Inactivated influenza vaccine effectiveness among department of defense beneficiaries aged 6 months-17 years, 2016–2017 through 2019–2020 influenza seasons

A test-negative case-control study was conducted to assess inactivated influenza vaccine effectiveness (VE) in children aged 6 months-17 years. The database was developed from the US Department of Defense Global Respiratory Pathogen Surveillance Program over four consecutive influenza seasons from 2016 to 2020. A total of 9,385 children including 4,063 medically attended, laboratory-confirmed influenza-positive cases were identified for VE analysis. A generalized linear mixed model with logit link and binomial distribution was used to estimate the VE. The adjusted VE for children was 42% [95% confidence interval (CI): 37–47%] overall, including 55% (95% CI: 47–61%) for influenza A(H1N1)pdm09, 37% (95% CI: 28–45%) for influenza A(H3N2), and 49% (95% CI: 41–55%) for influenza B. The analysis by age groups indicated that the adjusted VE in children aged 6 months-4 years was higher against influenza A(H1N1)pdm09 and influenza B, and comparable against influenza A(H3N2), compared to those in children aged 5–17 years. Further age-stratified analysis showed that the VE against any types of influenza was low and non-significant for children aged 6–11 months (33%; 95% CI:-2-56%), but it was high (54%; 95% CI: 34–67%) in children aged 12–23 months, and then declined linearly with increasing age. In conclusion, the inactivated influenza vaccination was moderately effective against influenza infection, based on the analysis from a large number of children aged 6 months-17 years over multiple influenza seasons.


Introduction
Seasonal influenza primarily causes human respiratory disease in all ages. Children, particularly those younger than 5 years of age, are at high risk of influenza infection and severe influenza-related complications including hospitalization and death [1,2]. The disease burden due to influenza infection is substantial. It was estimated that over six influenza seasons (2010-2011 through 2015-2016) in the United States, the total numbers of influenza-related symptomatic illnesses and hospitalizations among children have ranged from 2.8 to 10.7 million, and 11,000 to 45,000, respectively [3]. Moreover, influenza imposes a considerable economic burden to the healthcare system and society. Putri et al. [4] estimated the annual economic burden of influenza among children in the United States was $2.3 billion, including direct medical cost of $0.7 billion (30.8%) and indirect costs of $1.6 billion (69.2%). Influenza vaccination is considered the most effective way to protect against influenza viruses, combat influenza infection and lessen disease severity. The American Academy of Pediatrics recommends routine influenza immunization of all children without medical contraindications, starting at 6 months of age [5]. Every year, the Department of Defense (DoD) Global Respiratory Pathogen Surveillance (DoDGRS) Program performs routine respiratory pathogen surveillance among DoD service members and their beneficiaries, allowing annual estimates of influenza vaccine effectiveness (VE). Annual estimates of VE are necessary as the circulating influenza viruses differ from year to year. Nevertheless, in these previous DoDGRS studies [6][7][8][9], the VE has been assessed using data acquired in a single influenza season, in which all patients were categorized as children (6 months-17 years of age), adults (18-64 years of age), and the elderly(�65 years of age). Sample sizes might be too small to be able to accurately estimate VE within age categories, such as children aged <2 years. However, combining available data over multiple influenza seasons allows to provide pooled VE estimate by influenza type/subtype in age-specific category. The objective of our study was to use DoDGRS data to evaluate the effectiveness of inactivated influenza vaccine (IIV) against medically attended, laboratory-confirmed influenza in children 6 months to 17 years of age among DoD beneficiaries in the United States seeking outpatient care over four influenza seasons (2016-2017 through 2019-2020).

Study population
We used DoDGRS data over four influenza seasons from 2016-2017 to 2019-2020. In DoDGRS, all patients seeking outpatient medical care for influenza-like illness (ILI) clinical condition were selected. ILI was defined if the patient has (1) a fever greater than or equal to 38˚C and a cough or sore throat which presents within 72 hours after illness onset, or (2) physician determined ILI. Respiratory specimens were collected by nasopharyngeal wash or nasopharyngeal swab from ILI patients, and subject to testing via a multiplex respiratory pathogen panel, reverse transcription polymerase chain reaction (RT-PCR) and/or viral culture. Thus, the influenza viruses and other respiratory pathogens were identified and confirmed. Influenza virus type and subtype/lineage were identified when influenza virus tested positive. Patients who had received at least one influenza vaccine dose 14 days or more before the onset of an ILI, were considered vaccinated. Patients who otherwise had not received vaccination before the onset of an ILI, were considered unvaccinated. Vaccination status was verified through the records from the DoD Electronic Immunization Tracking System or self-reported questionnaire for each patient. Patients with an unknown vaccination status or type, or vaccinated <14 days prior to the onset of an ILI were excluded. Vaccine type in the present study was limited to IIV; thus, patients who received influenza vaccines other than IIV were excluded.
The patients in the present study were limited only to those aged 6 months-17 years from the DoDGRS program sentinel or participating sites in the United States.
subgroups (5-8 years, 9-12 years, and 13-17 years). The rationale for age classification was based on the fact that children younger than 5 years of age are at higher risk for influenza infection and severe influenza-related complications [1,2]. In addition, it was of interest to evaluate VE against influenza viruses in children under 2 years old with extra age stratification. Thus, children aged <2 years were further stratified into two age strata (6-11 months and 12-23 months).

