Vaccine effectiveness against influenza A(H3N2) and B among laboratory-confirmed, hospitalised older adults, Europe, 2017-18: A season of B lineage mismatched to the trivalent vaccine.

Abstract Background Influenza A(H3N2), A(H1N1)pdm09 and B viruses co‐circulated in Europe in 2017‐18, predominated by influenza B. WHO‐recommended, trivalent vaccine components were lineage‐mismatched for B. The I‐MOVE hospital network measured 2017‐18 seasonal influenza vaccine effectiveness (IVE) against influenza A(H3N2) and B among hospitalised patients (≥65 years) in Europe. Methods Following the same generic protocol for test‐negative design, hospital teams in nine countries swabbed patients ≥65 years with recent onset (≤7 days) severe acute respiratory infection (SARI), collecting information on demographics, vaccination status and underlying conditions. Cases were RT‐PCR positive for influenza A(H3N2) or B; controls: negative for any influenza. “Vaccinated” patients had SARI onset >14 days after vaccination. We measured pooled IVE against influenza, adjusted for study site, age, sex, onset date and chronic conditions. Results We included 3483 patients: 376 influenza A(H3N2) and 928 B cases, and 2028 controls. Most (>99%) vaccinated patients received the B lineage‐mismatched trivalent vaccine. IVE against influenza A(H3N2) was 24% (95% CI: 2 to 40); 35% (95% CI: 6 to 55) in 65‐ to 79‐year‐olds and 14% (95% CI: −22 to 39) in ≥80‐year‐olds. Against influenza B, IVE was 30% (95% CI: 16 to 41); 37% (95% CI: 19 to 51) in 65‐ to 79‐year‐olds and 19% (95% CI: −7 to 38) in ≥80‐year‐olds. Conclusions IVE against influenza B was similar to A(H3N2) in hospitalised older adults, despite trivalent vaccine and circulating B lineage mismatch, suggesting some cross‐protection. IVE was lower in those ≥80 than 65‐79 years. We reinforce the importance of influenza vaccination in older adults as, even with a poorly matched vaccine, it still protects one in three to four of this population from severe influenza.


| INTRODUC TI ON
In Europe, most countries recommend seasonal influenza vaccination for populations at risk of severe disease, such as older adults (aged 65 years and above), or those with co-morbid conditions like heart disease or diabetes. 1  Most (97%) of the B viruses typed were B/Yamagata. 3 Most of the hospitalised severe acute respiratory illness (SARI) cases (occurring mainly in older adults) were caused by infection with influenza B. 3 In February 2018, early influenza vaccine effectiveness (IVE) was estimated from five European studies, covering all ages and including both primary care and hospitalised patients. The IVE against influenza A(H3N2) from these studies across nine countries (including two multi-country I-MOVE/I-MOVE+ studies) ranged from −42% to 7%, with IVE of 36%-54% for influenza B and 55%-68% for influenza A(H1N1)pdm09. 4 The I-MOVE (Influenza-Monitoring of Vaccines in Europe) hospital network has provided annual IVE for participating hospitals across Europe since 2015. In 2017-18, we measured seasonal IVE against hospitalisation with influenza A(H3N2) and B among older adults (aged 65 years and over) in the European Union (EU) by age group, underlying co-morbid conditions, and vaccination in previous seasons.

| Setting and population
In 2017-18, the I-MOVE hospital network included 23 hospitals from 10 study sites in nine countries across the EU (one site in each of Croatia, Finland, France, Italy, Lithuania, the Netherlands, Portugal, and Romania, and two study sites in Spain). The study population comprised all community-dwelling elderly (≥65 years) admitted to these participating hospitals and diagnosed with SARI within the 7 days prior to swabbing. All study sites followed the same generic protocol, adapted to their setting. 5 Following local requirements, each local protocol was submitted to and approved by the ethics committee(s) in each site. Patient consent was an inclusion criterion for the study in each hospital, described in the shared protocol. 5

| Definitions and exclusions
Study methods have been described in detail elsewhere. 6 Briefly, this multi-centre study used a test-negative design (TND), in which cases were hospitalised SARI patients (ie with at least one systemic and one respiratory sign or symptom) with confirmed influenza A(H1N1)pdm09, A(H3N2) or B using reverse-transcriptase polymerase chain reaction (RT-PCR). Patients confirmed with influenza A(H1N1)pdm09 were excluded from analyses due to the small sample size within this population (N = 133). Controls were SARI patients who were influenza negative by RT-PCR. The study period, defined in each site, was from the onset week of the first to the last confirmed case by influenza (sub)type. Patients were considered vaccinated if they had SARI onset >14 days after vaccination. Those with onset within the 14-day period were excluded from analysis, and any patient vaccinated on or after SARI onset was recoded as unvaccinated. Any study site with <10 (sub)type-specific cases was excluded from pooled analysis.

