The covid-19 elimination debate needs correct data
BMJ 2020; 371 doi: https://doi.org/10.1136/bmj.m3883 (Published 06 October 2020) Cite this as: BMJ 2020;371:m3883
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Dear Editor
Accurate assessment of the chance of death following exposure to covid-19 is important, since this statistic has been used by governments to direct lockdowns in an effort to mitigate these deaths. In correspondence to the BMJ, professors Baker and Wilson, both proponents of New Zealand’s lockdown, have criticised our assertion that the infection fatality proportion (IFP) of covid-19 is similar to seasonal influenza.[1] They claim that the IFP for influenza is 0.039%, about six times lower than the IFP for covid-19 we cited, a corrected median value of 0.23%.[2]
Clearly, mortality is age-stratified from covid-19. The corrected median estimates of IFP for people aged lower than 70 years is currently 0.05%, [2] which, for the population less vulnerable to deaths, is similar to influenza. However overall estimates for covid-19 are higher, due to the higher fatality rate in elderly people.
Here, we explore in more detail the assessment of the IFP for the two viruses and focus on four issues we believe are important:
1. Modelled COVID-19 death has been overestimated
Early in the course of covid-19, the all-age IFP was high, estimated at 0.66% (credible interval: 0.39 to 1.33%).[3] This led to alarming projections of covid-19 deaths, if the infection was left to spread. These IFPs were obtained by adjusting case-fatality proportions by the proportions of returnees testing positive from repatriation flights from Wuhan. This led to 250,000 deaths overall being predicted for the UK,[4] justifying lockdowns. Yet observed fatalities in the UK now show that these models overestimated deaths by seven times.[5] Similar models in New Zealand predicted 80,000 deaths from the pandemic if severe lockdowns were not enacted.[6] Barnard et al. estimated between 12,600 and 33,600 deaths based on a ‘case-fatality ratio’ of 0.75% and 2% respectively— even with lockdowns.[7] We now know that these projections were too high, and that lockdowns are of questionable value for reducing per capita mortality.[8]
2. Estimates of IFP
For different diseases, the IFP is estimated in varying ways. For covid-19, it is the ratio of the cumulative count of clinically assigned covid-19 deaths to the number of infected people. Frequently, the number of infected people is estimated by the product of the prevalence of antibody positive cases and the population count. The prevalence of seropositive subjects is assumed to equate to a cumulative assessment of viral infections.
Baker and Wilson’s estimate for influenza is derived somewhat differently. The IFP for influenza is derived from a modelled annual influenza mortality rate,[9] divided by the influenza seropositive prevalence.[10] This is unlike that for covid-19. For this disease, deaths have been attributed individually based on mass testing carried out on an unprecedented scale for any respiratory disease. While on the face of it this should increase confidence in case numbers, history suggests this new testing regime, with sharpened focus on covid-19, is likely to overestimate mortality, as we will discuss.
3. Death ascertainment
New pandemics are often associated with biased changes to cause of death coding. For example, in the US in 1968-69, where doctors were aware of an impending influenza A (H3N2) winter, the number of deaths coded as influenza in the summer of 1968 increased sixteen-fold when compared with the summers in the years before and after the pandemic.[11] Yet no significant circulation of influenza was thought to have occurred during that summer. Due to this inaccurate death certificate recording for influenza, the authors statistically estimated these deaths, independent of death records, as has the data referred to by Baker and Wilson.
We see similar evidence for over counting of deaths in countries with high IFPs for covid-19. A notable example was England where it was impossible to recover from covid-19 once an individual had tested positive.[12] Reports from Italy have shown a similar bias in favour of covid-19 death early in that pandemic. After formal review of apparent covid-19 deaths only 12% of the previous figures were directly attributable to the new virus.[13]
Evidence for lower mortality comes from countries that have many covid-19 cases yet few deaths. At the time of writing, Singapore had 57,883 recovered cases and 28 deaths, yielding a case-fatality proportion of 0.05%. We believe, because of Singapore’s adherence to the case definition when assigning covid-19 deaths,[14] its numbers more reliably assess mortality, and illuminate the bias present elsewhere. Further, the denominator is large, likely due to aggressive testing. If serology were estimated, the IFP of this city state would likely be lower still.
