Fever is an important safety consideration after influenza vaccination in children. In 2010, an unexpected increase in childhood febrile convulsions after influenza vaccination in Australia, focused attention on influenza vaccine safety, particularly in children. At that time, 5–7 febrile convulsions/1000 doses occurred in children <5 years of age given trivalent influenza vaccine (TIV),1 necessitating a national suspension of the pediatric influenza vaccination program. Subsequent investigations revealed the events to be related only to bioCSL’s Fluvax and Fluvax Junior brands of TIV (bioCSL, Parkville, Victoria, Australia) which subsequently lost registration for use in children <5 years of age.2 Annual active safety monitoring in Australia,3 and plans for enhanced safety surveillance following annual strain changes as part of influenza vaccine licensure in Europe, reflect the increased focus on safety.4 , 5
After the 2010 events, we undertook a systematic review and meta-analysis of fever after influenza vaccination in children, to clarify expected rates of fever after TIV in children.6 This demonstrated pooled rates of fever ≥38°C of 6%–8% after first and second doses of inactivated TIV in children. However, the difficulty in comparing fever incidence between influenza vaccine trials became evident. Differences in study design and analysis, including definition of fever, its route of measurement and reporting differences (age ranges, fever after individual doses versus any dose), contributed.
The Brighton Collaboration has attempted to standardize adverse event reporting in clinical trials through use of recommended fever case definitions.7 However, there is inconsistent adoption of these guidelines. In addition, there remain many unanswered questions regarding fever after influenza vaccine in children, including its peak day of occurrence, rates of antipyretic use and the extent of interaction with immunogenicity of the vaccine.
Access to individual-level clinical trial data is seen as a key step toward transparency of research on vaccine development.8 GlaxoSmithKline (GSK) has been one of the first vaccine manufacturers to allow public access to deidentified individual-level data.9 This has obvious advantages particularly in investigating vaccine safety. We therefore used this opportunity to conduct a pilot study by obtaining access and analyzing individual-level clinical trial data from GSK pediatric influenza vaccine trials to investigate: (1) the pattern (eg, first day or day of peak fever) and frequency of fever after influenza vaccination in young children, (2) differences in fever rates due to variable analysis and reporting of fever, (3) the use of antipyretics and (4) the relation of fever and antipyretic use to each other and to vaccine immunogenicity. Given that all analyses were post-hoc, our primary aim was a descriptive analysis with hypothesis generation for future prospective research.
MATERIALS AND METHODS
Access to individual-level data from GSK vaccine trials was obtained through a data request website (clinicalstudydatarequest.com) in May 2014. We requested data from influenza vaccine trials in children identified on the website or through the ClinicalTrials.gov registry. We were provided access to 6 trials (quadrivalent and trivalent vaccine) but report in this paper on the 3 published trials with TIV, which had consistent designs with TIV in each study arm. The quadrivalent studies, in contrast, had TIV or non-influenza vaccine controls.
We focused on a primary pediatric trial10 (Pavia-Ruz et al; NCT00764790) because it had the largest number of vaccinated young children (6–35 months) among the studies provided; we then compared the findings with the 2 other TIV trials provided (NCT0038312311: 6 months to 17 years, NCT0098000512: 3–17 years). The primary study,10 was an observer-blind randomized controlled trial (RCT) that involved 3317 vaccinated children 6–35 months of age, randomized (1:1:1) into 3 groups, Fluarix (GlaxoSmithKline, Dresden, Germany) at 2 dosages (0.25 and 0.5 mL) in comparison to a licensed comparator, Fluzone (Sanofi Pasteur, Swiftwater, PA, 0.25 mL). The participant study diary recorded adverse events from day 0 (vaccination day) to 3 inclusive. Children received 1 dose of vaccine if previously primed (2 doses of vaccine in any previous year) or 2 doses, 28 days apart, if unprimed. Details for the other 2 studies are listed in Supplemental Digital Content 1, http://links.lww.com/INF/D6.
After independent review and approval of our study proposal, a data-sharing agreement was drafted by GSK and our study group outlining conditions of access, appropriate use of data, and privacy provisions. Access to datasets, dataset definitions and study protocols was provided by GSK. Data security was protected by the use of remote access to all data via a password-protected, multiple-step verification process. Our analyses were required to be conducted within the remote access environment, using SAS® Clinical Trial Data Transparency 4.5 (SAS Institute Inc., Cary, NC), with no ability to download or analyze data locally. However, results of analyses and statistical programming could be downloaded or uploaded pending approval of its content by GSK staff.
