Source: United Kingdom – Executive Government & Departments
The Office for National Statistics (ONS) have published the data from their Infection Survey pilot study, looking at COVID-19 in the UK.
This Roundup accompanied an SMC Briefing.
Prof Sir David Spiegelhalter, Chair of the Winton Centre for Risk and Evidence Communication, University of Cambridge, said:
“As a back-of-envelope calculation, the latest ONS survey suggests around 6.8% of 56 million people in England have been infected, which is around 4 million, and there’s been around 40,000 deaths in England linked to COVID. So this suggests that infection has carried around a 1% average mortality rate. Which is impressively close to the much-disputed estimate of 0.9% made by the Imperial College team back in March.”
Prof Daniel Altmann, Professor of Immunology, Imperial College London, said:
“This ONS dataset is helpful, especially the long-awaited antibody seroprevalence data. The antibody data was eagerly awaited to get a first snapshot of cumulative spread of this infection to date. What does all this tell us? The level of infections in the community (assessed by PCR) continues to rumble along at a low level. In terms of assessing your behaviour and risk-management as we shift out of lockdown, this means that, on average, every time you’re in a crowd of 400-500 people (such as a rush-hour tube platform), one person may be unknowingly shedding virus – you may want to keep some distance in case they’re the person pressed right up next to you. The sample tested for antibodies showed around 60 positives out of the first 885 blood samples, a first step to assessing our seroprevalence. Datasets are starting to emerge from many countries, few centres showing seroprevalence much above 10%. This small UK dataset is even lower than that. A few points to make about this: this picture of cumulative exposure, symptomatic and asymptomatic, suggests there hasn’t been a massive, hidden iceberg of infections – more a picture of around 130,000 infections per week since the start of the outbreak here, which would indeed build to a cumulative total of 5-7% of the population. The other obvious point to make is about herd immunity and antibody passports: mitigation of a 2nd wave would need around 60% seroprevalence, so this would not be achievable without an effective vaccine. Also, from this data, only a tiny minority of the population could obtain an ‘antibody passport’, notwithstanding the massive caveat that this dataset encompasses a range of antibody levels, some of which might confer protection, some not, meaning that an ‘antibody passport’ might not be a marker of future protection.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“It’s very useful to have these weekly results from the ONS COVID-19 infection survey. What’s particularly valuable about them is that the data come from a representative sample of the whole community population of England (though not people in hospitals, care homes, or other institutions). The people providing the data were not selected because of having disease symptoms, or having been tested in some way for the virus, or anything like that, so can give us a truer picture about infection rates than other data sources. But the weekly releases do have something of the character of the thousand and one nights – they tell you something new, but not the whole story, so you want to come back next week and hear more. This isn’t by any means a criticism of ONS and the other partners in the survey. I very much admire the transparency in which they are doing this work and telling us how they did it and what they have found so far. But the survey is still in its pilot phase, and the reason that the information is incomplete is that a lot of important data haven’t yet been collected – for reasons to do with how a survey like this must be conducted.
“Some new features in this week’s report, compared to last week’s, illustrate this. The new report contains data on antibody tests, carried out using blood samples from respondents which are then tested in a lab. The results indicate whether the person has antibodies to the SARS-2-CoV virus, which would have arisen because they had previously been infected with the virus. But it takes time to obtain the blood samples, test them, and report. So far they have results only from 885 such tests, of which 60 contained antibodies. So about 6.8% of the test samples contained antibodies, and one can infer that about 6.8% of the community population have antibodies and so must have previously been infected, though the margin of error around that (95% confidence interval) runs from 5.2% to 8.6%. More samples will be taken as the survey continues, and the accuracy will improve. But so far there’s no possibility of looking at how people who test positive on the swab test for a current infection move on to a positive antibody test on a blood sample, because it’s simply too early in this pilot stage to be able to measure that. And most of these people would have been infected before this survey began – because it’s been running only about a month, it takes something like 2 weeks after infection for the antibodies to develop, and other data indicate that the peak of infections occurred sometime before the survey started.
