select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
A preprint, non-peer reviewed results from the REACT-1 study, has reported on the possible latest COVID-19 R value and infection levels in the general population in England.
This Roundup accompanied an SMC Briefing.
Also comments on the Government announced R value of 1-1.2 for the UK.
Dr David Strain, Clinical Senior Lecturer, University of Exeter and chair of the British Medical Association’s Medical Academic Staff Committee, said:
“If anything, these latest figures from SAGE are likely to under-estimate the extent the virus is on the rise. Any estimate of the R-value above 1 is cause for concern, particularly considering that these figures do not include the potential impact of re-opening of schools. The difference between the R-values is likely to be a result of the different collection of data. The official government figures are predominantly based on those people who get tested through the NHS, Public Health England and the “Pillar 2” testing scheme. Therefore, they focus on individuals who have a clinical need to get tested either because of their own symptoms or because they have been in contact with a confirmed case (Through the Track and Trace service). This represent the burden of the disease in the country. On average it takes 5 days to develop symptoms, therefore this three-week snapshot tells us what was happening from a month ago until last week. The recent fiasco in the track and trace network, with individuals being sent for a 5 hour round trip in order to get their test, will have resulted in many people not making the journey. Thus the number of infected individuals will likely be under-estimated. The important omission in the SAGE report is the absolute numbers of patients. A 5% growth in cases, from a base rate of 10 patients per million compared to 2,000 patients per million clearly carries a different risk for an individual in that area. We need a clear indicator of a composite risk – the R value and the absolute cases – to generate a risk score in any region. This could easily be presented in a traffic light type manner, similar to what is already done for pollen counts to inform people with asthma.
“The REACT-1 study took a random sample of the population and explored how many people have the virus. Importantly, at the time of screening, 65% of participants who tested positive had not developed symptoms. Thus, this tells us more about what the virus is doing today. It makes the Prime Minister’s “rule of six” announcement even more timely in order to mitigate the potential impact of the school’s return last week and ensure that the education of our younger generation is protected.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
Comment on the Government announcement of R value as well as the REACT-1 study
“The latest estimate of the R number and the growth rate of SARS-CoV-2 infections (from the Government Office for Science (GOS) and SAGE) puts the range for R at 1.0 to 1.2, which doesn’t include any values below 1, and is therefore in accord with their stated view that the epidemic is growing. That’s hardly surprising in the light of other recent data. The growth rate range is estimated as -1% to +3%. If it’s as low as -1%, that would indicate that the number of new infections is shrinking by 1% a day, and if it’s +3%, that number would be growing by 3% a day. It may seem strange that the R range doesn’t include values that indicate a shrinking epidemic, but the growth rate range does. I think this is a feature of the rather complicated estimation process, and the fact that it’s difficult to estimate either of these quantities when the infection rate is fairly low, as it still is. All of these figures do indicate that the epidemic is likely to be growing, so actions are needed to make sure that the growth is controlled, and, I’d hope, reversed.
“The preprint from the REACT-1 study describes a very impressive programme of work, and the report contains many interesting and potentially important details. However, we must remember that all we have so far is a preprint, that has not yet been through peer review.
“To deal with its estimate of R first – the central estimate, based on the most recent round of swab tests (22 August to 7 September), is 1.7, with a range from 1.4 to 2.0 to take account of statistical uncertainty. That’s considerably higher than the GOS/SAGE figure. There are several possible reasons for that, including that there is a good deal of uncertainty in the REACT-1 estimate based only on the most recent round. The REACT-1 estimate based on data from the last two rounds rather than one is lower at 1.3, and more precisely estimated (range from 1.2 to 1.4). The GOS/SAGE estimate for R is based on more data than just the numbers of new infections – for example data on hospitalisations and deaths – which should improve its accuracy but also means that it can lag up to three weeks behind current new infections, so that the estimates published today might be most appropriately compared with the REACT-1 estimates over two rounds rather than just the latest round, and in that case the two sets of estimates do match better. But perhaps the REACT-1 R number estimate is a warning that things may be getting worse.