We conducted descriptive analyses of fever within the study groups and examined how published fever rates in the primary study were derived. We subsequently reanalyzed the study dataset to calculate fever incidence rates using the Brighton Collaboration definition of fever (≥38°C, recorded from any site) and by causality assignment (any fever versus vaccine-related fever). Relation to vaccination was marked yes/no in the datasets as assessed by original investigators with no further detail available. We assessed the timing of postvaccination vaccine-related fever by examining both the commonest day (0–3) of peak temperature and of first onset of fever. We examined the rates of antipyretic use, its peak day of usage and relationship to fever.
We assessed the association between fever, antipyretic use and immunogenicity, an analysis not performed in the original studies. We did this by using the “According To Protocol” subset within the study and pooling dose 1 and 2 data from all TIV study arms within each study. We examined the effect of fever presence and antipyretic use, as categorical variables, during days 0–3 after either dose, on postvaccination geometric mean titers (GMT) and geometric mean fold rise (GMFR). In the primary study,10 this was against H1N1 (A/Brisbane/59/07), H3N2 (A/Uruguay/716/07), and two antigenically similar B strains (B/Florida/4/06, B/Brisbane/3/07), as different B strains were used in the GSK and Sanofi products.
A multivariable general linear model was then used for each study to assess the effects of continuous (age and prevaccination titer) and categorical predictors (study group, number of doses, presence of fever and antipyretic use) on the log10 transformed postvaccination titer (dependent variable), as a measure of immunogenicity. For the categorical predictor variable of main interest (eg, fever or antipyretic use), the model calculates the ratio of the log10 transformed postvaccination titer if the predictor was present compared with a reference with the predictor absent, adjusting for all other predictors within the model. This result was back-transformed to obtain the GMT ratio. Lastly a pooled immunogenicity analysis was performed using the same model and combining participants from all 3 studies (but including study number and study group as predictor variables within the model, to maintain clustering of patients within studies).
The study was approved by the local Human Research Ethics Committee of the Sydney Children’s Hospital Network.
Our analysis of the Pavia-Ruz et al10 individual-level dataset revealed that the original investigators assessed fever based on differing definitions according to the route of measurement (oral/axillary/rectal, Supplemental Digital Content 2, http://links.lww.com/INF/D7). The published fever rate during day 0–3 postvaccination was 6.2%–6.6% of children, and our analysis showed this to be defined as grade 1 or higher classification (≥38.0°C by oral/axillary measurement or ≥38.5°C rectally). However, this was the rate after any dose and from any cause (Table 1). We reanalyzed the same data defining fever as ≥38.0°C by any route of measurement,7 and reported separately for dose 1 and dose 2; we found fever rates (any cause) were 3.2%–3.7% and 3.9%–5.0% for first and second doses, respectively, depending on study group. Vaccine-related fever rates were lower still at 2.7%–3.4% and 3.3%–4.1%, respectively, for first and second doses (Table 1).
In those who experienced fever ≥38°C (n = 102 dose 1, n = 82 dose 2) in the primary study,10 we found the onset of first fever was most often on day 1 and the peak of fever most commonly on day 3 for the first dose of vaccine but both measures occurred most frequently on day 0 with the second dose (Supplemental Digital Content 3, http://links.lww.com/INF/D8). As expected, fever observations judged to be vaccine-related occurred significantly earlier than vaccine-unrelated fever (mean: 1.49 vs. 1.86 days postvaccination, P = 0.04; data not shown). Antipyretic administration (paracetamol, ibuprofen or combination medications containing these drugs) was highest on the day of vaccination (day 0) for both doses of the vaccine with progressive decreases with each subsequent day (Supplemental Digital Content 3, http://links.lww.com/INF/D8).
