The Watering Eye and the Blocked Nostril: The Nerve Reflex Behind Cluster Headache and Migraine

What this video covers

• Why cluster headache and migraine involve a trigeminal-autonomic reflex, not just blood vessels
• Where the sphenopalatine ganglion sits in the pterygopalatine fossa and why it is reachable through the nose
• How transnasal, transoral, and image-guided approaches differ in precision and risk
• What the randomized data actually show, including one trial that missed its primary endpoint
• How long relief typically lasts — hours to days per block, sometimes weeks with a series — and why it is a bridging option, not a cure
• The material risks: bitter taste, palate numbness, nosebleed, nausea, fainting, and — with needle approaches — cheek numbness, bleeding from the maxillary artery, infection, temporary double vision, and rarely a seizure from anesthetic entering a vessel
• MEDICAL DISCLAIMER: This content is for educational purposes only and is not medical advice. It does not substitute for professional diagnosis or treatment. Always consult a licensed healthcare provider regarding your condition. Viewing this video does not establish a doctor-patient relationship.

It is one forty in the morning and you are awake again. Fourth night running, same side, behind the same eye. The pain does not throb — it bores. The eye streams. The nostril closes. The lid droops. You pace, because stillness is unbearable. Or your version is a one-sided migraine with a watering eye and a blocked nostril that everyone keeps calling sinus.

That combination — pain on one side, plus a watering eye and a closing nostril on that same side — is not incidental, and it is not your sinuses. It is a reflex, wired into a bony pocket behind your nose. This article explains what that reflex is, why it produces exactly the symptoms you feel, why it gets mistaken for sinus disease for years at a time, and what the published evidence honestly supports about quieting it.

The junction box behind your nose

Behind your nose, on each side, sits a narrow space called the pterygopalatine fossa — an inverted pyramid tucked in behind the maxillary sinus. Inside it is the sphenopalatine ganglion, sometimes called the pterygopalatine ganglion. It is roughly the size of a pea.

Call it a junction box rather than a nerve, because that is what it does. Two different cable bundles converge there:

• Sensory fibers from the maxillary division of the trigeminal nerve pass through it.
• Parasympathetic fibers arrive by way of the greater petrosal nerve, having come from the brainstem.

Here is the loop that matters. Pain traffic from the face enters the brainstem. The brainstem drives a parasympathetic limb of that reflex. That limb relays through this ganglion, and its outgoing fibers fire onto your tear gland and the lining of your nose, releasing vasoactive intestinal peptide and PACAP (pituitary adenylate cyclase-activating polypeptide) at the nerve endings.

That is the whole circuit: face → brainstem → ganglion → eye and nose.

Why the eye streams and the nostril closes

Now put the mechanism next to what you actually experience.

The streaming eye is the tear gland being switched on. The blocked, running nostril is the nasal lining being switched on. Neither one is infection, allergy, or inflamed sinus tissue. They are the efferent output of a reflex that your pain turned on — which is precisely why they arrive with the attack, sit on the same side as the attack, and leave when the attack leaves. Sinus disease does not keep that kind of schedule.

Two things you may have been told are wrong, and both matter.

First, the drooping eyelid and small pupil of a cluster attack are not parasympathetic overflow from this ganglion. They are a partial Horner sign — a sympathetic finding, the other autonomic system entirely. Attributing them to the sphenopalatine ganglion is a common and understandable error, but it is still an error.

Second, the vasodilation so often blamed for the pain is a correlate, not the engine. Blood vessels do widen during an attack. That widening rides along with the attack rather than causing it. The signal comes first.

This has a consequence you should hear before anyone reaches for a needle: quieting that ganglion can interrupt the autonomic limb of the reflex and the fibers passing through it — but it does not switch off the brainstem generator that started the attack. Every honest claim about this procedure follows from that one sentence.

Why this gets treated as a sinus problem for years

Consider what happens to a person whose headache arrives with a running nose and a blocked nostril. They get sent to the wrong organ, and they stay there.

A blocked nostril, a running eye, pain over one cheek — to a clinician working an eleven-minute slot from a template, that reads as sinus disease. Antibiotics follow. A scan is ordered and reads normal. The published case literature includes people who had sinus surgery, then more sinus surgery, then a tooth pulled — for a reflex that lives in neither the sinus nor the tooth.

The single most informative fact about you is usually the one nobody asks about: the tearing eye, and whether it is on the same side as the pain, and whether it comes and goes with the attack. That is not a hard question. It is a question that requires someone to have time.

None of this means the clinicians involved were careless. What a template cannot do is name a trigeminal-autonomic reflex, because nobody asked the question that would have surfaced it.

Before anything else: danger, then pattern

There is an order of operations here, and skipping it is how people get injected for the wrong problem.

