Systematic review of fever , febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccines in children

J Li-Kim-Moy (jean.likimmoy@health.nsw.gov.au)1,2, J K Yin1,2, H Rashid1, G Khandaker1,2,3, C King1,2, N Wood1,2, K K Macartney1,2,4, C Jones2,3,4,5, R Booy1,2,4,5 1. National Centre for Immunisation Research and Surveillance, The Children’s Hospital at Westmead, New South Wales, Australia 2. Sydney Medical School, The University of Sydney, New South Wales, Australia 3. Centre for Perinatal Infection Research, The Children’s Hospital at Westmead, New South Wales, Australia 4. Department of Microbiology and Infectious Diseases, The Children’s Hospital at Westmead, New South Wales, Australia 5. Marie Bashir Institute for Infectious Diseases and Biosecurity, University of Sydney, New South Wales, Australia


Introduction
is a common respiratory viral infection with a substantial disease burden in children younger than five years, of whom between nine and 45 per 10,000 need hospital admission each year in developed countries [1][2][3][4].Vaccination is the leading strategy to combat influenza.The recommendations for influenza vaccination have been progressively expanded and now include all healthy children aged six months and older in the United States (US) and several European countries [5,6].The United Kingdom's (UK) Joint Committee on Vaccination and Immunisation (JCVI) recommended vaccination of all children two to 17 years of age with live attenuated influenza vaccine (LAIV) from the 2013/14 season onwards, although implementation was being staggered, commencing with two and three year-old children in the first year [7].In Australia, TIV is funded nationally for any child older than six months with medical conditions predisposing to severe influenza, and in one state (Western Australia) also for healthy children aged six to 59 months [8].
In 2010, an unexpected and marked increase in fever and febrile convulsion (FC) rates in Australian children younger than five years was detected following receipt of the seasonal inactivated trivalent influenza vaccine (TIV).Influenza vaccination for children five years and younger was briefly suspended.The increase in FC (estimated to be between five and seven events per 1,000 vaccinated children) was related only to one brand of TIV, manufactured by bioCSL (Fluvax and Fluvax Junior) [9].Despite its subsequent deregistration for children younger than five years, public concerns about vaccine safety have persisted, leading to markedly lower influenza vaccine uptake, especially in Western Australia [10].Published data documenting the frequency and severity of fever after TIV in children are sparse.Furthermore, the age bands reported and fever cut-off values used vary widely, with limited application of standardised definitions such as those from the Brighton Collaboration [11].We therefore systematically reviewed the evidence for influenza vaccine safety in children to examine the rates of fever, FCs and serious adverse events (SAEs as per standard definition [12]) associated with contemporary TIVs.We also aimed to assess the effect of age, vaccine type (adjuvanted or not) and vaccine manufacturer on the frequency of these adverse events.

Methods
An electronic literature search, without language restriction, was performed using Medline, Embase, Cochrane Library databases, LILACS, SCOPUS, and Web of Science for studies published between January 2005 and March or April 2012.Our focus was on contemporary vaccines hence our restriction to this publication period.Both controlled vocabulary and text-word terms were used, including 'immunization', 'influenza vaccines', 'influenza, human', 'safety', 'fever', 'seizures, febrile', 'adverse event/effect', 'product surveillance, post-marketing', 'Guillain-Barré syndrome', together with 'child' or 'infant.'A listing of the specific databases, search strategy and coverage dates are available from the corresponding author upon request.In addition, a search was performed within Clinicaltrials.gov, a globally used registry, for phase 2, 3 or 4 clinical trials using TIV in a paediatric population.
We included randomised controlled trials (RCTs), non-randomised clinical trials (with or without a control group) and observational studies.Studies were included if they (i) involved the use of inactivated seasonal TIV, administered intramuscularly, in at least one study arm; (ii) involved healthy children up to 17 years of age; and (iii) presented safety data in an extractable format.Studies were excluded if they only involved children younger than six months or only populations with chronic illness and/or immunocompromise.We analysed data by age band, study design, vaccine type and vaccine manufacturer, where possible.Dose 1 and dose 2 data were analysed separately.Febrile convulsion rates and SAEs were noted, if documented.
The quality of RCT studies was assessed by examining bias using the Cochrane Collaboration's tool for assessing risk of bias [13]; non-randomised clinical trials were assessed by the Effective Public Health Practice Project (EPHPP) Quality Assessment tool, as this better encompassed variation [14,15].
Meta-analysis was conducted on fever data using the Brighton Collaboration case definition of ≥ 38 °C from any source (axillary, oral or rectal) [11].Due to variability in study methods and a lack of placebo-controlled studies, we conducted a proportion meta-analysis of fever rates using similar single-arm data from trials (StatsDirect statistical software version 2.7.9) to calculate pooled fever proportions.This method has been used previously in systematic reviews across different disciplines [16][17][18][19][20][21].A random effects model with the DerSimonian-Laird method was used to account for variability in study design and results.The I 2 statistic was used as a measure of heterogeneity of pooled estimates [13].
We conducted sensitivity analyses of meta-analyses to see if exclusion of high-risk RCTs, or those non-randomised clinical trials rated as weak, reduced heterogeneity.If heterogeneity was unchanged, then all available studies were used for analysis.

