Alternating patterns of seasonal influenza activity in the WHO European Region following the 2009 pandemic, 2010‐2018

Abstract Background Influenza virus infections are common and lead to substantial morbidity and mortality worldwide. We characterized the first eight influenza epidemics since the 2009 influenza pandemic by describing the distribution of viruses and epidemics temporally and geographically across the WHO European Region. Methods We retrospectively analyzed laboratory‐confirmed influenza detections in ambulatory patients from sentinel sites. Data were aggregated by reporting entity and season (weeks 40‐20) for 2010‐2011 to 2017‐2018. We explored geographical spread using correlation coefficients. Results There was variation in the regional influenza epidemics during the study period. Influenza A virus subtypes alternated in dominance, except for 2013‐2014 during which both cocirculated, and only one season (2017‐2018) was B virus dominant. The median start week for epidemics in the Region was week 50, the time to the peak ranged between four and 13 weeks, and the duration of the epidemic ranged between 19 and 25 weeks. There was evidence of a west‐to‐east spread across the Region during epidemics in 2010‐2011 (r = .365; P = .019), 2012‐2013 (r = .484; P = .001), 2014‐2015 (r = .423; P = .006), and 2017‐2018 (r = .566; P < .001) seasons. Variation in virus distribution and timing existed within reporting entities across seasons and across reporting entities for a given season. Conclusions Aggregated influenza detection data from sentinel surveillance sites by season between 2010 and 2018 have been presented for the European Region for the first time. Substantial diversity exists between influenza epidemics. These data can inform prevention and control efforts at national, sub‐national, and international levels. Aggregated, regional surveillance data from early affected reporting entities may provide an early warning function and be helpful for early season forecasting efforts.


| INTRODUC TI ON
Influenza virus infections are common and lead to substantial morbidity and mortality 1  To further inform evidence-based decision making for influenza preparedness and control, we characterized the first eight influenza seasons following the 2009 influenza pandemic in terms of virus distribution by type, subtype and lineage, age, timing, and geographical spread.

| Study design
We conducted a retrospective analysis of laboratory-confirmed influenza detections from sentinel surveillance of persons seeking care for a respiratory illness in an ambulatory setting (ie, seeking care at a general practitioner or primary healthcare facility).
We included data from 50 Member States in the WHO European Region which corresponded to 54 reporting entities with data from England, Northern Ireland, Scotland, and Wales (all in the United Kingdom) and Kosovo1 reported separately. We present data by reporting entity.

| Study population and data
Sentinel influenza surveillance is conducted in a representative subset of outpatient sites and coordinated by national (and occasionally sub-national) networks; samples should be collected from patients using a systematic sampling scheme with pre-defined influenza-like illness (ILI) and/or acute respiratory infection (ARI) case definitions.
Reporting entities collected and reported sentinel influenza surveillance data to TESSy each week: the number of patients seen or the population in the catchment area of sites; the number of patients presenting with ILI and/or ARI; and, of these, the number of samples collected and tested for an influenza virus, and the test results.
National influenza surveillance systems and case definitions varied by reporting entity. 5 Influenza virus detection was typically by reverse transcription polymerase chain reaction. 6 Reporting entity-week data for the period 2010 to 2018 were extracted from TESSy in August 2018. Influenza season was defined for the northern hemisphere as ISO week 7 40 in a given year to ISO week 20 in the following year. Reporting entity-week data from earlier than week 40/2010, later than week 20/2018 or not attributed to season weeks were excluded.