“The REACT-1 results are the second to be published today based on the results of swab tests on a sample of the population of England. The other was the regular weekly report on the ONS infection survey. The sampling approaches for the two surveys were different – REACT-1 has tested considerably more people, but, unlike the ONS survey, it does not generally include repeated tests on the same people. That means that it cannot track certain aspects over time so easily, including counts of new infections (as opposed to all infections). The ONS infection survey estimate for the rate of infections for the latest reporting week (30 August to 5 September) is that 7 people in every 10,000 were infected. The REACT-1 estimate, based only on round 4 of their sampling (22 August to 7 September) is that about 13 people in every 10,000 infected. That’s obviously higher than the ONS estimate. But taking the margins of statistical uncertainty into account, the figures aren’t as far apart as they look at first. ONS report that their data are consistent with rates between 5 and 10 infected people in every 10,000, and REACT-1 gives a range of between 10 and 16 per 10,000. So it’s just about possible that the difference between the two estimates is just due to statistical variability. If the true rate in the country is about 10 in every 10,000, that would be compatible with both the surveys. There are other possible explanations though. One is that the dates are different, and in particular the REACT-1 results include two extra days in September, at a time when other data have indicated that infection rates are increasing. There are other possible explanations connected with the details of the statistical analysis.
“A key point to bear in mind is that neither estimate puts the infection rate anywhere near where it was at the height of the epidemic in late March and early April. But we can’t be complacent – with all the indications from these surveys and other sources that infections are rising, constant surveillance and actions are necessary.”
Dr Yuliya Kyrychko, Reader in Mathematics, University of Sussex, said:
“The increase in today’s R number combined with an overall increase in the number of positive cases suggests that the disease was becoming more prevalent in the community not just last week but already at the end of August. Although the current data shows that the majority of outbreaks are still mainly localised, whilst during the summer, these were largely confined to very specific workplaces, now they become much larger and involve more general community transmissions. As the number of confirmed cases grows exponentially, linearly increasing the testing capacity will not be enough to quickly identify and stop the spread at localised level, so additional measures such as limiting the number of people allowed to interact will help to delay the rapid growth in cases.”
Dr Konstantin Blyuss, Reader in Mathematics, University of Sussex, said:
“The latest data on R estimates and from the comprehensive swab testing show a really worrying picture about the current situation of coronavirus epidemic in the UK. The first observation is that the infection appears to be growing in all regions of England, and increasing in all age groups. This means that it is not sufficient to focus on isolating or restricting contacts in any one group of people, because effectively the genie is out of the bottle. Another important aspect of these latest data concerns the fact that the R estimates are always lagging the actual data by a period from days to weeks, which means that the reality may indeed be even starker than these figures indicate, because effectively they may reflect a situation one or two weeks ago, and the growth of infection currently appears to be exponential.
“Another important issue concerns the robustness of the latest results in terms of testing of possible coronavirus cases. The latest data from the REACT survey shows that up to 65% of people who tested positive had absolutely no symptoms on the day of testing or the week prior to that, which further highlights the importance of having the possibility of testing asymptomatic people, who may be making a substantial contribution to disease transmission.”
Prof Azra Ghani, Chair in Infectious Disease Epidemiology, Imperial College London, said:
Comment on the Government announcement of R value as well as the REACT-1 study
“As has been evident in the increase in reported cases over the past few weeks, we are now entering a phase of increasing transmission as we head into the autumn. What is particularly concerning is the widespread nature of this increase, with infection no longer confined to localised areas or identified clusters. In this circumstance, broader national social distancing measures (such as those recently announced) are required to interrupt transmission and slow the growth of the epidemic. Such early action to curb spread is critical in order to prevent a subsequent increase in hospitalisations and to protect the NHS as we move into winter.
