Understanding herd immunity for influenza using archived sera from the UK and mathematical models

The 2009 pandemic highlighted some real problems with how scientists and governments monitor influenza. Up-to-now, they have counted numbers of people who go to their doctor or the hospital with flu-like symptoms. But this doesn't tell us about the mild cases.

Mild cases are important because of something called 'herd immunity'. During any epidemic, the number of new cases starts to go down when there are not enough susceptible people to keep 'the fire' of the epidemic going. People with mild or no symptoms are just as good at stopping the epidemic as people with symptoms. So, to be able to understand how big an influenza epidemic might be, we need to measure mild cases as well.

We can do this if we take blood and test for antibodies. In this project, we want to test anonymous blood samples for influenza that were discarded in the late 1990s and early 2000s. By doing 10s of 1000s of tests on these old samples, we can see exactly how useful the data on mild cases might be for future epidemics and we can answer interesting questions about different influenza strains. There were three different types of influenza circulating in the late 1990s and early 2000s. Two types of influenza A and one was influenza B. Two of those three are still circulating now (A/H3N2 and B). One of them was replaced by the pandemic strain (A/H1N1). We want to compare and contract these three types.

We will use mathematical models to analyze the data from the blood tests. Models help us to figure out why an epidemic occurred or was as big as it was, rather than just describing it. We can look at important basic questions like: is flu B less serious infection that influenza A; and if you get a mild infection of flu you can get infected again more quickly than if you get a serious infection?

Serological studies of the 2009 pandemic illustrated how patterns of infection were different to patterns of clinical cases. Therefore, standard approaches for the estimation of incidence (e.g. based on surveillance for influenza-like-illness) did not give a clear understanding of how the pool of susceptible individuals is being depleted at any point in time (herd immunity).

In order to fully characterize the transition from Sydney 1997-like to Fujian 2002-like viruses, we will measure the concentration of antibodies in the UK population for both H3 strains for all 6 seasons from 1998/1999 to 2003/2004. Our background knowledge for influenza A H1N1 and B is less complete than for H3N2. Therefore, we focus on a period when both H1N1 and B vaccine strains were changed, 1999-2000 to 2001/2002. To our knowledge, there are no population-wide serological surveys of influenza B.

We will use 3 analytical approaches: 1) we will classify each assay result as either positive or negative based on a titre threshold and then use logistic regression to test interesting hypothesis for the risk of infection; 2), we will incorporate serology as one of a number of data sources to infer parameters for strain-specific healthcare oriented dynamics model; 3) to better understand the relationship between our assay results and; infection, infectivity, susceptibility, and decay; we will propose the most parsimonious transmission model that can represent the full range of assay results and these fundamental processes.

The combination of data and models we propose will allow us to address hypotheses such as: absolute rates of infection with influenza B are lower than those with H3N2 and have a different pattern across age groups; infectivity at intermediate titres is age specific; and the use of pre and post seasonal titres would enable the detection of some strains being more severe than others by allowing the measurements of clinical outcomes against a 'true' measure of infection.

This project has been designed to ensure that the scientific discoveries we generate have immediate societal impact. More than half of the existing project team are employed by Health Protection Agency (HPA) and have, as part of their non-project duties, responsibility for the ongoing surveillance and modelling of influenza in the UK. Three of the four new full-time staff to be supported by this project will work for the HPA. When we make discoveries that improve our interpretation of existing data streams or justify additional serological testing, those discoveries will be made by individuals helping to set policy on a day-to-day basis.

Our request for 2 full time postdoctoral modellers and our request for Peter White's time are central to ensuring good impact for this project. By ensuring that the project supports an individual 100% within the HPA modelling unit, we are ensuring the highest possible likelihood that our discoveries will influence ongoing practice and improve influenza policy in the UK. As head of the modelling unit at the HPA, Peter spends a considerable portion of his time explaining the utility of infectious disease dynamics to non-scientists. As a champion for our project, we will ensure that our results are considered at the highest level.

Elizabeth Miller, Neil Ferguson, Maria Zambon and Steven Riley all sit on key advisory organizations at home and abroad. For example, Prof Miller is a member of the WHO SAGE committee for vaccine preventable diseases, Neil Ferguson was an advisor to the WHO pandemic influenza committee and Maria Zambon is head of the Infection and Immunity function at the health protection agency.

Therefore, because our project is embedded within the UK health policy infrastructure and our senior team is highly engaged with international health policy, we suggest that there is a strong likelihood that our results will have high impact by changing health policy. For example, we expect to demonstrate an immediate value for routine testing of discarded blood samples for currently circulating influenza strains. We will be ideally positioned to champion the adoption of such routine seroepidemiology for influenza in the UK and elsewhere.