Seasonality and periodicity of influenza virus in Chengdu
10-year surveillance showed periodicity transitions of composite influenza virus activity in Chengdu, a subtropical city in southwestern China. Composite influenza activity changed from semi-annual cycles during pre-pandemic seasons to annual cycles since the 2009 pandemic. Moreover, semi-annual cycles were observed again during 2014/15–2015/16 when A/H3N2 major antigenic drift groups, A/Switzerland/9715293/2013-like (SW13) and A/HongKong/5738/2014-like (HK14) [
25], predominated. These A/H3N2 strains dominated in 2014/15 in many countries and caused massive morbidity and mortality because of the antigenic variation [
26]. This seasonal pattern was different from the temperate regions and many tropical/subtropical areas with well-defined annual or semi-annual peaks [
3‐
5]. Virological data of influenza virus demonstrated that an annual cycle with one peak in winter was noticeable in the temperate region of China, including Beijing and Shandong province [
27,
28]. Influenza virus peaked in summer in Guangdong province in subtropical southern China [
29]. However, seasonal pattern was not determined in Chongqing which is adjacent to Sichuan province [
30]. A number of factors might drive influenza circulation patterns, including viral evolution, host susceptibility, demographic characteristics, and environmental conditions. An epidemiological study during the pre-pandemic period in Hong Kong found non-stationary seasonality of composite influenza virus activity, which might be caused by a lack of herd immunity when significant antigenic changes in A/H3N2 emerged and predominated [
31]. However, the seasonality and interaction of individual influenza subtypes/lineages remain to be investigated.
By analyzing the time series of individual influenza A subtypes and B lineages across age, the differences in seasonality among influenza subtypes/lineages and the different response of resident subtypes/lineages when a novel pandemic strain or new antigenic groups emerged and predominated were observed. Our study firstly found periodicity transitions of A/H3N2, which were similar to composite influenza activity. Semi-annual periodicity of A/H3N2 shifted to an annual cycle since the 2009 pandemic and became evident again coinciding with the emergence and predominance of A/H3N2 major antigenic drifts in 2014/15–2015/16. The periodicity transmissions of A/H3N2 have not been reported previously and are possibly related to the emergence and predominance of the novel pandemic subtype or major antigenic drifts in a largely susceptible population. Secondly, after the pandemic, the pre-existing seasonal A/H1N1 was replaced by A/H1N1pdm09 as observed in other studies [
12,
22]. Thirdly, A/H1N1pdm09 generally exhibited an annual cycle peaked in winter or early spring except for 2011/12 and 2014/15 when this virus subtype was not detected. ‘Skip-and-resurgence’ behavior of A/H1N1pdm09 was found during 2011/12–2012/13 in Europe and Eastern Asia, which might be related to the level of immunity as well as the number of susceptible individuals in the population [
32]. Our study found that the ‘skip-and-resurgence’ behavior recurred during 2014/15–2015/16. It has just been less than ten years since the emergence of A/H1N1pdm09. Therefore, it is not known whether more ‘skip-and-resurgence’ processes would occur in the future. Fourthly, both Victoria and Yamagata exhibited an annual cycle with a winter-spring peak, which was consistent with previous studies of combined influenza B lineages across China [
33]. In addtion, our study showed that the annual cycle of both Victoria and Yamagata did not change according to the significant antigenic variations of influenza A viruses during the pandemic and post-pandemic seasons.
In general, the co-circulation and interaction of A/H3N2, seasonal A/H1N1, A/H1N1pdm09, Victoria and Yamagata constituted the composite influenza activity. The mechanism of the discrepant seasonality has not yet been clarified. But viral evolution, host susceptibility, and environmental factors might drive influenza circulation activity [
3,
4,
34,
35]. The extent of periodicity transitions might be associated with the genetic differences and degrees of the antigenic change of the novel pandemic subtype or major antigenic variants, the antigenic relatedness of resident influenza viruses, and the level of host herd immunity. Moreover, similar seasonal patterns of influenza subtypes/lineages across age were identified, which might indicate that the factors driving influenza circulation are probably the same across different age groups.
