Infection with influenza viruses A (H1N1), A (H3N2), or B causes substantial human illness and excess deaths each year ( 1 , 2 ). Vaccination against seasonal influenza is the key control measure used in Europe to minimize illness and death. Antigenic mismatch between vaccine components and circulating viruses occurs every few years, requiring reformulation of the vaccine ( 1 ). In addition, suboptimal immunization in patient groups for which vaccine is recommended provides the rationale for use of antiviral drugs in the prophylaxis and treatment of influenza. M2 ion channel inhibitors (M2Is), amantadine and rimantadine, have been available since 1964, but adverse effects, rapid development of resistance, and lack of activity against influenza B have limited their usefulness ( 3 ). The introduction of neuraminidase inhibitors (NAIs), oral oseltamivir and inhaled zanamivir, which are active against both influenza type A and B viruses, was a major breakthrough in treatment and prophylaxis of influenza using antiviral drugs ( 4 ). However, prescription data indicate that they are not widely used in Europe (Figure 1); by contrast, in Japan during the 2003–04 season alone, ≈6 million NAI treatment courses were prescribed ( 5 ). Figure 1 Prescription data of oseltamivir treatment courses for Western Europe (in thousands); 12 months of data for each year 2002–2007 and through September for 2008. Data from the United Kingdom, the Netherlands, Switzerland, and Portugal are excluded because of negligible values. Data provided by IMS Health (www.imshealth.com), London, UK. Before the introduction of NAIs in 1999, and until 2007, 1, and therefore the peak incidence rate for 2007–08 is significantly higher than that for 2000–01.
#Data for seasons 2002–03, 2003–04, and 2004–05 were missing. Methods Clinical Influenza Activity The European Influenza Surveillance Scheme (EISS) actively monitored influenza activity from week 40 (October 1–7) of 2007 through week 19 (May 5–11) of 2008. EISS covers all 27 European Union countries plus Croatia, Norway, Serbia, Switzerland, Turkey, and Ukraine. In each country each week, 1 or several networks of sentinel general practitioners (GPs) reported rates of consultation for influenza-like illness (ILI) or acute respiratory infection (ARI) ( 15 – 17 ). ARI includes ILI and all other acute respiratory infections. For Croatia, Finland, Turkey, and Ukraine, no consultation data were available. Virologic Analysis Sentinel GPs involved in clinical data recording of ILI or ARI also send nasal, pharyngeal, or nasopharyngeal specimens from a subset of their patients to the National Influenza Centers (NICs) for virus detection and characterization by using a variety of genetic or phenotypic methods ( 18 – 20 ). The NICs also analyzed specimens and influenza viruses obtained from other sources (e.g., from nonsentinel GPs, hospitals, or institutions). For Cyprus and Turkey, no virus detection data were available. Antiviral Drug Susceptibility Monitoring Antiviral susceptibility data were generated either through the European Surveillance Network for Vigilance against Viral Resistance (VIRGIL) project at a single laboratory in London (UK Health Protection Agency) or directly by individual NICs by using methods described previously ( 14 , 21 ). Genetic analysis of virus isolates or clinical specimens was performed by using cycle-sequencing or pyrosequencing the NA gene, targeting the H275Y amino acid substitution in the N1 NA ( 22 ). The 50% inhibitory NAI concentration (IC50) of virus isolates was determined by using fluorescent or chemiluminescent enzyme assays ( 23 , 24 ). ORVs were defined as influenza viruses A (H1N1) with an IC50 >100 nmol/L for oseltamivir. Susceptibility to zanamivir was determined by using the same enzymatic method. Susceptibility to M2Is was determined by cycle-sequencing or pyrosequencing the M2 protein gene, targeting known resistance markers. Antiviral susceptibility data were not available for Cyprus, Lithuania, and Malta. Data Analysis To obtain United Kingdom estimates, clinical and virologic surveillance data and antiviral susceptibility data were totaled for England, Northern Ireland, Scotland, and Wales. A single web-based European database at the EISS password-protected website (www.eiss.org) was used to collect antiviral susceptibility data and linked patient demographic and clinical data ( 25 ). Updates on possible resistant viruses were provided at regular intervals to EISS members, the World Health Organization, and the European Centre for Disease Prevention and Control. The timing of the first week of continuous detection of influenza