Interim 2019/20 influenza vaccine effectiveness: six European studies, September 2019 to January 2020

Background Influenza A(H1N1)pdm09, A(H3N2) and B viruses were co-circulating in Europe between September 2019 and January 2020. Aim To provide interim 2019/20 influenza vaccine effectiveness (VE) estimates from six European studies, covering 10 countries and both primary care and hospital settings. Methods All studies used the test-negative design, although there were some differences in other study characteristics, e.g. patient selection, data sources, case definitions and included age groups. Overall and influenza (sub)type-specific VE was estimated for each study using logistic regression adjusted for potential confounders. Results There were 31,537 patients recruited across the six studies, of which 5,300 (17%) were cases with 5,310 infections. Most of these (4,466; 84%) were influenza A. The VE point estimates for all ages were 29% to 61% against any influenza in the primary care setting and 35% to 60% in hospitalised older adults (aged 65 years and over). The VE point estimates against A(H1N1)pdm09 (all ages, both settings) was 48% to 75%, and against A(H3N2) ranged from −58% to 57% (primary care) and −16% to 60% (hospital). Against influenza B, VE for all ages was 62% to 83% (primary care only). Conclusions Influenza vaccination is of continued benefit during the ongoing 2019/20 influenza season. Robust end-of-season VE estimates and genetic virus characterisation results may help understand the variability in influenza (sub)type-specific results across studies.


Introduction
All European Union (EU) countries and the United Kingdom (UK) recommend seasonal influenza vaccine for older adults and those at increased risk of influenza complications and severe disease, as well as for patients with chronic conditions [1]. In addition, universal childhood influenza is available in some countries in the World Health Organization (WHO) European Region, and was introduced incrementally in the UK in 2013/14 [2].
The 2019/20 influenza season started early in most countries of the WHO European Region, with influenza A(H1N1)pdm09 and A(H3N2) virus subtypes, as well as influenza B circulating throughout the region, although predominantly influenza A overall (69%) [5]. Despite this, some countries reported dominance of influenza B, with a few reporting co-dominance [5].
The I-MOVE (Influenza -Monitoring Vaccine Effectiveness in Europe) network has measured influenza vaccine effectiveness (VE) annually since 2008/09, with its partners including Denmark, Spain, the UK and many EU countries measuring VE through the I-MOVE multicentre studies. We summarise interim influenza VE estimates for the 2019/20 season from six studies (four single-country and two multi-country), with out-and in-patient (hospital) settings, in order to provide information for measures of influenza prevention and control for the remaining season. Results presented here also helped to inform the February 2020 WHO Vaccine Strain Selection Committee.

Study setting
The four primary care (PC) studies were conducted in Denmark (DK-PC), Spain (ES-PC), the UK (UK-PC) and through the EU I-MOVE multi-country network (EU-PC; eight of nine participating countries in this network having available data for interim analysis). The two hospital setting (H) studies were in Denmark (DK-H) and through the EU I-MOVE multi-country network (EU-H; two of 11 participating countries in this network having available data for interim analysis) ( Figure 1). In total, 10 countries contributed data to the interim influenza vaccine effectiveness results presented in this article.

Study design
The methods for all six studies have already been described [6][7][8][9][10]. The test-negative design [11] was used in all studies, although some studies varied in their patient selection and/or data collection (Table  1). Briefly, patients presenting to participating primary care settings with symptoms of influenza-like illness (ILI) or acute respiratory infection (ARI) were swabbed. For the hospital setting swabs were taken from those with symptoms of severe ARI (SARI). Three studies used an exhaustive or systematic selection of patients to swab (EU-H, ES-PC and EU-PC), while physicians' discretion was used to select patients for swabbing in the others (DK-H, DK-PC and UK-PC). Samples were tested by reverse transcription (RT)-PCR for generic influenza virus detection, type A subtyping and type B lineage determination. Cases were defined as patients with positive results by influenza virus (sub)type. Controls were defined as those with negative results. Vaccinated patients were defined as those having had the 2019/20 influenza vaccine at least 14 days before onset of symptoms (15 days for two studies: EU-PC and EU-H). Those vaccinated less than 14 days (less than 15 days for two studies) before symptom onset, or with unknown date of vaccination were excluded. Most study countries (six from EU-PC and Denmark) selected all or a random sample of influenza virus-positive specimens for haemagglutinin genome segment and/or whole genome sequencing. This was followed by phylogenetic analysis to determine clade distribution for potential impact on VE. In Spain, the ES-PC study (regions not included in EU-PC) sequenced an ad hoc sample of influenza viruses. In UK-PC, sequencing is done by two contributing surveillance schemes. One of these sequences all influenza viruses with sufficient genetic material (Ct value < 31) and all viruses derived from vaccinated cases. The other one sequences a subset only. Sequencing results were provided for both studies in Denmark together (DK-PC and DK-H).

