Tears collected from patients in the middle of a cluster-headache attack carried more of a pain-signalling peptide than the same patients' tears minutes earlier, and the increase was largest when the attack hurt the most.
That is the finding from a small study published in the European Journal of Neurology ↗ by a neurology group at University Hospital, LMU Munich. The peptide is calcitonin gene-related peptide, or CGRP, the same molecule that a wave of migraine drugs was built to block. The result is a data point in an argument that has run for years: does CGRP actually drive cluster headache, or is it a bystander?
Cluster headache is one of the most severe pain states in medicine. Attacks are short, strictly one-sided, and come with what neurologists call trigemino-autonomic signs on the painful side, including a watering eye, a drooping lid, and a running or blocked nostril. The watering eye is why tears are a plausible place to look. The trigeminal nerve endings that misfire during an attack sit close to the tear gland, so tear fluid offers a window on local nerve chemistry without a needle.
The Munich team, led by Katharina Kamm with senior author Ruth Ruscheweyh, screened 89 people with active episodic or chronic cluster headache and enrolled 18. On a study day, eligible patients took 0.9 milligrams of glycerol trinitrate under the tongue. Glycerol trinitrate is nitroglycerin, the same drug used for chest pain, and it is a well-established way to provoke a headache attack in a controlled setting so it can be measured. Tear fluid was sampled before the dose and again once an attack developed. It was then run through a commercial CGRP ELISA from Cusabio, a standard antibody-based assay that reports a concentration.
Sixteen of the patients developed a provoked attack. Across that group, tear CGRP rose from 2.34 nanograms per milliliter before the trigger to 3.70 nanograms per milliliter during the headache (p = 0.019), a difference unlikely to be chance. The more telling number is the split by severity. Patients who rated their attack at 5 or higher on a 0 to 10 pain scale showed a jump of about 2.65 nanograms per milliliter, while those with a milder attack barely moved, at 0.07 (p = 0.015). The signal scaled with the pain. That internal gradient is worth more than the headline average, because each patient's mild-versus-intense contrast is its own control.
Here is the tension the study lands in. CGRP-blocking antibodies like galcanezumab and fremanezumab reshaped migraine care. Galcanezumab even carries a US approval for episodic cluster headache. But when the antibodies were tested against cluster attacks the results were thin. As this section covered in June, the same antibodies barely beat placebo in cluster headache ↗. A biomarker that clearly rises with the attack, sitting next to a drug class that clearly underperforms against that attack, is a genuine puzzle. Measuring CGRP where it spikes is not the same as showing that removing it stops the pain. The Munich authors are careful to frame their work as support for CGRP's role in the pathophysiology, not as a treatment claim.
The caveats are real and the authors flag them. This is an exploratory study of 18 people at a single center, with only 16 usable attacks. The concentrations are in nanograms, far above the picogram plasma levels usually reported for CGRP. That gap reflects both the local sampling site and the particular assay, so the absolute figures should be read as relative, not as a portable reference range. The scatter is wide, with the during-attack mean of 3.70 carrying a standard deviation of 5.60. Attacks were chemically provoked rather than spontaneous. And because a watering eye is itself part of an attack, the tear signal is inseparable from the trigemino-autonomic response the pain sets off.
On peptidemodel, CGRP is hosted as alpha-CGRP ↗, the human form that acts through the calcitonin-receptor family ↗ to widen blood vessels and carry pain. It is the peptide that gepants and the injected antibodies were designed to silence, and it keeps turning up as a measurable marker: earlier work found CGRP shifts in migraine blood ↗ and in a rat brainstem circuit tied to facial neuralgia ↗.
What this exploratory study adds is narrow and concrete. In cluster headache, a peptide you can now sample from a tear rises as the attack builds and rises further as it worsens. Whether that makes CGRP the right thing to block in these patients, or just the easiest thing to watch, is the question the drug trials have not answered.