School‐based surveillance for influenza vaccine effectiveness during 2014‐2015 seasons in Hong Kong

Background Influenza imposes substantial healthcare burden in children, which can be prevented by vaccination. Influenza vaccination coverage varies widely among childhood populations worldwide, which has significant impact on herd immunity and usefulness of influenza vaccine. However, there are limited real‐life data on influenza vaccine effectiveness (VE) in children. Objective This prospective study aimed to investigate clinical spectrum of childhood influenza and VE in preventing influenza in Hong Kong children. Methods A total of 623 children were recruited from 15 kindergartens and primary schools. Parents completed a questionnaire on subjects’ health and influenza vaccination history. Flocked nasopharyngeal swabs (FNPSs) were collected in biweekly school visits during 2014‐2015 influenza seasons. Influenza A and B viruses were detected and typed by molecular assays. Results A total of 2633 FNPS samples were collected, with two or more samples being obtained from 607 (97.4%) of subjects. Thirty‐six (11.2%) subjects had influenza A or B in 2014, whereas all 19 (6.3%) subjects identified in 2015 had influenza A. Ninety‐nine subjects reported influenza‐like illness (ILI), and nine illness visits were arranged. Influenza vaccination was protective against ILI but not mild laboratory‐confirmed influenza by surveillance. Moderate overall influenza VE of 42%‐52% was observed for ILI, and subgroup analyses showed much higher VE for both ILI (70.9% vs 34.6%) and mild laboratory‐confirmed influenza (44.0% vs −6.2%) in school‐age children than preschoolers who were vaccinated within 12 months. Conclusions Mild laboratory‐confirmed influenza infection is common in children during influenza seasons. Influenza vaccination is effective against ILI but not mild infection identified by surveillance.

the study start, we contacted principals of selected kindergartens and primary schools to obtain their permission to join this study. Parents of eligible students in these schools then gave informed written consent for their children to participate in this surveillance. This study did not set any exclusion criterion for subjects in an attempt to recruit a representative study population. Based on local data 3

| Subject assessment
Parents completed a questionnaire that recorded subjects' demographics, pre-existing medical illnesses, and influenza vaccination history within 3 years. Our staff verified vaccination history against immunization cards or with responsible doctors. Subjects with the following criteria were considered vaccinated: (i) ≥14 days postvaccination; (ii) received two doses 28 days apart if vaccinated for the first time; or (iii) received at least one dose in a previous influenza season and one dose in the season under study. 5  Our staff phoned families every 2 weeks to remind subjects about the next surveillance visits and enquire whether they had ILI defined based on modified World Health Organization (WHO) case definition (ie, fever ≥38°C plus two of the followings: cough, sore throat, rhinorrhea, myalgia, headache) 10 and whether they recently received influenza vaccination. Parents were provided our contact phone number and encouraged to inform us as soon as their children developed ILI.
Our team attempted to arrange illness visit for all subjects with ILI during influenza seasons in early 2014 and 2015, and those agreed to attend illness visit returned to our outpatient clinic within 48 hours of ILI onset. During these visits, our nurse collected illness sample and recorded clinical features and concurrent medications.

| NPS collection and processing
Nasopharyngeal sample was collected using flocked swab (Copan Diagnostics, Corona, CA). 11,12 Swabs taken from both nostrils of a subject were placed in the same specimen bottle containing viral transport medium and transported within 4 hours at room temperature to virology laboratory, where swabs were discarded after vortexing for 20 seconds to release the cells. Viral transport medium was then centrifuged and the pellet resuspended in 1 mL buffered saline and stored at −80°C until analyses.  Viral load was expressed in the absolute copy numbers of influenza M gene determined from the standard curves generated from a standard plasmid with a known copy number in serial dilutions, which was included in the quantitative PCR simultaneously, as previously described. 13 Real-time PCR was conducted using SYBR Premix EX Taq master mix (Takara Kusatsu, Shiga, Japan), and the results were analyzed using PRISM 7900HT system (Applied Biosystems, Foster City, CA, USA).

