3.4 Melatonin

3.4.1 Why melatonin at all: the circadian rationale

This is the drug with the strongest mechanistic story and the weakest evidence. That mismatch is the whole point of this section.

The circadian biology of CH

Benkli B, Kim SY, Koike N, et al. Circadian Features of Cluster Headache and Migraine: A Systematic Review, Meta-analysis, and Genetic Analysis. Neurology 2023;100(22):e2224–e2236, DOI 10.1212/WNL.0000000000207240, PMID 36990725, PMC10259280. PROSPERO CRD42021234238. 72 studies. peer-reviewed This is the definitive quantitative account:

• 70.5% (3,490/4,953) of CH patients have a circadian pattern of attacks (16 studies)
• Clear peak between 21:00 and 03:00 — the 7 consecutive hours 21:00–03:00 comprise all 7 highest hourly values
• Single most likely hour: 02:00 (6.5%, 508/8,856)
• Peak months: March, April, September, October, November (53.8%, 1,882/3,495); most common season autumn (30.7%)
• Longitudinal stability: over ≥2 headache cycles, 52.5% retained the same pattern (n=139); over ≥5 years, 78.4% retained it (n=42)

The melatonin-specific findings

From the same meta-analysis (8 studies reporting corticosteroid or melatonin data):

• Nocturnal urinary melatonin LOWER than controls in 5 studies
• Plasma melatonin LOWER than controls in 1 study
• Corticosteroid blood levels higher than controls in 3 studies, unchanged in 1
• Association between CH and CLOCK SNV rs12649507 reported — but no association with rs1801260, rs11932595 or rs12649507 in other analyses (i.e. the CLOCK finding does not replicate consistently — contested)
• Lower REV-ERBα RNA in lymphoblasts of CH patients vs controls
• No association with PERIOD3 VNTR polymorphisms

Supporting mechanism from an earlier review (Gooriah R, Buture A, Ahmed F. Ther Clin Risk Manag 2015;11:1687–1696, DOI 10.2147/TCRM.S94193, PMC4646474): peer-reviewed a suppressed nocturnal melatonin peak is seen during the active phase of CH; melatonin production is regulated by a multisynaptic pathway from the suprachiasmatic nucleus to the pineal gland; smokers with CH have lower melatonin than non-smoking CH patients; corticosteroid administration increases nocturnal urinary 6-sulfatoxymelatonin (the stable melatonin metabolite); disordered circadian rhythms of cortisol, luteinizing hormone, growth hormone and prolactin; reduced TRH stimulation; lower testosterone during attacks; and PET demonstrating highly specific hypothalamic grey matter activation in both nitroglycerine-induced and spontaneous CH.

So the rationale is genuinely strong: CH is a disorder with a hypothalamic pacemaker signature, a 02:00 attack peak, seasonal equinox-linked bouts, and measurably suppressed endogenous melatonin. Replacing melatonin is a rational hypothesis, not a wellness fad.

3.4.2 Efficacy — the positive trial

Leone M, D’Amico D, Moschiano F, Fraschini F, Bussone G. Melatonin versus placebo in the prophylaxis of cluster headache: a double-blind pilot study with parallel groups. Cephalalgia 1996;16(7):494–496, PMID 8933994, DOI 10.1046/j.1468-2982.1996.1607494.x. peer-reviewed

• n=20 (18 episodic, 2 chronic), melatonin 10 mg nightly × 14 days vs placebo, parallel groups
• Attack frequency significantly reduced, p<0.03
• Analgesic use reduced, p<0.06 (i.e. did not reach conventional significance)
• 5/10 (50%) responders on melatonin vs 0/10 on placebo
• Neither of the two chronic patients responded
• No side effects

That last-but-one point is the one that matters most for a chronic daily patient, and it is almost never quoted: in the only positive randomised trial of melatonin in CH, both chronic patients failed to respond. Confirmed independently: “melatonin rapidly alleviated cluster attacks only in episodic cluster patients in the Leone et al. report. Two chronic cluster headache patients in the trial did not respond” (Peres MFP & Rozen TD, Cephalalgia 2001, DOI 10.1046/j.1468-2982.2001.00307.x). peer-reviewed

3.4.3 Efficacy — the negative trial

Pringsheim T, Magnoux E, Dobson CF, Hamel E, Aubé M. Melatonin as adjunctive therapy in the prophylaxis of cluster headache: a pilot study. Headache 2002;42(8):787–792, DOI 10.1046/j.1526-4610.2002.02181.x, PMID 12390642. peer-reviewed

