https://orcid.org/0000-0002-6995-4369
Introduction
Waning COVID-19 vaccine effectiveness (VE) and the capacity of the Omicron SARS-CoV-2 variant to evade pre-existing immunity have been major impediments to COVID-19 control efforts worldwide. In response, several countries have rolled out fourth-dose COVID-19 vaccination programmes.1–3 Published studies have reported relative VE (rVE) against Omicron for a recent fourth dose of an mRNA-based vaccine compared with a third dose administered at least 4 months earlier among older adults.1–3 One prior study has reported fourth-dose VE relative to no vaccination,3 which we term absolute VE (aVE). However, third- and fourth-dose aVE estimates for comparable time points were not included, meaning it was impossible to determine whether the fourth dose returned people to the same or greater protection as a third dose for the equivalent time post-vaccination. Accordingly, we aimed to estimate the aVE of three- and four-dose COVID-19 vaccine regimens at the same time post-vaccination.
Methods
We previously published a logistic regression equation that predicts aVE with waning over time for a third dose of an mRNA vaccine against symptomatic infection and hospitalization due to the Omicron BA.1 variant.4 We extended this equation to include age variation in VE, using a study assessing the relative change in VE across age groups.5 We then incorporated the additional outcomes of any documented infection (asymptomatic and symptomatic) and mortality using data reported by the United Kingdom Health Security Agency (UKHSA),6 establishing a logistic regression model to estimate third-dose aVE against Omicron BA.1 by clinical outcomes, age strata and weeks since vaccination. Details are provided in the Supplementary material (available as Supplementary data at IJE online).
Expected aVE values and their uncertainty intervals were estimated using 10 000 iterations of a Monte Carlo simulation propagating uncertainty using random draws from the covariance matrix of the logistic model fitted to UKHSA data; independent draws of the additional impacts of age, any infection and death in the extended logistic model; and independent draws from Magen et al.’s rVE central estimates and standard errors.
Results
Estimated aVE against Omicron by clinical outcome at 3 weeks following the receipt of third and fourth doses is shown in Figure 1. The aVE is higher against more severe outcomes for both three-dose and four-dose regimens. The estimated aVE against milder outcomes following a fourth dose appears to be higher than after a third dose at 3 weeks following vaccination, but uncertainty intervals overlap.
Vaccine effectiveness against Omicron infection-associated clinical outcomes at 3 weeks following third and fourth doses (cf. nil vaccination). Estimated absolute vaccine effectiveness (aVE) with 95% CIs against various clinical outcomes caused by the Omicron BA.1 variant of SARS-CoV-2 at 3 weeks following a fourth vaccine dose compared with those at 3 weeks following a third vaccine dose with the reference category of nil vaccination
Discussion
A fourth dose of the BNT162b2 vaccine appears to restore, if not boost even more (although uncertainty intervals overlap), the protection conferred by a third dose at an equivalent time post-vaccination. Our estimates are likely to be of interest to the public and policy makers weighing up the benefits and costs of a fourth dose in the context of waning immunity derived from triple vaccination who are asking the question: ‘Does a fourth dose return one to the same, or even higher, protection that one had shortly after a third dose?’
Our approach has limitations. First, due to the lack of reported data from the same jurisdiction, our use of UKHSA data assumes a third dose in Israel has the same effectiveness and waning as a third dose in the UK. However, the two most influential studies7,8 in the UKHSA report6 and the Israeli study by Magen et al.1 adjusted for similar potential confounders. We also account for the lack of data at comparable time periods by matching predicted third-dose aVE with reported fourth-dose rVE at an equivalent time (i.e. 3 weeks) since the last dose of vaccination. Whilst there may theoretically be residual regional and contextual variation between the UK and Israel that limit transportability of findings between jurisdictions, we believe any such bias will be small, and it is difficult to conceive of an alternative method to estimate aVE in the absence of any studies that directly compare individuals receiving third and fourth doses at equivalent times post-vaccination (cf. no prior vaccination or infection).
Second, our estimates concern VE against Omicron BA.1. The aVE against newer variants may drop due to immune escape, but nevertheless we expect the magnitude of this reduction to be consistent for both third and fourth doses—preserving the same pattern of findings we report here. Our findings may be used in conjunction with other VE estimates to support vaccine-related policy-making, albeit cautiously given the fourth-dose aVE is estimated using data from two specific contexts.
Ethics approval
This study utilizes publicly available data from the published literature and, as such, ethics approval is not required.
Data availability
Underlying data are available from the corresponding author upon reasonable request.
Supplementary data
Supplementary data are available at IJE online.
Author contributions
J.Z.—conceptualization of research ideas, formulation of research goals, development of methodology, data curation, application of statistical analysis, preparation and writing of original draft, review and editing of the published work and data visualization. J.S.—conceptualization of research ideas, formulation of research goals, development of methodology, review and editing of the published work and supervision. T.B.—conceptualization of research ideas, formulation of research goals, development of methodology, data curation, application of statistical analysis, review and editing of the published work, data visualization and supervision.
Funding
No funding was received for this paper.
Conflict of interest
The Population Interventions Unit is currently conducting separate VE research in Victoria, Australia that is funded by Moderna. Moderna had no involvement in the current study.
