Treatment decisions for migraine are worse than trial and error.
For the more than one in ten Americans who live with migraine, medicine has long offered little more than educated guessing — sorting patients by how often their pain arrives rather than why it arrives at all. Stanford researchers have now used functional brain imaging to identify two biologically distinct migraine subtypes, revealing that the brain's sensory processing patterns may matter far more than headache frequency in determining who suffers most and who might benefit from preventive care. Published in Cephalalgia in early 2026, the findings quietly challenge a treatment framework built on symptom counts rather than biology, and with it, the insurance logic that has long decided who deserves relief.
- Millions of migraine patients are denied preventive medication not because it wouldn't help them, but because they haven't crossed an arbitrary threshold of fifteen headache days per month — a bureaucratic line drawn without biological evidence.
- Stanford researchers scanned the brains of 111 migraine patients and found two distinct clusters: one whose brain activity resembled healthy controls, and one whose sensory processing systems appeared to be in a state of chronic overreaction to ordinary stimuli.
- The more severely affected cluster 2 patients experienced longer, more disabling attacks and greater suffering — yet had no more frequent headaches than cluster 1, exposing how badly the current chronic-versus-episodic classification misses the point.
- The team is now racing to translate imaging findings into accessible blood biomarkers and clinical criteria, so that a patient with five headache days a month can be identified as biologically high-risk before their condition deepens.
- If the research reaches clinical practice, it would force a reckoning not just in neurology offices but in insurance boardrooms — redefining who qualifies for treatment based on biology rather than a calendar count of bad days.
Migraine is among the world's leading causes of disability, yet for decades medicine has managed it almost entirely on the basis of what patients can describe. A person reporting fifteen or more headache days per month receives a chronic migraine diagnosis and access to preventive medication; anyone below that threshold is labeled episodic and largely left to manage attacks as they come. Insurance companies enforce this boundary strictly. "Right now, treatment decisions for migraine are worse than trial and error," said Stanford neurologist Robert Cowan. "It's darts in the dark."
Cowan's team chose to look beneath the symptom surface. They gathered brain imaging data from 111 migraine patients and 51 healthy controls, then used computational methods to find patterns without imposing assumptions. Structural MRI revealed little of note, but functional MRI — which tracks blood flow between brain regions — told a different story. Two distinct biological subtypes emerged from the data.
The first cluster's brain activity patterns resembled those of people without migraines, and their attacks tended to be milder. The second cluster showed abnormal connectivity between the brain's cortex and deeper structures, suggesting their sensory systems were chronically overreacting — interpreting ordinary light, sound, and touch as threats. These patients were older, endured longer attacks, and reported greater disability. Crucially, they had no more frequent headaches than cluster 1 patients. Frequency, it turned out, was an incomplete measure of suffering.
The implications cut directly against the current treatment framework. A patient with episodic migraine — fewer than fifteen headache days monthly — might carry the biological signature of cluster 2, meaning their attacks are more severe and disabling than those of a chronic migraine patient in cluster 1. Under existing guidelines, that person would still be denied preventive medication.
Published in Cephalalgia in March 2026 and led by data analyst Jaiashre Sridhar, the research now points toward a more precise future. Cowan's team is developing blood biomarkers and clinical criteria to identify subtypes without costly fMRI scans — so that a doctor could recognize a high-risk patient early and intervene before disability deepens. If that vision takes hold, it would reshape not only how neurologists treat migraine, but how insurers decide who deserves the chance to prevent it.
Migraine sufferers describe the experience in language borrowed from catastrophe: a brain on fire, an ice pick driven through the skull. More than one in ten Americans know this pain intimately. It arrives with nausea, with light becoming unbearable, with sound turning hostile. The attacks can render a person unable to work, unable to move, unable to think. Migraine ranks among the world's leading causes of disability, yet the medical system treats it like a guessing game.
For decades, doctors have diagnosed and managed migraine based entirely on what patients tell them. A person reports headaches on fifteen or more days per month, and they receive a diagnosis of chronic migraine—eligible for preventive medication. Someone with fewer headache days gets labeled episodic migraine and typically receives only acute treatment when attacks strike. Insurance companies enforce this boundary strictly, refusing to cover preventive drugs for those who don't cross the chronic threshold. "Right now, treatment decisions for migraine are worse than trial and error," said Robert Cowan, a clinical professor of neurology at Stanford Medicine who specializes in headache research. "It's darts in the dark."
