METHODS: We did a randomised, stratified, observer-blind, controlled, multicentre, phase 3 study at 89 sites in 12 countries in 2016-17 northern hemisphere and 2017 southern hemisphere influenza seasons. We enrolled community-dwelling male and female adults aged 65 years and older who were healthy or had comorbidities that increased their risk of influenza complications. We stratified eligible participants by age (cohorts 65-74 years and ≥75 years) and risk of influenza complications (high and low) and randomly assigned them (1:1) via an interactive response technology to receive either aQIV or a non-influenza comparator vaccine. We masked participants and outcome assessors to the administered vaccine. Personnel administering the vaccines did not participate in endpoint assessment. The primary outcome was absolute vaccine efficacy assessed by RT-PCR-confirmed influenza due to any influenza strain in the overall study population (full analysis set) from day 21 to 180 or the end of the influenza season. Vaccine efficacy was calculated on the basis of a Cox proportional hazard regression model for time to first occurrence of RT-PCR-confirmed influenza due to any strain of influenza. Safety outcomes were assessed in the overall study population. This trial was registered with ClinicalTrials.gov, NCT02587221.

FINDINGS: Northern hemisphere enrolment occurred between Sept 30, 2016, and Feb 28, 2017, and southern hemisphere enrolment between May 26, 2017, and 30 June 30, 2017. aQIV was administered to 3381 participants, who subsequently had 122 (3·6%) RT-PCR-confirmed influenza cases, and the comparator was administered to 3380 participants, who subsequently had 151 (4·5%) influenza cases. The majority, 214 (78·4%) of 273, were caused by influenza A H3N2. Most antigenically characterised isolates were mismatched to the vaccine strain (118 [85%] of 139). Vaccine efficacy was 19·8% (multiplicity-adjusted 95% CI -5·3 to 38·9) against all influenza and 49·9% (-24·0 to 79·8) against antigenically matched strains, when the protocol definition of influenza-like illness was used. The most common local solicited adverse event was injection site pain, reported by 102 (16·3%) of 624 participants in the aQIV group and 71 (11·2%) of 632 of participants in the comparator group. Deaths were evenly distributed; none were considered related to study vaccines. The safety profile for aQIV was similar to previously reported trials.

INTERPRETATION: The prespecified criterion for showing the efficacy of aQIV in older adults was not met during the influenza seasons with high amounts of vaccine strain mismatch. Vaccine efficacy was higher against influenza cases associated with higher fever, which represent more clinically significant disease.