U.S. flag

An official website of the United States government

NIH: Fogarty International Center NIH: Fogarty International Center
Advancing Science for Global Health
Advancing Science for Global Health
Home > Global Health Matters Mar/Apr 2024 > Charting the evolutionary course of influenza: Insights into epidemic dynamics Print

Charting the evolutionary course of influenza: Insights into epidemic dynamics

March/April 2024 | Volume 23 Number 2

Headshot of  David Spiro.David J. Spiro, PhD, is Director of the Division of International Epidemiology and Population Studies at the Fogarty International Center.

By David J. Spiro, PhD, director of the Division of International Epidemiology and Population Studies at the Fogarty International Center

A recent study in eLife, authored by Drs. Amanda Perofsky and Cecile Viboud from the Fogarty International Center, aims to better understand how antigenic drift—the accumulation of genetic changes—in influenza viruses affects the size and severity of seasonal epidemics each year.

Influenza is a cunning opponent. The flu, short for influenza, is a contagious respiratory illness that affects the nose, throat, and lungs. The viruses that cause the flu continuously accumulate genetic changes to escape detection by your immune system. This process, known as “antigenic drift,” is the reason why you can get sick with the flu each winter, even though you’ve been previously infected or vaccinated. To keep pace with antigenic drift, a network of scientists in over 100 countries monitor evolutionary changes in flu viruses throughout the year. With these data in hand, they convene each year at the WHO to update the flu vaccine, selecting strains they believe will match flu viruses circulating in the upcoming season. These efforts are not always successful, given that flu viruses evolve rapidly, and vaccine strains are selected up to six months in advance

In theory, flu viruses with increased antigenic drift make people more susceptible to infection, leading to more cases and earlier, larger, or more severe epidemics. However, evidence for this in epidemiological data is unclear. What has been well understood is that two influenza virus types, influenza A and influenza B, routinely co-circulate in humans and cause annual outbreaks in the U.S. One subtype of influenza A, named A(H3N2), experiences the fastest rates of antigenic drift, and causes more cases and deaths than other seasonal flu viruses.

What Perofsky, Viboud and their co-authors discovered is significant: Patterns of genetic changes in broad sets of epitope sites (small regions on the surface of antigens that are recognized by immune system components, such as antibodies or T cells) had stronger, more consistent relationships with different measures of flu epidemic dynamics than the "gold standard" serological assays used to measure how flu viruses change from season to season. These assays, laboratory tests that check for the presence of antibodies or other substances in blood samples, are used to develop yearly flu vaccines.

The authors also looked at how epitope changes in the flu virus' two major surface proteins—hemagglutinin (HA) and neuraminidase (NA)—could be used to predict things like epidemic intensity and virus transmissibility. They found that genetic changes of H3, the HA antigen of the A(H3N2) virus, were more strongly linked to larger epidemic sizes, higher viral transmissibility, more cases in adults than children, and a greater number of excess deaths caused by the A(H3N2) virus each season than changes in N2, the NA antigen of the A(H3N2) virus. Meanwhile, antigenic drift of N2 was more strongly associated with increased epidemic intensity ("spikier" epidemic curves) and fewer days from epidemic onset to peak incidence, a measure for the speed of an epidemic, than changes in H3.

Importantly, the researchers also discovered that the level of co-circulation of A(H1N1) viruses was more predictive of the size of A(H3N2) epidemics than viral evolution. So, although antigenic drift in both HA and NA contribute to differences in A(H3N2) epidemics across seasons, subtype interference (the co-circulation of different influenza viruses) is more important than viral evolution when it comes to shaping annual outbreaks.

Headshot of Amanda Perofsky Amanda Perofsky, PhD

Perofsky, Viboud, and colleagues from the Fred Hutchinson Cancer Center, Icahn School of Medicine at Mt. Sinai, U.S. Centers for Disease Control and Prevention (CDC), and WHO analyzed extensive epidemiological, genomic, and serological datasets spanning 22 flu seasons prior to the COVID-19 pandemic and studied how different factors contributed to variability in A(H3N2) epidemics in the U.S. across those years. The team began their painstaking analysis by identifying indicators of viral evolution based on serological assays measuring antigenic similarity between viruses and genetic changes in epitopes of HA and NA. Next, they compiled surveillance data on influenza types and subtypes across regions in the U.S., noting the characteristics of each epidemic. These data included the total number of flu cases and deaths, the timing of epidemic onsets and peaks, and the distribution of illness across age groups.

The authors assessed individual relationships between different indicators of HA and NA evolution across seasons and epidemic characteristics. They also looked at the possibility that A(H3N2) epidemics were affected by the circulation of other influenza viruses (A(H1N1) and B). They used machine learning models to estimate the relative importance of viral evolution, immunity, and subtype interference in predicting regional epidemic dynamics. Machine learning is a specific approach in artificial intelligence where models are trained to learn patterns and relationships in data and make predictions.

Headshot of Cécile Viboud. Cécile Viboud, PhD

Linking flu evolution to epidemics has been a focus of interest for the flu research community for decades, and Perofsky and colleagues' study of this question is the most comprehensive investigation to date. Nonetheless, it has limitations. First, their analysis is limited to one country with a temperate climate and annual epidemics, so its findings concerning flu subtype interactions may not be applicable to tropical and subtropical countries, where the seasonality of influenza is less defined. Second, their measures of influenza activity are derived from coarse, regional CDC data that are publicly available starting with the 1997-1998 flu season. State-level data are not available until 2009, while finer resolution data from electronic health records are not in the public domain at all. Third, public health measures to curb the spread of SARS-CoV-2 severely disrupted the transmission of influenza viruses throughout 2020 and 2021, and population immunity to influenza is expected to have decreased substantially during this period of low circulation. Because this study's data end before the COVID-19 pandemic, it's unclear if its modeling approach will be useful in projecting seasonal flu burden during the post-pandemic period.

Overall, the study's authors found that increased susceptibility to flu occurs during seasons with high antigenic drift. This study is also the first to link antigenic drift in NA to the disease burden, timing, and the age distribution of cases. Because HA elicits a stronger immune response than NA, the researchers had expected HA to demonstrate stronger relationships with seasonal incidence, so this outcome surprised the authors. Currently, NA content is not standardized in vaccines, even though anti-NA antibodies lessen the severity of flu infections.

This study provides support for the inclusion of NA in flu vaccines and highlights the importance of monitoring evolution in both HA and NA to inform vaccine strain selection and epidemic forecasting efforts. Finally, although this study is a retrospective analysis and concludes with the 2018-2019 season, its findings suggest that the size or intensity of future A(H3N2) outbreaks could be projected based on viral genetic changes and A(H1N1) incidence alone.

This study also provides a window into a nascent scientific field that is combining evolutionary analysis of large pathogen genomic datasets with machine learning-based analysis of epidemic dynamics. The goal of this pursuit is to seamlessly combine genomic and epidemiological data so that genomic surveillance can aid in predicting the size, severity, and timing of emerging infectious disease outbreaks. This work by Perofsky and colleagues emphasizes the importance of comprehensive long-term surveillance for answering key questions about the ecology of infectious disease outbreaks, even for pathogens as well studied as influenza, and is indicative of the greater predictive powers to come.

More Information

Updated April 12, 2024

To view Adobe PDF files, download current, free accessible plug-ins from Adobe's website.

Related World Regions / Countries

Related Global Health Research Topics