Improving influenza virological surveillance in Europe: strain-based reporting of antigenic and genetic characterisation data, 11 European countries, influenza season 2013/14

Influenza antigenic and genetic characterisation data are crucial for influenza vaccine composition decision making. Previously, aggregate data were reported to the European Centre for Disease Prevention and Control by European Union/European Economic Area (EU/EEA) countries. A system for collecting case-specific influenza antigenic and genetic characterisation data was established for the 2013/14 influenza season. In a pilot study, 11 EU/EEA countries reported through the new mechanism. We demonstrated feasibility of reporting strain-based antigenic and genetic data and ca 10% of influenza virus-positive specimens were selected for further characterisation. Proportions of characterised virus (sub)types were similar to influenza virus circulation levels. The main genetic clades were represented by A/StPetersburg/27/2011(H1N1)pdm09 and A/Texas/50/2012(H3N2). A(H1N1)pdm09 viruses were more prevalent in age groups (by years) < 1 (65%; p = 0.0111), 20–39 (50%; p = 0.0046) and 40–64 (55%; p = 0.00001) while A(H3N2) viruses were most prevalent in those ≥ 65 years (62%*; p = 0.0012). Hospitalised patients in the age groups 6–19 years (67%; p = 0.0494) and ≥ 65 years (52%; p = 0.0005) were more frequently infected by A/Texas/50/2012 A(H3N2)-like viruses compared with hospitalised cases in other age groups. Strain-based reporting enabled deeper understanding of influenza virus circulation among hospitalised patients and substantially improved the reporting of virus characterisation data. Therefore, strain-based reporting of readily available data is recommended to all reporting countries within the EU/EEA.

Influenza antigenic and genetic characterisation data are crucial for influenza vaccine composition decision making. Previously, aggregate data were reported to the European Centre for Disease Prevention and Control by European Union/European Economic Area (EU/EEA) countries. A system for collecting case-specific influenza antigenic and genetic characterisation data was established for the 2013/14 influenza season. In a pilot study, 11 EU/EEA countries reported through the new mechanism. We demonstrated feasibility of reporting strain-based antigenic and genetic data and ca 10% of influenza virus-positive specimens were selected for further characterisation. Proportions of characterised virus (sub)types were similar to influenza virus circulation levels. The main genetic clades were represented by A/StPetersburg/27/2011(H1N1)pdm09 and A/Texas/50/2012(H3N2). A(H1N1)pdm09 viruses were more prevalent in age groups (by years) < 1 (65%; p = 0.0111), 20-39 (50%; p = 0.0046) and 40-64 (55%; p = 0.00001) while A(H3N2) viruses were most prevalent in those ≥ 65 years (62%*; p = 0.0012). Hospitalised patients in the age groups 6-19 years (67%; p = 0.0494) and ≥ 65 years (52%; p = 0.0005) were more frequently infected by A/Texas/50/2012 A(H3N2)-like viruses compared with hospitalised cases in other age groups. Strain-based reporting enabled deeper understanding of influenza virus circulation among hospitalised patients and substantially improved the reporting of virus characterisation data. Therefore, strain-based reporting of readily available data is recommended to all reporting countries within the EU/EEA.

Background
Influenza virological surveillance data, including characteristics of circulating viruses, are collected to describe the annual occurrence of influenza virus (sub) types and lineages for selection of vaccine components for the following season. Virological surveillance also supports epidemic and pandemic preparedness with detection of emerging influenza viruses. European Union and European Economic Area (EU/EEA) countries report influenza surveillance data on a weekly basis during influenza seasons as part of the World Health Organization (WHO) Global Influenza Surveillance and Response System (GISRS) [1] to describe the antigenic character and genetic makeup of circulating viruses [2]. Surveillance at the EU/EEA level is carried out by the European Influenza Surveillance Network (EISN) and data are collected on a weekly basis in The European Surveillance System (TESSy) under the coordination of the European Centre for Disease Prevention and Control (ECDC) [3,4].