Rule out the dangerous causes first

These are worked up urgently, not treated with an elective block:

• Thunderclap onset — the worst headache of your life, peaking within seconds. That is an emergency tonight.
• Fever with a stiff neck.
• A first severe headache after age fifty.
• Any focal neurologic deficit — weakness, numbness, visual or speech change.
• Any headache after head trauma.

Then separate the primary headaches

The one-sided headaches that come with autonomic signs are a family — the trigeminal autonomic cephalalgias — and they are not interchangeable. Getting this wrong is not academic. Two members of the family are ended by a tablet.

• Cluster headache. Untreated attacks run 15 to 180 minutes, up to eight a day, one-sided, around or behind the eye, with restlessness — you pace rather than lie down.
• Migraine. Attacks run 4 to 72 hours, and you want to lie still.
• Paroxysmal hemicrania. Imitates short cluster attacks, but responds absolutely to indomethacin.
• Hemicrania continua. One-sided pain with the same autonomic signs that never fully switches off — and it too responds absolutely to indomethacin. Miss it and you have injected a headache that a tablet would have stopped.
• SUNCT and SUNA. Stabbing attacks lasting seconds to minutes, dozens or hundreds a day, responding to sodium-channel drugs such as lamotrigine — not to this ganglion.

The pattern is the diagnosis, and much of the pattern lives in information only you can collect. Start an attack diary tonight and write down the hour.

A block is not first-line, and nobody should pretend otherwise

If nobody has put you on high-flow oxygen through a non-rebreather mask, offered an injectable triptan, or started a genuine preventive, then the gap in your care sits upstream of any procedure. Fix that first.

A sphenopalatine ganglion block is for the person who has already done those things and is still losing their nights.

Three routes in, three different risk profiles

If a block is appropriate, understand that “the block” is really three different procedures.

Transnasal is the least invasive, and there are two versions that are not interchangeable. The older approach soaks a cotton-tipped applicator in anesthetic and parks it against the mucosa — cheap, common, and with no placebo-controlled data behind it. The version actually tested in both randomized trials used a small intranasal applicator delivering a measured 0.3 cc of 0.5% bupivacaine per side, past the posterior tail of the middle turbinate, where roughly a millimeter of mucosa separates the drug from the ganglion. No needle enters tissue with either version. You are entitled to be told which one you are getting.

Transoral passes through the greater palatine foramen in the palate.

Percutaneous enters below the cheekbone through the coronoid notch under live fluoroscopy, with contrast confirming spread. It is the most precise route and carries the highest risk, because the maxillary artery shares that space. Procedural detail on how each route is performed and what the visit involves is on the sphenopalatine ganglion block treatment page.

One disclosure belongs here, not in the fine print: no local anesthetic is FDA-approved for treating cluster headache or migraine at this ganglion, and the intranasal applicators are cleared to deliver medication rather than for a headache indication. This use is off-label, and you should hear that word before you consent, not after.

Patients also ask about diet and light. A ketogenic dietary pattern, omega-3-derived pro-resolving mediators, and 660 to 850 nanometer light are investigational in cluster headache and migraine — no randomized trial shows that any of them shortens a bout or aborts an attack. They can reasonably be tracked in a diary. They should not be allowed to displace oxygen, a triptan, or a tested preventive.

What the trials actually found — including the one that failed

This is the section most brochures leave out.

A small placebo-controlled pilot in chronic migraine showed promise, not proof. Cady and colleagues published a double-blind, placebo-controlled pilot in Headache in 2015: adults with chronic migraine received repeated transnasal bupivacaine or saline. Pain scores were lower with bupivacaine at 15 minutes, 30 minutes, and 24 hours. But 38 patients finished, allocated 2:1 — so the entire placebo comparison rests on 12 people — and it was run with a single commercial intranasal device. The authors’ own word was promise. Note also that the HIT-6 headache-impact improvement in that paper was a within-group change in the bupivacaine arm; the paper reports no between-group comparison, so no between-group superiority on that measure may be claimed. (Investigational-grade signal, not established efficacy.)

The 87-patient emergency department trial missed its primary endpoint. Schaffer and colleagues, in Annals of Emergency Medicine the same year, randomized 87 emergency department adults with undifferentiated acute headache — not criteria-confirmed migraine, not cluster. It is not a bigger version of the migraine pilot; it is a different population entirely. The primary endpoint failed: 48.8% versus 41.3% reached 50% pain relief at 15 minutes, an absolute difference of 7.5% with a confidence interval of −13% to +27.1% — an interval that crosses zero. Its 24-hour advantage was a secondary, multiplicity-unadjusted outcome and does not rescue a negative primary result. (Negative trial. It is reported here because it is real.)