4. Infection prevalence
As well as bias in the numerator, the denominator in covid-19 IFP calculations is likely to be lower than true infection counts, because positive antibody responses wane faster than for influenza.[15 16] This leads to underestimation of cumulative infection and consequently an overestimation of the IFP. Supporting evidence comes from levels of positive antibody tests halving after two months in a cohort of exposed health care workers from Nashville.[17] In contrast, high levels of influenza antibodies have been documented up to 28 weeks after vaccination in healthy adults in Maryland.[16] Further, evidence of exposure to covid-19 may be only detectable in specific T-cells (reactive to spike glycoproteins), rather than in antibodies alone.[18]
Other support for a low IFP for covid-19 come from studies which track serial antibody tests within individuals. For example, an eightfold increase in positive antibody prevalence in Tokyo occurred during summer, rising from 5.8% to 46.8%, yet little increase in fatality from the virus occurred.[19]
So, what is a reasonable IFP for covid-19? The overall corrected median IFP from 61 studies in a meta-analysis is 0.23%.[2] This agrees with a population serosurvey in Indiana.[20] These studies consider only seropositivity as an indicator of cumulative exposure to the virus. They have also assumed cause of death figures are accurate. So, once these factors have been considered, we believe that our comparison with seasonal influenza is not misleading. Since models that have incorporated higher IFPs have led to economically crippling lockdowns,[21] we believe scrutiny of these comparisons are vital and a reappraisal of the covid-19 IFP is overdue.
Simon Thornley, Section of Epidemiology and Biostatistics, The University of Auckland
Arthur J. Morris, LabPLUS, Auckland City Hospital
Gerhard Sundborn, Section of Pacific Health, The University of Auckland
Samantha Bailey, Clinical & Pharmaceutical Research Trust, 40 Stewart Street, Christchurch.
References
1. Baker MG, Wilson N. The covid-19 elimination debate needs correct data. BMJ 2020;371 doi: https://doi.org/10.1136/bmj.m3883
2. Ioannidis J. The infection fatality rate of COVID-19 inferred from seroprevalence data. Bull World Health Organ 2020
3. Verity R, Okell LC, Dorigatti I, et al. Estimates of the severity of COVID-19 disease. medRxiv 2020 doi: https://doi.org/10.1101/2020.03.09.20033357
4. Ferguson NM, Laydon D, Nedjati-Gilani G, et al. Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand. 2020. DOI 2020;10:77482.
5. National Health Service. COVID-19 Daily Deaths 2020 [Webpage]. Available from: https://www.england.nhs.uk/statistics/statistical-work-areas/covid-19-da... accessed 7/11 2020.
6. James A, Hendy SC, Plank MJ, et al. Suppression and mitigation strategies for control of COVID-19 in New Zealand. medRxiv 2020 doi: https://doi.org/10.1101/2020.03.26.20044677
7. Telfar Barnard LT, Wilson N, Kvalsig A, et al. Modelled Estimates for the Spread and Health Impact of Covid-19 in New Zealand: Revised Preliminary Report for the NZ Ministry of Health. Wellington: University of Otago, 2020.
8. Chaudhry R, Dranitsaris G, Mubashir T, et al. A country level analysis measuring the impact of government actions, country preparedness and socioeconomic factors on COVID-19 mortality and related health outcomes. EClinicalMedicine 2020;25:100464.
9. Khieu TQ, Pierse N, Telfar-Barnard LF, et al. Modelled seasonal influenza mortality shows marked differences in risk by age, sex, ethnicity and socioeconomic position in New Zealand. J Infect 2017;75(3):225-33.
10. Huang QS, Bandaranayake D, Wood T, et al. Risk factors and attack rates of seasonal influenza infection: results of the Southern Hemisphere Influenza and Vaccine Effectiveness Research and Surveillance (SHIVERS) seroepidemiologic cohort study. J Infect Dis 2019;219(3):347-57.
11. Thompson WW, Moore MR, Weintraub E, et al. Estimating influenza-associated deaths in the United States. Am J Public Health 2009;99(S2):S225-S30. doi: https://doi.org/10.2105/AJPH.2008.151944
12. Loke Y, Heneghan C. Why no-one can ever recover from COVID-19 in England – a statistical anomaly: Centre for Evidence Based Medicine; 2020 [Webpage]. Available from: https://www.cebm.net/covid-19/why-no-one-can-ever-recover-from-covid-19-... accessed 28/10 2020.