| Statistical analysis
Numerical data were expressed either in mean and standard deviation (SD) or median and interquartile range (IQR) as appropriate. The occurrence of laboratory-confirmed influenza or ILI in relation to vaccination was analyzed by logistic regression, adjusting for covariates including seasonality (month of study), subjects' age, sex, body mass index, and comorbid medical conditions. Surveillance data from all influenza seasons were combined during such analyses. Influenza VE was estimated by the "test-negative case-control" design according to published method. 14 As vaccine recipients may have a greater likelihood of seeking health care should they develop infections, this analytical approach adjusts implicitly for this confounding which would otherwise bias VE. All analyses were performed two-tailed using SPSS v.21 (Chicago, IL, USA), with 0.05 being the level of significance.    Table 2 summarizes the relationship between occurrence of ILI and subjects' personal, clinical, and vaccine factors. Subjects with ILI were younger than those without ILI (P<.001), but ILI was not associated with any environmental or clinical factor. Influenza vaccination at all time points was protective against ILI (P=.002-.022). Logistic regression confirmed such association for influenza vaccination within 3 years (odds ratio 0.49, 95% confidence interval 0.29-0.81; P=.005). None of the other factors enlisted in Table 2 was associated with mild laboratoryconfirmed influenza detected by surveillance (data not shown).

| Effectiveness of influenza vaccination
Sixty-three (18.9%) subjects who received influenza vaccination within 3 years had either IL-I or laboratory-confirmed influenza detected by surveillance, which was significantly lower when compared to 86 (29.7%) children who were not vaccinated (P<.005). Table 3   was designed to detect both of these A/H3N2 strains. Our government has included children aged 6 months to 5 years as a priority group in the influenza vaccination subsidy scheme. Nonetheless, a small local study found <20% uptake rate in preschool children. 7 In this study, 36%-44% of children received influenza vaccination annually ( Table 1). Uptake of influenza vaccination was also low among local healthcare workers in the postpandemic era. 29 Breaking barriers to accept influenza vaccination should be a public health priority in fighting against influenza outbreaks. Our findings of moderate overall influenza VE for ILI in children aged 2-12 years together with higher VE against both ILI and mild laboratory-confirmed influenza among the subgroup of school-age children may represent an additional strategy where the Government Vaccination Program can be expanded to cover children up to 12 years old.
We found that influenza vaccination was effective against ILI but not mild infection identified by 2-weekly surveillance during influenza seasons, which was surprising given that all the non-influenza causes of ILI would reduce the observed strength of influenza VE toward the null. One possible explanation relates to study power that we detected many more patients of ILI (n=99) than mild influenza infection (n=57).
On the other hand, influenza-infected children were reported to have increased susceptibility to co-infection by other respiratory pathogens, notably pneumococcus and staphylococcus, which increased the severity of their RTIs. [30][31][32] Thus, it is also possible that influenza vaccination prevented ILI by reducing the children with such co-infections.
Vaccinated and unvaccinated patients with influenza infections may exhibit different health care-seeking behavior. This study adopted the test-negative design to define influenza VE as it is less susceptible to bias due to misclassification of infection and to confounding by health care-seeking behavior relative to traditional case-control or cohort studies. 14,33 Different studies reported that influenza vaccination lowered subject hospitalization rate by 50%-66%. 5,19,20 In our research proposal, we targeted to recruit 560 children with evaluable study outcomes which was based on the assumptions of 40% influenza vaccination coverage, In a Taiwanese study, the presence of fever, cough and sneezing had the best specificity (77%) for laboratory-confirmed influenza. 35 Different ILI case definitions adopted in Europe, USA, and Taiwan had comparable accuracy in sensitivity and specificity, and clinical diagnosis of ILI was useful for providing valuable information for surveillance purpose. This study defined ILI by revised criteria that included other respiratory and constitutional symptoms. It is likely that our patients with ILI had influenza during influenza seasons. Another limitation relates to low participation rates of kindergartens (10/238) and primary schools (5/97), which raised concern if our subjects were recruited by convenience sampling. Nonetheless, we believe our surveillance data were generalizable because these kindergartens and primary schools were identified in sequential random batches from all in our target geographic regions that were registered under the Education Bureau. This study was also limited by the lack of surveillance data for subjects' household and class contacts. Most secondary cases in this surveillance were expected to have minimal influenza symptoms due to low viral load. Unless we collect surveillance samples from subjects' close contacts, we shall miss these secondary cases by only calling them for any ILI symptom.
In conclusion, mild laboratory-confirmed influenza was common among children during influenza seasons in 2014-2015. Moderate overall influenza VE was found for ILI, and subgroup analyses suggested higher VE for both ILI and mild laboratory-confirmed influenza in school-age children. Whether vaccination prevented influenza transmission within families or classes remained unanswered due to lack of reported secondary cases.

CONFLICT OF INTEREST
All authors declared no conflict of interest.