovered its ability to stand down, and not because a signal downstream of it was quieted. A matched input is the point of the word restore. Matching is a correspondence between the structure of the input and the pattern this particular body is holding.

It is a property of the reading taken beforehand rather than of the size of what is delivered. Magnitude is a separate axis altogether. A precise input delivered where the pattern is not organized is noise, however well delivered. Restoration would look like a cycle with more slack in it. A steep hormonal fall that costs less. A short night that no longer decides the following day.

The line runs across the instruments and not between them. A drug, a device, a course of movement and a conversation can each be delivered to quiet a signal or to widen a range. Which act is being performed is a question about the aim, and never about the tool in the hand. What sits on the restoring side of that line is thinner than what sits on the blocking side, and some of the thinness is that these trials keep counting bad days.

The autonomic reading

The beat-to-beat variation in heart rate is the cheapest available measure of how much braking influence the vagus is exerting, and the heart rate variability page teaches how it is taken. A meta-analysis led by the psychologist Julian Koenig pooled seven studies comparing people with headache disorders against healthy controls. The time-domain measure is simply how much the interval between beats varies from one beat to the next, and it came out significantly lower in the headache groups, a medium-sized effect.

The frequency-domain measure asks how much of that variation is happening at breathing speed. It pointed the same way when breathing was controlled and vanished when it was not, which tells you how much technique matters. Seven small studies, pooling headache disorders rather than isolating migraine. It is a signal that the reading is worth taking, not a settled quantity.

Stimulating the vagal brake from outside

The vagus is the braking half of the automatic nerves, and the vagus nerve page follows where it goes. A group led by the neurologist Stephen Silberstein randomized 59 people with chronic migraine, averaging 21.5 headache days a month. They received real or sham stimulation of that nerve at the neck for two months. Everyone then received the real device openly for six more months.

In the blinded phase the real device reduced headache days by 1.4 and sham by 0.2, a difference nowhere near significant. The blinded result was negative. Among the 15 who started on the real device and completed all eight months, the reduction was 7.9 days, with no control group and no blinding, so that larger number cannot be read as efficacy.

Exercise against a drug, and a draw

A Swedish group at the University of Gothenburg set three inputs against each other for three months. The arms were exercise for forty minutes three times a week, a recorded relaxation program, or a preventive drug titrated to the highest tolerated dose. Ninety-one patients were randomized.

Attack frequency fell by 0.93, 0.83 and 0.97 respectively, with no significant difference between the three arms. There was no placebo arm. Equivalence in a trial of 91 people does not establish that all three work. It establishes that this trial could not tell them apart.

That trial was built as a contest, and the model reads it as something else. Setting movement against a drug assumes the two are competing for one job. They are entrances. A drug enters this organization through receptor coupling. Movement enters it through the mechanoreceptors and the proprioceptive field. A relaxation recording enters through interoception and the coupling of the body's rhythms. Those are not the same act and they are not interchangeable.

They are inputs delivered at different points on one continuous loop, and a change introduced anywhere on that loop travels through the rest of it. None of that says any of the three worked, and this trial does not establish that they did. The claim is narrower and stranger. It is a claim about where an input lands, and it explains why disciplines with nothing physical in common keep getting tested against each other as rivals.

Attention training moved everything but the count

Eighty-nine adults were randomized to eight weeks of standardized mindfulness training or eight weeks of headache education, matched for class time and attention. On the primary outcome the trial failed. Both groups lost about two migraine days a month and the difference between them was not significant. On disability, quality of life, self-efficacy, catastrophizing and depression the mindfulness group did better and stayed better out to 36 weeks. Their rating of an experimentally applied pain fell by a third while the education group's rose.

Many secondary outcomes raise the multiple-comparison problem, and the team led by the neurologist Rebecca Erwin Wells called for a larger definitive trial. The two sets of numbers sit on opposite sides of the distinction this section is drawing. A count of bad days is what a quieted signal moves. Function, self-efficacy and catastrophizing are where a widened range would show, and this literature almost always calls the count primary.

Meissner found placebo responder rates ranging from 22 percent for pills to 58 percent for sham surgery

Then the finding that reads the whole field. A placebo is supposed to be an inert baseline that behaves identically everywhere. Karin Meissner, a medical psychologist in Munich, led a group that pooled the placebo arms of migraine prevention trials. They found 102 eligible trials and had usable placebo-arm data from 79 of them. The type of dummy treatment changed the result dramatically. A dummy pill produced a 22 percent responder rate. Sham acupuncture produced 38 percent. Sham surgery produced 58 percent.

Sham surgery beat a dummy pill by 36 percentage points in the same disease. Nothing pharmacological separates them. What separates them is what the system was told.

Trials of different modalities recruit different patients, and the finding does not identify which part of the surrounding context did the work. What it establishes is that the context is not zero. Meaning enters this organization the way a molecule does, through a different door onto one system. The input does not carry the outcome on its own. The system that meets it does, and that is the law the whole page has been running on.

