Global patterns of seasonal influenza activity, duration of activity and virus (sub)type circulation from 2010 to 2020

Abstract Background Seasonal influenza viruses undergo unpredictable changes, which may lead to antigenic mismatch between circulating and vaccine strains and to a reduced vaccine effectiveness. A continuously updated knowledge of influenza strain circulation and seasonality is essential to optimize the effectiveness of influenza vaccination campaigns. We described the global epidemiology of influenza between the 2009 A(H1N1)p and the 2020 COVID‐19 pandemic. Methods Influenza virological surveillance data were obtained from the WHO‐FluNet database. We determined the median proportion of influenza cases caused by the different influenza virus types, subtypes, and lineages; the typical timing of the epidemic peak; and the median duration of influenza epidemics (applying the annual average percentage method with a 75% threshold). Results We included over 4.6 million influenza cases from 149 countries. The median proportion of influenza cases caused by type A viruses was 75.5%, highest in the Southern hemisphere (81.6%) and lowest in the intertropical belt (73.0%), and ranged across seasons between 60.9% in 2017 and 88.7% in 2018. Epidemic peaks typically occurred during winter months in Northern and Southern hemisphere countries, while much more variability emerged in tropical countries. Influenza epidemics lasted a median of 25 weeks (range 8–42) in countries lying between 30°N and 26°S, and a median of 9 weeks (range 5–25) in countries outside this latitude range. Conclusions This work will establish an important baseline to better understand factors that influence seasonal influenza dynamics and how COVID‐19 may have affected seasonal activity and influenza virus types, subtypes, and lineages circulation patterns.


| INTRODUCTION
Influenza is a major respiratory disease that causes significant morbidity and mortality globally. 1,2 Influenza vaccination is the primary intervention to prevent influenza infections and their complications, thus deploying the vaccine efficiently and in a timely manner with regard to the timing of occurrence of influenza epidemics is critical in order to limit the burden of disease of influenza.
The World Health Organization (WHO) has established the Global Influenza Surveillance and Response System (GISRS) with the aim to conduct epidemiological and virological surveillance of influenza globally. 3 The GISRS is instrumental for the WHO influenza vaccine committee in order to advise on the antigenic composition of influenza vaccines for the next influenza season. 4,5 For the Northern hemisphere, influenza virus circulation patterns are reviewed every February, so that the vaccine can be distributed in early autumn, in advance of the winter influenza season. Recommendations for the Southern Hemisphere are usually issued in September, so that the vaccine can be deployed in advance of the immunization campaigns of March-April of the following year. The overall goal is to optimize vaccine formulation and timing of administration to obtain maximal vaccine effectiveness in all countries worldwide. This is particularly challenging, however, for tropical regions, where influenza viruses often circulate year-round and the periods of influenza activity can vary even between neighbouring countries or countries at the same latitude. [6][7][8] The timing of the vaccination is especially critical when considering that vaccineinduced immunity may wane during an influenza season, particularly in the elderly. 9 The picture is further complicated by the frequent, yet largely unpredictable changes in circulating influenza virus strains, which may lead to antigenic mismatch between circulating strains and vaccine strains and ultimately to a reduced vaccine effectiveness. [10][11][12] Finally, the regular seasonal dynamics can be greatly altered by the emergence of new pandemic influenza viruses, as was the case in 2009. More recently, Sars-Cov-2 virus circulation has caused a dramatic reduction in influenza circulation worldwide, most likely as a result of non-pharmaceutical interventions and reduction of global travel. 13,14 The aim of our study was to perform an updated analysis of the global epidemiology of seasonal influenza for the period 2010-2020, describing the patterns of circulation of the different influenza virus types, subtypes, and lineages, and determining the typical peak timing and duration of influenza epidemics for countries in temperate (Northern/Southern hemispheres) and tropical climate regions. While influenza activity is currently low, influenza will become a public health issue again but the timing and severity of future influenza outbreaks is hard to foresee at the moment.   The unit of analysis was the "season," which was defined as the calendar year (from week 1 to week 52/53) in countries situated in the Southern hemisphere and the intertropical belt, or from week 27 of a year to week 26 of the next year for northern hemisphere countries (where influenza epidemics, which typically occur in autumn and winter months, can often bridge two consecutive calendar years).
In what follows, the expression "season 2013" will therefore corre- were not used for Northern hemisphere countries). Countries with fewer than 50 influenza cases in a given season were excluded from the analysis of that season in order to enhance robustness to the study results. 17