Characteristics of randomised controlled trials
In the 18 randomised control trials (Table 1), a total of 22,484 subjects were enrolled, of whom 16,474 received TIV and had safety data collected.Multiple study designs were encountered in terms of comparison groups; for non-adjuvanted TIV, comparison with placebo was only found in one study [33].Five studies examined adjuvanted vaccines (MF59 or virosomal adjuvant) in at least one study arm [30,31,34,35,39].Classification of fever varied across studies, but a majority of studies [22,25,27,[29][30][31]34,35,37,38] provided data on fever ≥ 38 °C.We used these studies for meta-analysis of fever rate and one additional study [39], where we assumed a fever definition of ≥ 38 °C based on two similar studies by the same lead author [31,35].

Table 1a
Characteristics of randomised controlled trials included for analysis of fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children  Adj: adjuvanted; AE: adverse event; ATIV: adjuvanted trivalent influenza vaccine; FFU: fluorescence focus assay units; GSK: GlaxoSmithKline; LAIV: live attenuated influenza vaccine; SAE: serious adverse event; TIV: non-adjuvanted trivalent influenza vaccine.
a Where multiple doses were administered, fever is listed for the first dose.Rates are for the youngest age group within the study unless otherwise stated.

Adverse events following immunisation Fever
Pooled estimates of fever obtained using proportion meta-analysis of studies are shown in Table 3 and Table 4.

Non-adjuvanted vaccines in children six to 35 months of age
The pooled proportion estimate of fever was 6.7% (95% confidence interval (CI): 3.0-11.8)after first dose of TIV based on five eligible RCTs [22,29,31,38,39].None of these RCTs had a high risk of bias.Analysis of five nonrandomised clinical trials [40,41,47,49,53] provided higher first-dose fever estimates of 17.7% (95% CI: 11.3-25.2),largely due to the inclusion of two studies of bioCSL vaccines [47,53] that reported higher rates of post-vaccination fever.Rates after second doses are listed in Table 3 and Table 4.

Non-adjuvanted vaccines in children three to 17 years of age
There were only two eligible two-dose RCTs in this age group [29,39].The pooled proportion estimate of fever for children three years and older was 6.9% (95% CI: 5.2-8.7)for dose 1. Meta-analysis of non-randomised clinical trials revealed more fever, 15.1% (95% CI: 13.3-17.0),again due to the inclusion of studies using bioCSL vaccines [47,53].Second doses caused lower rates of fever.

Adjuvanted vaccines
Three RCTs used Fluad (Novartis), an MF59-adjuvanted vaccine which remains investigational and unlicensed in the paediatric age group, and included children aged from six months to 17 years [31,35,39].Two of these studies [31,35] had low risk of bias and one was medium risk [39].Point estimates of fever were higher than corresponding values for non-adjuvanted vaccines; however confidence intervals were wide due to the limited number of subjects.For children six to 35 months of age, first-dose pooled fever estimates were 11.9% (95% CI: 6.8-18.3).Data were more limited on children three years and older with pooled fever rates of 10.3% (95% CI: 1.1-27.0).Again, second doses elicited less fever.A small single non-randomised clinical trial reported fever rates of 16.0% for age 16-35 months, and 11.1% for age 36-48 months [48].