virus A and ORVs across Europe, both based on date of specimen collection, were analyzed by linear regression analysis using center longitude and center latitude of a country as explanatory variables. A maximum interruption of 1 week with no influenza virus A or ORV detection was allowed in estimating the first week of continuous detection. The average European delay between the first week of continuous detection of influenza virus A and of ORV was calculated as the average of the differences in number of weeks between both, by country. The analysis of temporal trends in the prevalence of ORVs in countries and for Europe was confounded by different levels of sampling in different countries ( 18 ), enhanced antiviral susceptibility testing in some countries, and lack of data on the proportion of ORVs for some or most weeks for several other countries. To ensure a more representative picture of temporal trends in the proportion of ORVs, a mixed effect logistic regression modeling approach ( 26 , 27 ) was used, which allows modeling of binomial proportions, i.e., a numerator and a denominator as a function of time, where the coefficients of this function are allowed to vary for each country around a mean value, combining data from all countries. If there are no observations or the denominator is small, the fit will shrink to its overall mean, and uncertainties increase. Three fractions were modeled: “ILI per population covered,” “influenza A virus detections per specimens tested,” and “A (H1N1) resistant per A (H1N1) tested.” By multiplying the first 2 fractions by the total population, we obtained the number of patients with ILI who had influenza A in a country. By dividing this number by the sum of the number of patients with ILI who had influenza A for all countries, we obtained the relative weights. By multiplying the weights with the prevalences of ORVs summed over all countries, we obtained the weekly European prevalences of ORVs. The modeled weekly prevalences of ORVs were subsequently used to calculate the average prevalence of ORVs by country and for Europe (Technical Appendix). We performed all statistical analyses by using the software package R version 2.8.0 ( 28 ). Box-and-whisker plot analysis was used to select viruses with outlying high IC50 values for further analysis ( 7 , 29 ). For oseltamivir outlier identification, all viruses defined as resistant for oseltamivir (IC50 >100 nmol/L) were first removed. Minor outliers were defined as values lying between the upper quartile (UQ) + 1.5 × interquartile region (IQR) and UQ + 3 × IQR; major outliers were defined as values lying above UQ + 3 × IQR, based on analysis of all viruses in a particular subtype over a particular winter season. Phylogenetic analysis of NA and hemagglutinin (HA) gene sequences used maximum parsimony (PAUP* version 4.0; Sinauer Associates, Sunderland, MA, USA). Sequences of ORVs and oseltamivir-sensitive influenza A (H1N1) viruses (OSVs) were chosen as representative of influenza viruses A (H1N1) isolated during the 2007–08 influenza season (i.e., weeks 40–52 of 2007 and weeks 1–19 of 2008) in different European countries and a few from other regions of the world and were compared with those of a few influenza viruses A (H1N1) isolated before the 2007–08 season, including sporadically isolated ORVs. GenBank accession numbers are listed in the Appendix Table. Results Seasonal Surveillance The 2007–08 influenza season in Europe was initially dominated by influenza viruses A (n = 10,720; 60% of all influenza virus detections). Influenza viruses B (n = 7,150; 40% of all influenza virus detections) became dominant in week 8 (Figure 2). Of the 5,984 (56%) influenza viruses A subtyped, 5,748 (96%) were H1, and 236 (4%) were H3. Overall, influenza virus detections peaked in week 6, in week 4 for influenza viruses A (H1N1), and in week 8 for influenza viruses B. Of the 2,136 influenza viruses A (H1N1) characterized antigenically, 97% were reported to be closely related to the vaccine strain A/Solomon Islands/3/2006, although half of these viruses were reported to be more closely related to A/Brisbane/59/2007, the vaccine strain recommended for the 2008–09 season ( 30 ). Figure 2 Total number of influenza virus detections, by type and subtype and by week, Europe, winter 2007–08. The first countries in Europe where influenza viruses A started to circulate continuously were France, Spain, Switzerland, and the United Kingdom in week 40. Spatial analysis of the timing of the first week of continuous detection of influenza viruses A across Europe (n = 30 countries) showed a west-to-east pattern: estimated parameter