Last month a 44-year-old woman sat across from me with a folder of old records and one sentence she had clearly rehearsed: “I’ve failed everything.” She had tried propranolol, amitriptyline, topiramate, candesartan, Botox and a CGRP injection. Her headaches came on 24 days a month. Her last neurologist had told her, kindly, that she might be “refractory.”
Here is the short answer I gave her, and the one I would give you: very few people are truly refractory, and most people who think they are still have something evidence-based left to try. Before accepting that label, I check a short list of things, and in my experience that list changes the plan more often than any new drug does.
“Resistant” and “refractory” are not the same thing
Doctors used to use words like “refractory,” “intractable” and “difficult to treat” interchangeably, which made the problem impossible to study. A 2020 European consensus separated two states. Resistant migraine means at least three classes of preventive medicine have failed, with at least eight debilitating headache days a month for three months in a row. Refractory migraine means all available classes have failed, over at least six months of the same burden.
Two details matter. First, “failure” includes stopping a drug because of side effects, not only because it did not work. Second, a “debilitating” day is one that stays disabling even when you used acute medicine early and at the right dose, including at least two different triptans unless they are unsafe for you. If you took your pill four hours into the attack, that day does not count the way you think it does.
My patient met the criteria for resistant migraine, not refractory. That distinction sounds small. It is not: it means the list of untried options was still long.
What I re-check before I believe “nothing works”
The commonest reason someone looks refractory is that something other than refractory migraine is going on. I go through the same review every time.
Is it really migraine? A side-locked daily headache can be hemicrania continua, which answers to indomethacin and to almost nothing else. Cluster headache is frequently mislabeled as migraine. A headache that began abruptly one day and never left may be new daily persistent headache. Constant dizziness may be vestibular migraine. I re-apply the diagnostic criteria from scratch rather than trusting the label in the chart.
Is something secondary hiding? Idiopathic intracranial hypertension (weight gain, a whooshing sound in the ears, blurred vision, so a look at the back of the eye matters), spontaneous low pressure (worse upright, better lying down), sleep apnea, and a neck source are the usual suspects. My patient had loud snoring and a BMI of 31, so we did a sleep study. It was normal, but I would have been wrong not to look. If this sounds familiar, I wrote about the intracranial pressure picture separately.
How many days a month are you taking acute medicine? She was using sumatriptan or ibuprofen on 16 days a month. That alone can keep migraine chronic, and withdrawal combined with a preventive often turns an “untreatable” picture into a treatable one. This is the single most common finding, and it is covered in my piece on how to stop rebound headaches.
Was each preventive a fair trial? An adequate trial is generally two to three months at a therapeutic dose; Botox needs two treatment cycles; CGRP antibodies need about three months, and up to six in chronic migraine when the early response is partial. Old records often show a drug stopped after three weeks at the starting dose. That was not a failed trial. See also how long a preventive really takes.
Then the quieter factors: depression and anxiety, insomnia, hormonal patterns, missed doses or injection technique, never having been offered cognitive behavioral therapy, relaxation or sleep work, and heavy or erratic caffeine.
What is left to try, even after several failures
Five large trials enrolled people in whom two to four preventive classes had already failed, testing erenumab, fremanezumab, galcanezumab, eptinezumab and atogepant. All were positive, with improvements similar to those in people who had never tried a preventive. So a person with resistant migraine who has never had a CGRP-pathway drug has not reached the end of evidence-based treatment.
If one CGRP drug has failed, failure of another is not guaranteed. Switching from a receptor antibody to one that binds the CGRP molecule itself, or the reverse, produced a meaningful response in roughly a third of people in small studies. One caution I always give: in one cohort nobody with daily headache responded to the switch, while about half of those with non-daily headache did. These data are observational, so I present them as reasonable practice, not as proven. I covered the practical steps in CGRP injections not working.
Other options with observational support include adding an antibody to Botox for chronic migraine (a chart review suggested a further three to four fewer headache days a month), and a short hospital stay when headache has become continuous. In hospital we can stop overused medication under supervision, give repeated intravenous dihydroergotamine with an anti-nausea drug (not safe with heart or vascular disease, or within 24 hours of a triptan), or use lidocaine infusions, add nerve blocks, and start a new preventive while the cycle is broken. These regimens rest on uncontrolled series, and I tell patients so.
What I do not recommend: surgery on “trigger points,” closing a PFO to treat migraine (three trials failed), or deep brain stimulation. Occipital nerve stimulation missed its main goal in the largest sham-controlled trial, and implant problems were common, so it stays a late, specialist-only option.
How it ended for my patient, and what to ask your doctor
We admitted her briefly, stopped the overused tablets, gave five days of IV dihydroergotamine with metoclopramide, performed bilateral occipital nerve blocks, switched her from the receptor antibody to a ligand antibody, kept the Botox, and started venlafaxine for depression she had never mentioned. She left with a written rescue plan for the emergency room: prochlorperazine, an occipital block, dexamethasone, and no opioids. Four months later she had 12 headache days a month, needed acute medicine on only 6, and was back at work part-time. Not cured, but a different life.
If you feel you have “failed everything,” bring your old records and ask four questions: Which preventive classes did I actually try, at what dose and for how long? Have I been counted for acute-medicine days per month? Has anyone re-checked the diagnosis and looked for a secondary cause? Have I had a CGRP-pathway drug, and if one failed, was switching considered? “There is nothing left with good evidence” is sometimes true. “There is nothing left” almost never is. If you are stuck, my guide on what doctors miss when a headache will not improve is a good place to begin.
From the textbook. Chapter 21, “Refractory and Resistant Migraine,” of my clinical textbook Headache Medicine covers this topic in full for doctors, with the evidence and decision pathways behind it. About the book →
Next steps: Take the free headache red-flags checklist to be sure nothing dangerous is being overlooked, or talk to a specialist about your history.
This article is for general education and is not a substitute for medical advice. Do not start, stop or change prescribed medicine without talking to your own doctor. Seek emergency care for a sudden, severe “worst ever” headache, weakness, speech or vision loss, fever with stiff neck, or confusion.