Post-vaccination fever, analysis by vaccine manufacturer
Fever estimates were calculated for Sanofi Pasteur, GlaxoSmithKline (GSK), Novartis, and bioCSL vaccines.Studies were grouped together, despite some variation in definition of fever, to maximise the number of studies evaluated.Data were analysed within age bands of six to 35 months and three to 17 years; data for dose 1 and 2 were analysed separately where possible.Data presented below covers non-adjuvanted vaccines.As MF59-adjuvanted (Novartis) and virosomal-adjuvanted (Berna Biotech) vaccines were produced by single manufacturers, corresponding data for adjuvanted vaccines are listed within the adjuvanted sections of Table 3 and  Table 4.

Non-randomised studies
Fever rates were relatively high in Sanofi studies after the first dose in young children aged six to 35 months (16.9%; 95% CI: 12.6-21.6),but lower in three to eight year-old children (0.4%; 95% CI: 0-2.4).GSK studies did not allow analysis by these age bands; the average childhood fever rate was 5.6% (95% CI: 2.9-9.1).

Serious adverse events (SAEs)
'Serious adverse events' were not routinely defined in studies but was we assumed them to be the standard definition commonly used in clinical trials [12].

Randomised studies
Using similar proportion meta-analysis of vaccinated study arms, we calculated an FC rate of 1.1 per 1,000 (95% CI: 0.51-1.9)using three large RCTs [26,32,39] (n = 7,439 children up to 59 or 71 months of age) that specifically reported FC as adverse events, and six RCTs (1,207 children aged up to 59 months) [22,25,29,31,34,38] that reported no related SAEs and by assumption, no FC.One of the three studies that reported on FC [32] included one vaccine-related seizure within a subset of 1,496 children aged 6-59 months (0.67 events per 1,000 children).Another study [26] reported two vaccine-related FCs among 4,173 children aged six to 59 months following TIV administration (0.48 events per 1,000).A third study [39], the only one incorporating a non-TIV control group, found similar FC rates in three study arms of non-adjuvanted TIV (2.82 per 1,000; n = 1,770), MF59 ATIV (2.59/1,000; n = 1,934) and active control vaccine (4.05/1,000; n = 988) in children six to 71 months of age.However, no comment was made if these FCs were causally related to vaccination.

Non-randomised studies
Two vaccine-related FCs were recorded in two nonrandomised clinical trials (in total 2,269 evaluable children, 854 aged between six months and three years) [47,53].Both studies used bioCSL TIV and had high rates of fever, particularly in younger vaccine recipients, compared with other non-randomised study results.Rates were not calculated due to the unavailability of denominator data within the susceptible age range.

Estimates of fever from unpublished clinical trial data (Clinicaltrials.gov)
Results from unpublished clinical trials are summarised in Table 5 and Table 6.Insufficient information on study methodology precluded detailed comparisons between studies.Temperature definitions were largely unavailable.There were five RCTs, of which three were double-blind RCTs (NCT00464672, NCT00764790, NCT00959049).One of these, an RCT (NCT00959049) which was unpublished at the time of our literature search [60], directly compared Afluria (bioCSL) with Fluzone (Sanofi) across several age bands.It was conducted in the US between September 2009 and May 2010 and defined fever as either ≥ 37.5 °C axillary or ≥ 38 °C oral.Afluria was associated with significantly higher rates of fever compared with Fluzone for first doses in children aged six to 35 months (37.1% vs 13.6%, respectively, p < 0.0001) and three to eight years (21.8% vs 9.4%, respectively, p = 0.0001).There were no significant differences in fever following second doses or after single doses in children aged nine to 17 years.
Fever rates in other RCTs ranged from 6.2 to 10.7% for children aged six to 35 months, 0-11.0% in children aged three to eight or nine years, and 0-3.8% in children aged nine or 10 to 17 years.Seven small non-randomised clinical trials were identified, all using Sanofi vaccine.Age ranges were variable, precluding detailed comparison.Fever rates varied widely (Table 6).

