Interim estimates of 2019/20 vaccine effectiveness during early-season co-circulation of influenza A and B viruses, Canada, February 2020

Interim results from Canada's Sentinel Practitioner Surveillance Network show that during a season characterised by early co-circulation of influenza A and B viruses, the 2019/20 influenza vaccine has provided substantial protection against medically-attended influenza illness. Adjusted VE overall was 58% (95% confidence interval (CI): 47 to 66): 44% (95% CI: 26 to 58) for A(H1N1)pdm09, 62% (95% CI: 37 to 77) for A(H3N2) and 69% (95% CI: 57 to 77) for influenza B viruses, predominantly B/Victoria lineage.

The 2019/20 northern hemisphere influenza season has been characterised by early co-circulation of influenza A and B viruses [1][2][3][4][5]. We report interim virological and vaccine effectiveness (VE) findings for the 2019/20 season from the community-based Canadian Sentinel Practitioner Surveillance Network (SPSN).

Study design
VE was estimated using a test-negative design as previously described [6]. Nasal/nasopharyngeal specimens were collected from patients presenting to sentinel sites in the provinces of Alberta, British Columbia, Ontario and Quebec. Patients who were at least 1 year of age and who presented within 7 days of onset of influenza-like illness (ILI) were eligible for inclusion in VE analyses. ILI was defined by self-reported fever and cough and one or more of arthralgia, myalgia, prostration or sore throat. Fever was not a requirement for adults aged ≥ 65 years old. Influenza vaccination status was based on self-(or parent/guardian) report of 2019/20 vaccine receipt ≥ 2 weeks before ILI onset. Specimens were tested for presence of influenza virus by real-time RT-PCR assays. Sanger sequencing of the haemagglutinin (HA) gene was undertaken on a convenience sample of original patient specimens. Amino acid substitutions at HA antigenic sites are hereafter specified in parentheses, those affecting the receptorbinding site as 'RBS' and changes associated with potential gain or loss of N-linked glycosylation as '+/− CHO'. Viral sequence data were deposited for reference into the Global Initiative on Sharing All Influenza Data (GISAID) platform (www.gisaid.org) under accession numbers EPI_ISL_41122-411846. Antigenic characterisation of a convenience sample of virus isolates was undertaken by haemagglutination inhibition (HI) assay using post-infection ferret anti-sera raised to egg-passaged influenza A and cell-passaged influenza B vaccine reference strains, conducted as previously described [6][7][8].
Adjusted odds ratios (OR) for influenza test-positivity between vaccinated and unvaccinated participants were derived using a logistic regression model. VE was calculated as (1 − adjusted OR) × 100%.

Influenza vaccine components and formulations
All influenza vaccines used in Canada were manufactured in eggs and inactivated. Overall and by province ≥ 74% of publicly-funded doses were quadrivalent except in British Columbia where 16% of doses overall were quadrivalent and targeted to children. In Ontario high-dose trivalent vaccine was publicly funded for elderly adults aged ≥ 65 years old.

Study period and influenza detection
The study period spanned specimen collection dates from 1 November 2019 (week 44) to 1 February 2020 (week 5) during which 2,808 specimens met inclusion criteria. Of these, 1,411 (50%) were influenza test-positive including 731 (52%) influenza A and 683 (48%) influenza B viruses, with three influenza A and B co-infections. Of the 715 influenza A cases of known Missing specimen collection dates were imputed as the date the specimen was received and processed at the provincial laboratory minus 2 days.

Participant characteristics
As in prior seasons [6,8], most (61%; 1,718/2,808) participants were adults 20-64 years old (Table 1). Among test-negative controls, 21% (295/1,397) had one or more comorbidities, which is comparable to last season's interim report (22%) and consistent with other surveillance data indicating > 20% of Canadians live with a major chronic disease [11]. Vaccination ≥ 2 weeks before ILI onset was reported by 29% (399/1,397) of controls overall and 26% (229/877) of those 20-64 years old, also similar to last season's interim report (27% and 24%, respectively) [8].  Additional adjustment for comorbidity (yes/no/unknown) and sex (male/female/unknown) did not alter any of the displayed VE estimates by more than 2% (absolute) except where specified. c Using a later study start date of 1 December 2019 did not alter any of the displayed VE estimates by more than 3% (absolute). d Excluding the province of British Columbia where a smaller proportion of doses distributed were quadrivalent, the VE estimate for all ages was unchanged and age-stratified estimates remained within 6% (absolute) of those displayed. e With additional adjustment for comorbidity (yes/no/unknown) and sex (male/female/unknown), VE was 14% (95% CI: −71 to 57). f Excluding the province of British Columbia, none of the influenza B VE estimates were higher and all remained within 5% (absolute) of those displayed.  clade 6B.1A5 of which 245 (86%) further sub-clustered with 6B.1A5A and 39 (14%) with 6B.1A5B (Table 2).
Of 87/551 (16%) A(H1N1)pdm09 viruses characterised by HI assay, 41 (47%) were antigenically distinct from the vaccine strain. Sequence information was available for 39/41 and all belonged to the 6B.1A5A sub-group bearing the new Sa substitutions.

