Effectiveness of influenza vaccine in preventing medically-attended influenza virus infection in primary care, Israel, influenza seasons 2014/15 and 2015/16

Introduction Influenza vaccine is recommended for the entire population in Israel. We assessed influenza vaccine effectiveness (VE) for the 2014/15 and 2015/16 seasons in Israel, for the first time. Methods: Combined nose and throat swab specimens were collected from patients with influenza-like illness (ILI) presenting to sentinel primary care clinics and tested for influenza virus by RT-PCR. VE of the trivalent inactivated vaccine (TIV) was assessed using test-negative case–control design. Results: During the 2014/15 season 1,142 samples were collected; 327 (28.6%) were positive for influenza, 83.8% A(H3N2), 5.8% A(H1N1)pdm09, 9.2% B and 1.2% A un-subtyped. Adjusted VE against all influenza viruses for this influenza season was −4.8% (95% confidence interval (CI): −54.8 to 29.0) and against influenza A(H3N2), it was −15.8% (95% CI: −72.8 to 22.4). For the 2015/16 season, 1,919 samples were collected; 853 (44.4%) were positive for influenza, 43.5% A(H1N1)pdm09, 57% B, 0.7% A(H3N2) and 11 samples positive for both A(H1N1)pdm09 and B. Adjusted VE against all influenza viruses for this influenza season was 8.8% (95% CI: −25.1 to 33.5), against influenza A(H1N1)pdm09, it was 32.3% (95% CI: (−4.3 to 56.1) and against influenza B, it was −2.2% (95% CI: (−47.0 to 29.0). Conclusions: Using samples from patients with ILI visiting sentinel clinics in Israel, we demonstrated the feasibility of influenza VE estimation in Israel.


Introduction
Influenza virus infection causes morbidity and mortality worldwide every year [1]. The most effective measure of preventing influenza is the influenza vaccine.
Israel is a country of 8.5 million people, with a Mediterranean and arid climate, located in the westernmost part of Asia. In Israel, influenza activity is seasonal, usually occurring from December through March [4] in patterns similar to those of Europe and those in neighbouring Jordan [5] and Egypt [6]. The outpatient sentinel influenza surveillance system was established in the 1996/97 season and has been conducted through primary care clinics throughout Israel. The sentinel clinics are located in all seven districts of Israel and are staffed by paediatricians, internists and family physicians. During the 2014/15 and 2015/16 influenza seasons, 23 and 26 sentinel clinics, respectively, participated in influenza surveillance. Israel's Ministry of Health recommends influenza vaccination for the entire population over the age of 6 months [7]. The trivalent inactivated vaccine (TIV), the quadrivalent inactivated vaccine (QIV) and live attenuated influenza vaccine (LAIV) are all registered for use in Israel. The inactivated vaccines against seasonal influenza are offered free of charge to all residents through clinics of the four national 'sick fund' organisations which are widely spread throughout the country and are similar to health maintenance organisations in the United States (US). TIV is the most widely used influenza vaccine in Israel.
In this study we evaluated, for the first time, VE against medically attended laboratory-confirmed influenza-like illness (ILI) in the community in Israel using our sentinel surveillance system. The predominant influenza strain in the 2014/15 season was a drifted influenza A(H3N2). Influenza A(H1N1)pdm2009 and influenza B co-circulated in Israel during the 2015/16 season. The overall influenza vaccination coverage in Israel during the 2014/15 and 2015/16 seasons was ca 21%. The vaccine coverage was ca 65%, 25% and 40% among individuals age 65 years of age and over, infants and children 6 months to 5 years, and individuals less than 65 years of age with chronic medical conditions, respectively [8,9].
We used the test-negative case-control design, a commonly used method for estimating influenza VE in studies that utilise sentinel surveillance systems [10].