Table 2b
Characteristics of non-randomised clinical trials included for analysis of fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children

Table 5a
Unpublished clinical trials from Clinicaltrials.govincluded for analysis of fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children  with TIV in healthy children, however, fever was significantly associated with TIV within the window between Day 1 and 14 (incidence rate ratio (IRR) = 1.71; 95% CI: 1.64-1.80).
One retrospective observational cohort study in children in Western Australia (WA) from 2010 reported on the rate of fever seen with bioCSL TIV [58].Data linkage of TIV-associated FC cases and vaccine exposure recorded in the Australian Childhood Immunisation Register, was added to data obtained from vaccine providers or primary caregivers.A high rate of FC, 3.3 per 1,000 vaccine doses, was documented during the 49-day vaccination programme, with 62 of 63 FC associated with bioCSL TIV, all occurring after a first dose, with a median time of 7 hours from vaccination to symptom onset.In children younger than five years, FCs were significantly more associated with bioCSL TIV than with Solvay's Influvac (p < 0.0001).
Subsequent to the reporting of excess FC rates post TIV in Australia, another VSD study was conducted in the US during the 2010/11 influenza season, examining Day 0 to 1 after TIV administration and examined 206,174 children aged six to 59 months who received at least one dose of vaccine [59].None received bioCSL vaccine as its recommendation had been removed.While the main finding was of increased FC with concurrent TIV and 13-valent pneumococcal conjugate vaccine (PCV13), adjustment for PCV 13 still yielded a statistically significant increase in seizures following TIV by itself (IRR = 2.4; 95% CI: 1.2-4.7).The risk difference estimate was maximal at 16 months of age with 12.5 vaccine-attributable seizures per 100,000 doses.