In the According To Protocol subset (n = 3065) of our primary study,10 the incidence of vaccine-related fever ≥38°C after either dose was 5.4% (n = 165). Antipyretics were administered in 61.2% of those who had fever versus 14.5% of those without fever (Supplemental Digital Content 4, http://links.lww.com/INF/D9). Overall there was a trend toward higher GMTs in children who experienced fever and lower GMTs in children given antipyretics (Supplemental Digital Content 5, http://links.lww.com/INF/D10). On univariate analysis, GMTs against B strains were significantly higher in those children who experienced fever versus no fever, but they were significantly lower against all strains in those subjects who received antipyretics versus no antipyretics. Assessing GMFR revealed similar findings with significantly higher GMFR against B strains in individuals with fever versus no fever, and significantly lower GMFR against A strains in those children who had received antipyretics versus no antipyretics.
The multivariable model (Table 2) showed that the presence of postvaccination fever during days 0–3, controlling for other predictors, was significantly associated with higher immunogenicity against both A/H1N1 and A/H3N2 strains with GMT ratios (reference: no fever) of 1.45 (P = 0.002) and 1.23 (P = 0.02), respectively. Conversely antipyretic use, during the same period, was associated with a significantly lower GMT ratios (reference: no antipyretic) for all strain analyses (ratios: 0.80–0.90, P = 0.003–0.04 dependent on strain).
The multivariable model revealed strong evidence of associations between immunogenicity and other variables; there was strong evidence that a higher prevaccination titer was associated with higher postvaccination titer (P < 0.0001) with a 3.5–4.3 fold increase in GMT ratio per unit increase in log10 prevaccination titer. There was also strong evidence of an association with age, with a GMT ratio of 1.05 per month increase in age, seen against all strains (P < 0.0001). Using GMFR rather than GMT, as the dependent variable, showed similar associations, directions of effect and P values, apart from a negative effect with increasing prevaccination titer (GMFR ratios: 0.35–0.43/unit increase in log10 prevaccination titer, P < 0.0001) (data not shown).
Comparison to Additional Pediatric Influenza Studies: Supplemental Digital Content 6, http://links.lww.com/INF/D11
We examined for similar findings in 2 other GSK pediatric influenza trials11 , 12 (details listed in Supplemental Digital Content 1, http://links.lww.com/INF/D6). Domachowske et al12 compared FluLaval TIV (GlaxoSmithKline, Quebec City, QC, Canada) and a comparator TIV in 2116 children 3–17 years of age. In those who had fever, the day of peak fever and first day of fever were slightly more common on the day after vaccination (day 1) but were also distributed across other days in the 4-day postvaccination period in this study (data not shown).12 The multivariable model indicated trends toward increased GMT ratios (1.13–1.26) in those with fever compared with no fever, and decreased GMT ratios (0.84–0.92) in those patients with antipyretic use compared with no antipyretic use. However, statistical significance was shown only with antipyretic use in the B strain analysis (GMT ratio: 0.84; P = 0.02).
Baxter et al11 compared Fluarix (GlaxoSmithKline, Dresden, Germany) with Fluzone (Sanofi Pasteur, Swiftwater, PA) in 3325 children, but the number of children in the immunogenicity subset was considerably smaller (n = 871, 6 months to <5 years of age). In contrast, this study found peak fever and first day of fever occurred most frequently on day 0 (data not shown). We found no relation between fever or use of antipyretics and immunogenicity in this study, possibly due to lack of power.
Combined Studies Analysis
When study participants from all 3 included studies were included in a pooled analysis (n = 5902), consistent strong associations between the presence of postvaccination fever (vs. no fever) and increased GMT ratios (1.21–1.39; P ≤ 0.01) were found for all strains (Table 3). Conversely postvaccination antipyretic use (vs. no antipyretic use) was associated with reduced GMT ratios (0.80–0.87; P < 0.0006) for all strains. Limiting the model to only those patients with no fever (n = 5624) did not change the reduced GMT ratios or significance associated with antipyretic use. However, after limiting analysis to only children with no antipyretic use (n = 4884), positive GMT ratios with presence of fever remained but statistical significance was no longer present, potentially due to the low numbers with fever (n = 89).
