Description of the condition
Viral respiratory disease imposes a heavy burden on society. The majority of viral respiratory disease (influenza-like illness (ILI)) is caused by many different agents which are not clinically distinguishable from one another. A variable proportion of ILI (7% to 15% on average) is caused by influenza viruses and is known as influenza (Jefferson 2009b).
Influenza is an acute respiratory infection caused by a virus of the Orthomyxoviridae family. Three serotypes are known (A, B and C). Influenza causes an acute febrile illness with myalgia, headache and cough. Although the median duration of the acute illness is three days, cough and malaise can persist for weeks. Complications of influenza include otitis media, pneumonia, secondary bacterial pneumonia, exacerbations of chronic respiratory disease and bronchiolitis in children. Additionally, influenza can cause a range of non-respiratory complications including febrile convulsions, Reye's syndrome and myocarditis (Wiselka 1994). Efforts to prevent or minimise the impact of seasonal influenza in the second part of the 20th century centred on the use of vaccines. Due to the yearly changes in viral antigenic configuration and the lack of carry-over protection from year to year, vaccination campaigns annually require a huge scientific and logistic effort to ensure production and delivery of that year's vaccines for high population coverage.
Description of the intervention
Currently there are three types of influenza vaccines: (1) whole virion vaccines which consist of complete viruses which have been 'killed' or inactivated, so that they are not infectious but retain their strain-specific antigenic properties; (2) subunit virion vaccines which are made of surface antigens (H and N) only; (3) split virion vaccines in which the viral structure is broken up by a disrupting agent. These vaccines contain both surface and internal antigens. In addition a variety of non-European manufacturers produce live attenuated vaccines. Traditionally whole virion vaccines are thought to be the less well-tolerated because of the presence of a lipid stratum on the surface of the viral particles (a remnant of the host cell membrane coating the virion, when budding from the host cell). Influenza vaccines are produced worldwide. Periodic antigenic drifts and shifts pose problems for vaccine production and procurement, as a new vaccine closely matching circulating antigenic configuration must be produced and procured for the beginning of each new influenza 'season'. To achieve this, the World Health Organization (WHO) has established a worldwide surveillance system allowing identification and isolation of viral strains circulating the different parts of the globe. Sentinel practices recover viral particles from the naso-pharynx of patients with influenza-like symptoms and the samples are swiftly sent to the laboratories of the national influenza centres (110 laboratories in 79 countries). When new strains are detected the samples are sent to one of the four WHO reference centres (London, Atlanta, Tokyo and Melbourne) for antigenic analysis. Information on the circulating strain is then sent to the WHO, who in February of each year recommends, through a committee, the strains to be included in the vaccine for the forthcoming 'season'. Individual governments may or may not follow the WHO recommendations. Australia, New Zealand and more recently South Africa, follow their own recommendations for vaccine content. Surveillance and early identification thus play a central part in the composition of the vaccine.
How the intervention might work
Every vaccination campaign has stated aims against which the effects of the campaign must be measured. Perhaps the most detailed document presenting the rationale for a comprehensive preventive programme was that by the US Advisory Committee on Immunization Practices (ACIP) published in 2006 (ACIP 2006). The document identified 11 categories at high risk of complications from influenza, among which are healthy adults 50 to 65 years of age and healthcare workers. The rationale for policy choices rests on the heavy burden which influenza imposes on the populations and on the benefits accruing from vaccinating them. Reductions in cases and complications (such as excess hospitalisations, absence from work, mortality and healthcare contacts) and the interruption of transmission, are the principal arguments for extending vaccination to healthy adults aged 50 to 65 years (ACIP 2006).
The 2009 ACIP document update recommends vaccination for three categories of healthy adults: “Annual vaccination against influenza is recommended for any adult who wants to reduce the risk of becoming ill with influenza or of transmitting it to others. Vaccination is recommended for all adults without contraindications in the following groups, because these persons either are at higher risk for influenza complications, or are close contacts of persons at higher risk:
persons aged > 50 years;
women who will be pregnant during the influenza season;
household contacts and caregivers of children aged below five years and adults aged > 50 years, with particular emphasis on vaccinating contacts of children aged under six months; and household contacts and caregivers of persons with medical conditions that put them at higher risk for severe complications from influenza” (ACIP 2009).
