Figure.

Correlates of protection for symptomatic SARS-CoV-2 infection by variants of concern

The plots show binomial generalised additive model covariate adjusted relative risk for real-time PCR (rtPCR) positive test by antibody marker level and stratified by variant of concern. (A) Relative risk of infection scaled to a reference value of 0·79 BAU/mL (manufacturer-defined positive cutoff index of 0·80 BAU/mL). The blue line indicates the regression point estimate with gray shading representing the 95% CI. The horizontal black dashed line indicates 0·25 relative risk and vertical black arrow the total anti-spike value at the 0·25 relative risk intercept, corresponding to an estimated 75% protection against the respective variant. To control for variable risk of pathogen exposure across the study population, covariates that are or might be associated with exposure were included in the model, including age, sex, month of sample collection, number of COVID-19 vaccine doses, days since last vaccine dose, urban versus rural setting, study site, and number of household residents. Given the non-linear relationship between log transformed antibody titre and risk of infection, the antibody titre covariate was modelled using two degrees of freedom. Case samples used in the models were all collected <5 days after symptom onset and were sequence-confirmed except BA.1, which includes all rtPCR-positive cases during the clearly delineated phase of BA.1 transmission. The number of rtPCR-positive/negative study participants per plot are 42/394 (mu), 84/474 (delta), 54/423 (BA.1), 17/288 (BA.2), and 19/288 (BA.4/5). (B) Plots represent the same generalised additive models, but risk of infection is referenced to a total anti-spike antibody titre of 500 BAU/mL. Unadjusted anti-spike antibody levels by rtPCR result and variant are shown in the appendix (p 3). BAU=binding antibody units.