Predominance of a Drifted Influenza A (H3N2) Clade and its Association with Age-specific Influenza Vaccine Effectiveness Variations, Influenza Season 2018-2019.

Background: Influenza A (H3N2) clade 3C.3a was the predominant influenza virus in Israel throughout the 2018-2019 season, constituting a drift from the influenza A (H3N2) vaccine. We estimated the end-of season vaccine effectiveness (VE) by age, among community patients with influenza-like illness (ILI), considering the hemagglutinin (HA) gene mutations and amino acid substitutions of influenza A (H3N2) viruses detected. Methods: Nose-throat samples were analyzed for the presence of influenza virus, type/subtype, and HA gene sequence. HA gene sequences and amino acid substitutions were compared to the influenza A/Singapore/INFIMH-16-0019/2016 (H3N2)-like 2018-2019 vaccine virus, and a phylogenetic tree was generated. Influenza VE against influenza A (H3N2) was estimated using the test-negative design. VE was estimated by age group and by 15 year moving age intervals. Results: In total, 90% of the influenza A (H3N2) viruses belonged to the 3C.3a clade, constituting a unique situation in the northern hemisphere. Adjusted all-age influenza A (H3N2) VE was -3.5% (95% CI: -51.2 to 29.1). Although adjusted VEs were very low among infants, children, and young adults, a VE of 45% (95% CI: -19.2 to 74.6) was estimated among adults aged ≥45 years old. Conclusions: The higher VE point estimates among older adults may be related to previous exposure to similar influenza viruses.


Introduction
In recent years, influenza A (H3N2) viruses have presented a significant challenge for vaccine selection. In fact, the recommendation regarding the 2019-2020 northern hemisphere influenza A (H3N2) vaccine component was made one month later than for the influenza A (H1N1)pdm09 and the influenza B vaccine components [1]. This challenge stems, for the most part, from the more frequent genetic changes that the influenza A (H3N2) subtype undergoes, as compared with the influenza A (H1N1) and influenza B viruses [2]. The last few influenza seasons witnessed an influenza A (H3N2)

Study Period and Population
Combined nose-throat samples were collected from ILI patients visiting the Israel Influenza Surveillance Network (IISN) 32 outpatient sentinel clinics. The study period lasted from September 30th 2018 to March 30 th 2019. ILI was defined as a temperature of ≥37.8 • C with at least one of the following symptoms: sore throat, coryza, cough, and muscle ache. Medical teams were granted discretion concerning the inclusion of other symptoms and signs that appeared relevant [3,8]. Demographic and medical information was collected from patients' medical records, including birth date, gender, symptom onset date, date of nose-throat sampling, chronic medical conditions associated with increased risk for severe influenza and complications, influenza vaccination status for the 2018-2019 season, as well as vaccination date and the type of vaccine administered. For children aged 6 months to 9 years, a second influenza vaccine dose (for those receiving the influenza vaccine for the first time) and its administration date were documented, when applicable.

Vaccine Effectiveness Study Design
VE was estimated for individuals ≥6 months of age by means of the test-negative design using the formula (1-OR) x 100 [8]. Sample suitability for VE analysis was determined as previously described [8]. Specifically, samples collected from patients >7 days after the onset of disease and <14 days following influenza vaccination were excluded from analysis. Patients with missing critical data, or unknown vaccination status were excluded as well. A patient was determined vaccinated if the 2018-2019 influenza vaccine was administered ≥14 days before the onset of ILI. For children 6 months to 9 years of age who received the influenza vaccine for the first time in their life, only those receiving two doses, with the second dose being administered ≥14 days prior to the onset of disease, were deemed vaccinated. Those children who were eligible for two vaccine doses, but received only one dose, or the second dose was administered <14 days prior to onset of illness, were considered partially vaccinated, and were excluded from the analysis.
VE was calculated for all ages and for specific age groups. In order to gain insight into the VE trend at various ages, adjusted VE was also calculated using moving 15 year intervals, beginning at the first year of life and progressing by 1 year increments [8]. A cubic spline function with knots located at five discrete 15 year age intervals was then applied [8].

Statistical Analysis
The Pearson's χ 2 test was applied for comparison of demographic and medical characteristics between cases and controls.
A univariate logistic regression model was applied to calculate the odds ratio for crude VE estimates, including the 95% confidence intervals. A multivariate logistic regression model was applied to adjust for age, gender, number of days from disease onset to sampling, calendar month of sample collection, and underlying chronic conditions placing patients at risk for severe influenza or complications. Wherever the number of patients was too small in a particular calendar month, instead of adjusting according to calendar month, adjustment was performed for the high influenza activity period (December, January, February) and low influenza activity period (October, November, March).
Statistical analyses were performed using SAS Enterprise Guide 7.1 (SAS Institute Inc.) and R version 3.5.2 (R foundation, Vienna, Austria) software.

Ethical Considerations
Influenza surveillance in Israel, including the laboratory confirmation, is implemented under the Israel Public Health Ordinance, and does not entail informed consent.