| Analyses
For regional aggregated analyses, the distribution of all available virological data derived from specimens taken from ILI or ARI cases in sentinel outpatient sites was summarized by influenza type, influenza A subtype, influenza B lineage, and season. The dominant circulating virus was defined for each season and system as ≥60% of influenza viruses, subtyped type A viruses and type B viruses with lineage, and codominance as the proportion of viruses circulating between 41% and 59%. For reporting entity-level analyses, data for a given reporting entity were excluded if, for a particular season, there were fewer than 50 specimens or <20 weeks of data submitted to TESSy. Ranges of seasonal, reporting entity level data were only described where there were at least five seasons of valid data, as per these criteria. Untyped influenza viruses, influenza A viruses not subtyped, and influenza B viruses not ascribed to a lineage were not included in the denominator for type, subtype, and lineage percentage calculations, respectively, for aggregated regional and reporting entity-level analyses.
Temporal analyses were conducted using the percentage of specimens taken in sentinel outpatient sites testing positive for an influenza virus (ie, percent positive) with aggregate data for all reporting entities in the region and by reporting entity. Timing of the start of the epidemic at both regional and reporting entity levels for a season was defined as the first of two consecutive weeks K E Y W O R D S Central Asia, Europe, influenza, surveillance with at least 10% positive specimens and the end of the epidemic as the last week with a percent positive of at least 10% or week 20.
The duration of the epidemic was defined as the number of weeks between the start and end weeks, inclusive. The peak of the epidemic was defined as the week with the highest percent positive.
The time from the start to the peak of the epidemic was defined as the number of weeks between the start and peak weeks, exclusive. The number of weeks with high influenza circulation during an epidemic was calculated as those with at least 40% positive (a threshold that is used operationally in routine regional surveillance practices and based on expert opinion). For percent positive to be valid, at least 10 specimens had to have been tested in a given week per reporting entity. Peak and end weeks and number of weeks above 40% positive were only valid if there was a prior epidemic start week defined.
To explore spread of influenza infections across the Region within a season, we calculated the correlation coefficient between the timing of the start week of the epidemic in a reporting entity and the reporting entity's geographical center defined as the rounded latitude and longitude of the geographical center of each reporting entity in decimal degrees. 8 A correlation (r) of 0-.19 was defined as very weak, .20-.39 as weak, .40-.59 as moderate, .60-.79 as strong, and .80-1 as very strong. 9 The threshold of significance was set at the 5% level and derived by applying linear regression to these data. Reporting entities with no defined epidemic start date were excluded from these analyses.
We conducted two sensitivity analyses: (a) given substantial geographical span across the Region, we removed the Russian Federation and re-ran the geographical correlation analysis to identify any changes ( Figure 1); and (b) we restricted data up to and including the regional epidemic start week each season to see whether the regional virus distribution at that time point was the same or different to that seen when looking retrospectively back on a full season. Reporting entity-specific summary data are available in Table S1.       Maximum weekly percent positive 51% 54% 57% 39% 55% 53% 53% 55% a No two successive weeks of at least 10 specimens and a percent positive of at least 10%.  subtypes. In addition, there was variation in the distribution of these virological characteristics at the reporting entity level between seasons (Table S1).
Across the Region, there was variation among reporting entities in the timing of epidemics within the same season ( Figure 5). The seasonal reporting entities' median timings differed from regional timings, although similar broad patterns between the two across seasons were observed (Table 2 and Figure 5). There was also variation in the timing of reporting entity-level epidemics between seasons (Table S1) (Table 4). There was no evidence of an association between latitude and reporting entity-level start of the epidemic for any season. Removing data from the Russian Federation did not alter the qualitative findings (Table S2).
When we restricted regional analysis to aggregated data up to the start of the epidemic, we found that the proportion of type A influenza viruses was greater compared to that using complete season    ternated each year, a pattern that was absent in similar regional data prior to the pandemic. 11 There was no evidence that the distribution of virus type or A virus subtypes correlated with epidemic start or peak We found evidence of a west-to-east spread across the Region for half of the seasons under consideration and this pattern has been observed previously when using pooled sentinel and non-sentinel (specimens taken for diagnostic purposes and derived from various sources including outpatient sites that are not part of sentinel networks, hospitals, outbreak investigations, long-term care homes, and closed facilities) data. 11 North-to-south spread across the Region has also been identified previously but this was not apparent in our analysis.

| D ISCUSS I ON
While we found that the proportional distribution of viruses using cumulative data up to the start of the epidemic was often dif- given season). It is possible that the regional data are biased by those countries submitting the majority of detection data and approaches to weighting reporting entity data for future similar analysis could be considered.
We have analyzed reporting entity defined sentinel surveillance data but it is possible that for some reporting entities diagnostic virological data from non-sentinel sources (with no systematic sampling scheme) have been erroneously included or sentinel surveillance systems had limited or no systematic sampling. In addition, in some sentinel systems, there may be non-systematic sampling and we are unable to stratify the virological data from these systems by presentation with ILI or ARI. Other  age-specific trends as the data were not available by age group.
Our findings highlight the substantial diversity between seasonal influenza epidemics in terms of virus distribution and level of activity, underscoring the challenges to accurately predicting the impact of a forthcoming influenza epidemic based on retrospective data in a large region. 36 Nevertheless, it is important to understand the nature of influenza epidemics at this level to help healthcare professionals at European and reporting entity-level target prevention and control strategies and ensure capacity to respond. We show that early data during a season are likely to portend virus dominance, which can be used to target risk communication and public health action. Furthermore, monitoring circulating influenza viruses is critical for tracking changes and assessing their match with vaccine strains, and for adapting vaccination and antiviral treatment strategies.

ACK N OWLED G EM ENTS
The authors sincerely thank Pernille Jorgensen for providing the map and all those clinical and public health professionals across the network who are involved in the collection, collation, and reporting of these data.

D ECL A R ATI O N S
The authors alone are responsible for the views expressed in this publication and they do not necessarily represent the views, decisions, or policies of the institutions with which they are affiliated.
Maps and terminology used in this publication do not imply any opinions on the part of ECDC and WHO or its partners about the legal status of the countries and territories shown or about their borders.