PREreview del 2024/25 end-of-season KP.2 vaccine effectiveness against COVID-19 hospitalization in older adults: a test-negative study in Quebec, Canada

Brief Summary

This study estimates the effectiveness of the 2024/25 KP.2 mRNA vaccine against COVID-19 hospitalization among adults 60 and older in Quebec, Canada, using a test-negative design linked to provincial administrative databases. The study period spans ten months, which is longer than most comparable studies, and both the fall 2024 and spring 2025 vaccination campaigns. Overall vaccine effectiveness was 34%, declining from 43% in the first eight weeks to negligible levels beyond 32 weeks. The study also provides the first published effectiveness estimate for the spring 2025 booster campaign, at 38%, based on limited uptake and 248 hospitalizations.

Strengths

1. The ten-month follow-up period answers a question that prior studies could not

Most comparable studies stopped at four to five months, leaving the question of when KP.2 protection reaches the floor unanswered for policy purposes. This study extends follow-up long enough to show that protection becomes negligible by month seven and remains undetectable through month ten. That distinction matters directly for booster timing decisions, since knowing that protection wanes is less actionable than knowing approximately when it does. With 2,644 hospitalizations contributing to the fall analysis, the estimates through the waning period are supported by a large enough event count to be reasonably precise.

2. The outcome definition reduces misclassification in a way that strengthens the validity of the VE estimates

Cases are defined as positive tests followed by hospitalization where COVID-19 is recorded as the reason for admission, not simply any hospitalization with an incidental positive test. This distinction is methodologically important and not universal across VE studies. Incidental COVID detection during admission for another cause can dilute apparent vaccine protection by inflating the case count with events the vaccine was not expected to prevent. By applying a stricter outcome definition, the study more precisely measures protection against severe COVID-19 disease specifically, improving the interpretability of the VE estimates relative to studies using looser definitions.

Major Concerns

1. The reference group includes individuals with prior vaccination history, which may attenuate VE estimates

Individuals are classified as unvaccinated if they received no COVID-19 vaccine or a non-KP.2 vaccine six or more months before specimen collection. This means the comparison group includes people with residual waned protection from prior vaccination series. Even modest residual protection in the reference group would compress the observed VE toward the null, making the KP.2 vaccine appear less effective than it truly is against a fully unprotected comparator. The six-month cutoff is a pragmatic threshold, not a biological one, and the authors do not test how sensitive the VE estimates are to this definition. A sensitivity analysis restricting the reference group to individuals with no prior COVID-19 vaccination history would provide a cleaner and more interpretable estimate.

2. Prior infection history is not accounted for through sensitivity analysis, and the directional bias may be opposite to what the authors suggest

The authors acknowledge that infection-acquired immunity among unvaccinated participants could bias VE estimates downward, but attribute this largely to limited self-awareness of infection outside hospital settings. However, Quebec's provincial administrative databases, which the authors link for vaccination, hospitalization, chronic disease, and laboratory records, include prior positive test records from hospital and public health laboratory settings. A sensitivity analysis excluding individuals with a prior documented positive test is likely feasible within the existing data linkage and would directly address whether hybrid immunity is unevenly distributed between groups. Importantly, if hybrid immunity is more prevalent among vaccinated individuals, which is plausible given higher health engagement in this group, VE could be inflated rather than attenuated, which is the opposite direction of the bias the authors describe. This possibility is not discussed.

3. The spring VE estimate is framed as a confirmatory finding but the precision does not support that interpretation

The spring VE of 38% carries a 95% confidence interval of 3% to 60% — a range statistically consistent with both negligible protection and protection nearly double the fall estimate. The authors describe spring VE as similar to fall VE at comparable time since vaccination, but that conclusion requires more precision than 248 hospitalizations and 2.3% booster uptake can provide. For a policy audience, particularly jurisdictions currently deciding whether to recommend spring doses, this framing risks overstating the strength of the evidence. The spring estimate is worth reporting as a preliminary signal, but should be presented explicitly as hypothesis-generating rather than confirmatory.

Overall Recommendation

This study makes a timely and meaningful contribution. The ten-month follow-up is the clearest picture so far of how long KP.2 protection lasts, and the strict outcome definition improves on many comparable studies. The fall VE findings are credible and directly policy-relevant. However, the heterogeneous reference group and the absence of a prior infection sensitivity analysis leave the estimates vulnerable to bias that is addressable within the existing data, and the spring VE conclusion is presented with more confidence than the underlying numbers support. Addressing these issues would strengthen the study's value for public health decision-making.

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.