Our study is the first to document independent access to, and use of, individual-level data from a manufacturer’s pediatric influenza vaccine trials. This has facilitated a closer examination of postvaccination fever, antipyretic use and their correlation with immunogenicity. Our study demonstrates the variability in results which occur due to minor differences in the definition of fever, methods of analysis and reporting of results. Our reanalysis of fever data from the primary study,10 using a standardized definition, found considerably lower fever incidence (3.2%–3.7% and 3.9%–5.0% for first and second doses, respectively) than in the original study report, if fever was reported after each dose rather than combining 1st and 2nd dose data. The fever rate was lower still if only vaccine-related fever was assessed, but this relied on accurate assessment by the original local investigators, something we could not verify. When fever occurred, we did not find a consistent pattern for which postvaccination day it was highest; however, we did find fevers marked as vaccine-related occurred significantly earlier than unrelated fevers.
Our study reinforces the importance of wider implementation of standardized fever reporting in clinical trials, using Brighton Collaboration definitions, to aid comparability of data from multiple trials.7 Given that increased fever after a certain TIV formulation may be a marker of a more reactogenic vaccine, the ability to compare fever incidence between formulations is important. Postvaccination fever rate is one indicator being used in annual early-season active safety monitoring of influenza vaccine in children in Australia.3
Antipyretic use after influenza vaccination was common with more than 1 in 6 young children being given antipyretics in the primary study.10 Its use was significantly higher in children who were febrile (61.2%) versus afebrile (14.5%, P < 0.0001). In contrast to fever, there was a clear pattern of most frequent administration of antipyretics on the day of vaccination in this dataset, which decreased with each subsequent day.
We found evidence of a significant association between increased immunogenicity and children who experienced fever at the time of vaccination in the primary study10 which only reached statistical significance for A strains in the multivariable analysis. However, our pooled analysis of the 3 trials demonstrated highly significant associations, for all strains, between postvaccination fever and up to 39% higher GMT; this may suggest a close role between fever and the body’s immunologic response to the influenza antigens within the vaccine.
Similarly, strong evidence of associations in the opposite direction was found between postvaccination antipyretic use (days 0–3), adjusting for all other factors including fever, and decreased immunogenicity against all vaccine strains in children. These associations were present in our primary study,10 and the strength of the association increased in the pooled study analysis. Up to a 20% decrease in GMT was found in children who had postvaccination antipyretic use. Unfortunately, antipyretic medications in the original dataset were not coded in a manner to allow determination of which class of medications were most responsible for this association.
Studies examining immunogenicity and fever or antipyretic use after influenza vaccination in children are scarce. One study examined factors affecting immunogenicity in an RCT of AS03B-adjuvanted split-virion versus nonadjuvanted whole virion 2009 pandemic H1N1 influenza vaccine in children 6 months to 10 years of age.13 That study found hemagglutination-inhibition (HI) titers to be 60% higher in both study groups in those with postvaccination fever ≥38°C. No reduction in immunogenicity was found with antipyretic use. However, vaccines in both study groups differed from the nonadjuvanted split-virion formulations which are currently more commonly used in modern vaccines and which we studied.
Our antipyretic findings were similar to studies of other routine childhood vaccines. An open-label RCT14 of prophylactic paracetamol administration for 24 hours after coadministration of 10-valent pneumococcal conjugate vaccine and hexavalent diphtheria-tetanus-acellular pertussis-hepatitis B-inactivated poliovirus types 1, 2 and 3-H influenzae type b vaccine, in 459 infants, found significantly lower GMTs of antibodies against all pneumococcal serotypes, diphtheria, tetanus and pertactin in the prophylactic paracetamol group. Some reductions in immunogenicity persisted before and after boosting approximately 1 and 3–4 years later, but these were not considered clinically significant at 4 years of age in a follow-up study.15
A systematic review of studies involving 5077 children <6 years of age, which compared prophylactic antipyretics with placebo after routine childhood vaccines, found no studies examining influenza vaccine.16 For other vaccines, it confirmed significant reductions in antibody responses to some antigens in the prophylactic paracetamol group; however, levels were still considered protective. Immunogenicity does not appear to differ in adults with or without prophylactic paracetamol, as shown in 2 placebo-controlled RCTs of influenza vaccination in health-care workers17 or elderly patients,18 respectively.