Statistical analysis
Prior to data assembly over four influenza seasons, a range of surveillance weeks in each influenza season was limited to November to April of the following year for the VE analysis, when approximately 10% or greater influenza positivity rate occurred, with an aim to minimize any potential bias due to high ratio of influenza negative to influenza positive that would typically occur earlier or later in the influenza season.
All data was combined by influenza season and performed analysis using generalized linear mixed model (GLMM) with logit link and binomial distribution. Influenza season was treated as a random effect in the model. The odds of influenza vaccination among children with laboratory-confirmed influenza positive (cases) were compared to the odds of influenza vaccination among children who were tested influenza negative (controls). The VE was calculated as (1-adjusted odds ratio) × 100%. All potential confounding factors, such as age, gender, specimen collection date, or geographical region were initially evaluated. Only those factors that changed the crude odds ratio by � 5% were included in the GLMM models to adjust VE. The point estimate of VE was considered significant when lower limit of the associated 95% confidence interval (CI) did not contain zero or negative value. In addition to overall VE estimated against any influenza viruses in entire children population, we estimated VE by influenza virus (sub)type in separate models [i.e., influenza A(H1N1)pdm09, influenza A(H3N2), or influenza B], and in stratified models by age category.
The characteristics of patients by the status of influenza testing result were compared by odds ratio and Chi-square test. The simple linear regression analysis was conducted to explore the potential relationship of VE by age [the average of each age subgroup (i.e., 6-23 months, 2-4 years, 5-8 years, 9-12 years, and 13-17 years]. A p < 0.05 was considered statistically significant. All analyses were conducted using SAS Enterprise Guide 7.1(SAS Institute Inc., Cary, NC).

Ethics
The present study was determined to be "Public Health Practice" and "Not Human Use Research", and was exempted from review by the DoD Air Force Research Laboratory's Institutional Review Board. Regulations or rules were followed to protect patient's identity and health information.

Patient characteristics
The characteristics of the patients overall and for each season are shown in Table 1

Confounding factors assessment
Among all potential confounders examined, month of specimen collected and age were the only variables that changed the crude odds ratio by � 5%. Therefore, month of specimen collected (i.e., one-month period from November to April of the following year) and the age (i.e., 6-23 months, 2-4 years, 5-8 years, 9-12 years, and 13-17 years) were included in the models to estimate overall VE and the VE stratified by influenza type/subtypes or age (sub)groups. Additionally, the age strata (i.e., 6-11 months and 1-2 years) was included in the model to estimate the VE for children aged 6-23 months.

