Licensed mRNA COVID-19 vaccines require booster doses to sustain SARS-CoV-2-specific responses, creating the need for novel, broadly immunogenic vaccines. We aimed to compare the immunogenicity, safety, and tolerability of ARCT-154-a self-amplifying mRNA vaccine against SARS-CoV-2 D614G variant-with the BNT162b2 (Comirnaty; Pfizer-BioNTech) mRNA vaccine when administered as a fourth-dose booster. This double-blind, multicentre, randomised, controlled, phase 3, non-inferiority trial, conducted at 11 outpatient clinical sites in Japan, enrolled healthy adults aged at least 18 years who had previously been immunised with two doses of an mRNA COVID-19 vaccine (BNT162b2 or mRNA-1273 [Spikevax; Moderna]) followed by a third dose of BNT162b2 at least 3 months before enrolment. Participants were randomly assigned, in a 1:1 ratio using an Interactive Response Technology system with a block size of four, and with stratification by age (18-64 years or ≥65 years) and by interval since last COVID-19 vaccination (<5 months or ≥5 months), to receive either ARCT-154 or BNT162b2 as a fourth-dose booster via deltoid intramuscular injection. Participants and investigators assessing outcomes were masked to group assignment. The primary objective, measured in per-protocol set 1 (consisting of participants with no evidence of previous SARS-CoV-2 infection who received their intended injection according to protocol), was to show that the immune response 28 days after the ARCT-154 vaccine was non-inferior to that of the BNT162b2 vaccine, measured in terms of both pseudovirus neutralising antibody geometric mean titre (GMT) ratios and seroresponse rates against the wild-type Wuhan-Hu-1 strain of SARS-CoV-2. Non-inferiority was declared when the lower limit of the 95% CI of the ARCT-154 to BNT162b2 GMT ratio exceeded 0·67, and when the lower limit for the difference in seroresponse rates exceeded -10%. Key secondary endpoints included the immune response against the omicron BA.4/5 subvariant, which was assessed for non-inferiority and superiority in per-protocol set 1. Safety was assessed in the full analysis set. This study was registered on the Japan Registry for Clinical Trials, jRCT 2071220080, and is ongoing. Between Dec 13, 2022, and Feb 25, 2023, we enrolled and randomly assigned 828 participants to receive ARCT-154 (n=420) or BNT162b2 (n=408) vaccines as a fourth-dose booster. In per-protocol set 1, the GMTs of surrogate neutralising antibodies induced against the Wuhan-Hu-1 SARS-CoV-2 strain in the ARCT-154 group (5641 [95% CI 4321-7363]) were non-inferior to those in the BNT162b2 group (3934 [2993-5169]) when measured at 28 days after boosting, with a GMT ratio of 1·43 (95% CI 1·26-1·63). Seroresponse rates were 65·2% (95% CI 60·2-69·9) in the ARCT-154 group versus 51·6% (46·4-56·8) in the BNT162b2 group, a difference of 13·6% (95% CI 6·8-20·5). GMTs against the omicron BA.4/5 variant on day 29 were 2551 (1687-3859) in the ARCT-154 group and 1958 (1281-2993) in the BNT162b2 group-a GMT ratio of 1·30 (1·07-1·58)-with seroresponse rates of 69·9% (65·0-74·4) and 58·0% (52·8-63·1). Both boosters were equally well tolerated. No treatment-related deaths were reported, nor were there severe or serious adverse events considered to be causally associated related to study vaccination. One serious adverse event, a foot deformity reported in a participant in the BNT162b2 group, was observed but determined not to have a causal relationship to the study vaccination. One severe adverse event, a case of abnormal hepatic function in the ARCT-154 group, was considered to be related to study vaccine. Adverse events of special interest for detection of myocarditis and pericarditis included chest pain (one case in the ARCT-154 group and three cases in the BNT162b2 group) and shortness of breath (two cases in the BNT162b2 group), all of which were considered to have a reasonable possibility of being related to vaccination. Local reactions were reported by 398 (95%) of 420 participants receiving the ARCT-154 vaccine and 395 (97%) of 408 participants receiving the BNT162b2 vaccine, and solicited systemic adverse events by 276 (66%) of those receiving the ARCT-154 vaccine and 255 (63%) of those receiving the BNT162b2 vaccine. Adverse events were mainly mild in severity, occurring and resolving within 3-4 days after vaccination. In adults who had previously received three doses of an mRNA COVID-19 vaccine, immune responses 28 days after an ARCT-154 booster dose were non-inferior to those observed after a BNT162b2 booster dose for the Wuhan-Hu-1 strain of SARS-CoV-2 and superior for the Omicron BA.4/5 variant. Increased immune responses at 28 days might provide increased likelihood of protection against these strains during this period and could also result in longer duration of protection. Further studies will assess the immunogenicity induced against more recent SARS-CoV-2 variants. Japanese Ministry of Health, Labour, and Welfare. For the Japanese translation of the abstract see Supplementary Materials section.