Influenza Virus Circulation and Characterization
During the 2018-2019 study period, 1476 combined nose-throat samples were collected from ILI patients. A total of 594 (40%) were positive for influenza. Of these, 451 (75.9%) were positive for influenza A (H3N2), 134 (22.6%) for influenza A (H1N1)pdm09, 5 (0.8%) for influenza B, and 4 (0.7%) were positive for both influenza A (H1N1)pdm09 and influenza B (Figure 1).  Table S2. A total of 16 substitutions were identified in all viruses. A total of 9 of those 16 substitutions belonged to known antigenic sites that have been associated with antigenic drift mutations [11][12][13][14], that is, antigenic sites A (one substitution), B (four substitutions), C (one substitution), D (one substitution), and E (two substitutions) (Table S2). Additional common AA substitution occurred in several viruses (Table S2). When compared to the egg-propagated vaccine virus, patient samples contained two additional AA substitutions, neither in an antigenic site. Figure 2A presents the phylogenetic tree of representative influenza A (H3N2) viruses. Figure

Vaccine Effectiveness against Influenza A (H3N2)
A total of 50 individuals were not eligible for VE analysis due to their age (<6 months old), 14 due to unknown vaccination status, 33 due to vaccination <14 days before symptom onset, 38 due to sampling >7 days after symptom onset, 6 due to missing critical data, and 21 children due to partial vaccination ( Figure 3). After exclusion of these 162 individuals, 1314 were found eligible for VE analysis ( Figure 3); among them, 570 (43%) were positive for influenza. Of these, 435 (76%) were positive for influenza A (H3N2), 131 (23%) for influenza A (H1N1)pdm09, and 7 for influenza B (1%). Three (0.5%) samples were positive for both influenza A (H1N1)pdm09 and influenza B ( Figure 3).
As a result of its predominance, VE was estimated against influenza A (H3N2). No VE analysis was carried out for influenza A (H1N1)pdm09 or influenza B due to the small number of cases. Table 1 presents the characteristics of influenza A (H3N2)-positive ILI patients (cases) and influenza-negative ILI patients (controls). All vaccinated individuals received the 2018-2019 inactivated quadrivalent influenza vaccine (QIV) [15].    In order to gain insight into VE trends at various ages, adjusted VE estimates were calculated according to moving 15 year age intervals, beginning at the first year of life and progressing by 1 year increments [8]. We then applied a cubic spline function with knots located at five discrete 15 year age intervals [8].
The adjusted moving influenza A (H3N2) VE estimates are shown in Figure 4A. The age intervals covering ages 0.5 to 47 years demonstrated adjusted VE with negative point estimates that were not statistically significant. The age intervals, starting at the 34-48 years interval and above, showed adjusted VE with positive point estimates. Specifically, the age intervals with the highest adjusted VE were 41-55 and 43-57 years, reaching VE point estimates of 81.5% and 84.3%, respectively. The age intervals of 42-56, 46-60, and 48-62 years old had adjusted VE point estimates between 73.4% and 75.7%. The adjusted VE of the age interval of 41-55 was statistically significant (81.5% (95% CI: 0.9% to 96.6%)), whereas other positive adjusted VE estimates were not statistically significant. The adjusted VE estimates in individuals aged 34 years and over were consistently positive, despite smaller sample sizes as compared with sample sizes of infants, children, and younger adults ( Figure 4A). Figure 4B shows the results of applying the cubic spline function to the VE point estimates data presented in Figure 4A.     [16][17][18][19][20]. In addition, although the 3C.3a clade constituted the minority of influenza A (H3N2) viruses in the beginning of the season in the USA, its proportion gradually increased throughout the 2018-2019 season [17], reaching 85% of influenza A (H3N2) viruses by the later part of the influenza season [21]. Similarly, the proportion of 3C.3a influenza A (H3N2) increased throughout the season in several Western European countries [1]. This phenomenon indicates that the circulation of influenza A (H3N2) clades can be dynamic even within the same season, and that the circulation of clade 3C.3a in Israel during the 2018-2019 season preceded that of other northern hemisphere countries.

