Early season co-circulation of influenza A(H3N2) and B(Yamagata): interim estimates of 2017/18 vaccine effectiveness, Canada, January 2018

Using a test-negative design, we assessed interim vaccine effectiveness (VE) for the 2017/18 epidemic of co-circulating influenza A(H3N2) and B(Yamagata) viruses. Adjusted VE for influenza A(H3N2), driven by a predominant subgroup of clade 3C.2a viruses with T131K + R142K + R261Q substitutions, was low at 17% (95% confidence interval (CI): −14 to 40). Adjusted VE for influenza B was higher at 55% (95% CI: 38 to 68) despite prominent use of trivalent vaccine containing lineage-mismatched influenza B(Victoria) antigen, suggesting cross-lineage protection.

The 2017/18 influenza season in Canada has been characterised by co-circulation of influenza A(H3N2) and B(Yamagata) viruses, the latter unusual so early in the season [1]. Most European countries are also experiencing simultaneous influenza A and B epidemics, with B(Yamagata) predominating [2], whereas the United States (US) has experienced a substantial epidemic due predominantly to influenza A(H3N2) [3]. Low vaccine effectiveness (VE) for the 2017/18 season has been anticipated following the interim report from Australia indicating VE of just 10% during its 2017 influenza A(H3N2) epidemic [5]. In the context of exclusive QIV use, Australia reported higher VE of 57% against co-circulating influenza B viruses [5]. Here we report interim 2017/18 VE estimates for influenza A(H3N2) and influenza B from participating provinces of the Canadian Sentinel Practitioner Surveillance Network (SPSN), where QIV comprised less than one third of vaccine doses distributed overall through the publicly funded campaign.

Vaccine effectiveness evaluation
VE was derived using a test-negative design [6][7][8][9]. Nasal/nasopharyngeal specimens and epidemiological data were collected from patients presenting within 7 days of onset of influenza-like illness (ILI) to community-based sentinel practitioners in Alberta, British Columbia, Ontario and Quebec. ILI was defined as acute onset of fever and cough and at least one other symptom including sore throat, myalgia, arthralgia or prostration. Fever was not a requirement for elderly adults 65 years of age and older. Vaccination status was based on patient and/or practitioner reporting of 2017/18 vaccination at least 2 weeks before symptom onset; patients vaccinated less than 2 weeks before onset or with unknown vaccination status/timing were excluded. Institutional review boards in each province provided ethical approval for the study.
Specimens collected from week 45 (starting 5 November 2017) to week 3 (ending 20 January 2018) were tested for influenza type/subtype by real-time RT-PCR at provincial public health reference laboratories. Sanger sequencing of the viral haemagglutinin gene was undertaken on a subset of original patient specimens collected up to 13 January 2018 to assess the contribution of genetic clades to VE estimates.
Odds ratios (OR) comparing test-positivity for influenza A(H3N2) or B between vaccinated and unvaccinated participants who were at least 1-year-old were calculated using logistic regression, adjusted for relevant covariates. VE was derived as (1 − OR) × 100%.
Most sequenced influenza A(H3N2) viruses belonged to genetic clade 3C.2a (213/229; 93%) and of these most (204/213; 96%) belonged to a single genetic subgroup of 3C.2a (denoted subgroup 3 by nextflu.org [10]), bearing antigenic site A substitutions T131K and R142K and antigenic site E substitution R261Q (Table 1). Overall 89% of influenza A(H3N2) viruses belonged to clade 3C.2a subgroup 3, which is similar to other surveillance observations from Canada (83%) ( Figure 2) and to recent reports from Europe [11]. However, this profile for the 2017/18 season is different from that found by the Canadian SPSN during 2016/17 or by Australia during its 2017 epidemic, when a greater mix of genetic variants contributed to interim analyses and only 14% and 7%, respectively, of influenza A(H3N2) viruses belonged to subgroup 3 ( Figure 2).