for longitude was 0.261 weeks per degree longitude (95% confidence interval [CI] 0.138–0.385, p = 0.001), and for latitude –0.108 weeks per degree latitude (95% CI –0.324 through 0.108, p = 0.366), with R2 = 0.32 for the linear regression fit. Antiviral Drug Susceptibility The estimated number of influenza viruses A (H1N1) among all detected influenza viruses A (n = 10,720) was 10,291 following extrapolation from the proportion of 96% influenza viruses A (H1N1) among all 5,984 subtyped influenza viruses A. Of the 10,291 influenza viruses A (H1N1), 2,949 (29%) were tested for antiviral susceptibility, 1,080 by both phenotypic assay (IC50) and sequencing, 601 by phenotypic assay alone, and 1,268 by sequencing alone. Of the 2,949 viruses tested, 712 (24%) were oseltamivir resistant either by presence of the H275Y substitution (n = 548) or an IC50 >100 nmol/L for oseltamivir (n = 463) (Figure 3). Correlation was 100% between sensitive phenotype (IC50 100 nmol/L) and the presence of Y275 (n = 299). OSVs (n = 1,218) had a median IC50 of 1.7 nmol/L for oseltamivir (range 0.1 nmol/L–23.2 nmol/L) and only 9 minor outliers (thresholds IC50 >12.0 nmol/L and 8.5 nmol/L and 100 nmol/L for NAI drugs (A. Lackenby et al., unpub. data), in concordance with results from worldwide surveillance ( 8 , 9 ). In 2007–08, influenza viruses A (H3N2) and B circulating in Europe remained sensitive to NAI drugs. This emergence of oseltamivir-resistant influenza virus A (H1N1) in Europe coincided with the dominant circulation of this virus subtype during the 2007–08 winter in Europe and the emergence of a new drift variant, A/Brisbane/59/2007 ( 30 ). Of the last 12 influenza seasons, influenza viruses A (H1N1) were dominant only in 2000–01, which included a new drift variant, A/New Caledonia/20/99 ( 20 ). In the other 10 seasons, influenza viruses A (H1N1) played a minor role, with influenza viruses A (H3N2) dominant in 9 seasons. Compared with 2000–01, peak incidence rates for ILI or ARI in 7 of 13 countries were similar or lower in 2007–08 (Table). In 6 countries, the peak incidence rates were significantly higher in 2007–08 than in 2000–01, but with a <2-fold difference in 5 countries and, in Spain only, a 4.8-fold difference. Both the 2000–01 and 2007–08 seasons were unremarkable in the overall clinical impact of influenza, with normal seasonal activity as measured by comparison of peak incidence rates for all seasons since 2000–01. Sporadically occurring A/New Caledonia/20/99-like ORVs with H275Y were detected during the 2006–07 season in the United Kingdom and United States but did not become epidemiologically important. Indeed, the genetic background plays a role in retaining the replication efficiency and pathogenicity of recombinant influenza viruses A (H5N1) and A (H1N1) after introduction of tyrosine at position 275 ( 33 ). Furthermore, other previously analyzed influenza viruses A (H1N1) with the H275Y mutation showed impaired replicative ability in cell culture and reduced infectivity and substantially compromised pathogenicity in animal models, compared with the corresponding wild-type virus ( 34 , 35 ). The coincidental emergence of H275Y with the circulation of the A/Brisbane/59/2007 drift variant may have favored the emergence of fit transmissible ORVs. This point is also illustrated by the emergence of A/Brisbane/59/2007-like ORVs in other parts of the Northern Hemisphere and their continued circulation during the 2008 Southern Hemisphere influenza epidemic season ( 36 – 38 ). Since the last quarter of 2007, ORVs have been detected in continents other than Europe, with proportions of ORVs varying from 100% in South Africa and Australia to <5% in Japan. Trend data are limited: a slight monthly increase was noted in China/Hong Kong and Japan; in Canada, the increase was similar to that in Europe, from 0% ORVs in November 2007 to 86% ORVs in April 2008 ( 36 ). Using modeling, we showed that the prevalence of ORVs increased in the European region from ≈0% at the start to 56% at the end of the season. The finding of a high prevalence of ORVs in the community and the overall temporal increase in resistance demonstrates that the previously documented reduced fitness of viruses bearing the H275Y mutation, ostensibly caused by structural and functional constraints ( 10 ), has been overcome in currently circulating influenza viruses A (H1N1). The results of Rameix-Welti et al. ( 32 ) suggest that a combination of specific amino acid substitutions have increased the affinity of the NA of recent influenza viruses A (H1N1) (ORVs and OSVs) for substrate. A better balance of NA and HA activities in ORVs