Cluster Headache: The Master Reference

⚠

Not medical advice. A privately maintained research summary, revised continuously, which may contain errors or superseded findings. Treatment decisions belong with a clinician who knows your history. Read the full notice.

What this is

Cluster headaches are hard to research. My team of helpers came back with a lot of information, which was wonderful. Having not studied in a scientific field since school, I assumed I’d be more than capable of parsing the data, constructing a mono-document, and understanding the complexity. I was wrong.

So with the assistance of further AI helpers, the disparate texts, trials, and stories became one big massive document. This document is my attempt to gather what is actually known about CH in one place, at full depth, for anyone who is interested.

Which is great. Except, I can only grasp the high level ideas, and I wanted detail! Then I thought - why not just make a retrieval system, so I can ask a language model to ELI5 without tapping into its un-citable training data and ruining an otherwise accurate answer! (In other words, if I ask the document a question, I only want it answering from the document, not from what it learnt in training) This was harder than I anticipated, and it’s still not perfect, but it’s pretty bloody close :)

In the document, every claim is tagged with its provenance, so the difference between a randomised trial and a message board is clear. Contradictions are kept intact, so where a 2025 meta-analysis and a Danish registry study disagree about suicide risk, both are included with the disagreement. The patient community is also taken seriously as a source of knowledge, as we’re the most motivated (and historically first) when it comes to new treatments and the like.

Facts

High-flow oxygen and subcutaneous sumatriptan for attacks; verapamil and occipital steroid injection for prevention. Treatment →

People Power

Psychedelics travelled from a 1998 message board to controlled trials at Yale. The vitamin D3 regimen is used by thousands and formally tested by almost nobody. Citizen science →

Unknowns

Why chronic cluster headache resists nearly everything episodic responds to; whether CGRP is raised or lowered; why East Asian cohorts look different. Open questions →

Next

PACAP is arguably the best-evidenced untried target in headache medicine. The live pipeline, the funding gap, and the watchlist of what to expect when. The frontier →

The Very Important Pattern

One finding runs through the entire treatment section: almost every therapy performs worse in chronic than in episodic cluster headache, and the figure usually quoted is the episodic one. CGRP monoclonal antibodies produced one marginal win in episodic disease and failed every chronic trial across three drugs. The best-evidenced neuromodulation device is commercially unavailable after its manufacturer collapsed. If you have the chronic form, read the statistics and results with that in mind.

How to use it

Read the executive summary first; it’s got the whole argument. Use search for anything specific — it covers the full text including tables. The knowledge graph shows how treatments, mechanisms and trials connect. The assistant answers from the text and links to the passage it used, and will tell you when the answer is not there.

The whole thing is downloadable as PDF, Word or Markdown, and there is no login, no tracking and nothing for sale. I hope this is useful, go do useful things with it.

Ben