The influenza A (H3N2) was the dominant influenza virus circulating in
The low all-ages influenza A (H3N2) VE estimate observed in Israel, was consistent with the genetic and antigenic analysis of these viruses. In addition, it was consistent with the all-ages influenza A (H3N2) clade 3C.3a VE estimates in the USA [21], Canada [20], and Europe [22].
Despite the low all-ages influenza A (H3N2) VE estimate in Israel, VE point estimates were higher in adults as compared with younger individuals (Table 2). Specifically, although the influenza A (H3N2) 2018-2019 vaccine was not effective in the 0-4, 5-17, and the 18-44 year old age groups, it demonstrated moderate adjusted VE point estimates in the older age groups ( Table 2). Estimation of VE for predetermined age groups is valuable; however, given the distinct character of such groups, it may not provide enough information regarding the trend of VE at various ages. Using the 15 year moving interval allowed us to evaluate VE at various ages despite the fact that our sample size may be smaller than that of larger countries [8,22,23]. This approach demonstrated the transition from negative to positive influenza A (H3N2) VE point estimates with age ( Figure 4A,B).
The difference in VE estimates between the younger and older age groups in our study is interesting. The low VE among infants, children, and young adults can be explained by the vaccine mismatch.
However, the higher VE point estimates among older adults may be explained by past exposure to similar influenza A (H3N2) viruses, which may have allowed the mismatched 2018-2019 vaccine to provide them with an immunological boosting effect. Furthermore, it has been proposed that influenza viruses that circulate during a person's childhood can confer lasting protection against new influenza viruses that belong to a similar phylogenetic group [24]. In this regard, ferrets that were vaccinated with a split-virion influenza vaccine following influenza A (H1N1) A/USSR/90/1977 infection earlier in their life showed less illness after infection with influenza A (H1N1)pdm09 A/California/07/2009 as compared with vaccinated ferrets who were not previously infected with influenza A (H1N1) A/USSR/90/1977 [25]. Moreover, it was recently shown that recurring exposures to influenza A (H3) in humans was associated with higher antibody titers, enhanced antibody affinity, as well as enhanced antibody avidity following influenza vaccination, as compared with individuals not previously exposed [26]. Additional research may be required to further explain the differences in VE point estimates between the younger and older age groups found in the present study.
Although year age group between the two seasons, may be explained by differences in prior exposure to similar viruses/antigens. However, future research may be required to further explain the differences in the adjusted VE results between these two seasons for the ≥65 year age group.
Recent studies by Skowronski et al. [20] and Kissling et al. [22] demonstrated statistically significant negative 2018-2019 influenza A (H3N2) VE estimates among the age ranges of 39-53 years old [20] and 32-54 years old [22], respectively. Skowronski et al. hypothesized that immune imprinting that occurred during childhood towards a common epitope of the hemagglutinin (HA) protected adults that did not receive the 2018-2019 influenza vaccine against the circulating 3C.3a influenza A (H3N2) viruses, whereas the immune response elicited by the 2018-2019 mismatched influenza A (H3N2) vaccine antigen interfered with this protection [20]. However, it is important to recognize that in addition to the imprinting that results from first infection, additional exposures can further shape the immune response towards influenza viruses [27].
In the present study, no statistical significance was noted for any of the age intervals showing negative VE point estimates. The negative VE point estimates observed here among the younger individuals (which were not statistically significant), were similar to smaller declines in VE estimates of younger birth cohorts observed in the study of Skowronski et al. [20]. Thus, the negative VE point estimates observed among the younger individuals in the present study can more likely be ascribed to the mismatch with the 2018-2019 influenza vaccine rather than to prior exposure to influenza viruses.
Children are thought to have the highest rates of illness and complications resulting from influenza [28,29]. In addition, children have been considered to play a key role in the spread of infections [30,31]. Vaccinating children and young adults against influenza was proposed in order to confer indirect protection to the population at large [32]. However, the low VE estimate among infants, children, and adolescents shown in our 2018-2019 analysis may pose a substantial challenge to the prevention of influenza in the community.
This study has several advantages. The predominance of the 3C.3a clade in Israel throughout the 2018-2019 season resulted in a sample size that enabled the analysis of VE trends by age against this clade. Additionally, vaccination status and vaccination dates of ILI patients included in our study were based entirely on medical records' data. It is also important to realize that Israel is the only country in the Middle East that performs yearly assessments of influenza vaccine effectiveness. Thus, these VE estimates may be relevant also to neighboring countries in the region. In this regard, influenza A (H3N2) circulated during the 2018-2019 season in Lebanon, Jordan, and the Palestinian authority, all of which share borders with Israel [33]. This study was limited by the unavailability of data regarding influenza vaccination in prior seasons. Although studies examining the effect of prior influenza vaccination on current season VE have shown variable results [34][35][36][37][38][39][40][41], most evidence to date suggests that influenza A (H3N2) VE estimates are highest when vaccination is administered during the season that is being evaluated [20,38].
Confidence intervals are used to describe statistical significance of VE estimates. In general, caution should be exercised when VE estimates are not statistically significant, as in such cases, the possibility of 'no protective effect' (zero VE) cannot be ruled out [42]. However, the use of statistical significance is being challenged by scientists around the world [43,44]. A discussion regarding this issue is necessary in order to determine how to interpret confidence intervals of VE estimates going forward.
The patterns of genetic evolution and divergence of influenza A (H3N2), as well as their impact on influenza VE, are likely to continue to challenge the selection of a suitable influenza A (H3N2) vaccine component. In the absence of a universal influenza vaccine, we support the concurrent inclusion of influenza A (H3N2) viruses from more than one clade in future influenza vaccines, as was recently suggested by others [45].

Conclusions
The genetic changes of the influenza A (H3N2) virus will likely continue to present a challenge for the selection of an optimal influenza A (H3N2) vaccine component. New solutions should be considered to overcome this challenge. Yellow cells represent AA substitutions that are common to all viruses; purple cells represent AA substitutions that were found in the majority of viruses; green cells AA substitutions that were found in the minority of viruses. Dots denote AAs that match the influenza A(H3N2) vaccine virus sequence.