The Element of Surprise: A New Way to Think about Migraine Triggers

A study based on information theory shows that a surprisal score can predict the likelihood of migraine.

People with migraine often report that factors like stress, certain foods, hormonal fluctuations, or exercise, just to name a few, can trigger their headaches. But understanding migraine triggers has been a challenge for the migraine field, especially since there are hundreds of possible triggers. Overall, research to date has provided inconsistent findings about which specific triggers are associated with headache.

But now, a new study from first author Dana Turner, Massachusetts General Hospital and Harvard Medical School, Boston, US, and colleagues suggests a different, more nuanced way to consider triggers, one that better captures the complexity of triggers than an approach that studies them in isolation.

Using methods from the field of information theory, a mathematical approach “for measuring the amount of information contained in communications, stored values, or variables,” as the researchers wrote in a previous paper, the team analyzed twice-daily diary entries over 28 days from 109 people living with migraine, who documented their migraine triggers and attacks. When the researchers quantified a “surprisal score” from those data – the degree to which a particular trigger surprised study participants – they discovered that this measure reliably predicted future headache.

Dawn Buse, Albert Einstein College of Medicine, New York City, US, a migraine researcher who has studied triggers but was not involved with the new work, called the study “creative and thoughtful,” as well as one that employed a “very sophisticated approach.” And, importantly, she said the work illustrates an evolution in thinking across the field about triggers.

“Instead of a single trigger, we are now thinking about these risk factors as cumulative, and we know they often travel together in life. They are contextual and part of a bigger pattern that stacks up, leading to a headache,” she told Migraine Science Collaborative in an interview. “This kind of surprisal modeling gives us a new way to think about these relationships and how you might go about changing some of the variables to change attack patterns.”

The study appeared in the November 2025 issue of JAMA Network Open.

A novel approach
Several years ago, Turner started considering how little the migraine field knew about how individuals with migraine come to believe that certain things trigger their attacks. In an email interview with Migraine Science Collaborative, she pointed to a growing belief that triggers interact with one another to kick off the neurobiological cascade of events leading to migraine.

“If this is true, then multiple variables must be studied at the same time, and this task is very challenging,” she said.

Since many people with migraine try to identify triggers by considering what atypical behaviors or experiences may have occurred before a headache, she and her colleagues, including Timothy Houle, also at Massachusetts General Hospital and Harvard Medical School, hypothesized that how surprising a trigger is might be as valuable as knowing how that trigger could lead to headache. The idea is rooted in information theory, which postulates that unusual events carry more salient information than common ones.

“Surprisal methods are a very promising way to think about many potential triggers at the same time,” she explained. “By using information theory, hundreds or even thousands of variables can be tracked and measured. This study illustrates how these measurements can be used to identify days when someone is experiencing things that are unusual for them [and then we can] examine the relationship between unusual events and experiences and migraine attack risk.”

Buse said the surprisal hypothesis makes sense. Consider a college student who experiences a migraine attack during finals week. That attack may be brought on by a combination of atypical behaviors and events.

“Not only does this person have the stress of finals, but they also probably haven’t kept to their normal sleeping schedule. They may not be eating well or adequately hydrating. They may not have been exercising, or had time for enjoyable activities like stress-reducing friend time or movie or sports time,” she said. “All these things can increase your risk of attack. It’s not just one thing triggering the headache; it’s a combination of all these things, plus the lack of protective activities, which raise the risk of an attack occurring.”

Information theory in action
To put the hypothesis to the test, Turner, Houle, and colleagues built on their previous work (see here and here), which found that surprisal was consistently associated with increased headache activity across a variety of different types of migraine triggers. They wanted to see if they could replicate this original association in a study with prospectively gathered data.

To do so, the research team recruited 109 participants – 102 female and 7 male (median age 35 years) – who had been previously diagnosed with migraine. Over a period of 28 days, 104 of those individuals completed electronic diary entries in the morning and evening of each day, capturing potential triggers, headache activity and severity, and medication use. The morning diary also included questions about sleep-related behaviors, late-night meals, weather influences, and mood. The evening diary asked participants about their mood, food and drink triggers, environmental exposures, weather influences, meal patterns or missed meals, as well as stress.

With these data points in hand, the researchers calculated a surprisal score that reflected the degree to which an experience was atypical for study participants. “This approach provides a within-person measure of how atypical a given day’s responses are relative to that individual’s behavioral patterns,” as the researchers wrote in their paper.

Next, they asked if there was an association between the surprisal score and whether a headache attack occurred within the subsequent 12 and 24 hours.

Results indicated that a higher surprisal score was significantly associated with increased headache risk within 12 hours (an odds ratio [OR] of 1.86, meaning that for each unit increase in mean surprisal, there were 86% increased odds of reporting a future headache), as well as within 24 hours (an OR of 2.15, reflecting more than twofold increased odds). The higher OR at 24 hours indicates a stronger association at that time point, compared to 12 hours.

The team also found a lot of variability among participants in how surprisal influenced headache risk, and when they adjusted for this, the ORs were lower (an OR of 1.56, and an OR of 1.88, at 12 and 24 hours, respectively).

Interestingly, the findings also revealed that recent context mattered. For example, the likelihood of a future headache attack was affected by the surprisal score that a person had on the previous day.

Turner said the results did not surprise her, and that the study replicated the results of the researchers’ previous work.

“Observing a similar relationship between surprisal and migraine attacks increases our confidence that surprisal may be a valuable tool for studying and managing attacks,” she said. “Our research suggests that it is the overall mix of unexpected events, stressors, and routine changes that might raise short-term migraine risk and not any single migraine trigger by itself.”

Better management of migraine
Turner added that, with these results, people living with migraine should feel encouraged to know that “their daily experiences matter.” She also acknowledged the study’s limitations, including its reliance on self-reported headache, and that the analyses did not include potential covariates like acute medication use, which could have affected the results. The group plans to continue analyzing its own data, as well as data from other sources, to further understand how interactions among different variables may affect the association between surprisal and future headache.

When asked what she would like others to take away from the results, Turner said she hopes the findings will encourage clinicians to view migraine risk as “shaped by the total level of unpredictability in an individual’s day, not just isolated triggers.”

“The surprisal score gives them a clearer way to measure this risk and communicate with patients about patterns and how sleep, stress, activity, and other factors combine to influence risk,” Turner said. “This perspective can help shift the conversation from ‘Which trigger caused this?’ to ‘What patterns do we see, and how can we support more consistent healthy patterns and behaviors?’”

Buse agreed. She added that she would like to see future studies look at how surprisal affects migraine attacks in less frequently studied populations, including those who are pregnant, children, adolescents, and older persons, to see whether the findings generalize to those groups.

In the meantime, for people with migraine, Buse said the new findings reveal the importance of continuing to regularly engage in protective activities like exercise, eight hours of sleep, proper hydration, and more.

“This reminds us of how important these normal, healthy routines really are,” she said. “They can get disrupted not just in stressful times, but in happy times like a vacation or a holiday. And these routines appear to be really soothing and comforting for the brain, helping to prevent attacks.”

Kayt Sukel is a freelance writer based in Kansas.

ID 155224095 © Chutima Chaochaiya | Dreamstime.com

Reference
Information-theoretic trigger surprisal and future headache activity.
Turner et al.
JAMA Netw Open. 2025 Nov 3;8(11):e2542944.