An mRNA flu vaccine beat a licensed shot by 34.5% — and made more people feel rough

Nucleoside-modified messenger RNA influenza vaccines had shown promising immunogenicity and an acceptable safety profile in earlier phase 1-2 work, and no human challenge trial of one had been reported [s1][s2]. On 19 November two trials arrived at once: a large phase 3 field trial in the New England Journal of Medicine and a small human challenge trial in NEJM Evidence, both testing a nucleoside-modified mRNA — modRNA — influenza vaccine from Pfizer [s1][s2]. The Journal ran an editorial the same day under the title "An Early Look at mRNA Vaccines to Prevent Influenza" [s3]. "Early" is the operative word.

The phase 3 trial

Healthy adults aged 18 to 64 were randomly assigned to receive either a quadrivalent modRNA influenza vaccine or a licensed inactivated quadrivalent influenza vaccine during the 2022–2023 influenza season, in the United States, South Africa and the Philippines [s1].

The comparator matters. This was not a placebo trial. The question was relative efficacy — whether the mRNA vaccine reduced the percentage of participants with laboratory-confirmed influenza associated with influenza-like illness at least 14 days after vaccination, compared against an already-licensed vaccine [s1]. The analysis tested both non-inferiority and superiority [s1].

A total of 18,476 participants were randomised: 9,225 to the modRNA vaccine and 9,251 to the control vaccine [s1].

Relative efficacy against influenza-like illness was 34.5% (95% CI, 7.4 to 53.9), based on 57 cases in the modRNA group and 87 in the control group — meeting the criteria for both non-inferiority and superiority [s1].

That confidence interval is the number to sit with. The point estimate is 34.5%, but the data are compatible with a relative benefit as small as 7.4% or as large as 53.9% [s1]. With 144 total cases across more than 18,000 people, the trial is precise enough to say the mRNA vaccine did better and not precise enough to say by how much.

The influenza B gap

Cases of influenza-like illness in the trial were caused by A/H3N2 and A/H1N1 strains, with almost no B strains [s1]. Non-inferiority of the antibody response on haemagglutination inhibition assay was shown for the influenza A strains but not for the B strains [s1].

So the efficacy result is, in practice, an influenza A result, and the immunogenicity data give a positive reason to be cautious about B rather than merely an absence of information. A season dominated by influenza B would be a different test, and this trial did not run it.

The reactogenicity trade

Reactogenicity was primarily mild or moderate in both groups, but consistently more common with the mRNA vaccine [s1]. Overall local reactions occurred in 70.1% of modRNA recipients versus 43.1% of controls, and overall systemic events in 65.8% versus 48.7% [s1]. Fever occurred in 5.6% of the modRNA group and 1.7% of the control group [s1]. Adverse event profiles were similar between the two groups [s1].

Roughly seven in ten people had a local reaction and two in three had a systemic one. For a vaccine given annually to healthy adults, that is not a trivial consideration — it is the kind of difference that shows up in uptake, not just in a safety table.

The challenge trial

The second paper took the opposite approach: infect people deliberately and watch closely [s2].

In a phase 2a double-blind trial, healthy adults aged 18 to 55 were randomly assigned 1:1 to the modRNA vaccine or a quadrivalent influenza vaccine 30 days before an influenza A/H1N1 challenge [s2]. Control participants came from a separate trial, were exposed to the same virus, and were not vaccinated [s2]. The per-protocol cohort comprised 55 modRNA, 48 QIV and 52 control participants [s2].

Symptomatic influenza occurred in 0% of the modRNA group, 4.2% of the QIV group and 26.9% of controls — a vaccine efficacy of 100.0% (95% CI, 75.2 to 100.0) for modRNA and 84.5% (95% CI, 43.4 to 96.0) for QIV [s2]. Febrile influenza occurred in 0%, 0% and 17.3% respectively, with efficacy of 100.0% for both vaccines [s2]. Viral load fell more in the modRNA group: the median difference from control in the area under the viral-load curve was −88.66 (95% CI, −261.95 to −1.35) for modRNA and −67.01 (95% CI, −254.13 to −0.97) for QIV [s2]. No serious adverse events were reported [s2].

A 100% efficacy figure from 55 vaccinated people against a single laboratory H1N1 strain is not a population-level claim, and the very wide confidence intervals say so. What the challenge trial adds is mechanistic: the mRNA vaccine did not merely reduce illness, it reduced how much virus people carried [s2].

What is still open

Both trials were funded by the vaccine's manufacturer; the challenge study was funded by hVIVO and Pfizer [s1][s2]. Both enrolled healthy adults — the phase 3 trial capped enrolment at age 64, and the challenge trial at 55 [s1][s2]. Neither studied older adults.

The unresolved questions are influenza B, older and immunocompromised populations, performance across seasons with different circulating strains, and whether a reactogenicity profile this much heavier than a standard shot is acceptable for an annual vaccine. Nothing here is a recommendation about which influenza vaccine anyone should receive.

Sources

• [s1] Efficacy, Immunogenicity, and Safety of Modified mRNA Influenza Vaccine, The New England Journal of Medicine, 19 November 2025. https://doi.org/10.1056/NEJMoa2416779
• [s2] Human Challenge Trial of a Nucleoside-Modified Messenger Ribonucleic Acid Influenza Vaccine, NEJM Evidence, 19 November 2025. https://doi.org/10.1056/EVIDoa2500087
• [s3] An Early Look at mRNA Vaccines to Prevent Influenza, The New England Journal of Medicine, 19 November 2025. https://doi.org/10.1056/NEJMe2514597