Why it is important to do this review
Given the very high cost of yearly vaccination for large parts of the population and the extreme variability of influenza incidence during each 'season', we carried out a systematic review of the evidence. To enhance relevance for decision-makers in the 2007 update of the review (Jefferson 2007) we included comparative non-randomised studies reporting evidence of serious and/or rare harms.
To identify, retrieve and assess all studies evaluating the effects (efficacy, effectiveness and harm) of vaccines against influenza in healthy adults we defined:
- efficacy as the capacity of the vaccines to prevent influenza A or B and its complications;
- effectiveness as the capacity of the vaccines to prevent influenza-like illness and its consequences; and
- harm as any harmful event potentially associated with exposure to influenza vaccines.
Criteria for considering studies for this review
Types of studies
Any randomised controlled trial (RCT) or quasi-RCT comparing influenza vaccines in humans with placebo or no intervention or comparing types, doses or schedules of influenza vaccine. Only studies assessing protection from exposure to naturally occurring influenza were considered.
Comparative non-randomised studies were included if they reported evidence on the association between influenza vaccines and serious adverse effects (such as Guillain-Barré or oculo-respiratory syndromes).
We defined as RCTs as studies in which it appears that the individuals (or other experimental units) followed in the study were definitely or possibly assigned prospectively to one of two (or more) alternative forms of healthcare using random allocation. A study is quasi-randomised when it appears that the individuals (or other experimental units) followed in the study were definitely or possibly assigned prospectively to one of two (or more) alternative forms of healthcare using some quasi-random method of allocation (such as alternation, by date of birth, or by case record number).
Types of participants
Healthy individuals aged 16 to 65 years, irrespective of influenza immune status. Studies considering more than 25% of individuals outside this age range were excluded from the review.
Types of interventions
Live, attenuated or killed vaccines or fractions thereof administered by any route, irrespective of antigenic configuration.
Types of outcome measures
Numbers and seriousness (complications and working days lost) of symptomatic influenza and influenza-like illness (ILI) cases occurring in vaccine and placebo groups.
Number and seriousness of adverse effects (systemic and severe). Systemic adverse effects include cases of malaise, nausea, fever, arthralgia, rash, headache and more generalised and serious signs such as neurological harms.
Local adverse effects include induration, soreness and redness at the site of inoculation.
Search methods for identification of studies
We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, 2010, issue 2) which contains the Cochrane Acute Respiratory Infections Group's Specialised Register, MEDLINE (PubMed) (January 1966 to June 2010) and EMBASE.com (1990 to June 2010).
Searching other resources
To identify further trials, we read the bibliographies of retrieved articles and handsearched the journal Vaccine from its first issue to the end of 2009. Results of handsearches are included in CENTRAL. In order to locate unpublished trials for the first edition of this review, we wrote to the following: manufacturers; first or corresponding trial authors of studies in the review.
Data collection and analysis
Review authors TJ and DR for the 2007 update and TJ, GB and LAA for the 2010 update independently applied inclusion criteria to all identified and retrieved articles. Four review authors (TJ, GB, LAA, EF) then extracted data from included studies on standard Cochrane Vaccines Field forms. The procedure was supervised and arbitrated by another review authors (CDP).
Selection of studies
One review author (AR) carried out an initial screening of retrieved citations. Subsequently two review authors (TJ, LAA) independently applied inclusion criteria to all identified and retrieved articles.
Data extraction and management
Four review authors (TJ, GB, LAA, EF) extracted data from included studies on standard Cochrane Vaccines Field forms. The procedure was supervised and arbitrated by another review author (CDP).
Assessment of risk of bias in included studies
Assessment of methodological quality for RCTs was carried out using criteria from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2009). We assessed studies according to randomisation, generation of the allocation sequence, allocation concealment, blinding and follow up. We assessed quality of non-randomised studies in relation to the presence of potential confounders using the appropriate Newcastle-Ottawa Scales (NOS) (Wells 2004). We used quality at the analysis stage as a means of interpreting the results. We assigned risk of bias categories on the basis of the number of NOS items judged inadequate in each study: low risk of bias - up to one inadequate item; medium risk of bias - up to three inadequate items; high risk of bias - more than three inadequate items; very high risk of bias - when there was no description of methods.