Study period and population
Influenza surveillance periods lasted in Israel from 28 September 2014 until 18 April 2015, and from 27 September 2015 until 16 April 2016. During these periods, sentinel clinic providers obtained combined nasal and throat swabs [11] from a convenience sample of patients meeting the ILI case definition. The kits of combined nasal and throat swabs were provided to all sentinel clinics by the Israel Center for Disease Control. ILI was defined as a temperature of 37.8˚C and over, accompanied by one or more of the following symptoms: coryza, sore throat, cough, and muscle ache [12]. Discretion was given to physicians to include other signs or symptoms considered relevant. Sentinel clinics were asked to send up to ten samples per week. Week and year Week and year

Molecular identification of influenza viruses in samples obtained from sentinel patients with influenza-like illness
The combined nasal and throat samples from sentinel ILI patients were kept at 4˚C in the upright position until transport. Samples were transported once a week

Study design
VE against influenza was assessed for individuals 6 months of age and over who received the TIV, using the test-negative case-control design [18,19]. VE was derived as (1−odds ratio (OR)) × 100, expressed as a n (%) a n (%) a n (%) a n (%) a n (%) a n (%) a n (%) b percentage. VE was estimated for influenza A and B together and for the specific influenza subtypes, for each of the two seasons. VE was not calculated for certain influenza types and subtypes if the total number of positive samples for the type or subtype was very low, not allowing at least 5 samples in each cell of the contingency table used for OR calculation. Individuals were considered vaccinated if they received the influenza vaccine 14 days or more before disease onset. For children between the ages of 6 months and 9 years receiving the influenza vaccine for the first time, only those receiving two doses, with the second dose given 14 days or more before illness, were included in the analysis. Individuals whose samples were taken more than 7 days after of onset of symptoms were excluded from analysis. Because relatively few individuals received LAIV or QIV during the two influenza seasons, we estimated VE for TIV only.

Statistical analysis
Percentages were compared using the Mantel-Haenszel chi-squared test. OR for crude VE calculation was performed using a univariate logistic regression model with no covariates. We adjusted for age group, sex, calendar week of sample collection, days from disease onset to swab and underlying chronic medical conditions (information regarding chronic conditions was available for the 2015/16 season only) using multivariable logistic regression. Sensitivity analysis was carried out to evaluate whether there was a difference between individuals who were swabbed on days 0-1 from symptom onset and individuals who were swabbed on days 2-7 from symptom onset. Statistical analyses were carried out using SAS version 9.4 (SAS Institute, Cary, NC, US).

Ethical consideration
Sentinel influenza surveillance in Israel, including the testing component, is conducted in accordance with the Public Health Ordinance enacted in Israel and does not require informed consent. Molecular characterisation of influenza strains isolated from patients was approved by the ethics committee at the Sheba Medical Center (1967-15-SMC), Tel Hashomer, Israel.

Study population 2014/15 influenza season
Of the 1,142 samples collected from patients with ILI, 127 samples were excluded due to missing vaccination status or vaccination dates, missing day of symptom onset, sampling more than 7 days after onset of symptoms, receipt of influenza vaccine less than 14 days before disease onset, or because of partial vaccination (children less than 9 years old require two vaccine doses) ( Figure 2). Of the 1,015 remaining ILI patients (316 cases and 699 controls), 10 received the LAIV ( Figure 2); all 10 were less than 13.5 years of age and excluded from the study. The remaining vaccinated patients received the trivalent injected egg-grown vaccine (split or inactivated) containing an A/California/7/2009(H1N1) pdm09-like virus, an A/Texas/50/2012(H3N2)-like virus and a B/Massachusetts/2/2012-like virus. The characteristics of the 1,005 samples that were included in the assessment are presented in Table 1.

2015/16 influenza season
Of the 1,919 samples collected from ILI patients, a total of 249 samples were excluded due to missing vaccination status or vaccination dates, missing day of symptoms onset, sampling more than 7 days after onset of symptoms, receipt of influenza vaccine less than 14 days before disease onset, partial vaccination (children less than 9 years old require two vaccine doses) and missing vaccine type (Figure 2). Of the 1,670 remaining ILI patients (788 cases and 882 controls), 12 received

Sensitivity analysis
Sensitivity analysis demonstrated substantial differences in VE estimates between individuals who were swabbed on days 0-1 days vs 2-7 days from disease onset in the 2014/15 and the 2015/16 seasons (

