A Better Way to Classify Migraine: Linking Symptoms to Patterns of Brain Activity

By Fred Schwaller | November 7, 2022 | Posted in

“Episodic” versus “chronic” migraine is not a biologically meaningful distinction, according to a new fMRI study.

According to current clinical criteria, people with migraine are said to have episodic migraine if they experience up to 14 headache days per month, or chronic migraine if they have headache on 15 or more days each month. A new analysis of brain imaging data now suggests that this episodic vs. chronic distinction does not stand up to scrutiny and instead offers a different way to group patients.

Using data from previous functional magnetic resonance imaging (fMRI) studies, researchers led by David Seminowicz, of the University of Maryland, Baltimore, at the time of the study but now at the University of Western Ontario, Canada, were able to relate a wide range of migraine symptoms to patterns of brain activity, finding little evidence that migraine should be divided into episodic vs. chronic subtypes. Instead, the data reveal four unique subtypes of migraine based on associations between symptoms and unique brain activity patterns.

The study also finds links between several migraine symptoms and brain activity patterns in major cognitive brain networks. And, patients with more disruption in those networks had better responses to mindfulness, a type of mind-body therapy.

“This is a great step forward, as it gives biological evidence of subcategories of migraine in patients,” said Nasim Maleki, a researcher at Harvard Medical School and Massachusetts General Hospital, Boston, US, who uses brain imaging to understand the pathophysiology of migraine but was not involved with the current study.

But, she cautioned, “fMRI data on its own isn’t enough to cluster patients based on clinical profiles. Future work needs to combine fMRI data with genetic and proteomic datasets to give a full biological picture.”

Whether a causal relationship exists between migraine symptoms and the patterns of brain activity identified in the current study must also be clarified, Maleki told MSC.

The study appeared in the Journal of Neuroscience on August 3, 2022.

Seeking something better
The limitations of classifying migraine according to headache days per month provided a spur for the new research.

“It isn’t really a helpful way to think about variability in patient experience,” said Samuel Krimmel, first author of the study and currently a postdoctoral fellow at Washington University in St. Louis, US.

“It’s based on headache frequency and doesn’t consider other symptoms. Even two patients who have the same number of headaches each month can present very different symptoms when you take a more holistic approach to diagnosis. For example, one patient might have sleep disturbances, while the other might sleep fine but experience anxiety,” Krimmel told MSC.

So the study aimed for a strategy that would link a broader range of migraine symptoms to what is known as resting-state functional connectivity. This refers to patterns of brain activity reflecting how different regions of the brain are functionally connected to each other at rest, in the absence of a person performing any tasks.

“Resting state functional connectivity is a way of looking at shared signals across distinct areas of the brain – both their temporal and spatial relationships,” Krimmel explained.

The underlying hypothesis of the study was that different migraine symptoms would change functional connectivity, even if a patient wasn’t experiencing symptoms at the time of measurement.

The authors would use a statistical approach, called canonical correlation analysis, that allowed them to study the relationships between multiple migraine symptoms and resting-state functional connectivity. This approach acknowledges the heterogeneous nature of migraine symptoms and attempts to root those symptoms in biology – patterns of brain activity, in this case.

Three dimensions
The team began by using data on resting-state functional connectivity that were available for 143 men and women with migraine who had been recruited for research investigating biomarkers of headache at the University of Maryland, Baltimore, or at Stanford University. The researchers also had comprehensive information about each patient’s symptoms.

Specifically, they had data for seven different symptoms, including the frequency of headache pain, the intensity of headache pain, how well patients slept, anxiety, depression, pain catastrophizing, and somatization (the physical manifestation of psychological or emotional distress).

The goal was to learn if the patients’ migraine symptoms were linked to specific brain activity patterns. Patients would receive scores that indicate the strength of the association between symptoms and those patterns.

The investigators found that there were three “dimensions of covariance” between symptoms and resting-state functional connectivity. That is, there were three unique constellations of migraine symptoms associated with specific patterns of functional connectivity.

The first dimension related to all seven symptoms. Here, those symptoms were associated with functional connectivity of frontoparietal and dorsal attention networks (brain areas involved with cognition) with other networks.

The researchers called the second dimension the inverse pain/anxiety dimension. Here, an inverse association between those two symptoms was related to functional connectivity of the default mode network – a group of brain regions that becomes active when a person is passive and not focused on what is happening around them – and the sensorimotor network (brain areas responsible for processing sensations into movements).

