How to Stop a Cluster Headache: Treatments That Work in Minutes

The attack arrives behind one eye, usually at the same hour of the night. The eye waters, the nostril runs, and the person paces because lying still is impossible. Forty-five minutes later it is gone, until tomorrow. Anyone searching for how to stop a cluster headache has usually already tried a painkiller that did nothing, because the tablet was still dissolving when the attack ended.

Chapter 11 of The Pained Brain, the video above, calls this the most painful headache in medicine and argues it is a problem of minutes, not weeks. This page adds the evidence behind each option, graded honestly, and the plan for the bout rather than the single attack. The book’s authors are Dr. Gurpreet Singh Padda, MD, MBA, MHP, and Dr. KrisJay Fucanan, MD.

The attack and the clock that sets it

Cluster headache is rare, with a lifetime prevalence of about 124 per 100,000 people. Attacks last 15 to 180 minutes, can come up to eight times a day, and stay on one side. Among 1,604 people who met the diagnostic criteria, the average attack scored 9.7 out of 10, and 72.1 percent gave it the maximum. The same respondents rated labor at 7.2.

Two biological systems run it. The timing comes from the hypothalamus, the brain’s master clock; a PET study of provoked attacks showed activation there on the side of the pain and only during the pain state. The pain is delivered through the trigeminal nerve and the parasympathetic wiring that passes through the sphenopalatine ganglion, a small knot of nerve tissue behind the nose. That relay is why the eye tears and the nose runs. It is also where one of the fastest treatments aims. The clock half of this story is in why pain is worst in the small hours, and the same circadian machinery runs through the gut’s own jet lag.

Why the diagnosis takes a decade

Pooled across 22 studies and 8,654 patients, the mean time from first attack to correct diagnosis is 10.43 years. That average is inflated by older cohorts. In one British series the delay fell from 22.3 years for onset in the 1950s to 2.6 years for onset in the 1990s, and in a Danish series it fell below one year after 2010. Onset before age 20 still carries a delay of 13.8 years.

The misroutes follow a pattern. In the Netherlands, 34 percent had consulted a dentist and 33 percent an ear, nose and throat surgeon first. Women are more often labeled migraineurs because they more often have nausea and less often the red eye. Yet half of all cluster patients have light or sound sensitivity, so those features cannot separate the two conditions.

A three-question screen covering intensity, duration and the tearing and nasal signs identifies cluster headache with 84 percent sensitivity and 89 percent specificity in its validation sample. The system does not ask it. Among emergency physicians in one survey, 92.9 percent said they place these patients in the lowest-acuity zone, and only 15 percent of family and emergency physicians felt adequately informed.

How to stop a cluster headache attack

• High-flow oxygen. In a randomized crossover trial, 100 percent oxygen at 12 liters a minute by face mask for 15 minutes ended or adequately relieved 78 percent of treated attacks, against 20 percent for air. Only 76 of 109 randomized patients contributed attacks, and the figure is per attack, not per patient.
• Injected sumatriptan. A 6 mg injection left 48 percent pain-free at 15 minutes against 17 percent on placebo, a number needed to treat of 3.3. The reviewers concluded that oral routes are not appropriate for an attack this short.

Opioids are on no recommended list, and patients know why. Among 2,193 people surveyed, oxygen was completely or very effective for 54 percent and opioids for 6 percent. Yet among 7,589 newly diagnosed patients in insurance claims, opioids were the most common prescription class in the following year, at 41 percent. Those given opioids were in the emergency department or hospital within the year at 53.3 percent against 26.8 percent for those given a recognized treatment without one. Sicker patients may get more opioids, so that is not proof of cause. The emergency room pattern is familiar from the mismatch chapter before this one.

The economic driver is access. Of 566 patients who eventually got oxygen, 36 percent had it within a month of diagnosis and 13 percent waited two to five years, because a physician did not believe in it or an insurer would not cover it. The mean wait for a new neurology appointment in the United States is 49.7 days. A six-week bout is over by then.

