Effect of seasonal influenza vaccination on influenza symptom severity among children in Hutterite communities: Follow‐up study of a randomized trial

Abstract Background We investigated whether influenza vaccination reduces symptom severity among children who develop laboratory‐confirmed influenza, and whether this association differed between influenza vaccine formulations. Methods We performed a retrospective cohort study using data from two blinded cluster randomized control trials of influenza vaccines in Hutterite colonies. In trial 1, children received trivalent inactivated influenza vaccine (TIV) or hepatitis A vaccine. In trial 2, children received trivalent live attenuated (TLAIV) or TIV. We assessed four outcomes (total number of symptoms, number of respiratory symptoms, number of systemic symptoms, and duration of symptoms) among children with PCR‐confirmed influenza. We utilized two‐sample t tests to quantify the relationship between vaccine group and outcome. We performed multivariable strain‐specific analyses, controlling for age and season. Results TIV vs. Hep A vaccine: Among vaccinated children, 200 confirmed influenza infections were observed across 3014 person‐seasons. Vaccine type (TIV vs. Hep A vaccine) did not significantly affect the number of respiratory or systemic symptoms, nor duration of symptoms (P > .05). TLAIV vs. TIV: Among 1186 children who received a study vaccine, 166 confirmed influenza infections were observed. TLAIV recipients experienced fewer total, respiratory, and systemic symptoms compared to TIV recipients (P < .05 for all). TLAIV‐associated attenuation of symptom severity was observed in influenza B or A/H1N1 infections, but not H3. Conclusions Seasonal influenza vaccine did not consistently attenuate symptom severity in the context of vaccine failure; however, TLAIV offered superior severity attenuation compared to TIV. Our results challenge the dictum that influenza vaccine reduces the severity of symptoms even when the vaccine fails to prevent influenza.


| INTRODUC TI ON
Seasonal influenza is a viral, acute respiratory illness that causes an estimated three to five million symptomatic cases and 290 000 to 645 000 deaths per year, globally. 1 Influenza usually manifests as a sudden onset of fever, cough, headache, malaise, rhinorrhea, and pharyngitis. 2 Illness lasts approximately one week and is typically self-limiting without the need for medical attention. Despite this, high-risk populations, such as young children, elderly individuals, people with chronic co-morbidities, and immunocompromised individuals, can experience severe influenza-related morbidity and mortality. 3 Annual vaccination confers protection against seasonal influenza infection. 4 When an individual is vaccinated against seasonal influenza, but develops a seasonal influenza infection, this is defined as a vaccine failure. 5,6 It follows that vaccine failures occur at a higher rate when vaccine effectiveness is low, or when the host cannot launch a protective antibody response. 6,7 Influenza vaccine effectiveness is variable and is often dependent on host (eg, age, health status) and epidemiological factors (eg, vaccine coverage, match to circulating strains). 8 The type of influenza vaccine administered may also affect vaccine efficacy. Live attenuated influenza vaccines (TLAIV) were shown in early studies to be more efficacious than inactivated influenza vaccines (TIV) in children 9,10 ; however, more recent studies have not supported this finding. 11 The US Centers for Disease Control and Prevention states that those vaccinated who develop a seasonal influenza infection may experience a milder course of disease as a result of vaccination. 12 This statement is supported by evidence from studies that show reductions in upper respiratory symptoms, overall symptom severity, 13,14 and hospitalization rates in vaccinated individuals where vaccine failure occurred. 15 Current evidence on vaccine-attenuated symptom severity in patients who present with influenza-like illness (ILI) has three main limitations. First, studies operationalize illness severity in different ways. Second, most studies have utilized testnegative design. In brief, test-negative studies typically identify a study population based on symptomatology and compare the characteristics (eg, severity of symptoms) of those who have a positive seasonal influenza test with those who do not; 16 however, there are limitations to test-negative studies. These studies can be vulnerable to substantial selection bias and confounding bias when estimating vaccine effectiveness. 17 Third, studies to date have focused largely on adult in-patient populations. 14,15 The relationship between seasonal influenza vaccination and symptom severity in the context of vaccine failure has been assessed using prospective data for children with influenza B infections, 6 but to date has not been validated more broadly across other influenza strains, or across influenza vaccine formulations.
Utilizing data from two cluster randomized trials of influenza vaccine in Canadian Hutterite communities, we sought to understand whether seasonal influenza vaccination attenuates the severity of symptoms in the context of vaccine failure. Furthermore, we investigated whether this association differed between those who received the live attenuated influenza virus vaccine (TLAIV) and those who received the inactivated influenza vaccine (TIV).