| Analysis
We estimated overall adjusted IVE as IVE = (1-OR a ) × 100 using a onestage analysis of pooled individual data from all sites, for influenza A(H3N2) and B. We used logistic regression, with study site as a fixed effect. In secondary analyses, IVE for both subtypes was estimated and stratified by (a) age group (65-79 years old and 80 years and over); (b) presence of at least two underlying chronic conditions (vs one or less); 12 Division for epidemiology of communicable diseases, Croatian Institute of Public Health, Zagreb, Croatia and (c) vaccination status over the current and previous two seasons, divided into five groups: vaccinated in this season only, vaccinated in either of the two previous seasons but not in this season, vaccinated in the current plus either of the previous two seasons, vaccinated in all three seasons, no vaccination in any of the three seasons the reference group). Penalised logistic regression was used where any stratum resulted in small enough numbers to lead to over-fit models (defined as a having fewer than 10 cases or controls per model parameter).
For most analyses, the odds ratio (OR) was adjusted (OR a ) by age, sex, symptom onset date, and presence of at least one, or the total number of underlying chronic conditions (none, one, two or more). The chronic conditions were selected from diabetes mellitus, cancer, heart/rheumatic/kidney/lung disease, being immunocompromised, or being obese. For each country, only chronic conditions for which vaccination was recommended were included. The OR a s for analyses stratified by individual chronic condition (where the total number of cases or controls with the condition was at least 50) were adjusted as stated above, but not by chronic condition. The variables onset date and age were modelled as continuous (using a restricted cubic spline with 3, 4 or 5 knots) or categorical terms, the choice of which for each model was determined by sample size and the Akaike's information criterion.
The categorical term for onset date, "phase," was determined for each site prior to pooling the data and comprised three categories: early (onset dates comprising up to and including the first 10% of cases), peak (representing 80% of cases) and late (for the final 10% of cases, with the latest onset date). The categorical term for age, "age group," comprised two categories: 65-79, and 80 years and over.
Sensitivity analyses included estimating IVE for those swabbed within 3 days and for those who were not on antivirals before swabbing.

| Influenza A(H3N2)
Cases and controls were similar in terms of median age (81 vs 80 years), sex (52% vs 53% being male) and vaccine status (60% vs 62% vaccinated in the current and 61% vs 62% in the previous season) ( Table 1). One vaccinated patient (<0.1%; a control) received quadrivalent, and the others received trivalent inactivated influenza vaccine. All vaccines were egg-propagated. Over 90% of both cases and controls had at least one underlying chronic condition (P = .360), the main one of which was heart disease, with almost two-thirds of both cases and controls affected (P = .283; Table 1). More controls than cases had lung disease (P = .027), but more cases than controls had rheumatic disease (P = .002; Table 1).
Adjusted IVE against influenza A(H3N2) was 37% (95% CI: −2 to 61) among patients vaccinated in this and either of the two previous seasons, 35% (95% CI: 12 to 52) among patients vaccinated in all three seasons, 22% (95% CI: −50 to 59) for those vaccinated in the current season only and 21% (95% CI: −23 to 50) among those vaccinated in either of the two previous seasons but not in the current.

| Influenza B
Cases and controls were similar in terms of median age (78 vs 79 years). A higher proportion of controls than cases were male (56% vs 48%; P = .008) and vaccinated in this or the previous season (59% vs 44% for both seasons; P < .001) ( Table 1). One vaccinated patient (<0.1%; a control) received quadrivalent influenza vaccine, and the rest received the lineage-mismatched trivalent vaccine. All vaccines were egg-propagated. Ninety-three per cent of both cases and controls had at least one underlying chronic condition (P = .527), the main one of which was heart disease, with almost two-thirds of both cases and controls affected (P = .124; Table 1). More controls than cases had lung disease (P < .001), rheumatic disease (P = .004) and kidney disease (P = .032; Table 1 for those who were not on antivirals before swabbing.  Spanish study with the same inpatient setting but in those >60 years found a similar result to ours (26%; 95% CI: −9 to 49). 25 Older adults (aged 65 years and above) are known to be at increased risk of severe influenza requiring hospitalisation, especially those with underlying co-morbid conditions. 26 Some of these conditions (such as cancer, diabetes, heart and lung diseases) are more prevalent in this age group, in particular cardiovascular disease, the most common chronic condition to affect older adults with influenza. 27 This was seen in our study, with almost two-thirds of patients with either type of influenza, as well as test-negative controls, having heart disease (Table 1), while the prevalence of each of the other co-morbid conditions measured was below 50%.

| D ISCUSS I ON
Vaccination against influenza has been described as having low or very low effectiveness among the older adult population, 28 as well as among patients with these co-morbid conditions. 29 Adjusted for study site, sex, age (modelled as a linear term) and onset date (modelled using a restricted cubic spline [RCS] with 5 knots). g Adjusted for study site, sex, age (modelled using a RCS with 3 knots) and onset date (modelled using a RCS with 5 knots).
h Adjusted for study site, sex, age (modelled using a RCS with 3 knots), onset date (three phases: early, peak and late) and presence of any chronic condition. i Adjusted for study site, sex, age (modelled as a linear term), onset date (modelled using a RCS with 5 knots) and presence of any chronic condition. AD is large between the prior vaccine and the current circulating strain. 41 Based on circulating strains this season 42 and recommended vaccines for this and the previous two seasons, 2

| CON CLUS ION
For the 2017-18 season among hospitalised older adults, IVE against influenza B was greater than that against A(H3N2), despite a trivalent vaccine and circulating B lineage mismatch, suggesting some cross-protection (as quadrivalent vaccine was used in <0.5% of this population). Antigenic changes due to egg-adaptation of the vaccine strain could have contributed to the low IVE against A(H3N2). Our results suggest lower IVE against both influenza A(H3N2) and B in those ≥80 years than in those aged 65-79 years. We reinforce the importance of influenza vaccination in older adults as, even in seasons using trivalent vaccine with circulating influenza B lineage mismatch and adaption of the egg-propagated vaccine virus, it remains preventive against severe influenza for at least one in four of this population.

ACK N OWLED G EM ENTS
This study received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 634446. The I-MOVE study team is very grateful to all patients, hospital teams, laboratory teams and regional epidemiologists who have contributed to the studies.

CO N FLI C T O F I NTE R E S T
None declared. At the time of the study, Ritva Syrjänen was a coinvestigator in pneumococcal studies (not related to this study), for which the Finnish Institute for Health and Welfare has received research support from GlaxoSmithKline Biologicals.