“The data reported today from the REACT-1 study and the government provide estimates of the reproduction number, R. Both are consistent in indicating that the R estimates are above the critical threshold of 1, and thus indicate increasing transmission; they differ however in the magnitude of this increase. This is due to the different data sources on which the estimates are based. The SPI-M (government) estimates are primarily informed by data from routine testing (Pillar 1 and Pillar 2) which are “responsive” i.e. they represent infections in those seeking testing on the basis of symptoms or through having been identified as a contact. In contrast, the REACT-1 study is based on sequential cross-sectional surveys of a large random sample of the population. There are strengths and weaknesses to both approaches to tracking the epidemic but the important takeaway message remains that we need to reduce transmission now to ensure that we do not enter into a new epidemic phase.“
Dr Thomas House, Reader in Mathematical Statistics, University of Manchester, said:
Comment on the Government announcement of R value as well as the REACT-1 study
“The REACT study is an important part of our response to coronavirus, providing valuable information on prevalence in different areas and population groups Unfortunately, the headline results in today’s press release from the Department of Health and Social Care seem to be designed to generate media interest rather than explain the evidence in a balanced way to a worried population.
“First, while REACT is indeed the research study with the largest number of participants, it involves relatively few positive tests meaning that the weight of evidence presented is not overwhelming and should also be compared to information from testing, hospitals and mortality as well as other studies.
“Secondly, the 7-8 day doubling time is only obtained through treating the most recent wave of the study in isolation – drawing on more data gives a figure of 2-3 weeks and on page 5 of the REACT report they say: “the epidemic increased between rounds 3 and 4, with a doubling time of 17 (13, 23) days corresponding to an R value of 1.3 (1.2, 1.4)” so this overlaps with the Government’s (1.0, 1.2) and is consistent with it – I think this is the correct number to take from REACT-1.
“Finally, the study does not directly address the question of the ‘rule of six’, which would require more detailed information on routes of transmission rather than a random sample design.”
Prof Rowland Kao, the Sir Timothy O’Shea Professor of Veterinary Epidemiology and Data Science, University of Edinburgh, said:
“The REACT-1 study is one of the most important epidemiological indicators of trends in COVID-19 infection because, unlike other markers, such as those based on results from the Pillars 1 and 2 of test and trace, it is largely unbiased and based on a large, representative population sample. Its key results confirm other findings that the resurgence of COVID-19 infections we are currently experiencing (in this study for England, but with similar findings for across the UK) is real and substantial. The estimate of R from this study for the current phase since 22nd August is high (1.7) and suggests that getting COVID-19 will require substantial measures – whether or not the most recent announcements of additional controls is sufficient will only be known in a few weeks time. It must also be remembered that the current data do not reflect the return of students to schools, the imminent return of students to universities, and the likely increase in indoor contacts that will result from cooling weather all of which may have further influences on increasing R.”
Dr Alexander Edwards, an Associate Professor of Biomedical Technology at the University of Reading said:
“This large study complements other ongoing testing schemes such as the ONS survey and the daily government statistics on testing of symptomatic people. It shows with great detail and with clear modelling overlaid with the raw test result data
“There are notable confirmations of several vital points, such as the high proportion (72% overall) of people who had swabs test positive in the community who don’t report symptoms at the time of testing. The overall trend in cases- falling earlier in the summer, then levelling off and starting to rise again into September- is very clear from this study.
“Given the high number of people picked up before they develop symptoms, it’s important to work out fast if rapid community testing could pick people up and help them self-isolate even before symptomatic, reducing the number of contacts more efficiently than current post-symptom testing (and subsequent track and trace).
“What is critical is that such testing programs are used to improve the way we reduce transmission- for example starting to understand what changes make the most difference to infection rates. This kind of community testing is clearly essential for such research, but there still remain gaps in working out how such testing could be incorporated into public health measures.
“For example, while self-swabbing at home is a vital research and surveillance tool, only an estimated 70-80% of infected individuals will test positive in this way (because a single swab test typically has around 70-80% sensitivity using RT-PCR), and so it can’t be assumed that every negative test confirms that individual is not infected. Likewise, it’s vital that each positive test is followed up through contact tracing and public health workers can support individuals in understanding what to do if they are infected.”