• n=9 (6 chronic, 3 episodic), single-blind, melatonin added to existing prophylaxis in patients with incomplete relief
• Chronic patients: 1 month baseline diary → 1 month melatonin → 1 month placebo. Episodic: placebo then melatonin.
• Primary endpoint mean headaches/day: no significant differences between melatonin, placebo and baseline months
• All secondary endpoints (analgesic consumption, % mild/moderate/severe headaches): no significant differences
• No side effects
• Authors’ conclusion: patients with chronic CH, or episodic CH uncontrolled on conventional therapy, “did not appear to gain therapeutically from adding melatonin to their usual treatment regimens”

The authors offer a face-saving hypothesis worth taking seriously: melatonin’s phase-shifting properties might mediate its effect in episodic CH, and treatment from the very beginning of the bout might be necessary to reset the circadian pacemaker. That is a testable claim that has never been tested. Unknown.

3.4.4 The contradictory case reports

Peres MFP & Rozen TD. Melatonin in the preventive treatment of chronic cluster headache. Cephalalgia 2001;21(10):993–995, DOI 10.1046/j.1468-2982.2001.00307.x. peer-reviewed

Two chronic CH patients, melatonin 9 mg at bedtime (9 rather than Leone’s 10 mg simply because 3 mg tablets were what was available in the US), added to verapamil.

• Case 1: 38-year-old man, 20-year history, 6–7 attacks/day despite verapamil 640 mg, headache reliably 1 hour after falling asleep, naps also triggered attacks, never a headache-free period >14 days, had failed steroid tapers, valproic acid 2000 mg, verapamil 640 mg and lithium 900 mg. On melatonin: significant improvement within 2 days, became headache-free, remained so for 6 months, sleep no longer disrupted, no side effects.
• Case 2: 40-year-old man, 8-year history, 3 attacks/day, one always 40 minutes after sleep onset. On melatonin: immediate pain relief, daytime and nocturnal headaches completely abated, headache-free for 8 months, did not want to discontinue.

The authors themselves note a placebo response or spontaneous remission could account for this, though they considered a therapeutic effect most likely, and called for a double-blind placebo-controlled trial in chronic CH. That trial has still not been done, 25 years later.

The conflict is direct and unresolved. Leone 1996: 2/2 chronic patients failed. Pringsheim 2002: 6 chronic patients, no benefit. Peres & Rozen 2001: 2/2 chronic patients became headache-free within days. n=10 chronic patients total across the entire randomised-ish literature. This is not enough evidence to conclude anything. Unknown.

3.4.5 Dosing and safety

• Table dose 10 mg; evidence level (+) = questionable; adverse events daytime sleepiness, headache, dizziness, hypothermia; contraindications depression, coagulation disorders; “very good tolerability” (PMC8748342). peer-reviewed
• Australian Prescriber: Grade 2C; start 4 mg, titrate to 8–10 mg (Aust Prescr 2022;45:15–20). peer-reviewed Note: in Australia, 2 mg modified-release melatonin (Circadin) is the registered product; 10 mg immediate-release generally requires compounding or import, which is a practical access wrinkle.
• Recommended as an option for patients who don’t tolerate lithium or verapamil (PMC8748342).

The honest summary of melatonin’s appeal: it is cheap, essentially free of serious side effects across every study cited above, and mechanistically well-motivated. It is also probably not very effective, and probably least effective in exactly the population reading this. Its main virtue is a superb risk-benefit ratio at low expected benefit — a reasonable add-on, a poor primary strategy.

3.4.6 Community view — the sharpest divergence in the chapter

Melatonin sits in the LOW self-reported efficacy clade in the largest citizen-science synthesis available, grouped with valproic acid, gabapentin, amitriptyline and propranolol (Rusanen et al. 2022, PMC9436841). citizen science

This is a notable divergence: melatonin has a positive randomised trial (Leone 1996) and yet patients rank it alongside drugs with no evidence at all. Three readings, all plausible: (a) patients are right and the 1996 pilot was a small-n false positive; (b) patients are mostly chronic or treatment-refractory, exactly the group Leone’s trial showed doesn’t respond; (c) patients are using 3–5 mg supermarket melatonin, not 10 mg at a fixed nightly time, and are effectively testing a different intervention. I cannot distinguish these from available data. Contested.