Fig. 1: Probability of dengue by baseline antibody measure.

The probability of disease is shown separately for baseline IgG ELISA (A–C) and baseline GMT measured by PRNT with standard reference (D–F) or mature clinical (G–I) strains. The probability of each disease outcome was modeled as a function of baseline antibody titer on both discrete and continuous scales. All continuous relationships were modeled using Poisson generalized additive models (continuous black line) with 95% confidence intervals (gray shading). Point estimates and confidence intervals correspond to predicted probabilities from logistic regression models. P-values were generated from the logistic regression models using two-sided tests comparing the dengue risk for each antibody bin with that in the naïve group. Only statistically significant p-values are shown. The dashed line indicates the disease probability in naïve individuals. DENV IgG ELISA was measured on all participants (n = 1,206), and GMTs were measured on random subsets using standard reference (n = 823) and mature clinical strains (n = 293). Inverse probability weighting was used to adjust for GMT subset size. All models were adjusted for age, sex, and enrollment site, and model estimates are shown for the average study participant (female, age 10, from Bogo). N for each row indicates the number of dengue cases and non-cases included in each analysis. Source data are provided as a Source Data file.