In July 2026, the journal Headache published a study I had quietly expected to see for years. Working from the world's two largest drug-safety databases, a French team reported a statistical association between COVID-19 vaccination and cluster headache - a signal that our community had been talking about, anecdotally, since the vaccines first rolled out. It is an important paper, and it deserves to be read the way its own authors wrote it: carefully. A signal is not a verdict. Here is what the study found, what it honestly cannot tell us, and - separately, and clearly labelled as mine alone - where it leaves me as a patient.
What the study actually did
The study, led by Elise Van Obberghen and colleagues at CHU Nice and CHU Lille, is a pharmacovigilance analysis (Van Obberghen et al., 2026). That word matters, because it defines both the strength and the limits of what follows. Pharmacovigilance is the science of monitoring spontaneous reports of suspected adverse drug reactions. It is very good at raising a flag. It cannot, on its own, prove that a drug caused anything.
The team ran two analyses. First, a disproportionality analysis on VigiBase, the World Health Organization's global database of individual case safety reports. The question here is simple to state: is cluster headache reported more often than you would expect after COVID-19 vaccination, compared with the background rate of all other drug–reaction pairs in the entire database? They measured this with a statistic called the information component (IC), and required its lower confidence bound (IC025) to sit above zero before calling it a genuine signal. To avoid being fooled by the sheer volume of COVID-vaccine reports, they set a deliberately strict threshold, analysing only vaccines with more than ten reported cases.
Second, because VigiBase strips out the case narratives, they turned to the French national database (the BNPV), where two neurologists could actually read the 55 French cluster-headache reports and judge, story by story, whether each one truly met the diagnostic criteria for cluster headache - and whether the vaccine appeared to have triggered a new case or worsened an existing one.
What they found
Across all COVID-19 vaccines, the signal was clearly positive: 2,275 cluster-headache reports where roughly 564 would have been expected (IC025 1.95). When they broke it down by individual vaccine, the picture was not uniform.
| Vaccine (type) | Reports | Signal (IC025) | Strength |
|---|---|---|---|
| Vaxzevria - AstraZeneca (viral vector) | 958 | 3.05 | Strongest |
| Comirnaty - Pfizer-BioNTech (mRNA) | 1,041 | 1.62 | Moderate |
| Spikevax - Moderna (mRNA) | 304 | 1.24 | Moderate |
| Jcovden - Janssen (viral vector) | 39 | 0.05 | Borderline / very weak |
The strongest signal by a wide margin belonged to Vaxzevria, the AstraZeneca viral-vector vaccine. But note the shape of the table: a strong signal appeared with a viral-vector vaccine and with the mRNA vaccines, while the other viral-vector product (Jcovden) barely registered. The authors are careful to say that neither vaccine technology clearly stands out as the culprit - signals showed up across both platforms.
The narrative analysis of the 55 French cases was, to me, the most human part of the paper. Of those cases, 24 (44%) looked like a new onset of cluster headache in someone with no prior history; 17 (31%) were a worsening or reactivation in someone who already had the disease; three were unclassifiable for lack of history; and 11 (20%) - a full fifth - did not actually fit a cluster-headache diagnosis at all on closer reading. The median time from vaccination to the reported attack was just three days.
That slight tilt toward new onset over worsening is genuinely unusual. As the authors point out, the classic provocateurs of cluster headache - alcohol, nitroglycerin, infused CGRP - only trigger attacks in patients who are already in an active bout. They do not create the disease in someone who never had it. If COVID-19 vaccines can do that, even rarely, it would make them behave differently from anything else we know of.
How might a vaccine even touch this?
Honestly: nobody knows, and the authors say so plainly. The mechanism is unknown. What they offer is a hypothesis rooted in inflammation. We know corticosteroids can break a cluster bout, and that they lower CGRP - a neuropeptide central to the attack - which points to an inflammatory component in the trigeminovascular system. The proposal is that the immune response provoked by any vaccination could, in a susceptible person, release prostaglandins and bradykinins, drive CGRP, and tip an already primed system over its threshold into an attack.
I want to be careful here, because this is exactly the kind of plausible-sounding chain that can harden into "fact" if it is repeated enough. It is a hypothesis. It has not been demonstrated. It is offered by the authors as a direction for future work, not a finding.
The limits - and why they matter
This is the part I most want people in our community to sit with, because it is where credibility is either kept or lost. The authors devote real space to everything their own study cannot do, and I think that honesty is what makes the paper worth trusting.
