We started estimating the SARS-CoV-2 reproduction number in the UK using Public Health England data in March 2020, contributing to the Scientific Pandemic Influenza Modelling subgroup (SPI-M) weekly consensus estimates. In December 2020, after B.1.1.7 was identified as a variant of concern, data were shared with the SPI-M research groups which allowed us to identify these infections. These were used by the London School of Hygiene and Tropical Medicine (LSHTM) team to develop estimates of increased transmissibility of B.1.1.7.
We started investigating epidemiological signatures associated with B.1.1.7, including the reproduction number, presentation, and outcome by S-gene status. On 13 January 2021, the LSHTM team presented an analysis to SPI-M which showed a concerning rise in mortality associated with B.1.1.7. However, given the mounting pressure on intensive care units, the risk of confounding resulting from unavoidable compromises in patient care1,2 meant SPI-M could not reach consensus that the new variant was more deadly, based on their work alone.
The matched cohort study is a technique used to estimate vaccine effectiveness, that we had recently employed for the Meningitis B vaccine.1 In this approach, participants are paired with equivalents on multiple characteristics. We decided to use this as a complementary analysis, aimed solely at answering the question of whether the mortality rate was different in B.1.1.7 versus previously circulating strains. By minimizing variation due to hospital pressures, we aimed to support or refute LSHTM’s initial findings.
Once we had decided on this study design, we had preliminary estimates within hours, which, to our dismay, showed a statistically significant increase in mortality risk due to B.1.1.7. Our estimates were higher than LSHTM estimates, with greater uncertainty. These two findings were combined with Imperial College estimates and analysis of admissions data from Scotland by the NERVTAG group,2 and discussed at SAGE, to alert UK government and NHS leaders of the emerging evidence that increased mortality. The process of rapidly conducting the analysis for NERVTAG to guide policy, was challenging, and the possibility of unidentifiable bias in our early results was real. However, if our study was to inform policy, it was better to have an uncertain answer than no answer at all.
An essential next step was to refine our preliminary analysis and make it reproducible. This was a period of rapid and intense collaboration between the different teams and resulted in a better understanding of the data, and how and why our estimates differed. In parallel, data continued to accrue, enabling us to test and revise assumptions made during initial analyses.
We needed to quickly communicate our findings to decision makers in other healthcare settings. We submitted a manuscript for publication despite the knowledge that patterns were continuing to emerge. Throughout the rapid review process we continued to update the data, and refine our research, with the help of the reviewers. It was encouraging to see that our conclusions were robust to additional data. Our convergence with the updated estimates of the team at LSHTM3 and emerging work from other research groups4,5 has supported this finding of increased risk.
The emerging threat of new variants of SARS-CoV-2 requires an agile approach to research, drawing from a broad range of tools and approaches. Developing timely evidence to guide national and international responses to the threat of new variants is not compatible with a prolonged design and implementation of a prospective cohort study. We were lucky that the B.1.1.7 variant was identifiable in Pillar 2 testing, and that enabled us and others to quickly repurpose existing information to answer emerging questions.
There is no substitute for the tried and tested hypothesis-driven approach to research, but in this case, pragmatic access to data enabled researchers from a wide range of institutions to work together and rapidly deliver robust evidence. There is significant value in this capability, but there is inherent tension between this approach and the principles of data protection and minimisation. This needs further discussion if researchers are to continue to develop evidence for pressing clinical and policy decisions in the future.
Robert Challen,1,2,3 Ellen Brooks-Pollock,3,4,5 Jonathan M Read,3,6 Louise Dyson,3,7 Krasimira Tsaneva-Atanasova,1,8 and Leon Danon3,5,8,9
- College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, Devon, UK.
- Somerset NHS Foundation Trust, Taunton, Somerset, UK.
- Joint Universities Pandemic and Epidemiological Research (JUNIPER) consortium.
- University of Bristol, Bristol Veterinary School, Langford, Bristol, UK.
- Bristol Medical School, Population Health Sciences, University of Bristol, Bristol, UK.
- Lancaster Medical School, Lancaster University, Bailrigg, Lancaster, UK.
- The Zeeman Institute for Systems Biology & Infectious Disease Epidemiology Research, School of Life Sciences and Mathematics Institute, University of Warwick, Coventry, UK.
- The Alan Turing Institute, British Library, 96 Euston Rd, London, UK.
- Department of Engineering Mathematics, University of Bristol, UK.
Competing interests: see full declaration on research paper.
- Rodrigues FMP, Marlow R, Simões MJ, et al. Association of Use of a Meningococcus Group B Vaccine With Group B Invasive Meningococcal Disease Among Children in Portugal. JAMA 2020;324:2187–94. doi:10.1001/jama.2020.20449
- Horby P. NERVTAG note on B.1.1.7 severity. NERVTAG (accessed 31 Jan 2021).
- Davies NG, Jarvis CI, Edmunds WJ, et al. Increased mortality in community-tested cases of SARS-CoV-2 lineage B.1.1.7. Nature 2021;:1–5. doi:10.1038/s41586-021-03426-1
- NERVTAG: Update note on B.1.1.7 severity, 11 February 2021. GOV.UK (accessed 14 Feb 2021).
- Grint DJ, Wing K, Williamson E, et al. Case fatality risk of the SARS-CoV-2 variant of concern B.1.1.7 in England. medRxiv 2021;:2021.03.04.21252528. doi:10.1101/2021.03.04.21252528