Effectiveness of seasonal influenza vaccines in children – a systematic review and meta-analysis

Aim To assess the efficacy and effectiveness of seasonal influenza vaccines in healthy children up to the age of 18 years. Methods MedLine, EMBASE, CENTRAL, CINAHL, WHOLIS, LILACS, and Global Health were searched for randomized controlled trials and cohort and case-control studies investigating the efficacy or effectiveness of influenza vaccines in healthy children up to the age of 18 years. The studies were assessed for their quality and data on the outcomes of influenza-like illness, laboratory-confirmed influenza, and hospitalizations were extracted. Seven meta-analyses were performed for different vaccines and different study outcomes. Results Vaccine efficacy for live vaccines, using random effects model, was as follows: (i) for similar antigen, using per-protocol analysis: 83.4% (78.3%-88.8%); (ii) for similar antigen, using intention to treat analysis: 82.5 (76.7%-88.6%); (iii) for any antigen, using per protocol analysis: 76.4% (68.7%-85.0%); (iv) for any antigen, using intention to treat analysis: 76.7% (68.8%-85.6%). Vaccine efficacy for inactivated vaccines, for similar antigen, using random effects model, was 67.3% (58.2%-77.9%). Vaccine effectiveness against influenza-like illness for live vaccines, using random effects model, was 31.4% (24.8%-39.6%) and using fixed-effect model 44.3% (42.6%-45.9%). Vaccine effectiveness against influenza-like illness for inactivated vaccines, using random effects model, was 32.5% (20.0%-52.9%) and using fixed-effect model 42.6% (38.3%-47.5%). Conclusions Influenza vaccines showed high efficacy in children, particularly live vaccines. Effectiveness was lower and the data on hospitalizations were very limited.

Influenza virus causes an acute respiratory tract infection, which results in considerable morbidity and mortality worldwide (1,2). Influenza A genus is classified into many different serotypes; H1N1 and H3N2 notably cause disease in humans. Population groups at particular risk of influenza include children, pregnant women, the elderly, and patients with co-morbidities (3). Prevention of influenza is a major health issue worldwide, but providing protection by vaccination is challenging because each season the disease is caused by different strains of the virus. Current recommendations for influenza vaccination include people over the age of 65, pregnant women, people with chronic respiratory illness, and those in long-stay residential care homes (4). Influenza vaccines are not typically recommended for use in healthy children. Vaccination of healthy children aged from 6 to 24 months has been advocated in the USA, since 2004, which has gradually included children up to 59 months, and the current recommendation is to vaccinate all children from 6 months up to 19 years -with particular emphasis on children under the age of 5 or with chronic illnesses (5). No influenza vaccines have been approved for use in children younger than 6 months (6).
There are two types of vaccine available -live and inactivated, both of which can be used in children. They contain three different virus strains: two influenza A (H1N1-like and H3N2-like) and one influenza B, which have to be administered annually. Live vaccines are administered via an intranasal spray, whereas inactivated vaccines are injected, either intramuscularly or intradermally. Inactivated influenza vaccines contain killed virus, which makes them safer for use in pregnant women and those with underlying medical conditions; they cannot induce mild signs and symptoms of influenza like live vaccines. Vaccines can be adjuvanted to improve their immunogenicity (7). Inactivated influenza vaccines have been successfully adjuvanted and their safety and immunogenicity proven for use in children (8). It has also been hypothesized that vaccinating children against influenza has indirect benefits, such as reducing influenza-related morbidity and mortality in the elderly population (9). This study aimed to assess which vaccines were most appropriate for use in healthy children from 6 months to 18 years of age, by critically and systematically reviewing the evidence on their efficacy and effectiveness and performing meta-analysis of the available information.

