Statistical analyses
The reproducibility of the assay was assessed from sera of four individuals, which were measured in each analysis run (thus providing six replicates), by assessing the coefficient of variation (CV) and intraclass correlation coefficient (ICC) for each antigen.
To meet the assumptions of regression-based methods, RSV and RV-A/B/C were normalized by using the Ordered Quantile Normalization method (21) in the pooled EGEA1 and EGEA2 measured samples.
To assess the associations between potential determinants and IgG responses to RSV and RV-A/B/C, multiple mixed linear regression models considering age, sex, BMI, tobacco smoking (passive in children, active in adults) and season of blood sampling were performed, adjusted for ever-asthma and allergic sensitization. Family correlation was considered as a random effect. The analyses were performed separately in the EGEA1 and EGEA2 populations to assess these associations in children and in adults, respectively.
Interaction between age and sex was tested by adding an interaction age*sex term to the previous models.
The main analysis and the analysis testing the age*sex interaction were then performed with stratification on asthma status. Indeed, due to the case-control design of EGEA cohort, collider bias can be an issue when investigating, as in our study, associations between factors (some of the personal factors and RSV and RV-specific IgG levels) that may both influence asthma. Results from separate analyses among participants with and without asthma may inform about the presence of such a bias.
Finally, among individuals with RSV and RV-specific IgG responses measured at both time points, associations between time elapsed during the two follow-ups and the evolution of IgG level over time was assessed using linear mixed models to account for repeated data, adjusted for age at first survey, sex, BMI, exposure to tobacco, season of blood sampling, allergic sensitization and ever-asthma.