There is uncertainty whether our findings, and those of others, on immunogenicity translate into clinically significant effects. However, the fact that influenza vaccine, unlike many routine childhood vaccines, is only moderately protective may mean that modest reductions in antibody response are more likely to correlate to less protection. Reduced immunogenicity in children with influenza vaccine is an important issue, as evidenced in our primary study10 where the study vaccine did not meet noninferiority criteria against the licensed comparator. This importance is compounded by the fact that a higher HI titer of 1:110 is thought to be required as a correlate of protection in children than the 1:40 titer used in adults.19
The effects of fever and antipyretics on immune responses may be related. Fever is a response to endogenous and exogenous pyrogens20–22 thought to be mediated through the production of endogenous cytokines including tumor necrosis factor-α, interleukin-1 and interleukin-6 and mediated through cyclooxygenase-2 ultimately altering the hypothalamic set point.21 , 23 , 24 Use of antipyretics alters this fever response through effects on cyclooxygenase-2 prostaglandin production,25 , 26 and some studies have suggested antipyretics may prolong symptoms of common viral infections, potentially through lower viral-neutralizing antibody production.27–29 Additionally, a study found that common antipyretics, in particular ibuprofen, reduced immunoglobulin M and immunoglobulin G antibody synthesis in stimulated human peripheral blood mononuclear cells.30 Studies, however, have yet to establish which cytokine responses are important for influenza vaccine immunogenicity in adults or children.31 , 32
Our study has obvious limitations; our analysis was retrospective and exploratory in nature with no correction for multiplicity of comparisons. However, our ability to conduct a pooled analysis across 3 studies allowed increased confidence in our findings. The clinical significance of any changes on vaccine immunogenicity is unknown and requires further study. Our analyses have been limited to the manufacturers within the included studies, although more vaccine manufacturers have subsequently provided a commitment to allow access to their data, including Sanofi and Novartis.33
Individual-level data access from pediatric influenza vaccine trials has revealed the potential for in-depth re-examination and reporting of data. Our findings of significant associations between fever and increased vaccine immunogenicity, and between antipyretic use and reduced immunogenicity in children after influenza vaccination warrant future prospective evaluation, given the common use of postvaccination antipyretics by parents.20 , 34 The benefit of open access to individual-level clinical trial data is demonstrated by this study, and such access will hopefully be an important stepping stone toward increased transparency by vaccine manufacturers and allow new discoveries from reanalysis of previous data.
We thank Elizabeth Barnes and Stephanie Knox for SAS programming and statistical advice.
1. Therapeutic Goods Administration. Overview of Vaccine Regulation and Safety Monitoring and Investigation into Adverse Events Following 2010 Seasonal Influenza Vaccination
in Young Children
. 2010. Available at: https://www.tga.gov.au/file/555/download
. Accessed September 11, 2012.
4. Wijnans L, Voordouw BA review of the changes to the licensing of influenza
vaccines in Europe. Influenza
Other Respir Viruses. 2016;10:2–8.
6. Li-Kim-Moy J, Yin JK, Rashid H, et alSystematic review of fever
, febrile convulsions and serious adverse events following administration of inactivated trivalent influenza
vaccines in children
. Euro Surveill. 2015;20:pii=.
7. Marcy SM, Kohl KS, Dagan R, et alBrighton Collaboration Fever
Working Group. Fever
as an adverse event following immunization: case definition and guidelines of data collection, analysis, and presentation. Vaccine. 2004;22:551–556.
8. Zarin DAParticipant-level data and the new frontier in trial transparency. N Engl J Med. 2013;369:468–469.
9. Nisen P, Rockhold FAccess to patient-level data from GlaxoSmithKline clinical trials. N Engl J Med. 2013;369:475–478.
10. Pavia-Ruz N, Angel Rodriguez Weber M, Lau YL, et alA randomized controlled study to evaluate the immunogenicity
of a trivalent inactivated seasonal influenza
vaccine at two dosages in children
6 to 35 months of age. Hum Vaccin Immunother. 2013;9:1978–1988.
11. Baxter R, Jeanfreau R, Block SL, et alA Phase III evaluation of immunogenicity
and safety of two trivalent inactivated seasonal influenza
vaccines in US children
. Pediatr Infect Dis J. 2010;29:924–930.
12. Domachowske JB, Blatter M, Chandrasekaran V, et alA randomized, controlled trial in children
to assess the immunogenicity
and safety of a thimerosal-free trivalent seasonal influenza
vaccine. Pediatr Infect Dis J. 2012;31:605–615.