Ethical statement
The planning, conduct and reporting of the studies was in line with the Declaration of Helsinki [12].

Statistical analysis
Each study computed VE by subtracting the ratio of the odds of vaccination in cases and controls from one, as a percentage (VE = (1 -odds ratio (OR) x 100%). All studies applied logistic regression to adjust their VE for measured confounding variables (Table 1). Studyspecific VE was estimated overall and where possible, by age group and target population (as defined locally in the various studies and study sites) against influenza A overall, A(H1N1)pdm09, A(H3N2), influenza B and B/ Victoria. For analyses with a small sample size, defined as those having fewer than 10 cases or controls per parameter, a sensitivity analysis was performed using Firth's method of penalised logistic regression (PLR) to assess small sample bias [13,14]. We considered a difference of > 10% between the PLR and original estimate to be an indication of small sample bias, and any such estimates were not included.

Results
From 30 September 2019 to 29 January 2020, in the primary care setting, there were 12,842 patients included in the DK-PC study (1,723 cases), 1,798 in ES-PC (955 cases), 2,987 in EU-PC (1,052 cases) and 2,548 in UK-PC (782 cases). In the hospital setting, there were 10,761 in DK-H (659 cases) and 601 in EU-H (129 cases).
Overall, 84% (4,466/5,310) of confirmed infections were influenza A virus-positive and 16% (844/5,310) were influenza B virus-positive, noting that the previously mentioned total refers to patients rather than infections. The proportion of subtyped influenza A viruses was 93% to 98% in EU-H, ES-PC, EU-PC and UK-PC, and 45% to 51% in DK-H/DK-PC. Most subtyped influenza A viruses were influenza A(H1N1)pdm09 (67-89%) in ES-PC, EU-PC and EU-H; this subtype comprised 46% to 47% in DK-PC/DK-H and 7% in UK-PC ( Figure 2). The proportion of influenza B viruses ranged from 4% in UK-PC to 30% in ES-PC ( Figure 2). The proportion of B viruses ascribed to a lineage was 27% overall (100% for EU-H, 70% for EU-PC, 23% for UK; lineage information not available in DK-PC, DK-H and ES-PC studies). Most of those that were ascribed to a lineage were B/ Victoria: 98% in EU-PC, 75% in UK and 67% in EU-H.

Sensitivity analyses
Results with small sample sizes were subject to sensitivity analyses, all of which gave similar results (absolute difference range: 1-6%).

Discussion
Our results for the 2019/20 influenza season in six wellknown influenza studies across Europe indicate that interim VE against any laboratory-confirmed influenza among all ages in primary care and hospital settings ranged between 29% and 61%, while VE was from 53% to 60% in vaccination target groups. The proportions of influenza (sub)types contributing to these overall results varied considerably by study.
Against influenza A (all subtypes) among all ages, point estimates for VE ranged between 30% and 60% in both types of settings, and they were slightly higher (49-62%) in the target groups for vaccination.  [15], and a little higher than those reported from the United States (US) (45%) [16], noting that proportions of influenza (sub)types and proportions of study participants contributing to age groups may be slightly different across all studies. For older adults (aged 65 years and over), the six studies presented here had VE point estimates of 43% to 66% across all settings. This is similar to estimates from Finland, Sweden and Canada for this age group, at 41% [17], 44% [18] and 60% [15], respectively. It is also similar for those aged 50 years and over in the US, at 43% [16], noting that underlying proportions of influenza (sub)types are likely to be different across all studies.
The 2019/20 interim VE against influenza A(H1N1) pdm09 was higher compared with the 2018/19 interim season estimates in most studies among all ages, except for the EU-PC study [19]. The VE against A(H1N1) pdm09 was also higher among those aged 65 years and older in the hospital-based studies in the 2019/20 season compared with the previous season [19]. The main circulating genetic clade in the 2019/20 studies is 6B.1A5A and to a lesser extent, 6B.1A5B, which although different from the vaccine strain, show good reactivity with ferret antiserum raised against the vaccine virus [20]. The 2019/20 interim influenza A(H1N1) b Includes three co-infections. These, and two patients from one site, were not included in the A(H3N2) analysis, but in the 'any influenza A' analysis.