The third dimension was the pain catastrophizing dimension. In this instance, catastrophizing – a pattern of distressing emotional and cognitive responses to pain – was associated with functional connectivity of the salience network (brain regions responsible for attention) to the default mode network and sensorimotor network, and with connectivity of the default mode network with other networks.

“These were extremely exciting findings for me, as they fit our model that migraine – like other chronic pain conditions – is also a disorder of cognitive networks,” Seminowicz told MSC.

Episodic versus chronic? No. Biotypes? Yes.
Next, the team tested the episodic versus chronic migraine distinction, using data from 166 patients. They asked if that distinction had any biological relevance in terms of the relationship between symptoms and functional connectivity.

Using their statistical technique, the researchers found no support for subtyping patients into episodic versus chronic migraine categories.

“In effect, our data shows that clustering based on headache days per month is not better than chance at explaining variability in migraine,” said Krimmel.

So the next goal was to identify a better way to subtype patients – one rooted in actual biology rather than on an arbitrary distinction based on monthly headache days. The investigators referred to these subtypes as “biotypes.”

Based on the functional connectivity scores from the three dimensions, four different biotypes emerged, each with different patterns of connectivity. The first biotype comprised those who had worse than average symptoms overall. Biotype 2 had above average catastrophizing. Biotype 3 had a pattern of symptoms opposite to all the other biotypes. And biotype 4 had below average pain and above average anxiety.

A biotype approach, the authors argue, would be much better than what the migraine field relies on now. For example, Seminowicz explained that people with globally worse migraine symptoms might only have eight headaches a month but have all their other symptoms be severe. This kind of patient would be diagnosed with episodic migraine under current criteria, but might be better classified as having chronic migraine, compared to someone who has 18 headache days per month but no other symptoms.

“This really argues that diagnosis and drug approval shouldn’t be stratified by current definitions of episodic and chronic migraine,” Seminowicz said.

Mind-body therapy: Precision medicine for migraine treatment?
The authors next tested whether their model could predict how people with migraine respond to mind-body therapy. Here, they included data from a clinical trial investigating the effect of 20 weeks of mindfulness therapy on migraine.

In that trial, Seminowicz and colleagues had reported that mindfulness caused a reduction in the frequency of headaches. Now, in the new study, there was also an association between the first dimension – the dimension related to all seven migraine symptoms examined in the study and resting-state functional connectivity of cognitive networks – and treatment outcome.

“We found that the lower your global symptom dimension score – meaning you had worse symptoms – and the more disruptive connectivity of cognitive networks you had – the better your outcome with mindfulness was,” Seminowicz said.

Though it’s still early days, Seminowicz is excited by the prospect that people with migraine could be selected for mindfulness therapy based on their brain activity patterns.

“This points to a type of precision medicine approach,” he said.

Looking ahead
One major limitation to this (and other) fMRI studies is that it’s uncertain if the brain patterns the researchers identified actually reflect something about migraine specifically.

“It’s possible here that the brain patterns are correlated with pain in general. Future work should look at the causality of brain patterns and migraine symptoms,” Maleki said.

Nevertheless, the team’s approach of linking clinical features of migraine to brain connectivity patterns results in a more complex model than current classification approaches. In doing so, it better reflects the heterogeneous nature of a disorder in which diverse symptoms vary among patients.

The authors hope that their analysis will open debate about how the field defines migraine and how doctors select patients for treatment. They also hope the study spurs researchers to combine data from brain imaging with unique genetic and proteomic information from patients, as Maleki suggested, to provide in-depth information about a migraine patient’s clinical profile. This could pave the way for a precision medicine-based approach to migraine treatment.

Krimmel and Seminowicz said that to truly address these big questions, the field needs to get better at data sharing.

“Sample size is very important for these kinds of datasets. Clinicians rely on these results to help patients, and results need to be from large sets of patients to cover all possible alternatives,” Seminowicz said.

Fred Schwaller, PhD, is a freelance science writer based in Germany. Follow him on Twitter @SchwallerFred


Three dimensions of association link migraine symptoms and functional connectivity.
Krimmel et al.
J Neurosci. 2022 Aug 3;42(31):6156-66.


Sign Up For An MSC Newsletter

Fred Schwaller is a science writer and communicator based in Berlin, Germany. Fred spent a decade in pain research during his doctoral degree at University College London, UK, and his postdoc at the Max Delbrück Centre in Berlin, Germany. After transferring to science communication in 2020, he has been writing and podcasting about life sciences and medicine, specializing in somatosensation and pain. Follow him on Twitter @SchwallerFred.



  • Content Types

  • Paper Categories

  • Paper Dates

Recent News

Recent Papers