Search of the literature
We aimed to identify all randomized control trials (RCTs) and cohort or case-control studies investigating the effectiveness or efficacy of influenza vaccines in reducing the incidence of influenza-like illness (ILI) or laboratoryconfirmed influenza in children. After elimination of duplicates, results of searches from all databases combined identified 2105 potentially relevant articles. The titles and abstracts were screened and studies measuring only the safety or immunogenicity of vaccines were excluded, and so were the studies on pandemic influenza vaccines, studies assessing efficacy or effectiveness of vaccines in adults and children with asthma, and review articles. This process left 122 articles that required a more detailed evaluation, 92 of which did not fulfill the inclusion criteria. The number of studies that met all of the inclusion criteria was 30 ( Figure 1).

Selection criteria
This review included RCTs, cohort studies, and case-control studies carried out in any country until the end of December 2011. Studies eligible for inclusion were those assessing any preparation of seasonal influenza vaccine, administered via any route to healthy participants up to and including the age of 18 years. For inclusion, studies had to have a comparison to a control group that received either an alternative vaccine, placebo, or no intervention. Outcome measures considered for inclusion were cases of influenza or ILI confirmed either clinically or by laboratory techniques, hospitalizations due to influenza, and preventive efficacy or effectiveness of vaccines. For the purposes of this review, efficacy was defined as the ability of the vaccine to reduce the number of laboratory-confirmed cases of illness caused by influenza infection. Effectiveness was defined as the ability of the vaccine to reduce clinical cases of symptomatic influenza-like illnesses.
We excluded the studies that assessed vaccines in children with chronic illnesses, such as asthma. Searches also pro-duced a considerable number of studies that did not record clinical outcomes, but rather defined the presence of infection using serological data, such as antibody titers. Those studies were also excluded from the assessment of effectiveness.

Data analysis
We extracted information on the main characteristics of the studies on live vaccines (Table 1) and inactivated vaccines ( Table 2) for all 30 studies. In order to judge the quality of the studies, further information about randomization, allocation concealment, intention to treat analysis, and blinding was extracted. We assessed the quality of studies using standard GRADE criteria (Supplementary Table S1) and extracted detailed definitions on the outcomes used in the studies (Supplementary Table S2). Potential confounding factors and any resulting adjustments were also recorded.
In the studies that reported overall vaccine efficacy or effectiveness (VE), calculating it as: VE = 100% -odds ratio (OR); or VE = 100% -risk ratio (RR), with 95% confidence intervals (CIs) provided, results were entered directly into the data extraction table. Wherever studies had reported OR or RR, the overall VE and CIs were calculated. In the studies that only reported the number of ILIs (or laboratory-confirmed influenza cases) in vaccinated individuals and controls, the OR was calculated and overall VE and CIs recorded. Studies were then classified by outcomes, to allow separate assessment of efficacy and effectiveness. Tables were created to compare the efficacy or effectiveness of live and inactivated vaccines ( Table 3-6). From each table, the data were used to perform 7 separate metaanalyses: for efficacy of live vaccines (based on similar antigen and per protocol analysis, similar antigen and intention to treat analysis, any antigen and per protocol analysis, and any antigen and intention to treat analysis), efficacy of inactivated vaccines (based on similar antigen, any analysis), effectiveness of live vaccines, and effectiveness of inactivated vaccines (supplementary Figures S1-S7). Two results were reported for each meta-analysis, based on random effects model and fixed effects model, and presenting pooled meta-estimates with 95% confidence intervals. For all computations, Stata 11.2 (StataCorp, TX, USA) was used.
One study (33) recorded the efficacy of inactivated vaccines in children up to 6 years old only, and presented the results by age group. The efficacy was consistently high in the 6 months to 1 year age group. Only one study (28) assessed the efficacy of an adjuvanted inactivated vaccine and compared it to a non-adjuvanted inactivated vaccine.
The adjuvanted vaccine showed a greater efficacy than control and also non-adjuvanted vaccine. The efficacy of the adjuvanted vaccine in this study was 80%, which is comparable to the efficacy of live vaccines. The effect of inactivated vaccines on reducing hospitalization was only recorded in one study (33), which showed the overall efficacy of 71% in reducing hospitalizations from influenza A infection in children aged 6 months to 6 years and of 72% for influenza B infection.

effectiveness of live and inactivated vaccines
Vaccine effectiveness against influenza-like illness for live vaccines, using random effects model, was 31 Figure S6-S7). Clearly, the lower values for effectiveness than for efficacy reflect the differences in outcomes used in the studies. The estimates of efficacy were based on studying the outcomes which influenza vaccines were designed to protect against, while for assessing effectiveness the outcomes were less specific.