Table 2b
Interim vaccine effectiveness (VE) against all laboratory-confirmed influenza, influenza A, A(H1N1)pdm09, A(H3N2) and B, by age group, target group for vaccination and by study, six European studies, influenza season 2019/20 pdm09 VE in Canada [15] and the US [16] was 44% and 37%, respectively. These overall results are lower than the DK-PC and ES-PC results, but are more comparable with the EU-PC results (48%). In the US, there was some indication of lower VE among younger adults (18-49 years) with VE being 5% [16]. None of the studies reported in this article indicate a much lower A(H1N1) pdm09 VE among younger adults. End-of-season overall results as well as clade/genetic variant-specific results and birth cohort-specific VE will help understand differences between studies at international level.
Unlike observed estimates in the 2018/19 season, in which three of five studies reporting A(H3N2) cases had VE below zero and all were below 50% [19], two of six 2019/20 interim study results were below zero, two were between 25% and 50%, and two were between 50% and 60%. The EU-PC and the ES-PC studies showed very different point estimates of A(H3N2) VE among all ages; however, there were only eight cases in common across these studies. There is considerable variation in the VE estimates against A(H3N2) presented here, which may go hand-in-hand with the considerable genetic diversity observed. While A(H3N2) viruses remain difficult to characterise antigenically [20], reports suggest that the 3C.3a circulating viruses were antigenically similar to the vaccine virus. While a higher VE against A(H3N2) among studies with high proportions of 3C.3a characterised viruses might be anticipated, this was not observed everywhere. The other viruses circulating belonged to the 3C.2a1b clade, most with additional substitutions. These viruses are more genetically distinct from the vaccine virus and therefore antigenically less similar. However, 2019/20 interim VE from Canada against A(H3N2) was 62%, and 94% of the 80 viruses sequenced in Canada belonged to the 3C.2a1b clade. While clade-specific VE estimates at the end of the 2019/20 season could elucidate whether different circulating clades of A(H3N2) viruses across the region explain the differences in VE against A(H3N2) observed here in these studies, some of the observed differences in interim VE may indicate that further explanations, such as immunological cohort effects [21], need to be investigated.
Recent vaccine seed A(H3N2) viruses have developed adaptions during propagation in eggs, potentially negatively affecting VE [22]. Cell-propagated vaccines were used in some studies, but there was insufficient data to estimate vaccine type-specific VE.
VE against influenza B was > 60% among all ages in primary care, with a lower VE among those aged 18 to 64 years in EU-PC. However, sample size was lower in this age group and the low VE may possibly be a result of random variation. The high overall VE was comparable with 2019/20 interim VE against influenza B in Canada (69%) [15]. In the US, overall VE against B/Victoria was lower, at 50% [16].
In studies where lineage was available, the vast majority of circulating influenza B belonged to the B/Victoria lineage, which was the lineage included in the trivalent vaccine but of another subgroup. The quadrivalent vaccine was used widely in Europe, however because of low circulation of B/Yamagata viruses, the VE of quadrivalent and trivalent vaccines could not be compared in the context of VE against trivalent lineage-mismatched influenza B. Among the sequenced B/Victoria viruses, 98% belonged to the subgroup 1A(del162-164)B, differing from the vaccine virus by a further amino acid deletion, and being antigenically different [3,23,24]. Nevertheless, human serology studies show some evidence of cross-reactivity between the vaccine and the circulating B/Victoria viruses [20,23,24].
The early start of the season in most European countries included in these six studies [5] resulted in higher incidence and greater precision for interim VE estimates than in 2018/19, although some studies did have lower sample size for some subgroups in this interim analysis. Each study used study-specific criteria to define if a sample size was too small to attempt to estimate VE. Where VE results were presented, and sample size was small, sensitivity analyses were used to address potential small sample bias where appropriate. Endof-season estimates will have higher sample size and provide more robust estimates. Residual confounding and bias are known limitations potentially present in all observational studies.
Vaccination remains the key successful method of influenza prevention, with one to two-thirds of all vaccinated individuals receiving protection during the 2019/20 influenza season. Promotion of influenza vaccination should be maintainted in line with national guidelines and recommendations in all European countries with ongoing influenza virus circulation. Given the variation in VE estimates against A(H3N2), it remains important that when national guidelines indicate neuraminidase inhibitors to be used, they are used regardless of vaccination status as prophylaxis and therapy where there is influenza A(H3N2) virus circulation to help prevent severe outcomes [1].
Bi-annual reports on influenza VE in prior and exist- The VE and antigenic studies at the end of the 2019/20 season will help to explain the differences in age-, subtype-and study-specific VE estimates presented here. In order to be prepared for the next season in the northern hemisphere, we should continue to monitor the genetic diversity of the 2020 southern hemisphere influenza viruses and their influenza VE. European IVE group: (i) Primary care and hospital sites at national/regional level: data collection, data validation, results interpretation, review of manuscript. (ii) Laboratories: virological data collection, validation and analysis, genetic characterisation, interpretation of results, review of manuscript.

European IVE group
Marta Valenciano, Alain Moren: study design, coordination of I-MOVE network, interpretation of results, contribution to manuscript writing.
Pasi Penttinen: study design, interpretation of results, review of manuscript.