AZD2816 is a variant-adapted COVID-19 vaccine that expresses the full-length SARS-CoV-2 beta variant spike protein but is otherwise similar to AZD1222 (ChAdOx1 nCoV-19). This study aimed to evaluate the safety and immunogenicity of AZD1222 or AZD2816 (or both) primary-series vaccination in a cohort of adult participants who were previously unvaccinated. In this phase 2/3, randomised, multinational, active-controlled, non-inferiority, immunobridging study, adult participants previously unvaccinated for COVID-19 were enrolled at 16 study sites in Brazil, South Africa, Poland, and the UK. Participants were stratified by age, sex, and comorbidity and randomly assigned 5:5:5:2 to receive a primary series of AZD1222 (AZD1222 group), AZD2816 (AZD2816 [4-week] group), or AZD1222-AZD2816 (AZD1222-AZD2816 group) at 4-week dosing intervals, or AZD2816 at a 12-week interval (AZD2816 [12-week] group) and evaluated for safety and immunogenicity through 180 days after dose 2. Primary outcomes were safety (rates of solicited adverse events occurring during 7 days and unsolicited adverse events occurring during 28 days after each dose) and immunogenicity (non-inferiority of pseudovirus neutralising antibody geometric mean titre [GMT], GMT ratio margin of 0·67, and seroresponse rate, rate difference margin of -10%, recorded 28 days after dose 2 with AZD2816 [4-week interval] against beta vs AZD1222 against ancestral SARS-CoV-2) in participants who were seronegative at baseline. This trial is registered with ClinicalTrials.gov, NCT04973449, and is completed. Between July 7 and Nov 12, 2021, 1449 participants were assigned to the AZD1222 group (n=413), the AZD2816 (4-week) group (n=415), the AZD1222-AZD2816 group (n=412), and the AZD2816 (12-week) group (n=209). Ten (2·6%) of 378 participants who were seronegative at baseline in the AZD1222 group, nine (2·4%) of 379 in the AZD2816 (4-week) group, eight (2·1%) of 380 in the AZD1222-AZD2816 group, and 11 (5·8%) of 191 in the AZD2816 (12-week) group had vaccine-related unsolicited adverse events. Serious adverse events were recorded in one (0·3%) participant in the AZD1222 group, one (0·3%) in the AZD2816 (4-week) group, two (0·5%) in the AZD1222-AZD2816 group, and none in the AZD2816 (12-week) group. Co-primary immunogenicity endpoints were met: neutralising antibody GMT (ratio 1·19 [95% CI 1·08-1·32]; lower bound greater than 0·67) and seroresponse rate (difference 1·7% [-3·1 to 6·5]; lower bound greater than -10%) at 28days after dose 2 were non-inferior in the AZD2816 (4-week) group against beta versus in the AZD1222 group against ancestral SARS-CoV-2. Seroresponse rates were highest with AZD2816 against beta (12-week interval 94·3% [95% CI 89·4-97·3]; 4-week interval 85·7% [81·5-89·2]) and with AZD1222 (84·6% [80·3-88·2]) against ancestral SARS-CoV-2. Primary series of AZD1222 and AZD2816 were well tolerated, with no emergent safety concerns. Both vaccines elicited robust immunogenicity against beta and ancestral SARS-CoV-2 with greater responses demonstrated when testing against SARS-CoV-2 strains that matched those targeted by the respective vaccine. These findings demonstrate the continued importance of ancestral COVID-19 vaccines in protecting against severe COVID-19 and highlight the feasibility of using the ChAdOx1 platform to develop COVID-19 vaccines against future SARS-CoV-2 variants. AstraZeneca.