Four migraine days a month is now the point at which doctors should offer preventive medication. That single number sits at the centre of a joint practice guideline published on 31 August 2026 by the American Academy of Neurology and the American Headache Society.[1]
It is the first full revision of their migraine prevention advice since 2012, and it arrives after a decade in which an entirely new drug class — the CGRP blockers — moved from experiment to everyday prescription.[2]
Who should be offered prevention
The panel makes three recommendations about starting treatment, all at evidence Level B. Clinicians should tell every migraine patient that effective preventive treatments exist for frequent attacks.[1]
Prevention should be offered to anyone with four or more migraine days per month, or four or more moderate-to-severe headache days per month. It should also be offered to anyone whose migraine causes substantial disability, regardless of the raw count.[1]
That second clause matters. A person with three ferocious attacks that wipe out entire days may need prevention more than someone with six mild ones, and the guideline now says so in plain language.
No single drug wins, so four properties decide
The guideline is unusually blunt on one point: no preventive medication has been shown to be clearly superior to the others for efficacy.[1]
Instead of ranking drugs, it asks clinicians to compare them across four properties — strength of the efficacy evidence, tolerability, long-term safety, and cost — and to weigh those against what the patient in front of them actually cares about.[1]
Where efficacy is the priority, the episodic migraine list runs atogepant, eptinezumab, erenumab, fremanezumab, galcanezumab, propranolol, topiramate and valproate. For chronic migraine, onabotulinumtoxinA replaces propranolol.[1]
Where tolerability is the priority, the list narrows to atogepant and the four CGRP monoclonal antibodies, with onabotulinumtoxinA added for chronic migraine. These are the newer agents, and they cause fewer of the side effects that make people quit.[1]
Where a patient is more worried about unknown long-term harms, the guideline points the other way — to propranolol and topiramate for episodic migraine, drugs with decades of accumulated use behind them.[1]

Pregnancy, older age and other special situations
Some of the strongest language in the document concerns reproductive planning. Patients of childbearing potential must be warned about fetal risk, and drugs with known teratogenic effects — divalproex sodium and topiramate — should be avoided where possible.[1]
During pregnancy the guideline puts non-drug approaches first: behavioural treatment, acupuncture, exercise and trigger management should be maximised before medication is considered. If a preventive is judged necessary, nifedipine is the first suggestion.[1]
For older adults, clinicians are told to check for vascular disease, drug interactions and reduced kidney or liver function, and to discuss the risk of low blood pressure and of sedation or confusion before prescribing.[1]
Two specific warnings apply to women. Valproic acid raises the risk of polycystic ovary syndrome, and topiramate above 200 mg a day can make hormonal contraception less reliable.[1]
When the painkillers themselves become the problem
Taking acute migraine medicine too often can generate headaches of its own, a pattern called medication-overuse headache. The guideline recommends offering preventive treatment to these patients rather than simply telling them to cut back.[1]
It also names the agents with actual evidence in this group: CGRP monoclonal antibodies, atogepant, onabotulinumtoxinA and topiramate. These should be tried before options that lack such evidence.[1]
Give it eight to twelve weeks before deciding
One recommendation is likely to change more conversations than any drug name. Clinicians should wait at least eight to twelve weeks at the tolerated dose before judging whether a preventive works, and twenty-four weeks for onabotulinumtoxinA.[1]
Response should be tracked with something more reliable than memory — a headache diary, or a scored tool such as HIT-6 or MIDAS. Attack frequency, severity, associated symptoms, quality of life and acute medication use all count.[1]
If there is no benefit at eight weeks, the dose should first be pushed to the maximum tolerated level. Only after twelve to twenty-four weeks of a fair trial does the guideline suggest switching to a different medication.[1]
Stopping is not automatic either. After six months of successful treatment the panel asks clinicians to discuss tapering, while warning that limited evidence points to more headache days and worse headache-related quality of life after discontinuation.[1]
What the guideline does not settle
The systematic review behind these recommendations covered evidence published through June 2024, so trials reported since then are not reflected in the grades.[3]
More importantly, almost all of the underlying trials compared a drug against placebo rather than against each other. That is precisely why the panel could not crown a winner — the head-to-head studies that would settle the question have largely not been done.[3]
Most recommendations carry a Level B grade, which signals moderate rather than high confidence. The guideline also says nothing about non-drug prevention outside pregnancy, because its remit was pharmacologic treatment only.[1]
| Aspect | Finding |
|---|---|
| Publication | Joint AAN/AHS practice guideline released 31 August 2026 in Neurology; first full update since 2012 |
| Threshold to treat | 4 or more migraine days per month, or 4 or more moderate-to-severe headache days per month (Level B) |
| Chronic migraine | Headache on 15 or more days per month for over 3 months, with migraine features on at least 8 of those days |
| Best efficacy evidence | Atogepant, eptinezumab, erenumab, fremanezumab, galcanezumab, propranolol, topiramate, valproate (episodic migraine) |
| Best tolerated | Atogepant and the four CGRP monoclonal antibodies; onabotulinumtoxinA added for chronic migraine |
| Longest safety record | Propranolol and topiramate for episodic migraine; onabotulinumtoxinA and topiramate for chronic migraine |
| Fair trial period | 8 to 12 weeks at the tolerated dose before judging efficacy; 24 weeks for onabotulinumtoxinA |
| Pregnancy | Avoid divalproex sodium and topiramate where possible (Level A); nifedipine is the first pharmacologic suggestion (Level C) |
| Evidence cut-off | Systematic review covered literature through June 2024; most recommendations graded Level B |
References
- Potrebic S, Tanveer S, Becker WJ, et al. Pharmacologic Treatment for Migraine Prevention in Adults Practice Guideline Recommendations. Neurology. 2026;107(7):e214881 (American Academy of Neurology and American Headache Society, 31 August 2026)
- AAN and AHS issue updated guideline on migraine prevention medications for adults. American Academy of Neurology press release, 31 August 2026
- Pringsheim T, Smith DB, Tanveer S, et al. Systematic Review of Pharmacologic Treatment for Migraine Prevention in Adults. Neurology. 2026;107(7):e218112
- Meglio M. American Academy of Neurology, American Headache Society Issue Updated Guideline on Migraine Prevention. NeurologyLive, 2 September 2026