The systematic review’s headline is narrower than it sounds. Ho and colleagues, in the Journal of Headache and Pain in 2017, reviewed 83 publications and found only 23 above case-series level. Their conclusion that the strongest evidence at this target lies in cluster headache is accurate — but it pools blocks with radiofrequency ablation and implanted neurostimulation, three different interventions with three different risk profiles, and the authors themselves note that the controlled studies were small and without replication. (Contested. Do not read a pooled statement as a block-only efficacy claim.)

Now the uncomfortable sentence, said plainly: there is no randomized, placebo-controlled trial of a sphenopalatine ganglion block, by itself, in criteria-confirmed cluster headache. Not one. The only placebo-controlled block data in a defined headache population is that 38-patient migraine pilot. Everything else is uncontrolled case series, or ablation and stimulator trials — different procedures, not evidence for an applicator.

The honest grade for this body of evidence is low. A block can still be a reasonable option for a carefully selected patient, offered with that grade stated out loud. When it helps, relief typically runs hours to days from a single block, and a series may extend that to weeks — but that duration comes from clinical experience and uncontrolled case series rather than from any controlled trial. It is a bridge that buys a window. It is not a cure.

Risks, and who should wait

Common and temporary: bitter taste, a numb palate, nosebleed, nausea, and a vasovagal drop in blood pressure.

Added by needle approaches: cheek numbness, maxillary artery puncture with hematoma, rarely bleeding behind the eye or double vision, infection, and seizure if anesthetic enters a blood vessel.

Deferred with: active nasal or sinus infection, therapeutic anticoagulation, or known local anesthetic allergy.

An honest picture of a real course

What follows is a composite — a picture assembled from many patients with this condition, not one person’s chart.

A man in his forties. Bouts arriving every spring and every autumn, several attacks a night, always the same side — and behind him a decade of treatment aimed at his sinuses and his teeth.

What helped first was not a procedure. It was a diagnosis, and a diary that made the timing undeniable within two weeks. What went in first was oxygen — a tank and a non-rebreather mask at high flow, which nobody had handed him in that entire decade — plus an injectable abortive and a preventive titrated against an electrocardiogram. The block came afterward, as a bridge while the preventive reached dose.

His attacks shortened. Not vanished. Shortened. The first block did almost nothing he could measure. The second, better placed, bought about two days. He had a nosebleed. The bout still ran most of its course. Had three blocks changed nothing durable, the right move would have been to stop rather than escalate.

This is a practice observation rather than a trial result. Individual results vary.

Frequently asked questions

Key takeaways

• The watering eye and blocked nostril of a one-sided headache are the output of a trigeminal-autonomic reflex relaying through the sphenopalatine ganglion behind your nose — not sinus disease.
• The drooping lid and small pupil of a cluster attack are a partial Horner sign, which is sympathetic, and the vasodilation of an attack is a correlate rather than its cause.
• A block can interrupt the autonomic limb of that reflex, but it does not switch off the brainstem generator — which is exactly why relief runs hours to days rather than permanently.
• The evidence is graded low: one 38-patient placebo-controlled migraine pilot, one negative 87-patient emergency department trial in undifferentiated headache, and no placebo-controlled block trial in criteria-confirmed cluster headache. The use is off-label.
• Oxygen, an injectable triptan, and a real preventive come before any procedure — and dangerous causes of headache get worked up, never injected.

Medically reviewed by Gurpreet Singh Padda, MD — Board Certified in Anesthesiology, Pain Medicine, Interventional Pain Management, Addiction Medicine, and Obesity Medicine. Last reviewed July 2026.

This article is educational and is not a substitute for evaluation, diagnosis, or treatment by a physician. Individual results vary. Do not start, stop, or change any medication without consulting your physician. To have a one-sided headache with autonomic features properly evaluated before any procedure is scheduled, call (314) 481-5000 or text (314) 886-5902.

References

1. Cady R, et al. A double-blind, placebo-controlled study of repetitive transnasal sphenopalatine ganglion blockade with Tx360 as acute treatment for chronic migraine. Headache. 2015;55(1):101-16. PMID 25338927. PubMed
2. Schaffer JT, et al. Noninvasive sphenopalatine ganglion block for acute headache in the emergency department: a randomized placebo-controlled trial. Ann Emerg Med. 2015. PMID 25577713. PubMed
3. Ho KWD, et al. Sphenopalatine ganglion: block, radiofrequency ablation and neurostimulation – a systematic review. J Headache Pain. 2017. PMID 29285576. PubMed

In an active cluster bout?

A sphenopalatine ganglion block takes only a few minutes, uses no needle through the skin and no sedation, and is often used during a bout to break the cycle. The Padda Institute offers same-day and emergency visits — no referral needed, and you do not need to be an existing patient. Call (314) 481-5000 or text (314) 886-5902.

Dr. Gurpreet Singh Padda, MD, MBA, MHP