- A signal is not causation. Disproportionality tells you something is being reported more than expected. It cannot tell you why. People may have developed cluster headache anyway, vaccine or no vaccine, with the timing being coincidence.
- The "Weber effect" and media attention. Few drugs in history have been watched as closely as the COVID-19 vaccines. Intense scrutiny inflates reporting all by itself - if everyone is looking for adverse events and encouraged to report them, more get reported, which can manufacture a signal that partly reflects attention rather than biology.
- One in five reports weren't even cluster headache. When neurologists actually read the narratives, 20% did not meet the diagnostic criteria. That is a sobering measure of how noisy database-only studies can be.
- Underreporting cuts the other way too. Cluster headache is rare and badly under-recognised, so many real cases never get reported at all.
- Small numbers, missing data. The qualitative analysis rested on 55 cases with frequently incomplete records, which the authors say makes it hard to generalise to the wider population.
And then there is the sentence I think every responsible write-up of this paper is obliged to quote, because the authors chose to include it:
“Our findings neither undermine the benefit/risk ratio of the vaccines nor the importance of COVID-19 vaccination.”
They mean it. A rare association with a rare disease, even if real, does not overturn the calculus that led public-health bodies to recommend these vaccines for the populations they were recommended for. Both things can be true at once: that this signal deserves serious follow-up with prospective studies, and that the vaccines' overall benefit–risk picture is a separate and much larger question this paper does not claim to settle.
Not medical advice. This article summarises one published study and shares a personal reflection. It is not a recommendation for or against vaccination, and nothing here should replace a conversation with your own doctor, who knows your history and can weigh the trade-offs that actually apply to you.
A personal note - mine, and only mine
Everything above is the science, reported as straight as I can make it. What follows is different in kind. It is one patient's reflection, offered honestly and held loosely, and I have deliberately walled it off from the evidence so no one mistakes my feelings for findings.
I will be transparent: I did not take the COVID-19 vaccine. Part of that was a specific fear - that it might trigger or worsen cluster, the very thing this paper now examines. I am not going to pretend I reasoned my way there from data that did not yet exist; it was closer to an instinct, and papers like this one land on that instinct in a particular way.
Part of it, too, was influence I respect. Pete Batcheller - the man behind the high-dose vitamin D3 regimen that has helped so many of us, and someone I regard as a mentor - was openly cautious about the vaccines, and I found I agreed with him. My own reasoning came through vitamin D. The more I learned about how that one molecule works - how the vitamin D receptor sits on the genome and helps regulate an enormous number of our genes, up and down, in concert - the more the sheer intricacy of it stopped me in my tracks. It left me with a deep respect for how finely tuned these systems are, and a corresponding humility about how much we still do not understand them.
From that place, the idea of instructing our own cells to manufacture a foreign protein gave me pause. Not as a scientific verdict - I am not qualified to render one, and mRNA vaccines have a real evidence base I am not dismissing - but as a personal intuition that we are very early in understanding this machinery, and that intervening in it so directly asks for a humility the moment did not always seem to have. That is a feeling, not a fact, and I want to name it as such.
It also fits how I think about treating cluster headache more broadly. I understand completely why someone reaches for every pharmaceutical tool available - CGRP antibodies, the lot. This disease is one of the most severe pains known to medicine, and I would never judge anyone for whatever it takes to survive it. That has to be said clearly, because the last thing I want is for a personal preference to read as a rule for anyone else. For me, though, I have come to see cluster as a sign of something knocked out of balance, and my instinct is to look first for the road back toward balance - through environment, habits, mindset and remedies that work with the body's own regulation rather than around it. A study like this one doesn't prove that instinct right. It just makes me feel less alone in holding it.
Where this leaves us
The grounded takeaway is small and solid. A large, carefully-conducted pharmacovigilance study has found a real statistical signal between certain COVID-19 vaccines - Vaxzevria most of all - and cluster headache, including some apparent new-onset cases. That signal is worth taking seriously, worth following up with the prospective studies the authors call for, and worth mentioning to a neurologist if your own bouts began in close proximity to a vaccination. It is not proof of cause, it is not a reason for panic, and by the authors' own words it does not overturn the case for vaccination in the populations it was meant for.
For a condition that has spent most of its history being disbelieved, having our shared observation examined properly and reported responsibly is, in its own quiet way, a kind of progress. Read the paper carefully. Hold the uncertainty. And make your own decisions with a doctor who knows you.