The ganglion behind the nose

The sphenopalatine ganglion sits just behind the middle turbinate, a few millimeters under the nasal lining. Sluder described it as a source of one-sided facial pain with tearing and congestion in 1908. Because it is so shallow, it can be bathed in local anesthetic through the nostril with a soft catheter and no needle through the skin. Onset is typically within 10 to 30 minutes. In a series of 489 such blocks in children and young adults with migraine, done under fluoroscopy, there were no immediate complications and each took less than 10 minutes. In a chronic migraine trial, the common effects were tearing in 30 percent and a bad taste in 26 percent, all short-lived.

Here is the evidence tier, stated once. In cluster headache itself, the controlled study is small: of 15 patients, 9 had attacks provoked with nitroglycerin, and anesthetic on the ganglion ended the pain in all of them, in 37.0 minutes with lidocaine against 59.3 with saline. A systematic review gives the block a Grade B on that study. No sham-controlled trial of the office block in spontaneous attacks has been run. The anatomy has been tested another way: an implanted stimulator on the same ganglion relieved 67.1 percent of attacks within 15 minutes against 7.4 percent for sham in 28 patients. The block is done awake, without sedation, and patients drive themselves home. SPG and stellate ganglion blocks for cluster headache and migraine describes the procedures in more detail.

Ending the bout: the bridge

Stopping one attack is the first job. Ending the run of attacks has stronger trials behind it. A single steroid injection around the occipital nerve at the back of the head left 11 of 13 patients attack-free for the following week against none of 10 on placebo. Three injections brought 20 of 21 patients under two attacks a day against 12 of 22. Pooled, the relative risk of pain freedom at one month is 4.86. A short oral prednisone course cut first-week attacks from 9.5 to 7.1.

These cover the gap while verapamil, the standard preventive, takes effect: about 1.7 weeks in episodic cluster and 5 weeks in chronic. The dose that works can disturb heart rhythm. In one audit, 19 percent of patients with an electrocardiogram on file had an arrhythmia, and 41 percent of those treated had no electrocardiogram at all. Dose changes and monitoring belong with your physician.

Episodic and chronic cluster split apart on newer treatments: an anti-CGRP antibody reduced weekly attacks by 8.7 against 5.2 on placebo in episodic cluster and missed its target in chronic cluster. When everything fails, an implanted occipital nerve stimulator helps 57.3 percent in pooled series. The terrain matters too, because sleep timing and light feed the hypothalamic clock; a small, unreplicated pilot of evening melatonin helped 5 of 10 patients against none on placebo.

What to bring to your next appointment

• Can we go through the three screening questions for cluster headache?
• What is my plan for an attack that works within fifteen minutes?
• Can the home oxygen paperwork start today?
• What is the transitional treatment for this bout while a preventive takes hold?
• If I am on verapamil, when is my next electrocardiogram?

Every trial cited here, with sample sizes and limits, is laid out in the Technical Supplement to Chapter 11. Why the fast, low-risk options are the ones the system pays least for is the subject of the chapter on misaligned incentives.

Frequently asked questions

What is the fastest way to stop a cluster headache?

High-flow oxygen by face mask and injected sumatriptan are the two treatments with the strongest trial evidence for ending an attack within about fifteen minutes. Tablets are too slow for attacks that often last under an hour. A transnasal sphenopalatine ganglion block is another fast option given in a clinic, with a thinner evidence base in cluster headache. See how cluster headache is treated in our clinic.

How can I tell cluster headache from migraine?

Cluster attacks are strictly one-sided, last 15 minutes to three hours, recur up to eight times a day, and come with tearing, a runny or blocked nostril on the same side, and restlessness. Migraine sufferers usually want to lie still. Nausea and light sensitivity occur in about half of cluster patients, so they do not rule cluster out. Learn the nerve reflex behind the watering eye and blocked nostril.

What is a sphenopalatine ganglion block, and does it hurt?

It places local anesthetic on the lining of the nose over a nerve relay that carries cluster and migraine pain, using a thin catheter passed along the nostril. There is no needle through the skin and no sedation. Most people notice a bitter taste, a numb throat or a watery eye for a short time, and the procedure takes minutes. Read what to expect from a sphenopalatine ganglion block.

Why don’t opioids work for cluster headache?