Figure 1
Detections and characterisations by influenza A virus subtype and surveillance system, by week of specimen collection, strain-based reporting of antigenic and genetic characterisation data, 11

D. Detections of A(H3N2) by sentinel and non-sentinel systems
Non-sentinel A(H3N2) Sentinel A(H3N2) Week/year Week/year Week/year

Number of viruses
Number of viruses

Figure 2
Phylogenetic and cluster analysis of available haemagglutinin 1 sequences, strain-based reporting of antigenic and genetic characterisation data for influenza viruses, 11  In 2004, strain-based reporting of influenza antiviral susceptibility with epidemiological, demographic and clinical information was introduced [5]. In the 2007/08 influenza season, this new system facilitated rapid assessment of the spread of former seasonal A(H1N1) influenza viruses showing clinical resistance to oseltamivir due to neuraminidase (NA) H275Y amino acid substitution [6].
Although there have been earlier studies on severity and its association with influenza subtypes [7][8][9][10], there is limited evidence of risk factors for severe influenza or influenza complications due to specific subtypes and viruses [11]. To assess the disease burden in different patient risk groups caused by influenza viruses of various (sub)types with particular antigenic and genetic characteristics, it is crucial from the public health perspective to have detailed information about the distribution of specific viruses in different risk groups. This study piloted the integrated collection of strain-based antigenic and genetic characterisation data and epidemiological, demographic and clinical information.
The objectives were: (i) to test the feasibility of collecting influenza virus strain-based antigenic and genetic data; and (ii) to assess the collected data and explore the benefits of non-aggregate strain-based reporting.

Data collection
Respiratory specimens were obtained in the participating countries as part of their routine influenza surveillance activities from week 40/2013 to week 39/2014. Sentinel general practitioners swabbed patients with ILI and/or another ARI, with most meeting the EU case definition for ILI and/or ARI [12], depending on the country's choice of syndrome under surveillance and following the nationally agreed sampling protocol. Non-sentinel specimens, mainly from hospital laboratories, were also included. All specimens were analysed for the presence of influenza virus, by real-time RT-PCR, at the local laboratory or the NIC. If specimens were first analysed at a local laboratory, all or a subset of influenza-virus-positive specimens or virus isolates were sent to the NIC for further analysis of subtype or lineage, antigenic characterisation by haemagglutination inhibition assay, and genetic characterisation by sequencing of haemagglutinin (HA) genes. All participating laboratories take part in regular external quality assessments of rapid detection, virus culture, antigenic and genetic characterisation and antiviral susceptibility analysis [13]. Within EISN, a target of characterising ca 10% of influenza detections has been agreed, although depending on predominant virus (sub)type and intensity of the epidemic, it is valid to characterise less than 10%. In addition, NICs sent smaller subsets of specimens and virus isolates to the WHO CC in London for more detailed characterisation. When selecting specimens for characterisation, laboratories were expected to include specimens with sufficient viral load, based on their resources from all (sub)types, from different age groups, surveillance systems, geographical locations and phases of the epidemic [14].
As part of the existing reporting scheme, countries reported weekly aggregate virological influenza surveillance and antigenic and genetic characterisation data to ECDC. Prefixed, coded reporting categories defined by WHO CC London were used for antigenic and genetic characteristics which included vaccine viruses and additional non-vaccine reference viruses with specific antigenic properties or specified HA amino acid substitutions and phylogenetic clade (see Table 1 for the categories).
In addition, for this pilot study, all EU/EEA countries were invited to submit antigenic and/or genetic characterisation data in strain-based format. The virus name, e.g. A/Netherlands/2245/2013, acted as a unique identifier and duplicated data from national and WHO CC sources were merged. The epidemiological data included variables: age, complication diagnosis, date of onset, exposure to antiviral drugs, sex, hospitalisation, immunocompromised status, outcome, probable country of infection and vaccination status. All data for the 2013/14 influenza season were extracted from TESSy on 15 January 2015. In addition, HA-gene sequences of viruses for which database accession numbers were reported were retrieved from the Global Initiative on Sharing All Influenza Data (GISAID) EpiFlu database.

