Fig. 2: Mediation analysis of antibody effects on infection and symptomatic infection through prior infection.

The first column depicts the diagrams of the mediation framework analysis. The second column presents the corresponding estimates for each antibody measure (homotypic-neutralizing, D614G-neutralizing, and spike-binding antibodies), displaying the total effect, average direct effect, and average causal mediating effect from any prior infection on protection against infection and symptomatic infection. Each mediation analysis was adjusted for age and prior vaccination. Protection effects are expressed as percentages ([1-RR] x100). Point estimates (means) are indicated by symbols (circle, triangle, square, or cross), with solid horizontal lines representing the 95% confidence intervals (CIs). Dashed line indicates null effect. Mediation effects were estimated using Poisson regression models, with statistical significance assessed via non-parametric bootstrapping with 1000 simulations and accelerated bias-corrected (BCa) 95% confidence intervals (two-sided tests). P-values are indicated for statistical significance, with ***, and * corresponding to P-values of <0.001, <0.010, and <0.050, respectively; no adjustments for multiple comparisons were applied; exact P-values are shown in Table S11. Data were derived from all 345 study participants (237 infected vs 108 uninfected, 134 symptomatic vs 211 not symptomatic), with one pre-exposure sample per participant. Neutralizing titers (ID₅₀) were obtained by averaging inhibition values at each dilution across technical triplicates of each sample and fitting a 5-parameter log-logistic model (see Methods).