Measures of treatment effect
Efficacy (against influenza) and effectiveness (against ILI) (effects) estimates were summarised as risk ratios (RR) and for the main findings risk difference (RD) within 95% confidence intervals (CIs) (in brackets after the summary estimate). Absolute vaccine efficacy (VE) was expressed as a percentage using the formula: VE = 1-RR whenever statistically significant.
Similar analyses were undertaken for other events, such as complications, hospital admissions and harms.
As the data on average time off work were reported as a continuous measurement, these results were expressed as differences in means and combined using the mean difference method. Caution should be exercised in interpreting these results as the data are very skewed.
Unit of analysis issues
Four different definitions of 'epidemic period' were found.
- The interval between the first and the last virus isolation in the community.
- The interval during which influenza virus was recovered from more than a stated percentage of ill subjects.
- The period during which an increase of respiratory illness more than a stated % was recorded.
- The winter period taken as a proxy for epidemic period.
The data were included regardless of the definition of epidemic period used in the primary study. When data were presented for the epidemic period and the entire follow up period, those which occurred during the former were considered.
An ILI case (specific definition) was assumed to be the same as a 'flu-like illness' according to a predefined list of symptoms (including the Centers for Disease Control and Prevention (CDC) case definition for surveillance), or 'upper respiratory illness' according to a predefined list of symptoms. When more than one definition was given for the same trial, data related to the more specific definition were included.
The laboratory confirmation of influenza cases found were:
- virus isolation from culture;
- four-fold antibody increase (haemagglutinin) in acute or convalescent phase sera; and
- four-fold antibody increase (haemagglutinin) in post-vaccination or post-epidemic phase sera.
When more than one definition was given for the same trial, data related to the more sensitive definition (for example, influenza) were included.
Dealing with missing data
For the initial version of the review we wrote to first authors and manufacturers to identify possible unpublished studies and missing data. The response was disappointing and we desisted from any further attempts.
Assessment of heterogeneity
Assessment of reporting biases
The main problem with influenza vaccines studies is their poor quality and discrepancies between data presented, conclusions and authors' recommendations. For example, an earlier review of 274 influenza vaccines studies in all age groups (including the studies in this review) showed an inverse relationship between risk of bias and direction of study conclusions. Conclusions favorable to the use of influenza vaccines were associated with higher risk of bias. In these studies the authors made claims and drew conclusions unsupported by the data they presented. In addition, industry funded studies are more likely to have favorable conclusions and be published in significantly higher impact factor journals and have higher citation rates than non-industry funded studies. This difference is not explained by either their size or methodological quality (Jefferson 2009a). The review found no evidence of publication bias.
Any interpretation of the body of evidence in this review should be made with these findings in mind.
We carried out a random-effects meta-analysis of efficacy and effectiveness data (Higgins 2009) but we did not perform a quantitative analysis of non-randomised studies.
The data and analyses tables were constructed according to the following criteria.
- Inactivated parenteral (intramuscular or subcutaneous) influenza vaccines versus placebo or no intervention (Analysis 01).
- Live aerosol vaccines (Analysis 02).
- Inactivated aerosol vaccines (Analysis 03).
For all three major comparisons, subgroup analyses were carried out according to the degree of matching with that year's WHO recommended content and with circulating viruses ("WHO recommended and matching" when known). WHO recommendations on content of vaccines have been published since 1973. Different dosages and schedules of the vaccine and the presence of different adjuvants were not compared and data from arms of trials comparing only vaccine composition or dosage were pooled in the analysis. Compliance of the study vaccine with the official antigenic content and potency recommendations was checked by reviewing WHO records when possible. In case of uncertainty due to ambiguity of wording used (in the oldest trials), the opinion stated by authors was taken into account. The compliance of a live attenuated vaccine with the recommendation was classified according to the antigenic comparability of the wild strains.