Data analysis
Feasibility of strain-based reporting was assessed through the pilot, looking at country-wide distribution among participating countries and data completeness. We also received comments on the feasibility of the reporting by questionnaire.
Detection and characterisation data were plotted by week of specimen collection over the influenza season (week 40/2013 to week 39/2014) and timing of aggregate and strain-based antigenic and genetic characterisations were compared between detections from both sentinel and non-sentinel data sources.
Nucleic acid sequences encoding the HA1 subunit were subjected to cluster analysis of maximum-likelihood phylogenetic trees using BioNumerics 7.5 software. Furthermore, encoded HA1 subunit sequences were checked for match to the signature amino acid substitutions of the genetic categories that individual viruses had been ascribed to. The resulting phylogenetic trees were checked for misattributed viruses, as well as for apparent clade patterns beyond the resolution of the categories provided in the TESSy reporting scheme. The European Reference Laboratory Network for Human Influenza (ERLI-Net) reference HA1 encoding sequence sets provided by WHO CC for the 2013/14 season were used as reference sequences in the analysis. To better understand the ongoing evolution of the viruses and in order to check for the presence of groups that predominated in the following season, two A(H3N2) and one B/ Yamagata-lineage ERLI-Net reference viruses defined for the subsequent 2014/15 season were also included: Extended virus characterisation was achieved by including antiviral susceptibility data in the analysis. To standardise interpretation and reporting of influenza virus susceptibility to the neuraminidase (NA) inhibitors (NAIs) oseltamivir and zanamivir, WHO-Antiviral Working Group definitions, based on half maximal inhibitory concentration (IC50),were used [15]. Raw IC50 data were converted into relative fold-change values compared with the median of all data by virus type or subtype and NAI to facilitate pooled analysis of the data from all laboratories [16]. As influenza B virus IC50 data varied widely between laboratories, the fold-changes for influenza B viruses were calculated by reporting laboratory. IC50 fold-change data were generated to classify the viruses as with normal inhibition (NI), reduced inhibition (RI) or highly reduced inhibition (HRI). Amino acid substitution data were analysed against published data on specific amino acid substitution in the M2 and NA proteins previously associated with resistance to adamantane M2 ion channel blockers and RI or HRI by NAIs (oseltamivir and zanamivir), respectively [17].
Patients were stratified into the following age groups: < 1 year, 1-5 years, 6-19 years, 20-39 years, 40-64 years and ≥ 65 years. Distribution of sex by age group was tested for significance using the nonparametric Kruskal-Wallis test. Distribution of genetic clades in different age groups was compared by Dunn's test (multiple pairwise comparisons using rank sums) with Bonferroni adjustment. The level of significance was set at p < 0.05.

Participating countries, data completeness and feasibility
Eleven of 30 EU/EEA countries participated in this pilot: Belgium, Finland, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain and Sweden. However, Belgium did not report patient age and sex, so Belgian cases were excluded from epidemiological analysis. Data completeness is shown in Table 2. All reporting laboratories found the reporting feasible and recommended the use of it to other laboratories in the questionnaire (data not shown).
Of the 1,633 viruses characterised, 516 (32%) were only characterised antigenically, 974 (60%) only genetically and 143 (9%) both antigenically and genetically (Table 1). For the latter, the antigenic and genetic characterisation data were consistent. The participating countries contributed unequally to the antigenic and genetic characterisation data. Germany submitted 300   Figure  2D) Of the 1,633 viruses with antigenic and/or genetic characterisation data, 678 (42%) were tested for neuraminidase inhibitor (NAI) susceptibility using genetic and/or phenotypic methods: 349 A(H1N1)pdm09, 264 A(H3N2), 54 B/Yamagata-lineage and 11 B/Victorialineage viruses. One A(H1N1)pdm09 virus carrying neuraminidase (NA) I223R amino acid substitution showed reduced inhibition (RI) by oseltamivir. Two others showed RI by zanamivir, only one of which was sequenced and shown to carry NA S247I substitution. One virus carried NA H275Y substitution which has been associated with highly reduced inhibition (HRI) by oseltamivir but it was not tested phenotypically. One A(H3N2) virus showed RI by oseltamivir and zanamivir and one by zanamivir only. Both viruses were sequenced but no amino acid substitutions previously or potentially associated with RI were identified. One B virus showed RI by zanamivir (sevenfold) but no amino acid substitution previously or potentially associated with RI was identified. For 80 cases with antiviral susceptibility data, antiviral treatment with oseltamivir up to 14 days before specimen collection was reported, including one case infected with an A(H3N2) virus showing RI by zanamivir. All other cases with indications of being infected with viruses showing RI or HRI by a NAI, for which antiviral exposure status was reported, had not received antivirals before specimen collection. One case infected with A(H1N1)pdm09 carrying NA S247N substitution was exposed to oseltamivir through a treated household contact.