Discussion
We report the pooled estimates of VE against medically attended, laboratory-confirmed influenza among children aged 6 months-17 years over four influenza seasons (2016-2017 through 2019-2020). Significant protection of IIV was found against any influenza (42%), while by influenza (sub)type, the protection was higher for influenza A(H1N1)pdm09 and influenza B, and lower for influenza A(H3N2) ( Table 2). Moreover, higher overall VE against any influenza viruses was found in children aged 6 month-4 years vs. 5-17 years. The age-stratified analysis indicated the overall VE estimates against any influenza and the VE estimates against A (H1N1)pdm09 declined linearly as age increased (Table 3 and Fig 1A and 1B).
Children aged <2 years are at high risk of developing serious influenza-related complications [25]. Generally, it is of public health interest to evaluate the effectiveness of influenza vaccination in this special pediatric population comprising infants and young children <2 years old of age. In the present study, there were >4000 identified cases of influenza over four consecutive influenza seasons. Our large sample size provided an opportunity to further compare age-specific VE by influenza (sub)type. It was shown that the adjusted overall VE (46%) in children aged 6-23 months was low but comparable, in comparison to the VE (ranging from 51 to 86%) against any laboratory-confirmed influenza reported previously in published studies [15,16,26]. Our findings demonstrated inactivated influenza vaccine was moderately effective against any influenza in children aged 6-23 months [27].
In the separate VE analysis for children aged 6-23 months, we found there were great differences of the VE estimates in children aged between 6-11 months and 1-2 years, with the exception of similar low and not-significant VE against influenza A(H3N2). For children aged 6-11 months, both overall VE and the VE against influenza A(H3N2) and influenza B were low and non-significant; moreover, 95% CI (6-76%) of the VE against influenza A(H1N1) pdm09 was wide, with lower limit for the mean being low close to zero. By comparison, higher overall VE and the VE against influenza A(H1N1)pdm09 and influenza B were observed in age 12-23 months. Consistently, several studies in Japan [26,[28][29][30], which performed VE analysis each season individually throughout the four seasons from 2013-2014 to 2016-2017, or via a combination of three-season data from 2013-2014 to 2015-2016, have shown low and nonsignificant effectiveness of influenza vaccination for children aged 6-11 months. Influenza vaccination has been recommended for infants starting 6 months old in the US [5]. Nevertheless, our findings support the view that influenza vaccination should not be strongly recommended for children aged 6-11 months [29,30]. Further research efforts should be made to elicit immune response in infants 6-11 months of age to influenza vaccination, thus providing more robust evidence to substantiate those findings.
The overall VE against any influenza declined linearly with increasing age (Fig 1A). The pattern of decreasing VE as age increases may suggest a role of immune history in shaping the vaccine response. There are evidences suggesting that repeated vaccination impacts vaccine protection against influenza virus, which might consequently alter VE estimates [31,32]. The underlying immunologic mechanisms for the potential vaccine interference are not well understood [33]. Shinjoh et al. [30] attributed possibly a low VE in children aged 13-15 years to their strong immunity, due to the past history of natural influenza infection and influenza vaccination. Indeed, children aged <2 years often receive influenza vaccination for the first time in their lives, while as they grow older, there are more chances for them to experience influenza infection or receive repeated influenza vaccination.
The DoDGRS program collects, compiles and analyzes outpatient influenza information from DoD beneficiary year-round to monitor global influenza activity. Influenza VE have been evaluated annually via the test-negative case-control design [6][7][8][9]. In those individual analysis, sample size might be too small to permit accurate VE estimates for specific age category of interest. All DoDGRS data were collected using same or similar methods during the 2016-2017 through 2019-2020 influenza seasons, which allowed pooling for estimation of VE. Moreover, as shown above, pooling data increased the available sample size, thus allowed for an age-stratified evaluation of VE by influenza (sub)type. Nevertheless, while pooling data from multiple influenza seasons increases statistical power for analysis, it may mask important variation at the individual season level [33]. Indeed, there are season to season variations in circulating viruses, vaccine match, and host-virus immunological interactions [32]. In the present study, we conducted the analysis to estimate pooled VE using generalized mixed models, with influenza season being considered a random effect to account for the heterogeneity across the influenza seasons. Using a random effect approach does statistically adjust for the heterogeneity; however, there may still be residual heterogeneity across seasons [34]. Although there are statistical methodological challenges arising from pooling data, pooling individual data over multiple seasons permits effectiveness analysis of IIV among children, specifically for targeted age category, thus exploring the potential relationship between the estimated VE with those virus and host factors influencing VE against influenza viruses.
The VE estimates in the present study have been adjusted, using month of specimen collected and age, identified from potential confounders such as covariates being used, gender, and geographical region, etc. as covariate in generalized linear mixed models. However, due to lack of randomization that is inherent in any observational study, it is difficult to rule out unmeasured confounding factors as a possible alternative explanation for the findings. For instance, we were unable to obtain all information on past history of vaccine status and influenza infection in previous seasons. Thus, the models used in the present study did not account for prior exposure to influenza by repeated influenza vaccination or natural influenza infection in previous seasons. Therefore, it would be valuable to collect such patients' data in future endeavors. In addition, our efforts to estimate the effect of vaccination rely on the DoD surveillance platform for data acquisition. There are limitations worth mentioning with regard to the DoD surveillance data collection. When ascertaining vaccine status based on the self-reported questionnaire, potential non-differential misclassification could occur. Moreover, the specimens collected from the sentinel and participating sites in DoDGRS were through routine outpatient clinical care. Before evaluating the validity of using such routinely collected data in VE analysis, possibility of potential selection bias could not be completely ruled out.
In conclusion, the present study demonstrated moderate effectiveness of inactivated influenza vaccination in children against medically attended, laboratory-confirmed influenza infections. We found the VE was higher against influenza A(H1N1)pdm09 and influenza B, but comparable against influenza A(H3N2) in children aged 6 months-4 years, when compared to those in children aged 5-17 years. Moreover, we observed low and non-significant VE against any influenza in children aged 6-11 months, which was consistent with the finding from the previous observational studies in Japan [26,[28][29][30]. Further research that includes assessment of host immune response to influenza vaccine in children 6-11 months of age is warranted, which would undoubtedly provide valuable information helping guide future vaccine policy recommendations.