The following outcomes were included in the comparisons.
- Cases of influenza (defined on the basis of a specific list of symptoms and/or signs backed up by laboratory confirmation of infection with influenza A or B viruses).
- Cases of ILI (clinically defined on the basis of a specific list of symptoms and/or signs).
- Hospital admissions.
- Working days lost.
- Local harms.
- Systemic harms.
- Severe/rare harms.
Hospital admissions rates were calculated as proportion of cases hospitalised for respiratory causes. Complications were considered as proportion of cases complicated by bronchitis, pneumonia or otitis. Working days lost in episodes of sickness absence regardless of cause were also considered. Only five trials used working days lost as an outcome measure and four of them measured the work absence in terms of difference of the average number of days lost in the two arms of the trial ( Analysis 1.7). These studies presented a value of standard error measured accordingly. The remainder (Nichol 1999a) expressed the work absence in terms of rate ratio and this does not allow the recalculation of the correct estimate of the standard error. Therefore this study was excluded from the pooled analysis.
Local symptoms are presented separately from systemic symptoms. Individual harms have been considered in the analysis, as well as a combined endpoint (any or highest symptom). All the data included in the analysis were used as presented by the authors in the primary study regardless of the number of drop-outs. This approach (complete case scenario) was decided because the majority of the studies did not present any attempt at using an intention to treat analysis nor mentioned the reasons for the loss to follow up and did not contain detailed information to allow estimations of the real number of participants.
Subgroup analysis and investigation of heterogeneity
Several trials included more than one active vaccine arm. Where several active arms from the same trial were included in the same analysis, the placebo group was split equally between the different arms, so that the total number of subjects in any one analysis did not exceed the actual number in the trials. As it was not possible to identify all sources of heterogeneity, we decided to carry out a sensitivity analysis on the results applying fixed-effect and a random-effects model to assess the impact of heterogeneity on our results. Finally, we carried out a separate analysis of trials carried out during the 1968 to 1969 (H3N2) pandemic.
Future updates of this review may include sensitivity analysis by funding source.
Description of studies
Results of the search
The first version of the review contained 20 studies (Demicheli 1999). The 2004 version added five more studies (Demicheli 2004). In 2007 we included 48 studies in all (Jefferson 2007). Some of them had more than two arms, comparing different vaccines, routes of administration, schedules or dosages and reported data from different settings and epidemic seasons. We split these studies into sub-studies (data sets). For the remainder of this review, the term 'study report' refers to the original study report, while the word 'dataset' refers to the sub-study. Details of the division of the reports of studies into data sets are given in the table of included studies. In this 2010 update we included two new trials (Beran 2009a; Beran 2009b). We excluded three new studies (Belongia 2009; Chou 2007; Khazeni 2009).
Overall, 25 data sets contributed data on efficacy/effectiveness (16 on inactivated parenteral vaccines, seven on live aerosol vaccines and two on inactivated aerosol vaccines), 12 on all effects (seven on inactivated parenteral vaccines, three on live aerosol vaccines and two on inactivated aerosol vaccines) and 20 on harms only (nine on inactivated parenteral vaccines, nine on live aerosol vaccines and two on inactivated aerosol vaccines) ( Table 1).
Included trials assessed three types of vaccine: inactivated parenteral, live attenuated aerosol and inactivated aerosol.
Thirty-four data sets of inactivated parenteral vaccine were included. Eigtheen data sets (12 study reports) provided data about efficacy or effectiveness (Beran 2009a; Beran 2009b; Eddy 1970; Hammond 1978; Keitel 1988a; Keitel 1988b; Keitel 1997a; Keitel 1997b; Keitel 1997c; Leibovitz 1971; Mixéu 2002; Mogabgab 1970a; Mogabgab 1970b; Powers 1995b; Powers 1995c; Waldman 1969a; Waldman 1969b; Weingarten 1988). They involved 34,573 participants: 18,557 in the vaccines arm and 16,016 in the placebo arms.
