Age-seroprevalence curves for the multi-strain structure of influenza A virus

Abstract

AbstractThe relationship between age and seroprevalence provides the simplest and least expensive approach to computing the annual attack rate of an infectious disease. However, many pathogens circulate as multiple serologically distinct strains, with no single assay able to determine seropositivity or seronegativity to an entire clade or family of co-circulating pathogens. An approach is needed to describe population exposure to an antigenically variable group of pathogens without focusing on any particular strain or serotype in the group. In this study, we focus on the two-subtype multi-strain taxonomy of human influenza A virus. We describe a data set of 24,402 general-population serum samples collected in central and southern Vietnam between 2009 to 2015, and assayed for influenza HA1 antibodies to eleven different strains of human influenza A (both H3 and H1 subtypes). We find that a principal components decomposition of the data results in the first principal component PC1 being an appropriate surrogate for seroprevalence (or composite antibody titer) which can be further decomposed for H1 and H3 contribution to the serological profile. Using this approach, we are able to provide the first ELISA-based standardized measurements of serology to reconstruct population exposure history, which correlates well with known influenza epidemiology. Annual attack rates in Vietnam are estimated at 25.6% (95% CI: 24.1% – 27.1%) for H3 and 16.0% (95% CI: 14.7% – 17.3%) for H1, with some variation in location-specific attack rates. The remaining principal components act as descriptors of influenza history and sort the population by birth year. The novel contributions of this analysis are (1) the introduction of dimensionality reduction on human antibody profiles to construct an age-seroprevalence relationship for an antigenically variable pathogen, (2) an analysis of >24,000 individuals, using >260,000 serological data points in total, allowing us to construct age-seroprevalence relationships with the precision of modern big data studies, and (3) a description of long-term attack rates in a non-vaccinated setting showing the natural history of influenza A virus in a densely-populated tropical country.