compared with OSVs may have contributed to the overall fitness and transmissibility of ORVs. However, growth curves conducted in tissue culture of pairs of ORVs and OSVs demonstrated no differences in growth kinetics or final virus yields. Therefore, changes in other genes also may be involved in the overall impact on the fitness of ORVs, for which whole genome sequencing is necessary. For Europe, no focal point of initiation of spread could be identified. The spread of ORV from west to east paralleled that of influenza virus A in Europe, and there was an average delay of 5.7 weeks for the appearance of ORVs after the start of influenza virus A circulation. However, the low R2 values for both patterns make definitive conclusions difficult to draw about the spatial spread of either influenza viruses A or ORVs. Several independent introductions into European countries of a sensitive and a resistant strain might explain the low R2 values. Estimating whether a global focal point exists from which ORVs emerged to spread to the rest of the world is not possible, but the fact that Japan, the country with the highest per capita use of oseltamivir ( 5 ), had relatively low levels of circulating ORVs during the 2007–08 influenza season is relevant and reflects the limited circulation of the clade 2B A/Brisbane/59/2007-like viruses belonging to the European cluster in this region ( 31 , 36 ). The close relationships between the NA sequences of most of the 2007–08 European ORVs and their segregation from those of OSVs suggest that resistance results in large part from the spread of a single variant. Phylogenetic analyses show that this is a property of clade 2B A/Brisbane/59/2007-like viruses and is not associated with emergence of another antigenic variant. However, identification of other resistant variants in the United Kingdom, some of which are more closely related to OSVs than to most ORVs (e.g., A/England/654/2007) indicates the independent parallel emergence of multiple resistant variants. This is emphasized by small distinct clusters of closely related ORVs in Japan that are related to European OSVs, whereas only a few of the Japanese ORVs belonged to the large European ORVs cluster ( 31 ). Resolution of the origin and frequency of emergence of ORVs and association with drug use clearly require substantially more intimate knowledge of the genetic relationships among OSVs and ORVs worldwide. Our observations suggest that the new genetic background of influenza viruses A (H1N1) that appeared in 2007 enabled the virus to develop oseltamivir resistance independently at several locations in the world. The combined effect of the relatively high level of circulation of influenza viruses A (H1N1) in Europe; the introduction of a new antigenic drift variant in a susceptible population, partly related to the lack of substantial influenza virus A (H1N1) circulation since the 2000–01 season; and the uncompromised transmissibility of the ORVs contributed to the epidemiologic success of the ORVs during the 2007–08 season. This phenomenon shows clearly that continuation of antiviral susceptibility monitoring and increasing capacity for timely response are essential ( 21 , 39 ). In addition, the appearance of viable transmitting ORVs is a reminder that the level of resistance to oseltamivir of seasonal or pandemic virus cannot be predicted, and therefore antiviral strategies should not rely on single drugs ( 40 ). Although oseltamivir remains a valuable influenza antiviral agent, the emergence of natural resistance shifts attention from oseltamivir to other antiviral agents and to improved vaccination (e.g., greater vaccination coverage, more immunogenic and broadly reacting vaccines) in the fight against seasonal and pandemic influenza. Supplementary Material Appendix Figure Fitted curves to the proportion oseltamivir-resistant viruses among influenza viruses A (H1N1) tested for resistance (both sentinel and nonsentinel) for all countries for which data were available for inclusion in modeling the European trend (see Figures 4 and 5). The x axes display the week in which the clinical specimens were collected (weeks 40-52 of 2007 and weeks 1-19 of 2008). The y axes display the percentage oseltamivir resistant influenza viruses A (H1N1). Plus signs indicate the actual determined proportions resistant A (H1N1) viruses; light gray region is the 95% confidence interval of the model. Appendix Table GenBank accession numbers of hemagglutinin and neuraminidase sequences used in the phylogenetic Analyses. Technical Appendix Statistical Analysis of Temporal Trends of Resistant Influenza A (H1N1) Viruses, Europe.