Reply to Ferdinands et al

To the Editor—We agree with Ferdinands et al that it would be premature to recommend delay of influenza vaccination until November. Although our study found that people vaccinated in November are better protected than people vaccinated earlier, we noted in the Discussion that delaying vaccination might reduce vaccine coverage and should not be recommended before careful consideration of potential harms as well as benefits [1].

We disagree with Ferdinands et al about the potential bias that was called to our attention by Dr Lipsitch [2]. This is an interesting bias—associated with “depletion of susceptibles”—but our findings cannot be attributed to it. This bias can arise as higher-risk individuals, called “susceptibles,” get infected and thereby depleted from the population still at risk. During influenza season the unvaccinated population loses more of its “susceptibles” (because it is unprotected by the vaccine) than does the vaccinated population. Consequently, influenza incidence eventually falls faster among the unvaccinated than among vaccinees, bringing these 2 incidence rates closer together. Thus, later in the influenza season, vaccine effectiveness (VE) can artifactually seem to wane (unless susceptibility is measured and adjusted for).

For this kind of bias to arise, the groups compared must have differential depletion of their susceptible individuals. In our vaccinee-only study, the groups compared were persons vaccinated earlier in the year vs those vaccinated later in the year. Before influenza begins to circulate, there is no depletion of susceptibles. After influenza begins to circulate—and if there really is no waning of VE—the rate of depletion of susceptibles among the vaccinees will be the same regardless of when they were vaccinated. In this no-waning scenario, there is no potential for bias related to depletion of susceptibles.

(If VE really does wane, then the early vaccinees would be less protected than the later vaccinees and would get more depleted of their susceptibles. In this case, the amount of waning would be underestimated rather than exaggerated.)

However, bias in the direction of exaggerating waning is potentially introduced by the inclusion of persons who are vaccinated after the influenza season is under way and have less time to be depleted. To address this concern, we restricted our analysis to the 90% of our study population that was vaccinated before 1 December (before the influenza season ever started in Northern California during the 2010–2016 study period), and our estimate of waning changed very little. (Our estimate of the increased odds of influenza with every 28 days since vaccination went from 1.16 to 1.18.)

Notes

Financial support. This work was supported by the National Institute of Allergy and Infectious Diseases, National Institutes of Health (grant number 1R01AI107721-01).

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.