Intended for healthcare professionals

Letters Covid antibodies from infection or vaccination

Overcoming spectrum bias for accurate SARS-CoV-2 seroprevalence estimates

BMJ 2021; 373 doi: https://doi.org/10.1136/bmj.n917 (Published 12 April 2021) Cite this as: BMJ 2021;373:n917
  1. Milo A Puhan, professor of epidemiology and public health1,
  2. Arnaud Chiolero, professor of public health2 3,
  3. Jan Fehr, professor of health and travel1,
  4. Stéphane Cullati, senior lecturer in epidemiology2
  5. On behalf of the Corona Immunitas Research Group
  1. 1Epidemiology, Biostatistics and Prevention Institute, University of Zurich, Hirschengraben 84, 8001 Zurich, Switzerland
  2. 2Population Heath Laboratory (#PopHealthLab), University of Fribourg, Fribourg, Switzerland
  3. 3School of Population and Global Health, McGill University, Montreal, Canada
  1. miloalan.puhan{at}uzh.ch

Repeated population based SARS-CoV-2 seroprevalence studies provide evidence on how immunity develops in a population. They are especially relevant now, as we are at a turning point with vaccines that greatly accelerate the development of immunity.

A surprisingly low seroprevalence of 14% was recently reported for the UK.1 The cumulative proportion of confirmed infected people in the UK (6.3%) is very close to that of Switzerland (6.6%).2 Assuming that only around one in 3-4 infected people are diagnosed,3 a seroprevalence of 20-25% would be expected, without people who have been vaccinated. This is consistent with what has been observed for different cantons of Switzerland in the Corona Immunitas programme.4 Switzerland is experiencing a slow start to its vaccination programme, and the effect of vaccination on seroprevalence estimates is negligible. In the UK, however, with 33 doses given per 100 people2 and the cumulative proportion of infected people, the seroprevalence should be higher than 14%.

One possible reason for this low estimate is the use of self-administered lateral flow tests in the React-2 study, so seroprevalence estimates were adjusted for the sensitivity (84.4%) and specificity (98.6%) of these tests.2 The test performance estimates were based on the analyses of clinically ill people with SARS-CoV-2 infection (cases) and pre-pandemic samples (controls), leading to a substantial spectrum bias.5 This is a major problem for population based studies as many people with past infections have had a mild or asymptomatic course, which is more difficult to detect than moderate to severe infections.

Commercially available tests based on venous blood miss up to 40% of infections,6 and lateral flow tests are clearly even less accurate and probably have a sensitivity lower than 84.4% in population based samples. In Switzerland, Corona Immunitas chose a sophisticated test, with test performance estimates based on a population based sample.6 This is more laborious and costlier but gives a more accurate picture of immunity development in a population.

Footnotes

This article is made freely available for use in accordance with BMJ's website terms and conditions for the duration of the covid-19 pandemic or until otherwise determined by BMJ. You may use, download and print the article for any lawful, non-commercial purpose (including text and data mining) provided that all copyright notices and trade marks are retained.

https://bmj.com/coronavirus/usage

References

View Abstract