Ten influenza seasons in France: distribution and timing of influenza A and B circulation, 2003–2013

Until a few years ago, inactivated vaccines against influenza were only available in France as a trivalent formulation, i.e. containing hemagglutinin of two strains of type A influenza virus (A[H3N2] and A[H1N1]) and of only one strain of type B influenza virus, belonging to either the B-Yamagata or B-Victoria lineage [1]. Type B influenza has evolved into two distinct lineages, Yamagata and Victoria, from which strains of both have been co-circulating since 2001 [2]. Recently, quadrivalent vaccines with a B strain from each lineage have been developed to reduce lineage mismatch of influenza vaccine. Quadrivalent vaccines have been added to the WHO recommendation in Northern hemisphere starting the season 2013–2014 [3, 4]. It is expected that this will result in an improved performance of influenza vaccination campaigns in terms of both public health and economic outcomes [5, 6].

Influenza variability, as exemplified above, calls for a detailed and continuous description of virus circulation and characteristics of seasonal epidemics. Comprehensive surveillance in general population that integrates clinical and virological data (including laboratory testing, strain typing and phylogenetic analysis of circulating viruses) remains a very useful tool to optimize any strategy for influenza prevention and control, and is a prerequisite to calculate vaccine effectiveness through observational studies.

Here, we aimed to provide a descriptive analysis of circulating influenza viruses in France during ten consecutive seasons (from 2003–2004 to 2012–2013) using data from a nationwide community-based sentinel influenza surveillance network. Results are provided on the distribution of specimens testing positive for influenza by virus type (A, B), subtype (A(H1), A(H3)) and lineage (B-Yamagata, B-Victoria) in each season; on the match between the influenza B lineage included in the vaccine and the dominant circulating B-lineage viruses; on the estimated medically-attended influenza incidence by virus type and subtype; and on the timing of influenza A vs. B epidemics.

This is a retrospective descriptive study including data collected during ten consecutive seasons (from 2003–2004 to 2012–2013) within the GROG (Groupes Régionaux d’Observation de la Grippe) influenza surveillance network in France.

An average 530 general practitioners and paediatricians participated every season in the GROG network (range: 498–608), of which an average 466 swabbed at least one ARI patient each season (range: 403–578).

Over the whole study period, 49,919 naso-pharyngeal specimens were collected and tested for influenza viruses (Table 1), which corresponds to around 3 % of all ARI patients seen by participating practitioners. The number of specimens that were collected each season ranged between 2,737 in 2003–2004 and 5,447 in 2012–2013 for seasonal influenza; it was 8,822 in the pandemic season 2009–2010. The total number of specimens that tested positive for influenza viruses was 16,312, of which 16,287 were used for the analysis (Table 1). The number of influenza cases reported per season ranged between 903 in 2005–2006 and 2,613 in 2012–2013; it was 3,105 in 2009–2010. The overall proportion of specimens that tested positive was 32.6 %, ranging between 18.6 % in 2005–2006 and 48.0 % in 2012–2013. The proportion of influenza virus positive samples that belonged to influenza type B ranged between <1 % (in 2003–2004, 2006–2007 and 2009–2010) and 62.3 % in 2005–2006. Overall, 23.7 % of influenza positive swabs belonged to type B from 2003–2004 to 2012–2013.

Overall, 85.9 % of influenza A viruses were subtyped: the subtypes A(H1) and A(H3) accounted for 51.6 % and 48.4 % of them, respectively. Influenza A(H1) and A(H3) were the most frequent influenza A subtype in four and, respectively, five seasons; the two virus subtypes co-circulated in 2012–2013 (Fig. 1). Influenza B accounted for at least 10 % of all influenza cases in five seasons, predominating in 2005–2006.

The lineage was characterized for 58.8 % of influenza B cases: the Yamagata and Victoria lineages accounted for 62.8 % and 37.2 % of them, respectively. In the five seasons where influenza B viruses circulated actively, B Yamagata and B Victoria were the most frequent lineage in two and, respectively, three seasons (Fig. 1). The influenza B lineage included in the vaccine and the dominant lineage were mismatched in three of these five seasons.

The average estimated weekly incidence of medically-attended influenza during the study period was 191 per 100,000 inhabitants for influenza virus subtype A(H3), 58 per 100,000 inhabitants for the seasonal A(H1), 203 per 100,000 inhabitants for the pandemic A(H1) viruses, and 127 per 100,000 inhabitants for influenza B (yearly average and median values are available in Additional file 2: Table S1. Influenza A(H3) viruses were responsible of three of the five highest peaks of estimated weekly influenza incidence, including the highest estimated value (2,004 influenza cases per 100,000 inhabitants in week 49 of 2003) (Fig. 2). Pandemic influenza A(H1) viruses caused the second highest peak during the study period: 1,695 influenza cases per 100,000 inhabitants in week 49 of 2009. The highest peak of weekly estimated incidence of influenza B was 1,210 per 100,000 inhabitants in week 7 of 2013. The epidemics caused by influenza B viruses seemed to be still ongoing when surveillance was discontinued (in week 15) in 2007–2008 and 2008–2009 (Fig. 2).

