Background: Stent retriever thrombectomy of large-vessel occlusion results in better outcomes than medical therapy alone. Alternative thrombectomy strategies, particularly a direct aspiration as first pass technique, while promising, have not been rigorously assessed for clinical efficacy in randomised trials. We designed COMPASS to assess whether patients treated with aspiration as first pass have non-inferior functional outcomes to those treated with a stent retriever as first line. Methods: We did a multicentre, randomised, open label, blinded outcome, core lab adjudicated non-inferiority trial at 15 sites (ten hospitals and four specialty clinics in the USA and one hospital in Canada). Eligible participants were patients presenting with acute ischaemic stroke from anterior circulation large-vessel occlusion within 6 h of onset and an Alberta Stroke Program Early CT Score of greater than 6. We randomly assigned participants (1:1) via a central web-based system without stratification to either direct aspiration first pass or stent retriever first line thrombectomy. Those assessing primary outcomes via clinical examinations were masked to group assignment as they were not involved in the procedures. Physicians were allowed to use adjunctive technology as was consistent with their standard of care. The null hypothesis for this study was that patients treated with aspiration as first pass achieve inferior outcomes compared with those treated with a stent retriever first line approach. The primary outcome was non-inferiority of clinical functional outcome at 90 days as measured by the percentage of patients achieving a modified Rankin Scale score of 0–2, analysed by intent to treat; non-inferiority was established with a margin of 0·15. All randomly assigned patients were included in the safety analyses. This trial is registered at ClinicalTrials.gov, number: NCT02466893. Findings: Between June 1, 2015, and July 5, 2017, we assigned 270 patients to treatment: 134 to aspiration first pass and 136 to stent retriever first line. A modified Rankin score of 0–2 at 90 days was achieved by 69 patients (52%; 95% CI 43·8–60·3) in the aspiration group and 67 patients (50%; 41·6–57·4) in the stent retriever group, showing that aspiration as first pass was non-inferior to stent retriever first line (pnon-inferiority=0·0014). Intracranial haemorrhage occurred in 48 (36%) of 134 in the aspiration first pass group, and 46 (34%) of 135 in the stent retriever first line group. All-cause mortality at 3 months occurred in 30 patients (22%) in both groups. Interpretation: A direct aspiration as first pass thrombectomy conferred non-inferior functional outcome at 90 days compared with stent retriever first line thrombectomy. This study supports the use of direct aspiration as an alternative to stent retriever as first-line therapy for stroke thrombectomy. Funding: Penumbra.