Bayesian re-analysis of the DANFLU-1 trial.

Modin D., Johansen ND., Granholm A., Claggett BL., Nealon J., Samson S., Loiacono MM., Harris RC., Larsen CS., Jensen AMR., Landler NE., Solomon SD., Landray MJ., Gislason GH., Køber L., Sivapalan P., Jensen JUS., Biering-Sørensen T.

DANFLU-1 was an open-label, pragmatic feasibility trial which randomized persons aged 65 to 79 years to high-dose inactivated influenza vaccine (HD-IIV) or standard-dose inactivated influenza vaccine (SD-IIV). The trial found that HDIV was associated with a reduced incidence of death and hospitalization for influenza or pneumonia as compared to SDIV. Bayesian analysis offers a framework for probabilistic interpretation of trial data and provides a method for incorporating prior information into the analysis. This study presents a post-hoc, Bayesian re-analysis of the DANFLU-1 trial. We conducted a Bayesian re-analysis of the DANFLU-1 trial, which randomly assigned 12,477 adults (65-79 years) 1:1 to HDIV or SDIV during the 2021/2022 season. The trial used Danish nationwide registers for data collection including baseline and follow-up data. This re-analysis applied neutral non-informative, evidence-based, and neutral skeptical priors. The evidence-based priors were informed solely by randomized trials published before DANFLU-1. Relative vaccine effectiveness (rVE) with 95% credible intervals (CrI), and posterior probabilities were estimated using Bayesian log-binomial regression models. Probabilities of rVE >0%, 10% and 20% were estimated. The findings were consistent across different priors. There was a greater than 95% probability of any benefit (i.e. rVE >0%) for all-cause mortality and hospitalization due to pneumonia/influenza, regardless of the prior used. For pneumonia/influenza hospitalization, the probabilities of rVE >10% were at least 95% with the non-informative and evidence-based priors, while it was 93.2% with the skeptical prior. For all-cause mortality, the probabilities of rVE > 10% ranged from 91.1% to 98.4% across priors. For the remaining outcomes, including cardiorespiratory hospitalization and any hospitalization, the probabilities of of rVE >10% ranged from 25.0% to 59.0% across priors. This Bayesian re-analysis of DANFLU-1 demonstrated robust results, with high probabilities of any benefit (rVE >0%) for all-cause mortality and hospitalization due to pneumonia/influenza. We also found high probabilities of an rVE > 10% for both outcomes, indicating robust findings supportive of clinical benefit. As a feasibility trial, the findings warrant further Bayesian investigation of adequately powered trials.

DOI

10.1080/21645515.2025.2550050

Type

Journal article

Publication Date

2025-12-01T00:00:00+00:00

Volume

21

Keywords

Vaccine, bayesian analysis, high dose influenza vaccine, influenza, standard dose influenza vaccine, trial, Aged, Female, Humans, Male, Bayes Theorem, Denmark, Hospitalization, Influenza Vaccines, Influenza, Human, Pneumonia, Vaccine Efficacy, Vaccines, Inactivated

Permalink More information Close