“Another new feature this week is that the report makes comparisons between different population groups (for example males and females, different ages, patient-facing or resident-facing health and care workers compared to everyone else), not on the basis just of people with positive tests for virus in their throat and nose swabs in a single two week period as in last week’s report, but in those who tested positive over the whole time since the survey began. That makes the comparisons more precise – but the survey has been obtaining data for only just over a month (since 26 April) and many comparisons are still not very precise. They will get more precise as the weeks pass.
“But the increase in the amount of data available does allow ONS to report what I think are the most interesting new findings. These relate the symptoms that people report they have had, to whether they tested positive for a current SARS-2-CoV infection. You can look at that two ways – if a person reports symptoms, what’s the chance they test positive, or, if a person tests positive, what’s the chance they reported symptoms? Looking at people who reported a symptom on the day of the swab test, only 2.6% of them tested positive in the test. Even of those who reported the symptoms most closely associated with COVID-19 (cough and/or fever and/or loss of smell or taste) on the test day, 6.7% tested positive – so most people with those symptoms did not test positive for this virus. There are plenty of other reasons for the symptoms, of course, and there’s also some uncertainty about these exact figures because the amount of data so far is not all that large. That will improve over time. Arguably, looking the other way is even more interesting. Consider all the people who had a positive test for virus in their throat and nose in the survey so far. How many of them reported symptoms? It turns out that 30% of them did, either on the day that they were tested or on some earlier or later day during the survey period so far. That would indicate that most of people who tested positive, about 70% of them, did not report any symptoms at all. These figures are subject to considerable uncertainty – so far, of all the people tested, there have only been 80 positive tests in all, and that’s not big enough to provide precise estimates of these percentages. Confidence bounds for the 30% figure run from 20% to 43%, and the amount of data is too small to compare different types of symptoms. Again that position will improve as the survey goes on – but there are potential implications for the effectiveness of tracking and tracing the contacts of people with symptoms, if high numbers of infected people don’t actually have symptoms anyway, or if many people with symptoms don’t actually have the infection.”
Prof Lawrence Young, Professor of Molecular Oncology, Warwick Medical School, University of Warwick, said:
“There have been a wide range of estimates of the seroprevalence of SARS-CoV-2 in different populations around the world with estimates ranging from 5% to over 20%. This result of 6.78% testing positive for antibodies to SARS-CoV-2 is consistent with these estimates.
“ This is the initial result from a cross-sectional study of 885 individuals as part of the COVID-19 Infection Survey being run by the ONS and University of Oxford. The antibody test being used targets the spike protein of the virus and only provides an initial estimate for the percentage of the private-residential population testing positive. The document accompanying the study states that ‘We do not report on the prevalence rate within the analysis sections of this bulletin. To calculate the prevalence rate we would need to know the rates of false-positive and false-negative test results. Our initial analysis of modelling different false-positive and false-negative rates indicate only slight increases to uncertainty in the estimates. As we complete more sensitivity analysis, we will present these results in a future release.’ The recent PHE evaluation of antibody tests showed high levels of specificity and sensitivity but these tests target a different virus protein – the nucleocapsid.
“It demonstrates who has previously had the infection and therefore indicates that the majority of the population are unlikely to have been infected.
“ We still don’t know whether testing positive for SARS-CoV-2 antibodies means that you are protected from re-infection. However, data from previous coronavirus infections and from initial clinical studies using convalescent plasma from recovered COVID-19 patients as a therapy, suggests that these antibodies may provide some protection. More studies are needed to confirm this possibility and to determine how long any immune protection lasts.”
Coronavirus (COVID-19) Infection Survey pilot: England, 28 May 2020
All our previous output on this subject can be seen at this weblink:
Prof Daniel Altmann: No conflicts to declare
Prof Kevin McConway: “Prof McConway is a member of the SMC Advisory Committee, but his quote above is in his capacity as a professional statistician.”
Prof Lawrence Young: No conflicts of interest
None others received