| Study design and data collection
We performed a retrospective cohort study using data from two blinded cluster randomized control trials conducted across Hutterite colonies in Alberta, Saskatchewan, and Manitoba between 2008 and 2015. Both studies assessed the impact of immunizing children against seasonal influenza on community-level infection rates.
Study 1 was a three-year cluster randomized control trial where children within colonies received either TIV or hepatitis A vaccine as control. Throughout the three-year trial, children aged 36 months to 15 years who received a study vaccine (hepatitis A or TIV), as well as other community members who did not receive study vaccine, were followed prospectively. Children randomized to a vaccination group were followed for a total of 3014 person-seasons (1554 person-seasons for TIV, 1460 person-seasons for hepatitis A vaccine), and non-vaccine recipients were followed for a total of 7971 personseasons across 65 colonies in Alberta, Manitoba, and Saskatchewan.
Individuals with high-risk conditions (eg, immunocompromise, pregnancy, diabetes) were also vaccinated with TIV at baseline regardless of age and followed prospectively. Study 2 was a three-year cluster randomized TIV-controlled study, where children within colonies received either TLAIV or TIV.
All participants were blinded to their colony's vaccine allocation. As above, the study followed children aged 36 months to 15 years who received study vaccine, community members who did not receive a study vaccine, and individuals with high-risk conditions that were vaccinated at baseline. Over three seasons, 1186 children who received a study vaccine and 3425 community members who did not receive study vaccine across 52 Hutterite colonies in Alberta and Saskatchewan were followed forward. The complete methodology and results of the Hutterite community randomized trials have been published previously. 11,18 In both study 1 and study 2, the beginning of the follow-up period for each influenza season was defined by the identification of >1 polymerase chain reaction (PCR)-confirmed cases in two consecutive weeks in a sentinel site. Follow-up for each influenza season concluded in a given health region after two consecutive weeks with zero laboratory-confirmed influenza cases reported.
Throughout the follow-up period, research nurses assessed vaccinated and non-vaccinated participants twice a week. Using a standardized survey, nurses collected data on self-reported signs and symptoms of ILI, including fever (≥38.0°C), cough, nasal congestion, sore throat, headache, sinus problems, muscle aches, fatigue, ear ache, or chills. The exact date of symptom onset was recorded.
Nasopharyngeal specimens were obtained from any participant reporting 2 or more symptoms and were sent for real-time reverse transcriptase polymerase chain reaction for viral genotyping.

| Inclusion criteria
Study 1 was conducted over three influenza seasons between 2008 and 2011. Of the study participants followed prospectively, our analyses included children aged 36 months to 15 years of age from the original randomized control trials who (a) had a PCR-confirmed seasonal influenza or pandemic (pH1N1) influenza infection during the study period and (b) received either the study influenza vaccine or a hepatitis A vaccine (control). We used November 1 as an age cut-off at 15 years. We included all PCR-confirmed influenza infections that were observed within the study period.
If more than one confirmed infection occurred in the same participant over the three influenza seasons studied, each infection was considered an independent observation in the cohort. We as-

| Outcomes
We identified two main outcomes to quantify the course of disease: number of symptoms and duration of symptoms. Given the potential lag between symptom onset and laboratory confirmation of influenza infection, we used symptom data collected by study nurses 7 days prior to the nasopharyngeal PCR sample collection date and assessed symptom data up to 14 days following the positive PCR result.
We counted the total number of individual symptoms reported during the 21-day window. We then divided symptoms into two categories: respiratory (consistent with either upper and/or lower respiratory tract infection) or systemic. Total symptoms, respiratory symptoms, and systemic symptoms have been shown to correlate with cytokine titers in nasal lavage samples. 19 In our analyses, respiratory symptoms included cough, sore throat, runny nose, sinus issues, and ear infection. Systemic symptoms included fever, headache, muscle aches, and fatigue. Symptomatic days where the participant reported at least one symptom did not have to be consecutive over the 21-day follow-up period and were summed to calculate duration of symptoms.