The attack is usually over before an oral opioid is absorbed, and the pain is driven by a nerve reflex that opioids do not switch off. In a survey of more than two thousand patients, only 6 percent rated opioids completely or very effective, against 54 percent for oxygen. Patients given opioids after diagnosis also used emergency and hospital care more often. Our view on medication as a bridge, not a destination.

Can a nerve block end a cluster headache bout?

A steroid injection around the greater occipital nerve at the back of the head is the best-supported way to shorten a bout. In one randomized trial, 11 of 13 patients were attack-free for a week after a single injection, against none on placebo. It covers the weeks a daily preventive needs to start working. See the nerve blocks we perform and what each one targets.

Sources

1. Burish, M. J., Pearson, S. M., Shapiro, R. E., Zhang, W., & Schor, L. I. (2021). Cluster headache is one of the most intensely painful human conditions: Results from the International Cluster Headache Questionnaire. Headache, 61(1), 117–124. https://doi.org/10.1111/head.14021
2. Van Obberghen, E. K., Fabre, R., & Lanteri-Minet, M. (2025). Cluster headache diagnostic delay and its predictors: a systematic review with a meta-analysis. The Journal of Headache and Pain, 26(1), 71. https://doi.org/10.1186/s10194-025-02001-7
3. Parakramaweera, R., Evans, R. W., Schor, L. I., Pearson, S. M., Martinez, R., Cammarata, J. S., Amin, A. J., Yoo, S. H., Zhang, W., Yan, Y., & Burish, M. J. (2021). A brief diagnostic screen for cluster headache: Creation and initial validation of the Erwin Test for Cluster Headache. Cephalalgia, 41(13), 1298–1309. https://doi.org/10.1177/03331024211018138
4. Cohen, A. S., Burns, B., & Goadsby, P. J. (2009). High-flow oxygen for treatment of cluster headache: a randomized trial. JAMA, 302(22), 2451–2457. https://doi.org/10.1001/jama.2009.1855
5. Law, S., Derry, S., & Moore, R. A. (2013). Triptans for acute cluster headache. Cochrane Database of Systematic Reviews, 2013(7), CD008042. https://doi.org/10.1002/14651858.CD008042.pub3
6. Pearson, S. M., Burish, M. J., Shapiro, R. E., Yan, Y., & Schor, L. I. (2019). Effectiveness of Oxygen and Other Acute Treatments for Cluster Headache: Results From the Cluster Headache Questionnaire, an International Survey. Headache, 59(2), 235–249. https://doi.org/10.1111/head.13473
7. Choong, C. K., Ford, J. H., Nyhuis, A. W., Joshi, S. G., Robinson, R. L., Aurora, S. K., & Martinez, J. M. (2017). Clinical Characteristics and Treatment Patterns Among Patients Diagnosed With Cluster Headache in U.S. Healthcare Claims Data. Headache, 57(9), 1359–1374. https://doi.org/10.1111/head.13127
8. Costa, A., Pucci, E., Antonaci, F., Sances, G., Granella, F., Broich, G., & Nappi, G. (2000). The effect of intranasal cocaine and lidocaine on nitroglycerin-induced attacks in cluster headache. Cephalalgia, 20(2), 85–91. https://doi.org/10.1046/j.1468-2982.2000.00026.x
9. Ho, K. W. D., Przkora, R., & Kumar, S. (2017). Sphenopalatine ganglion: block, radiofrequency ablation and neurostimulation – a systematic review. The Journal of Headache and Pain, 18(1), 118. https://doi.org/10.1186/s10194-017-0826-y
10. Ambrosini, A., Vandenheede, M., Rossi, P., Aloj, F., Sauli, E., Pierelli, F., & Schoenen, J. (2005). Suboccipital injection with a mixture of rapid- and long-acting steroids in cluster headache: a double-blind placebo-controlled study. Pain, 118(1-2), 92–96. https://doi.org/10.1016/j.pain.2005.07.015

Medically reviewed by Dr. Gurpreet Singh Padda, MD, MBA, MHP, board certified in anesthesiology, pain medicine, interventional pain management, addiction medicine and obesity medicine. Adapted from Chapter 11 of The Pained Brain by Dr. Padda and Dr. KrisJay Fucanan, MD. Last reviewed September 2026.

Dr. Gurpreet Singh Padda, MD, MBA, MHP