Migraine patients who started a new class of preventive drug had no more strokes than patients who started Botox injections. If anything, they had slightly fewer, though the numbers were too small to be sure.
That is the finding of a real-world safety study published July 2 in Cephalalgia ↗, which compared the two most common preventive treatments for migraine on the question that worries doctors most about drugs acting on blood vessels: do they raise the risk of stroke.
What the drugs are
The newer class is a group of antibodies aimed at calcitonin gene-related peptide, or CGRP, a small signaling peptide that dilates blood vessels and carries pain signals during a migraine attack. Four of these drugs are on the market: erenumab, fremanezumab, galcanezumab, and eptinezumab. Some block the peptide itself, one blocks its receptor. Because CGRP relaxes blood vessels, blocking it carries a theoretical risk. Less dilation could, in principle, tip a vulnerable brain toward an ischemic stroke, the kind caused by a blocked vessel rather than a bleed. Regulators flagged that concern when the drugs launched, and it has hung over the class ever since.
The comparison drug was onabotulinumtoxinA, better known as Botox, which has been used to prevent chronic migraine since 2010 and does not act on the CGRP pathway. Using an already-accepted treatment as the yardstick, rather than untreated patients, is the cleaner design here. People who get any migraine prevention differ from people who get none, and comparing a new drug against no treatment smuggles those differences into the result.
What they found
The team drew on the NIH All of Us Research Program, a diverse national research database, and identified adults with migraine who started one of the two treatments between January 2018 and September 2023. The main comparison held 1,581 people who started a CGRP antibody and 947 who started Botox. They tracked ischemic stroke and transient ischemic attack, the brief warning-stroke that resolves on its own, counting only events that happened more than 90 days after a patient started treatment.
There were 14 stroke or warning-stroke events in the CGRP group and 22 in the Botox group. After statistical weighting to make the two groups look alike across 23 baseline characteristics, the hazard ratio came out at 0.52 (95 percent confidence interval 0.26 to 1.05). In plain terms, the CGRP patients were about half as likely to have one of these events, and the range of statistical uncertainty stopped just short of the no-difference mark. The result cleared the study's prespecified bar for non-inferiority. The upper edge of that uncertainty range stayed below a hazard ratio of 1.5. That 1.5 was the threshold the authors set in advance, the largest increase in risk they would tolerate before calling the drug worse.
The honest reading is narrow. The study was built to answer one question, whether starting a CGRP antibody carries a large jump in stroke risk versus Botox, and the answer is no. It was not built to prove the two are equally safe, and it cannot rule out a small harm. Thirty-six events total is a thin foundation, and the authors say so.
The checks that make it credible
Two design choices are worth naming. The authors ran a negative control: fractures, an outcome that neither drug should affect. If healthier people were quietly sorting into one group, you would expect the fracture rate to differ too. It did not (hazard ratio 1.18, not statistically significant), which argues against the hidden-healthy-patient bias that plagues this kind of database study. And the broader measure of major cardiovascular events, which folds in heart attacks and cardiovascular death, landed right at no difference (hazard ratio 1.07) and was inconclusive for non-inferiority. The reassuring signal is specific to stroke, not a blanket claim about the heart.
CGRP itself is one of the peptides peptidemodel tracks as a target, and the tension in this paper is the peptide biology made clinical. A molecule that relaxes blood vessels is a good thing to block for migraine and a plausible thing to worry about for stroke. This study is the first sizeable head-to-head to put a number on that worry, and the number is small. Larger studies with more events will decide whether it stays that way.