| Statistical analyses
For both study 1 and study 2, we compared demographic characteristics of children and adolescents aged 36 months to 15 years randomized to each vaccine group using Pearson chi-square testing for categorical variables, and Welch two-sample t tests for continuous variables. For both studies, we stratified each cohort by vaccine group and used Welch two-sample t tests, assuming unequal variances between groups, to assess the association between vaccine allocation and infection severity outcomes. Given the presence of repeat infections within individuals, we assessed the robustness of our results to the assumption that all infections are independent by performing a sensitivity analysis, using only the first infection for each participant.
To assess the influence of influenza stain type on the relationship between vaccination and infection severity, we performed multivariable strain-specific analyses for influenza B, H3, H1N1, and pH1N1, controlling for age. We included a categorical interaction term for influenza season to assess whether the relationship between influenza strain type and each infection severity outcome changed based on the antigenic match of the influenza vaccine to the circulating strains each year.  Table 2.

| Study 1: TIV vs Hepatitis a vaccine
We did not observe a significant effect of vaccine group on duration  (Table 3). For seasonal H1N1, all infections occurred in Season 1 among those who were randomized to the hepatitis A vaccine group; therefore, we were unable to assess the effect of vaccine group and season on our outcomes of interest.

| Study 2: TLAIV vs. TIV
Over three influenza seasons between November 2012 and May 2015, participating individuals were monitored for influenza    disease when compared to those randomized to TIV, resulting in fewer symptoms overall. Furthermore, compared to children rand- of TIV was associated with a higher risk of pH1N1 illness and a higher risk of requiring medical attention, compared to those with pH1N1 who did not receive TIV. 23 The biological mechanism of this risk is not completely understood. 24  categorized as "mild." 25,26 In contrast to the aforementioned approaches, we did not measure self-reported symptom severity.

| D ISCUSS I ON
Instead, we used objective binary measures to quantify symptom presence/absence among participants, and defined severity as the summed number of symptoms, or symptom duration. It is possible that the presence of multiple symptoms may not be perceived as increased infection severity. Further analyses are required to determine whether our severity measure correlates with patients' self-reported perception of infection severity. A second limitation of this analysis is that the randomized population was limited to healthy children. Further investigation is required to prospectively validate the association between influenza vaccine type and subsequent infection severity among other patient populations, such as adults and high-risk children.
One key strength of this analysis is the use of data from a blinded randomized control trial. These data include detailed demographic information, laboratory data, and daily symptom data collected prospectively for all study participants. By extracting data from a trial that required blinding and equivalent follow-up in both intervention groups, our analysis was less vulnerable to bias, compared to studies that use a test-negative approach to assess the severity of symptomatic influenza. 16 In test-negative studies, patients typically present to an emergency department with ILI. Patients then receive a laboratory influenza test to confirm diagnosis and are then questioned about their seasonal influenza vaccination status. Bias may arise in these studies due to unmeasured confounders associated with both exposure and outcome, or due to the effect of vaccination on healthcare-seeking behavior. 17 Our use of RCT data in our analysis minimized the risk of both selection bias and bias due to confounding factors, allowing us to objectively estimate influenza severity in the context of vaccine failure.
In conclusion, our study shows that in a community setting, healthy children vaccinated against influenza do not experience significant vaccine-associated symptom attenuation, compared to those who received a control vaccine. When comparing influenza vaccine types, TLAIV does appear to provide superior vaccine-associated attenuation of symptom severity than TIV for influenza A/ H1N1 and influenza B. The relationship between vaccination and symptom severity in those who develop seasonal influenza appears to vary by demographics, clinical context, and flu strain. The widely held dictum that people who receive the influenza vaccine experience less severe symptoms does not appear to be generalizable to all patient populations and may vary from season to season based on vaccine match. Therefore, clinical setting and patient-specific factors must be considered when evaluating the potential protective effect of the influenza vaccination in the context of vaccine failure.