Vaccine Effectiveness Against Covid-19 - What Can The Covid-19 Infection Survey Tell Us?

25th October 2021

This article if from the Office for National Statistics written by Sarah Crofts is the Head of Analytical Outputs for the COVID-19 Infection Survey.

As well as providing estimates on infection levels and antibodies, the Coronavirus (COVID-19) Infection Survey (CIS) collects information from participants on their vaccination status, providing some insight into vaccine effectiveness on variants, including Delta.

Our monthly visits to the same participants over an extended period allow us to see who becomes infected over time.

Sarah Crofts explains more about our work on vaccine effectiveness:

We have seen the landscape of the pandemic shift in recent months, a picture which has included fluctuating infection levels across the UK, combined with more vaccines issued to younger people and planned boosters for older people. So, the burning question remains: ‘how effective are the vaccines?'

This is where data from our survey can help. There are different ways to address the question of vaccine effectiveness. A couple of months ago our academic partner, Professor Sarah Walker (University of Oxford), published analysis on vaccine effectiveness, which is summarised in our blog.

This showed that two doses of vaccine provided the best protection and that the highest protection was found among those who had experienced a past infection followed by two vaccines. It also showed that vaccines were slightly less effective against the Delta variant compared with the Alpha variant.

Today we have published an extension to this work using a different method. This latest analysis, based on a Poisson regression model, includes additional data on infections and the date of infections from NHS Test and Trace (T&T) for survey participants from England.

Including the Test and Trace data means we can estimate the day on which someone first became infected more accurately than relying on survey data alone. We compare the infections we find to ‘all days at risk', the period of time (in days) that someone could have become infected. This improves our statistical power to identify the effects of vaccination on a person's risk of testing positive. Adding an additional two weeks of data from the Delta dominant period to our analysis provides further power.

The earlier work used a ‘logistic regression model' which was a faster method, focused on producing results of vaccine effectiveness as quickly as possible once the Delta variant was dominant. However, it was based only on test results from survey visits which are mostly monthly and may have missed some infections between monthly visits or identified some infections late.

Therefore, our latest results are based on a more powerful analysis to detect the effects of vaccination on a person’s risk of becoming infected.

Both methods provide broadly the same results. They both show that vaccination (particularly two doses) reduces the risk of testing positive. Additionally, vaccines were more effective at reducing the risk of becoming infected and developing symptoms during periods when Alpha was the main variant than when Delta became the main variant. The Pfizer-BioNTech and Oxford AstraZeneca vaccines showed a similar level of protection as prior natural infection now that virtually all infections are the Delta variant.

The main difference we noticed between the two analyses was that the Poisson model showed a slightly lower vaccine effectiveness now Delta is the main variant compared to the results of the logistic regression model. The likely reasons relate to the time periods included in the analyses. The most recent Poisson model includes results from participants who had experienced a longer average time period since their second vaccination. If there is any change in vaccine effectiveness over time (such as vaccines wearing off) then it is more likely to have been identified in the latest analysis. Additionally, when we estimate the risk of infection among vaccinated people compared to unvaccinated people, we need test results from each group of people. Over time, more people have received vaccinations and so there are smaller numbers of unvaccinated people to compare with.

In conclusion, today’s analyses on vaccine effectiveness builds on previous research and confirms that two doses of a vaccine reduces the risk of testing positive for Covid-19 infection. The risk of testing positive for those who had received two doses of either the Pfizer-BioNTech or the Oxford AstraZeneca vaccines was lower during the Alpha variant than the Delta variant periods of the pandemic.

Our technical article presents these results and describes the methods used in greater detail.