BACKGROUND In four randomised trials, human papillomavirus (HPV)-based screening for cervical cancer was compared with cytology-based cervical screening, and precursors of cancer were the endpoint in every trial. However, direct estimates are missing of the relative efficacy of HPV-based versus cytology-based screening for prevention of invasive cancer in women who undergo regular screening, of modifiers (eg, age) of this relative efficacy, and of the duration of protection. We did a follow-up study of the four randomised trials to investigate these outcomes. METHODS 176 464 women aged 20-64 years were randomly assigned to HPV-based (experimental arm) or cytology-based (control arm) screening in Sweden (Swedescreen), the Netherlands (POBASCAM), England (ARTISTIC), and Italy (NTCC). We followed up these women for a median of 6·5 years (1 214 415 person-years) and identified 107 invasive cervical carcinomas by linkage with screening, pathology, and cancer registries, by masked review of histological specimens, or from reports. Cumulative and study-adjusted rate ratios (experimental vs control) were calculated for incidence of invasive cervical carcinoma. FINDINGS The rate ratio for invasive cervical carcinoma among all women from recruitment to end of follow-up was 0·60 (95% CI 0·40-0·89), with no heterogeneity between studies (p=0·52). Detection of invasive cervical carcinoma was similar between screening methods during the first 2·5 years of follow-up (0·79, 0·46-1·36) but was significantly lower in the experimental arm thereafter (0·45, 0·25-0·81). In women with a negative screening test at entry, the rate ratio was 0·30 (0·15-0·60). The cumulative incidence of invasive cervical carcinoma in women with negative entry tests was 4·6 per 10(5) (1·1-12·1) and 8·7 per 10(5) (3·3-18·6) at 3·5 and 5·5 years, respectively, in the experimental arm, and 15·4 per 10(5) (7·9-27·0) and 36·0 per 10(5) (23·2-53·5), respectively, in the control arm. Rate ratios did not differ by cancer stage, but were lower for adenocarcinoma (0·31, 0·14-0·69) than for squamous-cell carcinoma (0·78, 0·49-1·25). The rate ratio was lowest in women aged 30-34 years (0·36, 0·14-0·94). INTERPRETATION HPV-based screening provides 60-70% greater protection against invasive cervical carcinomas compared with cytology. Data of large-scale randomised trials support initiation of HPV-based screening from age 30 years and extension of screening intervals to at least 5 years. FUNDING European Union, Belgian Foundation Against Cancer, KCE-Centre d'Expertise, IARC, The Netherlands Organisation for Health Research and Development, the Italian Ministry of Health.