Prof Paul Hunter, Professor in Medicine, UEA, said:
“The latest results from the REACT -1 study confirm what has become clear from the daily reports of testing in and that is that the number of cases of COVID-19 in the community in England are increasing and are increasing really rapid. Given that this is a prospective study of a random selection of the population its weighted results are unlikely to be biased by variation in demand for tests. As such this study is likely to give one of the least biased of all estimates of current infections prevalence rates in England. It is important to know that this is prevalence (the proportion of people yielding a positive result) and not incidence (the number of new infections/day) so will appear to be rather higher than would be calculated from daily reports.
“Of particular concern is that the R value for this study is now 1.7 quite a bit higher than other recent estimates and cases numbers may be doubling every 7.7 days. Although the rate of increase and the R value varies from one region to another, cases seem to be increasing in all regions of England. The weighted prevalence over the days 22 Aug to 7 Sep was higher then the sampling periods in June/July and July/Aug but not quite as high as in May. However, given the rapid increase in reported cases since Sunday then it is likely that the real population incidence now is as high as it has been since May. With a doubling rate of just over a week it may not be many days before case numbers exceed those seen in May.”
Dr Simon Clarke, Associate Professor in Cellular Microbiology, University of Reading said:
“This is a massive blow to the government’s strategy to contain the spread of COVID-19, and suggests that the recent uptick in cases is not just because of greater testing. It’s likely that the coronavirus is circulating more freely out in the community again, meaning we are likely to need greater restrictions on our lives to push the transmission rate back down.
“If the R0 number is as high as 1.7, as this data suggests, then we could be at risk of being almost back at square one in terms of our ability to contain the virus. It suggests that the way we are currently living our lives, even with many restrictions still in place, is not enough to keep R0 below 1, and that more stringent measures and behaviours will be needed.
“These data also show that almost two-thirds of the people who were found to currently have the virus, had no symptoms at all. This is good news for them, but means that it could be increasingly difficult to try and contain the virus through a system that relies on ‘test and trace’ of people who show symptoms. If the majority of people who have the virus have no idea they are ill, they are likely to be passing it on to other people in their households and elsewhere, without them even knowing it.
“Young people are more likely to have the coronavirus, but this doesn’t necessarily mean their behaviour is to blame for it – it might just suggest they are more likely to do the kind of jobs and legitimate activities that means they are more vulnerable to picking up the virus. The stricter ‘rule of six’ may help prevent transmission between young people tempted to socialise in larger groups, but these are unlikely to be the only routes of infection.
“While this study is only preliminary, it is a well-designed survey, carried out by some of the country’s leading experts in the field, with a large number of participants. By its nature, it needs to be published quickly to provide the government, health agencies and wider population with the most up-to-date snapshot of how prevalent COVID-19 is in the wider population.”
Dr Zania Stamataki, Viral Immunologist, University of Birmingham, said:
“This detailed study clearly shows how valuable it is for us to continue to monitor infections using testing, and how contact tracing can actually help identify other infected people for isolation even in the absence of symptoms.
“The concerning increase in cases over September compared to the summer months is not unexpected given fewer restrictions, but the Health and Social Care Secretary is right: there are simple actions that each of us can take immediately to limit transmission and we must take personal responsibility.
“There is a real risk that the surge in cases, even amongst the young, will soon translate to increased hospitalisations of the vulnerable.
“We must all take action to limit transmission because we still cannot predict with certainty who will succumb to disease and who will suffer from debilitating long covid for months to come; it is not a simple matter of shielding our elderly.
“The study found that 65% of positive samples came from asymptomatic or presymptomatic people, which shows how important it is to be vigilant as individuals because we don’t know our covid status for certain even if we feel fine. The best thing to do is act like we are potential carriers and keep our distance and wear face coverings in public places to protect those around us.”
Preprint: ‘Resurgence of SARS-CoV-2 in England: detection by community antigen surveillance’ by Steven Riley et al.
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/covid-19
Declared interests
Prof Kevin McConway: “I am a member of the SMC Advisory Committee, but my quote above is in my capacity as a professional statistician.”
None others received.