BACKGROUND: S44819, a selective GABA α5 receptor antagonist, reduces tonic post-ischaemic inhibition of the peri-infarct cortex. S44819 improved stroke recovery in rodents and increased cortical excitability in a transcranial magnetic stimulation study in healthy volunteers. The Randomized Efficacy and Safety Trial of Oral GABA α5 antagonist S44819 after Recent ischemic Event (RESTORE BRAIN) aimed to evaluate the safety and efficacy of S44819 for enhancing clinical recovery of patients with ischaemic stroke. METHODS: RESTORE BRAIN was an international, randomised, double-blind, parallel-group, placebo-controlled, multicentre phase 2 trial that evaluated the safety and efficacy of oral S44189 in patients with recent ischaemic stroke. The study was done in specialised stroke units in 92 actively recruiting centres in 14 countries: ten were European countries (Belgium, Czech Republic, France, Germany, Hungary, Italy, Netherlands, Poland, Spain, and the UK) and four were non-European countries (Australia, Brazil, Canada, and South Korea). Patients aged 18-85 years with acute ischaemic stroke involving cerebral cortex (National Institute of Health Stroke Scale [NIHSS] score 7-20) without previous disability were eligible for inclusion. Participants were randomly assigned to receive 150 mg S44819 twice a day, 300 mg S44819 twice a day, or placebo twice a day by a balanced, non-adaptive randomisation method with a 1:1:1 ratio. Treatment randomisation and allocation were centralised via the interactive web response system using computer-generated random sequences with a block size of 3. Blinding of treatment was achieved by identical appearance and taste of all sachets. Patients, investigators and individuals involved in the analysis of the trial were masked to group assignment. The primary endpoint was the modified Rankin Scale (mRS) score 90 days from onset of treatment, evaluated by shift analysis (predefined main analysis) or by dichotomised analyses using 0-1 versus 2-6 and 0-2 versus 3-6 cutoffs (predefined secondary analysis). Secondary endpoints were the effects of S44819 on the NIHSS and Montreal Cognitive Assessment (MoCA) scores, time needed to complete parts A and B of the Trail Making Test, and the Barthel index. Efficacy analyses were done on all patients who received at least one dose of treatment and had at least one mRS score taken after day 5 (specifically, on or after day 30). Safety was compared across treatment groups for all patients who received at least one dose of treatment. The study was registered at ClinicalTrials.gov, NCT02877615. FINDINGS: Between Dec 19, 2016, and Nov 16, 2018, 585 patients were enrolled in the study. Of these, 197 (34%) were randomly assigned to receive 150 mg S44819 twice a day, 195 (33%) to receive 300 mg S44819 twice a day, and 193 (33%) to receive placebo twice a day. 189 (96%) of 197 patients in the 150 mg S44819 group, 188 (96%) of 195 patients in the 300 mg S44819 group, and 191 (99%) patients in the placebo group received at least one dose of treatment and had at least one mRS score taken after day 5, and were included in efficacy analyses. 195 (99%) of 197 patients in the 150 mg S44819 group, 194 (99%) of 195 patients in the 300 mg S44819 group, and 193 (100%) patients in the placebo group received at least one dose of treatment, and were included in safety analyses. The primary endpoint of mRS at day 90 did not differ between each of the two S44819 groups and the placebo group (OR 0·91 [95% CI 0·64-1·31]; p=0·80 for 150 mg S44819 compared with placebo and OR 1·17 [95% CI 0·81-1·67]; p=0·80 for 300 mg S44819 compared with placebo). Likewise, dichotomised mRS scores at day 90 (mRS 0-2 vs 3-6 or mRS 0-1 vs 2-6) did not differ between groups. Secondary endpoints did not reveal any significant group differences. The median NIHSS score at day 90 did not differ between groups (4 [IQR 2-8] in 150 mg S44819 group, 4 [2-7] in 300 mg S44819 group, and 4 [2-6] in placebo group), nor did the number of patients at day 90 with an NIHSS score of up to 5 (95 [61%] of 156 in 150 mg S44819 group, 106 [66%] of 161 in 300 mg S44819 group, and 104 [66%] of 157 in placebo group) versus more than 5 (61 [39%] in 150 mg S44819 group, 55 [34%] in 300 mg S44819 group, and 53 [34%] in placebo group). Likewise, the median MoCA score (22·0 [IQR 17·0-26·0] in 150 mg S44819 group, 23·0 [19·0-26·5] in 300 mg S44819 group, and 22·0 [17·0-26·0] in placebo group), time needed to complete parts A (50 s [IQR 42-68] in 150 mg S44819 group, 49 s [36-63] in 300 mg S44819 group, and 50 s [38-68] in placebo group) and B (107 s [81-144] in 150 mg S44819 group, 121 s [76-159] in 300 mg S44819 group, and 130 s [86-175] in placebo group) of the Trail Making Test, and the Barthel index (90 [IQR 60-100] in 150 mg S44819 group, 90 [70-100] in 300 mg S44819 group, and 90 [70-100] in placebo group) were similar in all groups. Number and type of adverse events were similar between the three groups. There were no drug-related adverse events and no drug-related deaths. INTERPRETATION: There was no evidence that S44819 improved clinical outcome in patients after ischaemic stroke, and thus S44819 cannot be recommended for stroke therapy. The concept of tonic inhibition after stroke should be re-evaluated in humans. FUNDING: Servier.