Safety and efficacy of occipital nerve stimulation for attack prevention in medically intractable chronic cluster headache (ICON): a randomised, double-blind, multicentre, phase 3, electrical dose-controlled trial

Background: Occipital nerve stimulation (ONS) has shown promising results in small uncontrolled trials in patients with medically intractable chronic cluster headache (MICCH). We aimed to establish whether ONS could serve as an effective treatment for patients with MICCH. Methods: The ONS in MICCH (ICON) study is an investigator-initiated, international, multicentre, randomised, double-blind, phase 3, electrical dose-controlled clinical trial. The study took place at four hospitals in the Netherlands, one hospital in Belgium, one in Germany, and one in Hungary. After 12 weeks’ baseline observation, patients with MICCH, at least four attacks per week, and history of being non-responsive to at least three standard preventive drugs, were randomly allocated (at a 1:1 ratio using a computer-generated permuted block) to 24 weeks of occipital nerve stimulation at either 100% or 30% of the individually determined range between paraesthesia threshold and near-discomfort (double-blind study phase). Because ONS causes paraesthesia, preventing masked comparison versus placebo, we compared high-intensity versus low-intensity ONS, which are hypothesised to cause similar paraesthesia, but with different efficacy. In weeks 25–48, participants received individually optimised open-label ONS. The primary outcome was the weekly mean attack frequency in weeks 21–24 compared with baseline across all patients and, if a decrease was shown, to show a group-wise difference. The trial is closed to recruitment (ClinicalTrials.gov NCT01151631). Findings: Patients were enrolled between Oct 12, 2010, and Dec 3, 2017. We enrolled 150 patients and randomly assigned 131 (87%) to treatment; 65 (50%) patients to 100% ONS and 66 (50%) to 30% ONS. One of the 66 patients assigned to 30% ONS was not implanted and was therefore excluded from the intention-to-treat analysis. Because the weekly mean attack frequencies at baseline were skewed (median 15·75; IQR 9·44 to 24·75) we used log transformation to analyse the data and medians to present the results. Median weekly mean attack frequencies in the total population decreased from baseline to 7·38 (2·50 to 18·50; p<0·0001) in weeks 21–24, a median change of –5·21 (–11·18 to –0·19; p