Heterogeneous age distributions of influenza subtypes/lineages
Our study found that the highest positive rate of each subtype/lineage was in school-aged children for all study seasons combined. However, age distribution patterns varied by subtype/lineage and by season. Interestingly, age groups with higher positive rates for A/H3N2 shifted from school-aged children and adults in 2014/15 to adults and the elderly in 2015/16. The same age shift behavior of A/H3N2 has not been reported elsewhere but was observed for A/H1N1pdm09 from the 2009/10 pandemic season to 2010/11 in this study and some previous studies [
12,
36]. School-aged children and working adults have high social contact rates, so they are likely to experience the highest incidence of infection during the initial phase of a pandemic in a susceptible population [
37,
38]. The age shift of A/H1N1pdm09 was proposed to be due to the age heterogeneity in the pre-existing immunity obtained from the first wave of the pandemic [
35,
37]. It is noted that the age shift of A/H3N2 occurred when A/H3N2 major antigenic drift groups, SW13 and HK14, emerged and predominated in the Chengdu population during 2014/15–2015/16. Therefore, the same scenario explaining the age shift of A/H1N1pdm09 might also be applicable for A/H3N2.
Implication for influenza prevention and control measures
Understanding the epidemiological characteristics of influenza virus are critical for directing prevention and control measures. Since school-aged children and working adults were likely to have highest detection rates of infection for the emergence of novel pandemic strains or major antigenic drifts, it might be beneficial to prioritize influenza vaccination and implement interventions (e.g., school-closure) for school-aged children and working adults to prevent and control influenza outbreaks and pandemics in Chengdu [
39,
40]. Influenza vaccine has been introduced in China since 1998 [
41], but the immunization program has not been carried out. There is growing interest in establishing routine immunization programs nationwide [
42]. Currently, only trivalent inactivated influenza vaccine (TIV) is available in the market in China. Every year, TIV with updated vaccine strains recommendated by the WHO for the influenza season, generally from November of one year to April of the following year, in the northern hemisphere is applied in Chengdu. Differences in seasonality of influenza subtypes/lineages would complicate the optimization of vaccination campaigns. Therefore, except for following the timing of vaccination which is typically recommended for the Northern Hemisphere in winter, A/H3N2 activity should be accommodated. The timing of vaccination recommendated by the WHO for the influenza season in the southern hemisphere, generally from May to October of one year, might be incorporated. In addition, our study found that the two influenza B lineages co-circulated during most seasons. Accordingly, it is suggested that quadrivalent influenza vaccine (QIV) instead of TIV should be used to provide protection against both influenza B lineages.
Our study has some limitations that might motivate future research. Firstly, our virological surveillance data focused only on patients with ILI from sentinel hospitals, which might be improved by selecting a random sample of people with influenza-like symptoms from both hospitals and the community. Secondly, during the ten-year surveillance, only a subset of ILI cases were tested, typed, and subtyped, which might get some age groups over sampled and might not represent all of the ILI cases. To minimize the potential bias, the first one or two (the first three or four since 2013) nasopharyngeal swabs on each day of the weeks were collected as previously described [
33]. Moreover, the sampling strategy might have a minor impact on our study because we used the detection rates but not the number of positive cases of subtypes/lineages to assess the pattern of influenza activity as previously described [
12,
40]. We used the detection rates of influenza virus instead of ILI cases to describe influenza activity due to the following reasons. Firstly, other respiratory viruses, like human rhinovirus, human parainfluenza virus, respiratory syncytial virus, human adenovirus, and so on, might cause clinical symptoms similar to influenza [
43]. Therefore, the characteristics of ILI data might reduce the reliability for describing the pattern of influenza epidemics. Secondly, the ILI data are not able to reveal influenza activity of different types and subtypes/lineages. Lastly, our data only cover three pre-pandemic and six post-pandemic influenza seasons. On-going surveillance is necessary to reveal age-specific influenza seasonality after the pandemic.