Cowan and his team decided to look beneath the surface. Rather than starting with assumptions about how migraine works, they collected brain imaging data from 111 migraine patients and 51 people without migraines, then let computational methods find patterns in the numbers. They used two types of MRI: structural imaging, which maps the brain's physical anatomy, and functional MRI, which tracks brain activity by measuring blood flow between regions. When the computer sorted the data into clusters, functional imaging proved far more revealing than structure alone.
Two distinct biological subtypes emerged. The first cluster resembled the control group in their brain imaging patterns and experienced milder headaches overall. The second cluster showed something different: abnormal blood flow patterns between the brain's cortex and deeper structures. These patients' brains appeared to be overreacting to sensory input—light, sound, touch—triggering pain in response to ordinary daily experiences. Cluster 2 patients were older, experienced longer-lasting attacks, and reported greater disability from their condition. Yet they had no more frequent headaches than cluster 1 patients. The frequency of attacks, it turned out, told only part of the story.
This finding challenges the entire framework that currently governs migraine treatment. A person might have episodic migraines—fewer than fifteen headache days monthly—but possess the biological signature of cluster 2, meaning they suffer more severe, disabling attacks than someone with chronic migraine in cluster 1. Under current guidelines, that episodic patient would be denied preventive medication, even though they might benefit enormously from it. "Many episodic migraine patients could benefit from preventive treatment," Cowan said, but insurance barriers keep those medications out of reach.
The research, published in the journal Cephalalgia in March 2026 and led by data analyst Jaiashre Sridhar, opens a path toward more precise treatment. Cowan's team is now working to develop blood biomarkers and clinical criteria that could identify which patients belong to each biological subtype without requiring expensive fMRI scans. The goal is straightforward but ambitious: to tell a doctor that a patient with five headache days per month has the biological profile of cluster 2 and would genuinely benefit from starting preventive medication now, before their condition worsens or their disability deepens.
If that vision becomes reality, it would upend not just how doctors think about migraine, but how insurance companies decide who deserves treatment. The current system sorts patients by frequency. The new biology suggests that severity and sensory processing matter more. For millions of Americans living with migraine, that distinction could mean the difference between suffering through attacks and preventing them altogether.
Citas Notables
Current treatment decisions for migraine are worse than trial and error—it's darts in the dark.— Robert Cowan, MD, clinical professor of neurology, Stanford Medicine
Many episodic migraine patients could benefit from preventive treatment, but insurance companies often won't cover it for those who don't meet the chronic migraine threshold.— Robert Cowan, MD
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that cluster 2 patients have the same frequency of headaches as cluster 1 but worse outcomes?
Because frequency is what insurance companies and guidelines use to decide who gets preventive medication. You could have five migraines a month that leave you unable to function, but be denied treatment because you don't hit fifteen days. The biology says that's wrong.
So the brain imaging is showing something the patient's own description can't capture?
Exactly. A patient can tell you how often they hurt and how bad it hurts. But they can't tell you that their brain is processing sensory information abnormally, that their cortex and subcortical regions are communicating in a way that makes them vulnerable to pain. The imaging reveals the mechanism.
Why would a brain evolve to trigger pain from everyday sensations?
It wouldn't, not normally. Pain is protective—it makes you retreat from danger. But in cluster 2 patients, the system seems to have gone haywire. The brain is treating ordinary light or sound as a threat when it isn't.
If fMRI is expensive, how will this actually change treatment for regular patients?
That's the next phase. They're looking for clinical clues—maybe certain age ranges, or patterns in how long attacks last, or other markers in blood work—that would correlate with the imaging subtypes. If they find those, a doctor could identify cluster 2 patients without the scan.
And then what happens?
Then insurance might have to cover preventive medication for people who don't meet the old chronic migraine threshold. The biology would override the frequency rule.
How close are they to that?
They're working on it now. But this is early. The study was 111 patients. They need to validate it in larger groups, develop those clinical criteria, and show that treating cluster 2 patients preventively actually works better. It's a beginning, not an answer yet.