DISCuSSIoN
This review showed a high efficacy of influenza vaccines against influenza infection, but a considerably lower effec- tiveness. There was a slight difference in favor of live vaccines in efficacy, but hardly any in effectiveness. Two studies demonstrated the efficacy of live vaccines in children under 2 years of age (22,27), which was an important contribution to available information on this age group. Expectedly, both live and inactivated vaccines were more efficacious against infection with strains of influenza virus that were antigenically similar to strains contained in the vaccine (22,(26)(27)(28), highlighting the need for good strain-matching for the vac-cines that are being delivered in large programs. Adjuvanted inactivated vaccines showed efficacy almost as good as that of live vaccines, although this result came from only one study (28). These conclusions are in line with previously conducted systematic reviews on influenza vaccines (41), suggesting that there is a growing agreement about the key technical indicators, and that the future research challenges will be more focused on achieving more equitable impact of the available preventive interventions (42). This report has several limitations. We performed metaanalyses of the studies that were matched for the type of vaccine (live or inactivated), antigen used (similar or any), and analysis used (per protocol or intention to treat). Still, studies that were grouped together varied in study designs, case definitions, and age ranges applied (Supplementary material). Moreover, we excluded several papers in Russian (43-49), because it was not possible for us to assess full text articles and we already felt that we had enough evidence. For a more in depth review, we could also consider wider aspects of influenza vaccination, such as the cost-effectiveness and safety of implementing vaccination programs for different age groups, the effect that vaccination may have on reducing absence from school and parents' absence from work, and acceptability of vaccine to providers and end-users. There have been studies researching the indirect beneficial effect that vaccinating children has on morbidity and mortality from influenza in the elderly (9), which could be another interesting research area.
In many studies, there was a lack of information on the composition of the vaccine used and its matching to the circulating strains of influenza viruses for that season. This is vital as it inevitably will have a significant effect on the efficacy of vaccines. Influenza vaccines will not have effect against antigenically dissimilar viruses. There are also some limitations imposed by the nature of research available in the area of influenza vaccines. There were no studies that directly compared live and inactivated vaccines, although such studies would allow an analysis of the efficacy and effectiveness of the vaccines in a comparable context. There was a considerable heterogeneity in the outcome definitions used in the studies, particu-larly with reference to ILI definitions. Studies also differed according to methods of collecting data on ILIs; in some studies cases were diagnosed by general practitioners or pediatricians on visits to a clinic and in others according to parental reports of symptoms.
The main strength of this review, compared to other relevant literature, is that it summarized the available information on this issue and provided an evidence base from which further research could be conducted. It exposed differences in the quality and methodological approaches to studies, such as the definition of outcomes and study designs and gaps in information provided for certain age groups. It strengthened the existing knowledge, providing a summary of the data on the effect of live and inactivated vaccines in preventing influenza in children, in whom it showed comparable results to healthy adults. Areas for future research have also been highlighted, such as the direct comparison of the efficacy of live and inactivated vaccines, more systematic research into benefits of each type of vaccine for different age groups, and the need for a wider evidence base supporting potentially promising adjuvanted inactivated influenza vaccines (50).
Funding received from Bill and Melinda Gates Foundation.
ethical approval Not required.
Authorship declarations The study was initiated by HN, SC, and IL. SC collected and coded the data. IL, SC, and HN performed data analysis. The study was conceptualized by HN, HC, and IR. DP and RF supervised the study and participated in the explanation and discussion of the results. The manuscript was drafted by SC, and reviewed and revised by IL. All authors read and approved the final manuscript.
Competing interests All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organization for the submitted work; no financial relationships with any organiza-