Non-response in chronic migraine: separating residual afferent drive from drive-independent persistence

INTRODUCTION: Around a third of patients with chronic migraine do not benefit from any available preventive drug, and the same predictors recur whether the failed agent was a calcitonin gene-related peptide (CGRP) ligand antibody, a receptor antagonist, onabotulinumtoxinA or candesartan. AREAS COVERED: PubMed and ClinicalTrials.gov were searched from 2015 to 2026 for non-response, resistant and refractory migraine and for PACAP, amylin and potassium-channel targets, with reference lists hand-searched. Two explanations come first, neither biological: a percentage-reduction endpoint behaves differently at ceiling headache frequency, and acute medication overuse can produce failure of every class. What survives is compatible with two mechanisms rarely distinguished: afferent drive continuing through a channel the drug does not block, which PACAP and amylin make concrete, or persistence of the central pain state after drive has fallen. Allodynia measured between attacks rather than during them separates responders from non-responders, a first handle on the second mechanism. EXPERT OPINION: Emerging targets are assigned to one mechanism or the other, with the result that would overturn each. Trials of PACAP-directed and amylin-directed agents recruited patients with prior failure of any class rather than of the CGRP pathway, leaving the claim most often made for these drugs untested.