Sex and age
The majority of the 1,547 cases for which age was reported by 11 countries were adults aged 20-64 years (53%  Figure 3A).
Hospitalisation status and influenza virus subtype were reported for 1,147 (70%) of 1,633 cases. Of these, 672 cases were reported from non-sentinel sources and included reporting from 10 countries (Finland, Germany, Greece, Ireland, Italy, the Netherlands,  (13) Household member given or not given antivirals before collection of specimen 75 (5) Norway, Portugal, Spain and Sweden). Patient age and virus subtype/genetic clade information were available for 325 hospitalised patients from Finland, Greece, Ireland, the Netherlands, Norway, Portugal, Spain and Sweden ( Figure 3B). Influenza subtypes and genetic clades associated with hospitalisation differed between age groups. Hospitalised cases in the 6-19 years age group and ≥ 65 years of age were most frequently infected by A/Texas/50/2012-like A(H3N2) viruses, 8/12 (67%; p = 0.0494) and 58/111 (52%; p = 0.0005), respectively ( Figure 3B). All other hospitalised cases were infected in higher proportions by A/ St Petersburg/27/2011-like (H1N1)pdm09 viruses, with rates of infection in children 6-19 (p = 0.0306) and adults ≥ 65 (p = 0.0011) years of age being significantly less than in 40-64 year olds.

Vaccination status
Vaccination status was known for 798 of the 1,633 cases from all 11 countries; 130 (16%) had been vaccinated with the influenza vaccine for the 2013/14 influenza season. Among these, there were 400 (50%) males and 396 (50%) females (two cases with unknown sex). Vaccination coverage ranged from 4% in children 1-5 years of age to 45% among those ≥ 65 years of age. None of the infants < 1 year of age had been vaccinated. Vaccination status and hospitalisation was known for 712 patients. Among 139 hospitalised cases, 34 (24%) had been vaccinated against influenza. Of those vaccinated and hospitalised, 20 had an A(H3N2), 12 an A(H1N1)pdm09 and two a B/Yamagata infection. Of the 16 fatal cases for which vaccination status was known, three had been known to be vaccinated. Two of these cases were infected by A(H3N2) and one by A(H1N1) pmd09 virus. Due to limited data completeness for outcome and vaccination status, no statistical analysis was performed.