13. Newey S. Coronavirus: Is Covid-19 really the cause of all the fatalities in Italy? Stuff.co.nz: Stuff; 2020 [Webpage]. Available from: https://www.stuff.co.nz/national/health/coronavirus/120443722/coronaviru... accessed 29/10 2020.
14. Geddie J, Aravindan A. Why is Singapore's COVID-19 death rate the world's lowest [Webpage.]. Reuters.com: Thomson Reuters; 2020 [News report]. Available from: https://www.reuters.com/article/health-coronavirus-singapore-explainer-i... accessed 28/10 2020.
15. Weis S, Scherag A, Baier M, et al. Seroprevalence of SARS-CoV-2 antibodies in an entirely PCR-sampled and quarantined community after a COVID-19 outbreak-the CoNAN study. medRxiv 2020 doi: https://doi.org/10.1101/2020.07.15.20154112
16. Clements ML, Murphy BR. Development and persistence of local and systemic antibody responses in adults given live attenuated or inactivated influenza A virus vaccine. J Clin Microbiol 1986;23(1):66-72.
17. Patel MM, Thornburg NJ, Stubblefield WB, et al. Change in Antibodies to SARS-CoV-2 Over 60 Days Among Health Care Personnel in Nashville, Tennessee. JAMA 2020 doi: https://doi.org/10.1001/jama.2020.18796
18. Braun J, Loyal L, Frentsch M, et al. SARS-CoV-2-reactive T cells in healthy donors and patients with COVID-19. Nature 2020:1-5. doi: https://doi.org/10.1038/s41586-020-2598-9
19. Hibino S, Hayashida K, Ahn AC, et al. Dynamic Change of COVID-19 Seroprevalence among Asymptomatic Population in Tokyo during the Second Wave. medRxiv 2020:2020.09.21.20198796. doi: https://doi.org/10.1101/2020.09.21.20198796
20. Blackburn J, Yiannoutsos CT, Carroll AE, et al. Infection fatality ratios for COVID-19 among noninstitutionalized persons 12 and older: results of a random-sample prevalence study. Ann Intern Med 2020
21. Joffe A. COVID-19: Rethinking the Lockdown Groupthink. Preprints 2020 doi: https://doi.org/10.20944/preprints202010.0330.v2
Competing interests: ST and AM have provided paid advice to Auckland International Airport Ltd related to health risks associated with covid-19.
Estimates of the covid infection fatality rate
Dear Editor,
The covid infection fatality rate risk or ratio (IFR), the proportion of those infected with covid who subsequently die, is of crucial importance in assessing how risky it would be to relax covid restrictions and thereby allow a greater infection rate. Michael Baker (BMJ 2020;371:m3883) suggests a figure of at least 0.65%, whereas Morris et al. respond with a figure of 0.23%. Naturally, there will be some variation across the world due to various factors, most notably the age stratification of the population. Nevertheless, some general estimates of a global figure might be useful, and a reality check on these estimates could be done using national data.
Using the above authors’ estimates with the UK data it is hard to see how the CFR for covid could be as low as the Morris et al. suggestion of 0.23%. According to the official figures of the UK government (https://coronavirus.data.gov.uk/), to date there have been 144,810 deaths in the UK within 28 days of a positive covid test. It could be argued that some of those who died within the 28 day period might have died of other causes in any case. On the other hand, particularly early in the pandemic when testing was not generally available, many covid deaths will not be have been included in this statistic, so the true figure for covid deaths may be a lot higher. Indeed, there are reported to have been 167,927 deaths with COVID mentioned on the death certificate. In any case, the mortality peaks coinciding with waves of covid infection cannot be denied.
144,810 covid deaths with a CFR of 0.23% would require that 93% of the population (including children and those self-isolating) had been infected. A figure this high is hard to believe. Currently 88.6% of the UK population have been vaccinated against covid with one dose of vaccine and over 80% have had two doses. Yet for the last five months the infection rate in the UK has been remained somewhat high with around 40,000 new positive covid test results every day. It is difficult to see how this could be the case if 93% of the population have already had the disease. By contrast, a CFR of 0.65%, as suggested by Baker, would require that 33% of the UK population have already been infected with covid. A third of the population still seems a high proportion of the population to have already had covid, but it is more plausible than 93%.
Competing interests: No competing interests