August 9, 2021 - 10:00am

It’s fair to say that my contact with the world of public media over the past eighteen months hasn’t been without incident. When I look back at the first interviews I gave on Covid-19, including to UnHerd in May 2020, it is plainly evident that I still had no idea how politicised and nasty this controversy would become — I was still happy to share early hypotheses, not realising that they would be systematically collected and deployed to try to destroy my professional credibility, including by people I had previously respected and admired.

From the Sunday Times to the Guardian, senior MPs to Owen Jones, commentators have been lining up for months to pour scorn on my analysis. Even Tom Chivers, the Science Editor of UnHerd, (to which I am grateful for giving me the space to respond), writing in the Times this past weekend, singled me as the example par excellence of a “wrong” scientist:

Professor Sunetra Gupta, the Oxford epidemiologist, said in May 2020 that the coronavirus might kill one person in every 10,000 it infected. Since 36,000 people had died in the UK when she made that claim, that would have implied that 360 million British people — five times as many as exist — had had the disease.
- Tom Chivers, The Times

But this is not the whole story. In the UnHerd interview which he is referring to, I had already discussed my paper of March 2020 which had hypothesised an Infection Fatality Rate, or IFR, in the UK in the region of 0.1%, and the interviewer Freddie Sayers then asked what I now believed it to be? “I think it would definitely be less that 1 in 1000 and probably closer to 1 in 10,000,” I replied, and then corrected myself within seconds: “probably not 1 but perhaps 5 in 10,000.”

Obviously, it was unwise to make such estimates in a media interview and I have since learned to be much more guarded. However, the figure itself is not quite as outrageous as the journalist evidently believes.

There has been a huge amount of confusion during this pandemic around terms such as “Infection Fatality Rate” and “Herd Immunity Threshold”, with journalists drawn to the simplified idea that you can determine these by simply adding up published totals since the start of the pandemic. In reality, both these numbers are constantly fluctuating, depending on the vulnerability of a particular population (age, changing levels of immunity) and the conditions at the time (season, behaviour).

If IFR is to mean anything it must be attached to a particular wave of an infection at a particular time in a particular population. In May 2020 in the UK, particularly in London and the South East, it was indeed probable, and in my opinion remains so today, that around half the population had been exposed to Sars-CoV-2 and the Herd Immunity Threshold had been reached for that environment, which is why deaths and infections were coming down so rapidly. In this scenario, around 35,000 deaths for around 35m people would have indicated an IFR of around 1 in 1,000.

If, as I suspected at the time, the official death figures were being over-counted by as much as 50%, that figure would reduce further to nearer 5 in 10,000, or 0.05%.

Looking back, I now think those estimates were somewhat overly optimistic, but not outrageously so. I still believe that the most plausible explanation for the turnaround in that first wave was immunity, and that a large portion of the population had been exposed to the virus by that point. The logic of a highly infectious respiratory virus arriving into a new population leads powerfully to that conclusion, and we know the pathogen was present in the UK at the latest in January and infecting people at will throughout February and into March.

Contrary to most journalists’ understanding, the relatively low levels of antibodies found in subsequent seroprevalence studies do not disprove this hypothesis. Firstly, we know it is possible to be infected without developing antibodies, and to fend off an infection through other arms of the immune system; second, it now appears that antibodies decay rapidly, meaning it is not a useful marker of exposure. Indeed, something I got very wrong was my presumption based on other coronaviruses that immunity would last at least for around 2 years; it now seems that for Covid-19, infection-blocking immunity, either from vaccination or natural infection, may only last for 3-6 months (although hopefully much more durably against severe disease and death). The combination of this and the changing season and behaviour affecting the HIT combined to allow a second wave to get going in the autumn of 2020 — as I predicted would happen at the time.

Looking back at that early interview, I would advise my younger self that not every question a journalist asks needs an answer in the form they want, and would have counselled a less headline-making response along the lines of “single figures for IFR are not helpful as it is entirely contingent on population and environment”. It wouldn’t have made such a good headline — although to be fair, the headline of the UnHerd article, “Covid-19 is on the way out”, is another thing that, while easy to lampoon, was perfectly true for the UK at the time.


Dr Sunetra Gupta is a professor at Oxford University, an epidemiologist with expertise in immunology, vaccine development, and mathematical modelling of infectious diseases