Discussion
Our study summarises fever and FC data from multiple clinical trials, reporting group (not individual) safety outcomes following TIV receipt.Using published RCT data, we have found a reassuringly low pooled rate of fever ≥ 38 °C after non-adjuvanted TIV, which was similar to most non-bioCSL vaccines in observational studies conducted during 2010 when safety concerns arose due to bioCSL TIV [61][62][63].
Limited pooled data on investigational MF59-ATIV showed higher fever rates compared with non-adjuvanted vaccines.However in the two RCTs [31,39] with direct comparison of MF59-ATIV and TIV, fever rate differences were non-significant between adjuvanted and non-adjuvanted vaccine groups, apart from a subset of children aged 36 to 71 months in one study where the MF59-ATIV recipients had higher fever [39].The same RCT [39] found no differences in fever rate between MF59-ATIV and TIV in younger children aged six to 35 months.However, it also recorded the highest fever rates in the non-adjuvanted arm for this age group (13.3% and 13.4% for doses 1 and 2, respectively) relative to all other non-adjuvanted vaccine study arms in our meta-analysis; this may have contributed to the absence of observable difference in fever between MF59-ATIV and TIV.In addition, the European Medicines Agency (EMA) raised concerns, after site inspections, that this study was not conducted in accordance with guidelines on good clinical practice (GCP), and therefore did not grant marketing approval for the Novartis MF59-ATIV used [64,65].
Non-randomised clinical trials were of lower quality, often being uncontrolled.Pooled fever estimates for non-adjuvanted vaccines were higher than those from RCTs, probably due in part to the inclusion of reactogenic bioCSL vaccines [47,53], although other manufacturers' vaccines also recorded higher fever rates than in RCT studies.
A recent systematic review of fever by Kaczmarek et al. following dose 1 of inactivated TIV, reported a similar rate (8.0%) for any fever in children aged six to < 36 months after non-adjuvanted TIV, using weighted average weekly risk [66].However, our study, by using a proportion meta-analysis method, allowed inclusion of a broader range of studies.We used the Brighton Collaboration's fever definitions (≥ 38 °C) and analysed fever in a number of additional settings: adjuvanted vaccine studies, older children (36 months and older), fever after second doses of vaccine and by vaccine manufacturer.
Most non-bioCSL brand TIVs had low rates of fever in RCT analyses.However, bioCSL TIVs had significantly higher fever after first doses in children aged six months to eight years, across three studies conducted from March 2005 through to May 2010, particularly in an RCT (NCT00959049) comparing bioCSL's Afluria and a comparator TIV [60], subsequently published after our literature search and review (Table 6).Observational studies from 2010 in Australia and New Zealand documented similar findings comparing bioCSL TIV to other manufacturers [58,62].
Our findings on SAE and FC rates are considerably limited by the absence of studies using within-study placebo controls, which precludes calculation of true vaccination-related rates.However, analysing TIVvaccinated arms, we found that vaccination-related SAEs were uncommon.Our calculated FC rate from published RCT data (no bioCSL studies available) was 1.1 per 1,000 children six to < 72 months-old and vaccinated with non-adjuvanted TIV.However, it was unclear in one study if all FC reported were causally related to TIV [39]; the actual rate may be lower.The same study showed no difference in FC rates between TIV and the non-TIV, active control arm [39].We could not calculate FC rates in the clinical trials with bioCSL vaccine, but two observational studies conducted since 2010 reported FC rates of 3.5-4.4/1,000doses for bioCSL Fluvax/Fluvax Junior compared with no FCs after 4,720 doses of Solvay vaccine (Influvac) or 3,213 doses of non-bioCSL TIV [58,62].Furthermore, a 2010 investigation by the Therapeutics Goods Administration (TGA)  into bioCSL vaccine found FC rates of 5-7 per 1,000 doses [9].
Based on one study, MF59-ATIV was associated with 2.59 FCs per 1,000 vaccinated children aged six to 71 months, but this was not significantly different to control groups (non-adjuvanted TIV or active control vaccine) [39].Further study of adjuvanted vaccines is warranted to investigate their safety profile, in terms of fever and FC.
Despite an observational study reporting a link between the 2010/11 US non-bioCSL TIV and FC on Day 0 to 1 [59] (mostly with concurrent PCV13), the absolute risk of TIV-related FC appeared low overall (a maximum of 12.5/100,000 doses), less than the risk seen after measles-mumps-rubella (MMR) vaccine (33/100,000) and similar to the risk after 13-valent PCV (13.7/100,000) [59,67].A subsequent study of the 2011/12 US influenza season confirmed elevated fever after concurrent TIV and PCV13 on Day 0 to 1 and listed fever rates after TIV alone similar to our findings at 7.5% in children aged six to 23 months [68].
Proposed explanations for higher fever rates with bioCSL vaccines have included 2010 TIV strain changes and manufacturing methods.Investigations by bioCSL concluded that their method of manufacture retained more virus components due to less splitting of virus, compared with other manufacturers, and that characteristics of the three viruses included in the 2010 vaccine elicited an excessive immune response in young children [69,70].However, all manufacturers used the same new strains in formulating the 2010 southern hemisphere vaccine without eliciting increased fever or FCs.
These results highlight the differences in the propensity to febrile events that may exist between different companies' TIVs.The single RCT (NCT00959049) comparing bioCSL TIV with a comparator vaccine in children most clearly demonstrates these important differences.This study was conducted in 2009/10 but only recently published in 2014 [60].It was not yet completed when the bioCSL TIV problem emerged in April 2010.Access to individual level data of this study would offer valuable insights into fever following receipt of TIV.
The lack of clearly presented, publicly available, comparable data regarding the safety of influenza vaccines, particularly in young children, has been emphasised in a previous systematic review of influenza vaccination [71].Few of the studies we examined were eligible for that systematic review due to the lack of placebo controls.Without such placebo-controlled studies, the true rate of adverse events due solely to TIV is difficult to ascertain accurately.Such studies are difficult to justify ethically as more and more countries recommend universal influenza vaccination of healthy children.Our study addressed as much data as possible, with sensitivity analyses, to provide the most comprehensive information by which to compare vaccines.
Limitations of this study are acknowledged, including the difficulty of comparing studies that have different methodology.By examining studies involving healthy children, we have maximised the comparability of studies, but the findings may not apply to children with chronic illness for whom TIV is specifically recommended.The majority of fever analyses showed substantial heterogeneity; I2 values ranged from 0% to 95.6% with most being larger than 50%.Bias assessment revealed that the majority of randomised studies had low to moderate risk of bias.A randomeffects model for pooled fever estimates was used to provide an accurate estimate across variable studies.Our sensitivity analysis was not able to identify specific sources of heterogeneity based on assessments of study quality, but underlying study variability is the most likely cause.
Our analysis did not specifically take into account differing follow-up periods.Solicited AE follow-up periods longer than 48 hours result in the possibility of unrelated fever being captured.This highlights the need for consistent reporting in studies of post-vaccination fever rates occurring within specific timeframes, particularly the first 24 hours.Lastly, most pooled fever estimates involved overlapping confidence intervals, meaning that the point estimates of fever must be compared cautiously.However, where possible, we have compared similar types of vaccines, within set age ranges, and included studies that used Brighton Collaboration definitions of fever.

