Our analysis shows that the speed of vaccination is key to reducing the impact of an influenza pandemic. Even with a low effectiveness vaccine, initiating vaccination 3 months after the declaration of a pandemic would lower the disease burden compared with initiating a higher effectiveness vaccine at 6 months, with 23–94% incremental benefits across health outcomes and VEs. While moderately and highly effective vaccines could flatten the pandemic curve if administered from 3 months, none of the scenarios evaluated could flatten the curve if administered from 6 months. Acute and ICU bed availability would also be less constrained under the early vaccination scenario, particularly with higher effectiveness vaccines, but administration of a vaccine at 6 months would not be able to prevent a surge in demand above bed availability thresholds in a severe pandemic, irrespective of VE.

The findings of this analysis echo those of models of the COVID-19 pandemic evaluating the impact of speed of vaccination and VE in controlling pandemics. In a study by Kim et al. which used a dynamic susceptible-infected-recovered-deceased model to simulate the spread of SARS-CoV-2 and VE estimates ranging from 60–95%, a lower effectiveness vaccine was found to achieve a lower infection attack rate than a high effectiveness vaccine, if the lower effectiveness vaccine could be distributed more quickly¹¹. Similarly, in a study Paltiel et al., which used a susceptible, exposed, infected, recovered (SEIR) framework to model the spread of SARS-CoV-2 through the population, vaccine benefits were mostly dependent on how quickly and broadly a vaccine could be distributed, rather than the VE¹².

There are several hurdles to production of pandemic influenza vaccines. In the 2009 influenza pandemic, a vaccine became available ~6 months after emergence of the pandemic virus, but supply was limited by manufacturing constraints¹³. During the pandemic, influenza vaccines were manufactured using the same techniques as approved seasonal influenza vaccines, thereby simplifying and accelerating the approval process¹⁴. However, egg-based techniques used for seasonal influenza vaccine manufacture are not readily upscalable and are subject to available egg supply, thus potentially creating a bottleneck in pandemic vaccine preparations¹⁵. Since the 2009 pandemic, newer technologies (e.g., mRNA-based and adenovirus vector vaccine platforms) have emerged, which have been used to rapidly develop vaccines during the COVID-19 pandemic, and could play an important role in a swift roll-out of mass vaccination in future influenza pandemics. Additionally, based on evidence that speed of vaccination is key, even in the event of a low effectiveness vaccine, pre-pandemic vaccines which contain potential pandemic strains but have not been matched to a particular pandemic virus could have a potential role in preparedness efforts. While these vaccines would be less effective than specifically designed vaccines they would be readily available at the start of the pandemic and therefore could be an important tool in flattening the curve of a future pandemic^16,17.

In addition to the speed of vaccination, choosing which groups to vaccinate as a priority will potentially influence the impacts of the pandemic. During the COVID-19 pandemic, frontline healthcare workers and older adults, who were at increased risk of severe outcomes, were prioritized for vaccination, with later vaccination of healthy younger adults and children¹⁸. Previous influenza pandemics have not necessarily affected the same high-risk groups as seasonal influenza, where older adults and young children have the greatest disease burden and risk of hospitalization or death¹⁹. In the 1918 pandemic, healthy adults aged 20–40 years were at highest risk of severe outcomes, and in the 2009 pandemic, adults ≥65 years experienced relatively mild symptoms compared with younger adults²⁰. School-aged children contribute significantly to the transmission of influenza, therefore vaccination of this age group as a priority may also help to flatten the pandemic curve, even if they are not at increased risk of severe symptoms^21,22. While not directly evaluated in the current model, vaccination of this age group may result in a faster and greater impact on transmission, and could also potentially help flatten the curve should vaccines not be available until 6 months into the pandemic.

As with all modeling studies, this analysis has limitations. Firstly, the model was limited by the input parameters included. It was assumed that a pandemic influenza virus would emerge in the Southern hemisphere and spread to the US, but this may not be the case as seen in the 2009 A/H1N1pdm09 pandemic. It is likely that the conclusions would still apply irrespective of transmission patterns and place of emergence, unless the virus emerged in the US and spread rapidly (e.g., during the winter season). In this case it would be likely that the majority of people would have been infected prior to the initiation of vaccination at 3 months, unless a total lockdown was effectively imposed and maintained until initiation of vaccination. Without the lockdown measures, the majority of deaths and hospitalizations would then occur prior to the initiation of vaccination, thereby limiting the potential effects of a mass vaccination campaign in preventing much of the pandemic disease burden, irrespective of whether it was initiated at 3 or 6 months. Secondly, we assumed a J shape for the distribution of mortality, with slightly increased mortality in young children and highest mortality in older adults. While this is commonly seen in seasonal influenza epidemics, previous pandemics (e.g., the 1918 pandemic) have shown an increased mortality rate in healthy adults²⁰, therefore a W shaped mortality may be more appropriate, depending on the nature of the pandemic virus and similarities to past viruses. Use of a W shaped mortality distribution would change the absolute number of deaths across the age groups but would most likely not alter the overall conclusions. Thirdly, we assessed the impact of NPIs in a no vaccination scenario, however, it is more likely that this would be used in conjunction with a subsequent vaccination campaign, thereby effectively halting transmission until vaccination could be initiated. Additionally, for simplicity, we assumed the NPI to be a total lockdown whereas it could include a variety of methods with varying impact on disease transmission. Fourthly, we did not assume that there was a stockpiling strategy which included availability of an off-the-shelf low VE vaccine which could be utilized directly after declaration of a pandemic. Fifthly, we modeled a pandemic with only one wave. Additional pandemic waves, e.g., as seen in the recent COVID-19 pandemic, would potentially alter the results, although the degree of impact would depend on a number of factors that we felt could not be accurately represented in the model, including rates of waning natural/vaccine-induced immunity, degree of match to the original pandemic strain, and the frequency and coverage achieved by future additional campaigns. Finally, the estimates presented in this model are based on a one-dose pandemic influenza vaccine; faster initiation of mass vaccination would be needed in the case of a two-dose schedule but would depend on VE after a single dose and the interval needed between doses.

In summary, our analysis has demonstrated the importance of rapid initiation of mass vaccination during a future influenza pandemic, with speed of vaccination playing a more important role than VE on population-level health outcomes. Preparedness exercises such as stockpiling potential pre-pandemic vaccines, as well as pre-emptive collection of data from newer vaccine manufacturing platforms, such as mRNA vaccines, will be paramount for ensuring a rapid and effective response in a future influenza pandemic.