Influenza B
VE against influenza B was 69% (95% CI: 57 to 77) ( Table 1). Of 683 influenza B viruses detected by the SPSN and contributing to VE analyses, 260 (38%) were characterized as B/Victoria lineage by sequencing (one other virus was characterized as B/Victoria lineage by HI assay). Virtually all (259; 99%) sequenced B/Victoria-lineage viruses belonged to clade V1A.3 (Δ3) characterised by a triple deletion at amino acids 162-164 in the 160-loop. All 259 viruses belonged to the V1A.3B sub-cluster with K136E in the 120-loop with most (196/259; 76%) also bearing other 120-loop substitutions. Just one virus belonged to the same B/ Victoria clade as the trivalent vaccine strain (V1A.1 (Δ2)) and one other sequenced virus belonged to the same B/Yamagata-lineage as the quadrivalent vaccine strain (clade 3) ( Table 2). Fifty-eight viruses were HI-characterised: all but one was distinct from the cellpassaged vaccine strain.

Discussion
Interim results from Canada's SPSN indicate that the 2019/20 influenza vaccine has provided substantial protection against medically-attended influenza illness during a season characterised by an approximately equal mix of influenza A and B viruses, a substantial proportion of which were genetically and antigenically mismatched to vaccine.
Almost all influenza B viruses belonged to the B/ Victoria lineage which has not otherwise contributed much since the 2015/16 season [1,18]. Children are most affected by influenza B, particularly B/Victorialineage viruses [19,20], and this may be evident in the over-representation of children 1-19 years old among unvaccinated influenza B cases (307/623; 49%) compared with controls (284/998; 28%) or with the population of SPSN provinces (20%) overall [21]. Whereas the 2019/20 vaccine is a double deletion V1A.1 (Δ2) strain, virtually all viruses collected and sequenced by the SPSN were instead triple deletion V1A.3B (Δ3) variants, as also noted from Europe [3] and the United States (US) [4]. The majority of B/Victoria-lineage viruses HI-characterised by the SPSN (57/58), and otherwise in Canada (157/173; 91%) [1] have also been antigenically distinct from the vaccine strain. Notwithstanding that vaccine mismatch, we found substantial VE of 69% overall and 77% in children. As previously highlighted, influenza B immuno-epidemiology is complex with cohort effects and cross-lineage interactions that may also play a role in vaccine protection [18,19,[22][23][24].
Most but not all A(H3N2) viruses successfully characterised by HI assay to date in Canada (35/41; 85%) [1] and in Europe (11/17; 65%) [3] have been antigenically distinct from the egg-adapted vaccine strain, and in the US most (39/69; 57%) have also been distinct from the cell-passaged vaccine strain based upon focus reduction assay [4]. In that regard, the VE of 62% we report may be unexpected. Effectiveness of the 2019/20 clade 3C.3a vaccine against predominant 3C.2a1b viruses is higher than observed for the 2018/19 clade 3C.2a1 vaccine against late-season A(H3N2) viruses overall (17%) or in clade-specific analyses against co-circulating 3C.2a1b (27%) or 3C.3a (−32%) viruses [25]. Antibody induced to clade 3C.3a may be more cross-reactive than that of antibody induced to clade 3C.2a [26,27], and recent parallel substitutions shared between 3C.3a and 3C.2a1b/T135K viruses (e.g. A138S, F193S) may further contribute. An immunological cohort effect (i.e. imprint-regulated effect of vaccine; I-ReV) was hypothesised last season to explain the paradoxical negative VE for the 3C.2a1 vaccine against 3C.3a viruses, notably among adults 35-54 years of age [25,28,29]. Whether the I-ReV hypothesis may also apply, but in reverse, to explain this season's protective VE for 3C.3a vaccine against 3C.2a1 viruses requires greater sample size to explore. We highlight that only once previously in the past decade (2011/12) has the SPSN reported an overall VE exceeding 50% for A(H3N2) viruses [6]. As such, and particularly noting the limited sample size of A(H3N2) cases, our interim estimate of 2019/20 A(H3N2) VE requires cautious interpretation pending further end-of-season evaluation.
Limitations of the current analysis include its observational design for which residual bias and confounding cannot be ruled out. Sample size considerations preclude further stratification (e.g. by additional age and/ or genetic sub-groups, or prior vaccination history) but will be attempted end-of-season. Our analyses reflect specimens and data collected as at 1 February 2020 but may change towards the end of the ongoing epidemic.

Conclusions
The 2019/20 VE reported by the Canadian SPSN suggests that, among non-elderly individuals, about six of 10 cases of medically-attended febrile respiratory illness due to influenza might have been prevented by vaccination. Such substantial vaccine protection despite antigenic mismatch, notably to circulating influenza A(H3N2) and B/Victoria viruses, invites exploration of other factors potentially contributing to VE.