The influenza B epidemics tended to take place a few weeks later than the influenza A epidemics in the five seasons where both virus types accounted for at least 10 % of all reported influenza cases (Table 2). In particular, influenza B peaked on average 3.8 weeks later than influenza A during such seasons. Instead, there were no differences between influenza A and B viruses in the average duration of the periods of intense circulation.

We presented data on the circulation of influenza viruses in France from 2003–2004 to 2012–2013 based on information collected within a countrywide, community based sentinel influenza surveillance network. While influenza A viruses circulated all seasons, influenza B viruses circulated (i.e., accounted for at least 10 % of all influenza cases, according to the definition that we used throughout this paper) in half of the seasons. Influenza B accounted for around one fourth (23.7 %) of all influenza cases in France from 2003–2004 to 2012–2013 and caused more than 50 % of all influenza cases in two of ten seasons (2005–2006 and 2012–2013). No such historical data have been published in France but these results are in line with data of comparable quality from other European countries [12], for the totality of Europe [13‐17], and for countries outside Europe as well [18, 19], despite differences in how the national influenza surveillance systems are organized in different countries.

The influenza virus A(H3) subtype circulated more often and was responsible for epidemics that were more intense (i.e., with a higher incidence peak) than those caused by the A(H1) subtype. This was certainly true before the pandemic of 2009–2010; our data on seasons 2011–2012 and 2012–2013 seem to confirm this pattern, although there is a need to extend the surveillance for a few years to be assured that the situation is now similar to what it was before the pandemic. Again, this is consistent with data from other countries, both in Europe [13‐17] and outside [18‐21].

When influenza B viruses circulated, one lineage was usually responsible for the great majority of all influenza B cases in the season. The dominant circulating B virus and the B virus in the vaccine were mismatched in three out of five seasons. Predictions about which B lineage will dominate in an upcoming season have largely failed worldwide during recent years [2]. Our data thus suggest that a quadrivalent influenza vaccine, including the two B lineages, may probably represents a useful tool to reduce the burden of influenza B, which in some seasons has a high incidence with significant public health and economic impact [18, 22‐24].

We showed that influenza B peaks on average 3.8 weeks later than influenza A when both virus types circulate in the same season. Usually, the protection conferred by the vaccine against influenza covers the whole influenza season but it has been reported that this protection may decline quite rapidly (3–4 months after vaccination) within a given season especially in the elderly [25‐27]. This may have important public health implications in seasons where both virus types circulate and cause epidemics that are spaced from one another by about one month, as in 2008–2009. In such seasons, it is possible that vaccinated frail people have poor residual protection against influenza B viruses, especially when B virus circulation peaks from late February onwards.

A major strength of our study is the fact that it relies on data that originate from a representative network of sentinel primary care practitioners spread over the whole French territory and therefore giving a strong picture of the circulation of influenza viruses in the general population. Our study has some limitations as well. The decision on which ARI patients to swab was left to the practitioners, which may have introduced some bias in the study. Our data include only three seasons after the season 2009–2010, which makes it difficult to establish with certainty whether the circulation pattern of influenza viruses has changed after the pandemic. The transition to the exclusive use of RT-PCR since the 2009–2010 season may have increased the influenza positivity rate of ARI patients of post- compared to pre-pandemic seasons. In some seasons, the influenza circulation was still ongoing when surveillance was discontinued in week 15. This occurred more often for influenza B viruses, as circulation of the latter tends to take place a few weeks after that of influenza A viruses. For instance, the influenza B epidemic was still ongoing on week 15 in seasons 2007–2008 and 2008–2009. Looking at Fig. 2, one can even hypothesize that the peak of influenza B occurred after week 15 in 2008–2009. As a consequence, it is likely that the proportion of influenza cases caused by type B viruses is somewhat underestimated in our study. The surveillance scheme should be extended by at least four weeks to include late epidemics caused by influenza B viruses and have a more precise picture of the circulation of influenza in the country. Another limitation is the lack of information on the age distribution of influenza cases, overall and by virus type and subtype/lineage. Finally, as in all GP-based influenza surveillance systems, our results most likely under-estimate the actual incidence of influenza, as not all influenza patients seek care when having ARI symptoms.

Influenza type B virus is responsible of a substantial share of all influenza cases, up to over 50 % in seasons where influenza A viruses circulate with low incidence. The ability to predict the dominant lineage has been poor during recent years, which led to frequent influenza B vaccine mismatches. In this context, in 2013, the WHO recommendations included a second influenza B strain allowing countries to decide to recommend a trivalent or a quadrivalent influenza vaccine, which is expected to reduce the number of influenza cases and influenza-related hospitalizations and deaths. The epidemiology of influenza is however constantly changing under the effect of various driving forces [28, 29]: this makes it critical that influenza surveillance systems are in place and collect reliable and timely data on the circulation of influenza viruses, to allow health policy-makers and planners to optimize the strategies for prevention and control of influenza. It is also a key to guide research efforts to develop and adapt well-tailored influenza vaccines capable of reducing the influenza burden of disease in all age groups.