Antigenic variation of influenza viruses and its impact on seasonal transmission of influenza : a multi-disciplinary approach in incorporating environmental and genetic measurements in an influenza epidemiology study in Hong Kong

In tropical and subtropical regions, influenza shows a less clear seasonal pattern than in temperate regions, with multiple peaks observed throughout the year. Understanding the mechanism of influenza outbreaks could facilitate the timely adoption of public health response and control measures. However the complicate seasonality of influenza in warm climate brings difficulty in modelling and forecasting the virus peaks. Previous studies have explored the impacts of meteorological factors and human behaviour on influenza seasonality, but there is no studies so far that have investigated the impact of another potential seasonal factor, antigenicity change of influenza viruses, likely due to the lack of consecutively collected genetic and antigenicity data. Besides seasonality, antigenic variation of influenza virus (AVI) has also been associated with disease burden for influenza, but evidence has also been limited. In my study I was trying to address the following research questions: 1) Is antigenic change associated with influenza disease burden? 2) Does antigenic change of influenza coincide with the surge of influenza seasonal activities? 3) Is antigenic change positively associated with virus transmission rates? To assess the association between annual disease burden and antigenic change, I first calculated annual antigenic changes of influenza for the seasonal dominant subtype A(H3N2). Then I used a Poisson regression model to calculate annual rates of excess mortality and hospitalization attributable to influenza in Hong Kong from 2001 to 2012. In the period 2001-12, antigenic changes of subtype A(H3N2) were found positively and highly correlated with both excess mortality and hospitalization rates (per 1,000,000 population) associated to A(H3N2) virus in cool seasons (October to March), with Spearman correlation coefficients from 0.655 to 0.809 for different disease. To answer the second question, I utilized a systematic sample of influenza A cases collected in Hong Kong to explore the temporal trend of antigenic changes. During 2013-14, a total of 2,115 samples were collected. Influenza A positive specimens were subsequently sub-typed into A(H3N2) and A(H1N1)pdm09 and were systematically selected for sequencing. Amino acid changes were found for both subtypes, but none of them were identified as antigenic drifted strains from the WHO vaccine composition strains in HI tests. No significant difference in antigenic variations was observed across age and gender, although further studies with longer and larger age-specific sequence and antigenicity data are warranted for better understanding the evolution of circulated influenza A virus particularly in tropical and subtropical regions. The association of antigenic changes with seasonal influenza transmission rates, which were measured by time-varying reproduction numbers Rt during 2013-14, was subsequently explored by comparing classical linear models with and without antigenic changes included as one of driving factors. The estimated reproduction numbers generally had a reverse linear relationship with the environmental factors for temperature and relative humidity in A(H1N1)pdm09 and for temperature in A(H3N2). My findings suggest that incorporation of antigenic variations and environmental factors could improve the performance of the mathematical model for influenza transmission rate, and potentially could contribute to a better prediction of seasonal influenza outbreaks.published_or_final_versionPublic HealthDoctoralDoctor of Philosoph