Real-time estimation of immunological responses against emerging SARS-CoV-2 variants in the UK: a mathematical modelling study
Timothy W Russell,Hermaleigh Townsley,Joel Hellewell,Joshua Gahir,Marianne Shawe-Taylor,David Greenwood,David Hodgson,Agnieszka Hobbs,Giulia Dowgier,Rebecca Penn,Theo Sanderson,Phoebe Stevenson-Leggett,James Bazire,Ruth Harvey,Ashley S Fowler,Murad Miah,Callie Smith,Mauro Miranda,Philip Bawumia,Harriet V Mears,Lorin Adams,Emine Hatipoglu,Nicola O'Reilly,Scott Warchal,Karen Ambrose,Amy Strange,Gavin Kelly,Svend Kjar,Padmasayee Papineni,Tumena Corrah,Richard Gilson,Vincenzo Libri,George Kassiotis,Steve Gamblin,Nicola S Lewis,Bryan Williams,Charles Swanton,Sonia Gandhi,Rupert Beale,Mary Y Wu,David L V Bauer,Edward J Carr,Emma C Wall,Adam J Kucharski
DOI: https://doi.org/10.1016/S1473-3099(24)00484-5
2024-09-11
Abstract:Background: The emergence of SARS-CoV-2 variants and COVID-19 vaccination have resulted in complex exposure histories. Rapid assessment of the effects of these exposures on neutralising antibodies against SARS-CoV-2 infection is crucial for informing vaccine strategy and epidemic management. We aimed to investigate heterogeneity in individual-level and population-level antibody kinetics to emerging variants by previous SARS-CoV-2 exposure history, to examine implications for real-time estimation, and to examine the effects of vaccine-campaign timing. Methods: Our Bayesian hierarchical model of antibody kinetics estimated neutralising-antibody trajectories against a panel of SARS-CoV-2 variants quantified with a live virus microneutralisation assay and informed by individual-level COVID-19 vaccination and SARS-CoV-2 infection histories. Antibody titre trajectories were modelled with a piecewise linear function that depended on the key biological quantities of an initial titre value, time the peak titre is reached, set-point time, and corresponding rates of increase and decrease for gradients between two timing parameters. All process parameters were estimated at both the individual level and the population level. We analysed data from participants in the University College London Hospitals-Francis Crick Institute Legacy study cohort (NCT04750356) who underwent surveillance for SARS-CoV-2 either through asymptomatic mandatory occupational health screening once per week between April 1, 2020, and May 31, 2022, or symptom-based testing between April 1, 2020, and Feb 1, 2023. People included in the Legacy study were either Crick employees or health-care workers at three London hospitals, older than 18 years, and gave written informed consent. Legacy excluded people who were unable or unwilling to give informed consent and those not employed by a qualifying institution. We segmented data to include vaccination events occurring up to 150 days before the emergence of three variants of concern: delta, BA.2, and XBB 1.5. We split the data for each wave into two categories: real-time and retrospective. The real-time dataset contained neutralising-antibody titres collected up to the date of emergence in each wave; the retrospective dataset contained all samples until the next SARS-CoV-2 exposure of each individual, whether vaccination or infection. Findings: We included data from 335 participants in the delta wave analysis, 223 (67%) of whom were female and 112 (33%) of whom were male (median age 40 years, IQR 22-58); data from 385 participants in the BA.2 wave analysis, 271 (70%) of whom were female and 114 (30%) of whom were male (41 years, 22-60); and data from 248 participants in the XBB 1.5 wave analysis, 191 (77%) of whom were female, 56 (23%) of whom were male, and one (<1%) of whom preferred not to say (40 years, 21-59). Overall, we included 968 exposures (vaccinations) across 1895 serum samples in the model. For the delta wave, we estimated peak titre values as 490·0 IC50 (95% credible interval 224·3-1515·9) for people with no previous infection and as 702·4 IC50 (300·8-2322·7) for people with a previous infection before omicron; the delta wave did not include people with a previous omicron infection. For the BA.2 wave, we estimated peak titre values as 858·1 IC50 (689·8-1363·2) for people with no previous infection, 1020·7 IC50 (725·9-1722·6) for people with a previous infection before omicron, and 1422·0 IC50 (679·2-3027·3) for people with a previous omicron infection. For the XBB 1.5 wave, we estimated peak titre values as 703·2 IC50 (415·0-3197·8) for people with no previous infection, 1215·9 IC50 (511·6-7338·7) for people with a previous infection before omicron, and 1556·3 IC50 (757·2-7907·9) for people with a previous omicron infection. Interpretation: Our study shows the feasibility of real-time estimation of antibody kinetics before SARS-CoV-2 variant emergence. This estimation is valuable for understanding how specific combinations of SARS-CoV-2 exposures influence antibody kinetics and for examining how COVID-19 vaccination-campaign timing could affect population-level immunity to emerging variants. Funding: Wellcome Trust, National Institute for Health Research University College London Hospitals Biomedical Research Centre, UK Research and Innovation, UK Medical Research Council, Francis Crick Institute, and Genotype-to-Phenotype National Virology Consortium.