Conclusions and recommendations
This review provides a generally reassuring assessment on the safety of most TIVs which have low rates of fever or serious adverse events.There is, however, evidence that the bioCSL brand vaccines have been associated with higher rates of fever than comparable vaccines.This cannot be ascribed to the change in vaccine strains alone as the 2010 TIV made by other manufacturers was not highly reactogenic.
Although Tse et al. [59] found an association between early post-vaccination FCs and US 2010/11 non-bioCSL TIVs, containing strains identical to the 2010 southern hemisphere TIV, the risk was low and comparable to other routine immunisations.
We advocate prompt reporting and publication of clinical trial safety data for influenza vaccines.This is even more pertinent with the impending adoption of quadrivalent influenza vaccines (QIV) containing an additional influenza B strain, to ensure that reactogenicity is not increased.Closer scrutiny of the safety of each new season's vaccine formulations in children, for example through a period of active surveillance after TIV release each season, may facilitate the early detection and rapid response to any future safety signals to minimise future impacts on the health of vaccinees and maintain confidence in immunisation programmes.
The EMA is heading in this direction with requirements from 2014 to 2015 for vaccine manufacturers to implement systems for yearly enhanced safety surveillance to rapidly detect clinically significant changes in the frequency or severity of expected reactogenicity of influenza vaccines [72,73].
Furthermore, we believe public availability of individual-level data (of precise levels of fever over time) from both past and future vaccine trials as well as the use of standardised study methods, through stricter adherence to Brighton Collaboration case definitions and reporting recommendations for adverse events, is essential to enable effective comparison both between vaccines and over time.

Erratum *
The statement of conflict of interest was omitted in the original publication and added on 25 June 2015.In Table 4, a line was added between the data for GSK and BioCSL.

Figure
FigureResults of literature search for fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children, and studies analysed interval; NA: not applicable.a Overall fever estimate calculated from studies using 38 °C fever definition for non-adjuvanted and adjuvanted vaccine analyses.Analysis by vaccine manufacturer used any fever definition.Random-effects proportion meta-analysis performed.b I 2 not calculated due to low numbers of studies.c Single study data.No meta-analysis performed.d Calculated confidence interval of a single proportion.e Dose 1 and 2 treated as separate groups within analysis.
Characteristics of non-randomised clinical trials included for analysis of fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children [57]ions between TIV vaccination and any MAE, including FCs, were found.Another retrospective cohort study examined children aged 24 to 59 months in the US Vaccine Safety Datalink (VSD) over four influenza seasons (2002-06)[57].Risk of fever and SAEs was examined in 66,283 children (91,692 doses).Similar case-crossover analysis showed no SAEs associatedTable 2a

Table 3
Pooled estimates of fever proportions from randomised controlled trials of inactivated trivalent influenza vaccine in children

Table 4
Pooled estimates of fever proportions from non-randomised clinical trials of inactivated trivalent influenza vaccine in children*

Table 5b
Unpublished clinical trials from Clinicaltrials.govincluded for analysis of fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children AE: adverse event; GSK: GlaxoSmithKline; NR: not recorded; RCT: randomised controlled trial; SAE: serious adverse event; TIV: non-adjuvanted trivalent influenza vaccine.a Studies published after our literature search and review.

Table 6
Fever estimates from unpublished trials identified at Clinicaltrials.govfollowing administration of inactivated trivalent influenza vaccine in children a Only data on intramuscularly administered vaccine group was used.b Only groups with full dose were examined.Data from groups with half dose are not presented.c Only groups with TIV administered alone are listed.

Table 7
Characteristics of observational studies included for analysis of fever, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza vaccine in children AE: adverse event; CI: confidence interval; FC: febrile convulsion; OR: odds ratio; PCV13: 13-valent pneumococcal vaccine; TIV: trivalent influenza vaccine.