Longitudinal survival methods for biomedical research

Our research is focused on survival methods which incorporate longitudinal or genomic data. We will focus on the following two main projects.
Correlates of protection for Covid-19
Brief description
Vaccination with a Covid-19 vaccine produces an immune response against Covid-19 infection, including the production of antibodies. The levels of antibodies found in the blood decays over time, as immunity wanes. Feng et. al (2021) reported the association between vaccine-induced antibody levels 28 days after a second vaccination of the ChAdOx1 nCoV-19 (Oxford-AstraZeneca) vaccine, and subsequent relative risk of Covid-19 infection compared to a control group. We propose to jointly model the waning antibody levels over time with the relative risk of Covid-19 infection.
Potential impact
The research could have the following impact:
1. Provide more information on what antibody level is needed to prevent infection. This could help regulators decide whether to licence new Covid-19 vaccines based on their immunology results.
2. Provide more information on the waning of immunity. This could help inform policy makers on when a booster dose may be required. It could further help provide information on the levels of immunity in the population and so the risk of future increases in infections, hospitalisations, and deaths.
Aims and objectives
1. To produce a Bayesian joint model for immune markers after Covid-19 vaccination such as antibody titres, and the time until Covid-19 infection. The model should account for any left censoring of immune marker measurements due to a lower limit of detection. It should further allow multiple antibody assays to be modelled jointly, along with the relative risk of infection.
2. To apply the model to data from the ChAdOx1 nCoV-19 vaccine trials COV002 and COV003. To output summaries showing how:
a. instantaneous antibody levels affect the instantaneous relative risk of Covid-19 infection (primary analysis),
b. vaccine efficacy changes over time since second vaccination (secondary analysis),
c. antibody levels change over time since second vaccination (secondary analysis).
Novelty of the research methodology
Standard joint models for longitudinal and survival data assume the survival component depends solely on the longitudinal process. This problem is more complicated - the survival component depends on the longitudinal process (antibody decay) and on the changing background exposure rates to Covid-19 for the subjects, due to the evolving epidemic. We will develop a novel method to incorporate this changing rate of exposure in the joint model. This might be done by jointly modelling the infection rates over calendar time in the control group and the vaccine group alongside the longitudinal antibody process in the vaccine group. To our knowledge no-one has previously developed a joint model which incorporates changing risk of disease in the population.
It may also be possible to adjust the model to allow for changing vaccine efficacy due to a change in the circulating variant of Covid-19 over time.
The project will be in collaboration with researchers from the Oxford Vaccine Group (OVG), Department of Paediatrics, University of Oxford. The project will report data from the COV002 and COV003 trials run by OVG. Oxford University has entered into a partnership with AstraZeneca for further development of ChAdOx1 nCoV-19, who were one of the funders for the COV002 and COV003 trials. AstraZeneca will review the data from the study and the final manuscript before submission, but the academic authors will retain editorial control.
This project falls within the EPSRC Statistics and applied probability research area in the Mathematical Sciences theme, and the Optimising Disease Prediction, Diagnosis and Intervention research area in the Healthcare Technologies theme.