| Circulation patterns of influenza viruses
For each country and season, we determined the proportion of influenza cases that were caused by either virus type (A and B). We then calculated its median value, and the proportion of seasons in which either influenza virus type accounted for ≥80%, ≥50 to 80%, ≥20 to 50%, or < 20% of all reported influenza cases, in each country and for countries belonging to each latitudinal area and WHO region. Furthermore, for each season in each country we determined the proportion of influenza cases that were caused by each type A virus subtype (H1N1, H3N2, and other/unsubtyped) and type B virus lineage (Victoria, Yamagata, and uncharacterized). The non-parametric Kruskal-Wallis test was applied to compare median proportions between countries belonging to different latitudinal areas or WHO regions.

| Peak timing and duration of influenza epidemics
The determination of the timing of the primary and secondary peak of influenza circulation was conducted by analysing country-specific influenza times-series using the EPIPOI software. 18 Because our objective was to define the "typical" timing of the epidemic peak, we included only countries with five or more seasons with at least 50 reported influenza cases. The season 2020 (i.e., the calendar year for countries in the Southern hemisphere and the intertropical belt, and the season 2020-2021 for Northern hemisphere countries) was not included in this analysis because it was particularly atypical due to the COVID-19 pandemic. In addition, because EPIPOI needs the same number of time points (in our case weeks) to operate, we excluded all 53rd weeks in our dataset (which only applied to years 2014 and 2020). EPIPOI proceeds by first detrending the country-specific time series using a quadratic polynomial, and it then works out the periodic annual function (PAF) of the time series by summing up the annual, semi-annual, and quarterly harmonics as obtained by Fourier decomposition. The typical timing of the peak corresponds to the peak month of the PAF. The amplitude of the PAF is obtained as the ratio of the wave height to the peak value and can be interpreted as an estimate of the intensity of seasonality of influenza epidemics in a given country (i.e., as a measure of how much influenza cases in a given country tend to occur over a short period of time instead of being distributed over the entire season). Owing to how it is calculated, the amplitude can sometimes take values above 100%, particularly when the time-series fall to zero for a certain number of weeks, for example, in summer months of temperate countries. (for more details, refer to Alonso and McCormick, 2012 19 ).
The duration of the influenza epidemic in each season was defined using the average annual percentage (AAP) method, according to which the duration is defined as the shortest stretch of consecutive weeks that account for at least 75% of all influenza cases that were reported in the season. 20 As for the determination of the typical timing of the epidemic peak, this analysis was conducted by including only the 122 countries that have data available for five or more seasons with ≥50 reported cases, and after discarding the season 2020 because of the COVID-19 pandemics.

| Software
All analyses were conducted using Stata version 15 (Stata Corp, College Station, TX) and the freely available analytical software EPIPOI. 18,19 3 | RESULTS

| Circulation patterns of influenza viruses
In total, data were available for 149 countries ( (Table S1). The median percentage of influenza cases that were caused by type A viruses was 75.5%. More in detail, the proportion of all reported influenza cases that were caused by type A viruses was ≥80% in 42.3% of seasons, ≥50% and <80% in 42.3% of seasons, ≥20% and <50% in 13.7% of seasons, and <20% in only 1.7% of seasons. These figures varied somewhat geographically (p value for difference of the median proportion of A cases between latitudinal areas <0.001) ( Table 1). The median proportion of influenza A cases and the percentage of seasons in which type A influenza viruses caused ≥80% cases were lowest in countries of the intertropical belt (73% and 38.8%, respectively, based on data from 66 countries and 528 seasons), and highest in countries in the Southern hemisphere (81.6% and 55.6%, based on data from six countries and 63 seasons). When grouping countries according to the six WHO regions (   Table S4.

| Peak timing and duration of influenza epidemics
The association between the median duration of the influenza epidemics and the country's latitude is depicted in Figure 5, and the complete data are available in Table S4. By visually inspecting  influenza epidemics and the country latitude for countries whose centroid was situated at latitudes above 40 N (the more northern the country, the longer duration of influenza epidemics, p value <0.001).

| DISCUSSION
We Colour bar represents the intensity of influenza incidence activity, from high (red) to low (blue). Monthly incidence counts were standardized annually, and shown as the proportion of the maximum number of cases in a month for that country and period.