13. Andrews NJ, Walker WT, Finn A, et alPredictors of immune response and reactogenicity to AS03B-adjuvanted split virion and non-adjuvanted whole virion H1N1 (2009) pandemic influenza
vaccines. Vaccine. 2011;29:7913–7919.
14. Prymula R, Siegrist CA, Chlibek R, et alEffect of prophylactic paracetamol administration at time of vaccination
on febrile reactions and antibody responses in children
: two open-label, randomised controlled trials. Lancet. 2009;374:1339–1350.
15. Prymula R, Habib A, François N, et alImmunological memory and nasopharyngeal carriage in 4-year-old children
previously primed and boosted with 10-valent pneumococcal non-typeable Haemophilus influenzae protein D conjugate vaccine (PHiD-CV) with or without concomitant prophylactic paracetamol. Vaccine. 2013;31:2080–2088.
16. Das RR, Panigrahi I, Naik SSThe effect of prophylactic antipyretic administration on post-vaccination
adverse reactions and antibody response in children
: a systematic review. PLoS One. 2014;9:e106629.
17. Aoki FY, Yassi A, Cheang M, et alEffects of acetaminophen on adverse effects of influenza vaccination
in health care workers. CMAJ. 1993;149:1425–1430.
18. Gross PA, Levandowski RA, Russo C, et alVaccine immune response and side effects with the use of acetaminophen with influenza
vaccine. Clin Diagn Lab Immunol. 1994;1:134–138.
19. Black S, Nicolay U, Vesikari T, et alHemagglutination inhibition antibody titers as a correlate of protection for inactivated influenza
vaccines in children
. Pediatr Infect Dis J. 2011;30:1081–1085.
20. Sullivan JE, Farrar HCFever and antipyretic use in children
. Pediatrics. 2011;127:580–587.
21. Kohl KS, Marcy SM, Blum M, et alBrighton Collaboration Fever
Working Group. Fever
after immunization: current concepts and improved future scientific understanding. Clin Infect Dis. 2004;39:389–394.
22. Roberts NJ Jr.Impact of temperature elevation on immunologic defenses. Rev Infect Dis. 1991;13:462–472.
23. Roth J, De Souza GEFever induction pathways: evidence from responses to systemic or local cytokine formation. Braz J Med Biol Res. 2001;34:301–314.
24. Dinarello CAInfection, fever
, and exogenous and endogenous pyrogens: some concepts have changed. J Endotoxin Res. 2004;10:201–222.
25. Botting RMMechanism of action of acetaminophen: is there a cyclooxygenase 3? Clin Infect Dis. 2000;31(suppl 5):S202–S210.
26. Bushra R, Aslam NAn overview of clinical pharmacology of Ibuprofen. Oman Med J. 2010;25:155–1661.
27. Doran TF, De Angelis C, Baumgardner RA, et alAcetaminophen: more harm than good for chickenpox? J Pediatr. 1989;114:1045–1048.
28. Graham NM, Burrell CJ, Douglas RM, et alAdverse effects of aspirin, acetaminophen, and ibuprofen on immune function, viral shedding, and clinical status in rhinovirus-infected volunteers. J Infect Dis. 1990;162:1277–1282.
29. Tapiainen T, Heininger UFever following immunization. Expert Rev Vaccines. 2005;4:419–427.
30. Bancos S, Bernard MP, Topham DJ, et alIbuprofen and other widely used non-steroidal anti-inflammatory drugs inhibit antibody production in human cells. Cell Immunol. 2009;258:18–28.
31. Mastalerz-Migas A, Pokorski M, Kiliś-Pstrusińska K, et alCytokines and toll-like receptors in the immune response to influenza vaccination
. Adv Exp Med Biol. 2015;836:35–40.
32. Zedda L, Forleo-Neto E, Vertruyen A, et alDissecting the immune response to MF59-adjuvanted and nonadjuvanted seasonal influenza
vaccines in children
less than three years of age. Pediatr Infect Dis J. 2015;34:73–78.
34. Bilenko N, Tessler H, Okbe R, et alDeterminants of antipyretic misuse in children
up to 5 years of age: a cross-sectional study. Clin Ther. 2006;28:783–793.