The antibodies that reshaped migraine prevention do not clearly work on cluster headache, the condition many neurologists call the worst pain a person can feel. A meta-analysis published June 6 in the Journal of Neurology ↗ pooled four trials and 655 patients and could not find a benefit that holds up.
The drugs are CGRP monoclonal antibodies. CGRP, short for calcitonin gene-related peptide, is a 37-amino-acid peptide ↗ that trigeminal nerves release to widen blood vessels and carry pain signals. Lab-made antibodies mop it up before it can act. In migraine that approach worked well enough to launch a class of drugs (Emgality, Ajovy, Aimovig) and cut monthly headache days for a lot of people. Cluster headache runs on the same peptide, so the obvious bet was that the same antibodies would help there too.
They mostly did not. Across the four trials, patients on a CGRP antibody had 0.81 fewer cluster attacks per week than patients on placebo. That sounds like something until you see the spread. The confidence interval, the range the true effect is likely to sit in, ran from 2.24 fewer attacks to 0.62 more, which means the real effect could be a modest help or could be nothing at all (P = 0.265). Split by type, neither episodic cluster headache (brutal bouts that come for weeks then lift) nor chronic cluster headache (the version that never fully goes away) reached significance on its own.
The one drug that ever cleared a trial here is galcanezumab, sold as Emgality, the only CGRP antibody the FDA approved for episodic cluster headache. Its episodic study was positive. Fremanezumab missed in episodic cluster, and trials of both drugs in the chronic form were stopped early when interim looks showed they were not going to work. Pool the winner with the losers and the average washes out.
The placebo did a lot of the work
The more telling number is the placebo response. Between 27 and 53 percent of patients improved on the dummy injection. That is why two standard ways of counting responders disagreed. The odds of a 50 percent or greater drop in attacks were not statistically better on the drug (odds ratio 1.39, P = 0.159), but a second method more sensitive to a high baseline did cross the line (risk ratio 1.25, P = 0.017). When half of your placebo arm gets better on its own, a real drug effect is hard to see, and a borderline statistic can flip depending on how you slice it.
The authors also ran a Bayesian analysis, which estimates how probable an effect is rather than only testing it against zero. It put the chance that the antibodies do anything at all at 81 percent. It put the chance that they cut attacks by two or more per week, the kind of drop a patient would actually feel, at 17.9 percent. That gap is the whole story. Probably some effect. Probably not an effect big enough to matter.
A trial sequential analysis, which asks whether enough patients have been studied to settle the question, agreed. For a target benefit of 2.5 fewer attacks a week, the trials had already enrolled more than enough people (136 percent of the number needed) without ever crossing the line that would prove the drug works. Enrolling more patients at that effect size would not rescue it.
CGRP sits on peptidemodel as alpha-CGRP ↗, the human-type peptide these antibodies are built to neutralize. We have written before about a CGRP antibody acting in the brain in ways its label never described ↗, and the cluster-headache result lands in the same register. A peptide-targeting drug whose biology is clean and whose clinical payoff stays murkier than the mechanism promised. Cluster headache and migraine share a messenger. They do not, it turns out, share a treatment.