Restoring versus masking in migraine, and how to tell

The signature prediction is bidirectional restoration: a matched input should move a dysregulated variable toward the middle from either side, beyond sham. That is what separates an input that restores regulation from one that masks a symptom. Migraine adds four predictions of its own, each testable with instruments that already exist.

The test does not grade the reading. It grades the intervention, and it says which of the two things that intervention is doing.

The bidirectional test, specified

The signature the model predicts is bidirectional restoration. A drug is directional by construction, because it acts on one mechanism in one direction and lowers a value in everybody who responds. A genuinely restored regulator behaves differently. It moves a dysregulated value toward the middle from whichever side that person is on, so the same input brings a high value down and a low value up.

Migraine makes that test harder than usual. Sensory thresholds in migraine are displaced in one direction only. Nobody walks into a clinic because their light threshold is too high. So the bidirectional test cannot be run on the threshold itself.

It has to be run on a variable with two sides: autonomic balance, heart rate variability, sleep timing, or the recovery slope after a challenge. Assemble a mixed group on such a variable and take the tone measure before any outcome is known. Let that measure name the site, in writing, before anything at all is delivered.

Then deliver a matched input to half of each side, and a sham matched for force, contact time and attention to the other half. The prediction is convergence in the treated arms beyond what the sham arms show, with the spread narrowing around the middle.

A uniform shift in one direction marks the input as one that pushes the output, helping whichever side it happens to point at and carrying the other side further from the middle. Convergence no greater than sham marks it the same way. The model predicts one thing further that no account built from independent set points expects. A single matched input should move several separately regulated measures toward the middle in the same person.

Four further predictions belong to this condition specifically.

Four predictions specific to migraine

One. The migraine brain between attacks separates from a healthy brain on dynamic measures under blinded recording, and that separation confirms the central claim here. That test is live right now, and the blinded studies disagree with one another. Two from Trondheim found normal habituation, and the second of them sorted every session by cycle phase in advance.

One from Italy and Belgium blinded the analysis instead of the recording and found the deficit intact. The claim currently survives on discomfort thresholds between attacks, on the raised messenger between attacks, and on the brainstem relay response. It is held exactly as strongly as those survive.

Two. Measures recorded before an attack carry information about when it will arrive, and that lead time confirms the cycle claim. The evidence for it is real and small: the anticipation wave, the climbing relay response, and the daily scanning that dated the warning phase to the last 48 hours.

Three. No single messenger blockade abolishes migraine in everyone, and that residual confirms this reading over a single-lynchpin account. The trial numbers say so plainly.

Four. The readings gathered here as one variable have to share one underlying factor, which is what makes the model a claim rather than a label, and compensation decides how far each reading moves. Record four things in the same patients on the same day. The discomfort threshold for light and sound on a good day. The beat-to-beat structure of the heart rhythm.

The phase relationship between the brain's slow and fast rhythms, meaning whether they keep step with one another the way a bass line keeps step with a melody. The time taken to return to baseline after a challenge. The model predicts they will share a common underlying factor. Loading together on that factor establishes them as one organization rather than separate quantities that happen to be abnormal together.

Every one of those recordings can be made today with instruments that already exist. What is missing is a habit rather than a technology. Migraine is almost always measured on the day it hurts. Everything above argues that the day it does not hurt is the day the disorder is visible.

How migraine relates to the rest of the library

Migraine is the library's clearest case of a condition that lives between its own attacks, and each neighboring page carries one part of that claim.

Three foundations of tone do the heaviest work here.

• Gain is the volume claim generalized: how loudly any nervous system answers, and why an answer that will not turn down makes ordinary light hurt.
• Oscillation is the carrier of tone, and the migraine cycle is an oscillation with too little slack in it.
• Prediction is the forecasting brain in every body, and migraine is what happens when the forecast starts running the household.
• What each attack costs over years belongs to load and to the time course, where chronification is worked out.

The condition pages divide the territory.

• Headaches owns the pain machinery itself: the fibers that wrap the vessels and membranes of the head, and the spreading wave of aura.
• Pain explains why pain is an output the brain constructs rather than a signal it receives, the premise the triptan-timing result runs on.
• Sleep carries the regulation that reliably ends attacks here while failing as a trigger.
• Vertigo assembles balance from the inner ear, the eyes and the joints, and vestibular migraine is that assembly caught in the migraine cycle.
• Why recovery differs states the law behind every messy trigger study on this page: the same input lands differently on differently organized people.

Two more pages complete the instrument set. Heart rate variability is the instrument behind the autonomic reading, which runs lower in headache disorders. Women's health carries the hormonal rhythm across the lifespan, of which the two-day, 40 percent estrogen fall is migraine's narrow slice.

Frequently asked