Discussion
In most other interim analyses by the Canadian SPSN, type B viruses comprised less than 10% of influenza detections, whereas in 2017/18, they were identified in an equal proportion with influenza A(H3N2) [7][8][9]. Although the reasons for an earlier influenza B onset are unclear, Canada experienced a substantial influenza A(H3N2) epidemic in 2016/17 that may have altered population immunity and the overall 2017/18 influenza A(H3N2) contribution [9].
Our 2017/18 interim VE for influenza A(H3N2) is more comparable to the 2017 southern hemisphere interim VE of 10% (95% CI: −16 to 31) reported from Australia [5]. Differences in virological and participant profiles, as well as the stage of the epidemic, have to be taken into account when comparing VE estimates across studies. Working-age adults comprised the majority of  All influenza vaccine manufacturing in Canada is eggbased. Mutations that arise from egg adaptation of the vaccine strain may affect VE, an issue also identified for the current season's A(H3N2) vaccine component [17,18]. In Canada this season, antigenic characterisation of influenza A(H3N2) viruses has only been presented in relation to a cell-propagated version of the vaccine reference strain; characterisation against an egg-based version has not been reported [1]. Among the small subset of Canadian viruses that could be successfully characterised, all were considered antigenically similar to the cell-propagated vaccine strain [1]. Conversely, where relatedness to the egg-propagated version of the vaccine strain has been specifically explored elsewhere, more variability has been identified, with a greater proportion of viruses considered antigenically distinct from the egg-propagated version [5,11,19].
We found higher VE of 55% (95% CI: 38 to 68) against influenza B despite prominent use of TIV containing a B(Victoria) antigen that was lineage-mismatched to almost exclusively B(Yamagata) viruses. Approximately 70% of vaccine doses distributed in SPSN provinces during the 2017/18 season were TIV, albeit with regional variation that will be explored in end-of-season analyses. Substantial cross-lineage VE for influenza B has been observed previously [20], including during the prior 2016/17 season in Canada when VE against lineage-mismatched influenza B using the same B(Victoria) TIV component was 73% (95% CI: 52 to 84) [14] and QIV comprised an even smaller proportion (< 25%) of vaccine doses distributed. Our estimate for the current season is comparable to the interim VE of 57% (95% CI: 41 to 69) for influenza B reported from Australia, despite exclusive use of QIV in that country [5].
Other agent-host and immuno-epidemiological interactions, including birth cohort effects induced by differential prime-boost exposures, may also play a role in VE [21]. The effect of prior vaccination history was Sequencing of the haemagglutinin gene was attempted on a subset of available influenza-positive original patient specimens contributing to interim 2017/18 vaccine effectiveness evaluation. Sequencing was successful for 229 of 236 (97%) influenza A(H3N2) specimens and 233 of 246 (95%) influenza B specimens (collection dates: 5 November 2017 to 13 January 2018). a Subgroup name as assigned by nextflu.org [10]. b Five of six viruses in this subgroup had additional substitutions T135K (associated with loss of a potential glycosylation site) + R150K + R261Q; the sixth virus had N122D (associated with loss of a potential glycosylation site) + N171D + S262N. c 44 of 204 (22%) viruses in this subgroup had an additional substitution K92R and 28 of 204 (14%) had A212T. d All viruses in this subgroup had additional substitutions G78D + Y94H + V182I; two viruses also had N122D (associated with loss of a potential glycosylation site). e All but two viruses in this subgroup had additional substitutions T135K (associated with loss of a potential glycosylation site) + E62G + R142G; one virus also had N122D (associated with loss of a potential glycosylation site). One of the other two viruses had T135N. f All but one of the viruses in this subgroup had additional substitutions L172Q + M251V; one virus had M251V without L172Q. g Five of six viruses in this subgroup had a deletion at position 162-163. The number of participants with unknown sex or comorbidity are shown in table but excluded from the denominator for calculating percentages. a Vaccination status based on patient and/or practitioner report; defined as receipt of 2017/18 seasonal influenza vaccine ≥ 2 weeks before symptom onset. Patients vaccinated < 2 weeks before onset or with unknown vaccination status/timing were excluded. b p value for comparison of influenza A(H3N2) cases to negative controls. Differences were compared using the chi-squared test or Wilcoxon rank-sum test. c p value for comparison of influenza B cases to negative controls. Differences were compared using the chi-squared test or Wilcoxon rank-sum test. d p value for comparison of vaccinated participants to unvaccinated participants. Differences were compared using the chi-squared test or Wilcoxon rank-sum test. e Includes chronic co-morbidities that place individuals at higher risk of serious complications from influenza as defined by Canada's National Advisory Committee on Immunization (NACI), including: heart, pulmonary (including asthma), renal, metabolic (such as diabetes), blood, cancer or immunocompromising conditions, conditions that compromise management of respiratory secretions and increase risk of aspiration, or morbid obesity (body mass index ≥ 40). f Missing specimen collection dates were imputed as the laboratory accession date minus 2 days, the average time between specimen collection and accession dates among specimens with complete information for both values. g Participants who received seasonal 2017/18 influenza vaccine < 2 weeks before ILI onset or for whom vaccination timing was unknown were excluded from the primary analysis. They are included here for assessing vaccination regardless of timing for comparison to other sources of vaccination coverage.
not assessed here owing to sample size limitations, but will be explored as part of the end-of-season analyses.

Conclusions
As reported from Australia for the 2017 southern hemisphere vaccine, interim estimates from Canada for the 2017/18 northern hemisphere vaccine indicate low VE of less than 20% against influenza A(H3N2), notably among working-age adults. While the influenza A(H3N2) epidemic continues, adjunct protective measures should be reinforced to minimise the associated disease burden in high-risk individuals [22]. Interim 2017/18 VE estimates against influenza B are higher at 55% despite prominent TIV use, suggesting crosslineage protection.