Background: High-risk human papillomavirus (hrHPV) testing on self-collected samples has potential as a primary screening tool in cervical screening, but real-world evidence on its accuracy in hrHPV-based screening programmes is lacking. Methods: In the Netherlands, women aged 30–60 years invited for cervical screening can choose between sampling at the clinician's office (Cervex Brush) or self-sampling at home (Evalyn Brush). HrHPV testing is performed using Roche Cobas 4800. We collected screening test results between January 2017 and March 2018 and histological follow-up until August 2019. The main outcome measures were mean cycle threshold (Ct) value, cervical intraepithelial neoplasia (CIN) grade 3 or cancer (CIN3+) and CIN grade 2 or worse (CIN2+). Findings: 30,808 women had a self-collected and 456,207 had a clinician-collected sample. In hrHPV-positive women with adequate cytology, Ct values were higher for self-collection than clinician-collection with a mean Ct difference of 1·25 (95% CI 0·98–1·52) in women without CIN2+, 2·73 (1·75–3·72) in CIN2 and 3·59 (3·03–4·15) in CIN3+. The relative sensitivity for detecting CIN3+ was 0·94 (0·90–0·97) for self-collection versus clinician-collection and the relative specificity was 1·02 (1·02–1·02). Interpretation: The clinical accuracy of hrHPV testing on a self-collected sample for detection of CIN3+ is high and supports its use as a primary screening test for all invited women. Because of the slightly lower sensitivity of hrHPV testing on a self-collected compared to a clinician-collected sample, an evaluation of the workflow procedure to optimise clinical performance seems warranted. Funding: National Institute for Public Health and the Environment (the Netherlands) and the European Commission.