Discussion
In this pilot study, TESSy was used to capture influenza virus strain-based antigenic and genetic characterisation data allowing phylogenetic analysis and reporting on the demographic information, outcome, vaccination status, immune status and the probable country of origin of the characterised viruses at the European level for the first time. Strain-based data analysis was feasible based on good data completeness for variables such as virus subtype, patient age and sex. Large and small countries from northern, southern and western parts of EU/EEA reported data and the target set for detailed characterisation of 10% of the viruses detected was achieved.
Although the distribution of (sub)types in our study was not exactly the same as the distribution in all EU/ EEA countries [18], all (sub)types were covered both in our aggregate and strain-based data. We recognised from past years' data that the proportions of different virus types/subtypes/lineages as well as the dominant type/subtype/lineage can vary between countries each season.
This pilot study showed that characterised viruses were congruent with guidance on targeted sampling for further characterisation: the data reported covered all age groups and had no sex bias. However, in our data, A(H3N2) viruses were slightly overrepresented among those selected for characterisation (16% vs 10% for all subtypes). A(H3N2) viruses have proved difficult to characterise antigenically in recent years [19] and therefore greater effort has been put into their characterisation.
In this pilot study, influenza virus types and subtypes did not affect the sexes differently, but did differ across age groups: A(H1N1)pdm09 viruses predominated in younger adults in the 20-64 years of age group as during the 2009 pandemic and in infants < 1 year, while A(H3N2) viruses predominated in patients aged ≥ 65 years, school-aged children and teenagers. Although vaccination status was reported, completeness was low for underlying diasease and immune status, and therefore no conclusions could be drawn on a possible effect of vaccination on the age distribution.
It will be of interest to follow the trend for age distribution among hospitalised cases over several seasons to better understand the age-distribution of influenza infection associated with severe infection by ( [21]. In the hospital surveillance data, many of the influenza viruses are reported without subtype and therefore no exact comparison is possible. Overall, only 41 (3%) of the 1,633 viruses characterised were from fatal cases which does not show a bias of the data towards fatal case specimens being characterised. An earlier analysis of the 2013/14 season showed that fatal outcomes occured mostly in adults > 40 years of age [21]; this pilot study showed the highest number of deaths in those ≥ 65 years of age. Based on our limited data on severe infection, hospitalised cases affected by A(H3N2) virus infection were mostly schoolaged children and the elderly, in line with the results of the meta-analysis for seasonal influenza [11].
Limitations of this study were that: only 11 of the 30 EU/EEA countries agreed to participate, and only three submitted data with indication of hospitalisation status with both non-hospitalised and hospitalised cases as most laboratories do not have the clinical information; and NICs aim for good representativeness of specimen selection but acknowledge selection biases and constraints in terms of: (i) characterisation of more A(H3N2) viruses as these viruses are currently drifting rapidly and have become more difficult to culture and characterise than A(H1N1)pdm09 viruses; (ii) capturing enough type B viruses to inform vaccine composition recommendations; (iii) increased interest in hospitalised and severe cases/deaths; (iv) limited resources and therefore focus on start, middle and end of season; (v) influenza surveillance systems may underestimate the cases in both ends of the age span due to healthcare seeking behaviour and sampling at outpatient clinics.
The extension of the existing antiviral strain-based reporting scheme with genetic and antigenic characterisation data was welcomed and supported by the pilot countries and it strengthens EISN as virological data reported can be subjected to more detailed analysis inclusive of the associated demographic and clinical information. We consider this as a substantial improvement over the previous aggregate reporting of antigenic and genetic categories only. Strain-based reporting also enabled early 2014/15 and 2015/16 influenza season analysis including HA phylogeny [22,23]. Through more traditional hospital surveillance, only virus subtype information related to hospitalisation has been reported by eight countries [24], but now genetic clade can be associated with information on hospitalisation.
We recommend the strain-based reporting to all EISN laboratories and we also recommend that laboratories continue to select specimens for characterisation across subtypes, geographic location and age groups, related to indicators of clinical status. The same principles as for selecting specimens to be sent to WHO CCs for detailed characterisation and informing vaccine composition recommendations may be adopted for national specimen selection [14]. Further, detailed reporting may allow greater definition of risk groups and support targeted vaccination and antiviral treatment strategies, e.g. if data on underlying conditions are included. The data should be combined with available hospital surveillance data as they may provide new ways of looking into vaccine effectiveness that has been low for A(H3N2) viruses in recent years [25].
The interplay between clinicians, epidemiologists and virologists collecting this type of data with public health specialists is crucial to ensure an even more representative sampling scheme for virus specimens. This will help to provide data for better estimates of risk factors associated with influenza.

* Authors' correction
The percentage marked with an asterisk was corrected, to 62%, at the request of the authors on 13 October 2016.
Sweden: The Swedish regional laboratories and participants of the Swedish sentinel programme who provided samples for surveillance and further characterisation.
Finally, we thank the authors and the laboratories which originated and submitted the sequences from the Global Initiative on Sharing All Influenza Data's (GISAID) EpiFlu Database on which this research is based. All submitters of data may be contacted directly via the GISAID website, www. gisaid.org.