Dengue remains a growing global health concern characterized by four co-circulating serotypes (DENV-1 to DENV-4). Interpretation of cross-sectional serosurveys is frequently limited by antibody cross-reactivity between serotypes, which complicates inference of past exposure and serotype-specific transmission. The authors developed a modelling framework designed to jointly estimate serotype-specific force of infection (FOI) and cross-reactivity using cross-sectional serological data. The stated objective was to expand what can be inferred from routine serosurveys without relying on difficult-to-collect longitudinal post-infection antibody measurements.
The framework is explicitly flexible to the granularity of available serological data. It can be fitted using binary serostatus alone or using binary serostatus in combination with quantitative antibody titres. This flexibility allows application to a wide range of cross-sectional study designs and enables interrogation of both population-level parameters (serotype FOI and cross-reactivity) and, when titres are available, individual-level exposure histories.
In simulation experiments, the model effectively recovered the serotype-specific FOI and cross-reactivity parameters under varying conditions. Simulated scenarios covered a range of endemicities, sampling designs, and sample age coverage. These validation exercises demonstrated the model's robustness to common variations in survey structure and supported its use on real-world cross-sectional datasets.
The authors applied the framework to annual cross-sectional serosurveys conducted in Vietnam between 2013 and 2017. The model produced serotype-specific FOI estimates and reconstructed population immune profiles by age, time, serotype, and number of past infections. The Vietnam analyses served as an empirical demonstration that the model can extract serotype-level transmission signals and exposure histories from routine cross-sectional data.
When applied to the Vietnam data, the estimated serotype-specific FOI patterns were comparable in overall magnitude to estimates produced by analyses that require supplementary longitudinal post-infection antibody measurements. The authors highlight that such longitudinal data are difficult to collect, and that their cross-sectional approach provides a practical alternative for recovering similar-scale FOI estimates from routine surveys.
The analysis revealed asymmetric cross-reactivity between the infecting serotype and heterologous antibody-response serotypes following primary infections. In contrast, post-primary infections were associated with high cross-reactivity. These findings describe differences in the antibody response depending on whether an infection is primary or follows prior exposure, with implications for interpreting serological assays in multi-serotype settings.
Using the jointly estimated FOI and cross-reactivity parameters, the model reconstructed detailed population immune profiles. These profiles were stratified by age, calendar time, infecting serotype, and number of prior infections. The reconstruction resolved probable single-serotype and multi-serotype exposure histories that are not directly distinguishable from observed seroprevalence alone, thereby enhancing the epidemiological interpretability of cross-sectional serology.
From the reconstructed immune profiles, the authors estimated age-specific susceptibility to secondary infection in the Vietnam population. Susceptibility to secondary infection peaked at approximately age 10 for each serotype. Averaged across individuals aged 1–30 years, susceptibility to secondary infection was estimated at roughly 25% for DENV-1 to DENV-3 and around 30% for DENV-4.
Incorporating quantitative antibody titres into the model allowed for inference of individual-level infection histories. Despite extensive cross-reactivity, the analysis detected exposure-history signals in titre profiles, indicating that titres can add discriminating information beyond binary serostatus and enable more resolved reconstructions of past infections.
The authors argue that improved modelling of cross-sectional serology can substantially increase the actionable information that serosurveys provide. By jointly estimating serotype-specific transmission and cross-reactivity, and by reconstructing immune profiles by age and exposure history, cross-sectional data can support risk assessment, burden estimation, and planning for vaccination strategies without requiring large longitudinal cohorts.
All data and code used in the study are publicly available at the authors' GitHub repository: https://github.com/chenyn226/dengue-cross-reactivity. The study used only openly available cross-sectional serosurvey data published prior to the analysis (doi: 10.1371/journal.pntd.0012568), and no additional institutional review board or ethics committee approval was required. The authors declared no competing interests. The manuscript is posted as a preprint and has not been certified by peer review; therefore the findings should not be used to guide clinical practice without further validation.