Discussion
This was the first study in Israel to evaluate influenza VE in the community using the test-negative case-control design. In this study we evaluated two seasons with different circulation of influenza viruses. We found that in the 2014/15 season, which was characterised by the predominance of a drifted strain of influenza A(H3N2), the TIV was not effective. These findings are consistent with reported VE against community influenza A(H3N2) from the UK [18,23], Canada [24], the US [25][26][27], Navarra, Spain [28] and Austria [29] in the same season.
These results are also consistent with the findings of an Israeli study demonstrating that sera from Israeli individuals vaccinated with the 2014/15 injected split virus vaccine had reduced ability to neutralise the drifted influenza A(H3N2) virus [21]. Although the 2014/15 influenza vaccine was not effective against outpatient influenza A(H3N2) in our study, several studies demonstrated a better VE against preventing influenza A(H3N2)-associated hospitalisations [30,31], reaching 43% in one study [30].   [35]. Although, the VE of 76.5% (95% CI: 41.9 to 90.5) for those 2 to 17 years of age in the UK can be attributed, at least in part, to the use of LAIV, the VE results of individuals between the ages of 18 and 64 in the UK who mostly received the TIV [35], suggest a different mechanism of protection.
The difference in VE results against influenza A(H1N1) pdm09 between the two age groups examined in our study is interesting. The vaccine was moderately effective in adults, a finding consistent with studies from the UK and US [32,35]. However, the low VE against influenza A(H1N1)pdm09 in those less than 18 years of age in Israel differed from VE estimates in the same age group in those same countries [32,35]. Of particular note is the difference from the US study, which showed good VE among children receiving the injectable form of the influenza vaccine (while low VE estimates were found among children receiving the LAIV) [32]. It is worth noting the different vaccination rates among the adult and child controls in our study in 2015/16. While the vaccination rate among adult controls was 20.6%, it was 9.7% among control children (Table 4). By comparison, in the 2014/15 season the vaccination rates were similar in adult and children controls (18.6% and 16.5%, respectively) ( Table 3). The low rate of vaccinated controls in those less than 18 years of age may have resulted in reduced precision in VE determination. Our study has several limitations. Our sample size was initially relatively small. However, we were able to add additional sentinel clinics for the 2015/16 season and increase the sample size by ca 40%. In addition, in both seasons we had a small number of participants in certain age groups. For this reason, we estimated VE for age strata of 18 years of age and over and less than 18 years of age. Further increasing our sample size in the future will allow stratification into additional age groups, particularly for seasons in which several influenza types and subtypes co-circulate. For the 2014/15 season we did not collect data regarding chronic medical conditions. However, this information was obtained during the 2015/16 season.
We used convenience sampling to select patients presenting with ILI, similar to other studies in the field [3,18], which may have biased our sample. However, our sentinel clinics represented all seven districts in the country, and therefore, the sample likely represented larger Israeli society with respect to geography and population groups.
To evaluate whether our VE estimates might be influenced by the time elapsed from disease onset to swabbing, we conducted a sensitivity analysis. Our analysis demonstrated substantial differences in VE between patients with samples obtained 0-1 days after disease onset compared with patients with samples collected 2-7 days after disease onset. To account for these differences, we adjusted for days from disease onset to swab in our VE estimation.
We did not have information regarding influenza vaccination in previous years among ILI patients. Thus, we were unable to measure VE considering previous vaccination status. Influenza vaccination in preceding years may affect vaccine effectiveness results. Previous influenza vaccination has been associated with a negative effect on VE in several studies [24,28,39].
In conclusion, using data from a community-based influenza surveillance system in Israel, we estimated VE of inactivated influenza vaccine in the Israeli population for the first time. Our study suggests that VE may vary by season, by influenza type and subtype as well as by age. Our results, and those of others, support the need for continued efforts of estimating influenza VE, both in outpatient and inpatient settings. These efforts are necessary in order to better understand the factors that affect influenza VE and to optimise vaccine composition and use. Timely VE estimates are of paramount importance for the yearly decision regarding influenza vaccine composition.