Seven data sets (five study reports) reported both effectiveness and harms data (Bridges 2000a; Bridges 2000b; Mesa Duque 2001; Nichol 1995; Powers 1995a; Waldman 1972b; Waldman 1972d). The population sample of these consisted of 4227 participants: 2251 received the vaccine and 1976 received the placebo.
The remaining nine data sets (nine studies) with inactivated parenteral vaccines assessed harms outcomes only and were carried out on 2931 participants (Caplan 1977; El'shina 1996; Forsyth 1967; Goodeve 1983; Phyroenen 1981; Rocchi 1979a; Saxen 1999; Scheifele 2003; Tannock 1984). In this last group, 1560 participants were immunised and 1371 received the placebo.
Live aerosol vaccines were tested in 19 data sets.
Seven data sets (three studies) reported efficacy/effectiveness outcomes (Edwards 1994a; Edwards 1994b; Edwards 1994c; Edwards 1994d; Sumarokow 1971; Zhilova 1986a; Zhilova 1986b). Altogether 29,955 participants were involved: 15,651 in vaccines and 14,304 in the placebo arms. Three data sets (three studies) provided effectiveness and harms data (Monto 1982; Nichol 1999a; Rytel 1977), 5010 individuals in all; 3290 in vaccines arms and 1720 in placebo. Nine data sets (eight studies) reported harms data only (Atmar 1990; Betts 1977a; Evans 1976; Hrabar 1977; Keitel 1993a; Keitel 1993b; Lauteria 1974; Miller 1977; Rocchi 1979b): 630 in the vaccinated and 344 in the placebo arms; 974 observations in total.
Six data sets with inactivated aerosol vaccine were included.
Two data sets provided data on efficacy or effectiveness only (Waldman 1969c; Waldman 1969d). The total number of subjects was 1187: with 950 who were vaccinated and 237 who received placebo.
Two data sets (one study) evaluated efficacy/effectiveness and harms (Waldman 1972a; Waldman 1972c) with a total population of 487: 389 in the vaccine arms 389 and 98 in the placebo arms.
Two trials (two studies) reported data on harms outcomes only (Boyce 2000; Langley 2005), with a total population of 151,120 in the vaccine arms and 31 in the placebo arms).
Two studies with live aerosol vaccine (Reeve 1982; Spencer 1977) each one data set) could not be introduced in the harms analysis (secondary effects) because data did not allow quantitative analysis (systemic and local harms were reported as given cumulative in Spencer 1977 and data were not clearly reported in Reeve 1982).
Ten studies (eight of which were comparative non-randomised studies) investigated possible associations between influenza vaccines and serious harms.
Atmar 1990 (respiratory function), DeStefano 2003 (multiple sclerosis and optic neuritis), Kaplan 1982 (Guillan Barrè Syndrome (GBS)), Lasky 1998 (GBS) Mastrangelo 2000 (cutaneous melanoma), Mutsch 2004 (Bell's palsy), Payne 2006 (optic neuritis), Scheifele 2003 (oculo respiratory syndrome), Shoenberger 1979 (GBS), Siscovick 2000 (cardiac arrest).
Included studies are described in the relevant table.
We excluded 92 studies (see Characteristics of excluded studies table).
Risk of bias in included studies
Thirty-three studies were properly randomised, seven stated that the allocation method was quasi-random and two studies were field trials. Three non-randomised studies were at high risk of bias (Kaplan 1982; Mastrangelo 2000; Siscovick 2000), one was at medium risk of bias (Mutsch 2004) and two were at low risk of bias (Atmar 1990; Lasky 1998).
In the included trials, allocation concealment was adequate in 10, inadequate in four, unclear in 26 and not relevant in two.
Assessment was double-blinded in 23 studies. Five studies were single blind and twelve did not mention blinding. Thirty-three studies were properly randomised, seven stated that the allocation method was quasi-random and two studies were field trials.
Incomplete outcome data
Few studies reported information on influenza circulation in the surrounding community, making interpretation of the results and assessment of their generalisability difficult
The harms dataset from randomised studies is small. The trial authors appear to regard harms as less important than effectiveness assessment. For example, in the trials by Beran et al (Beran 2009a; Beran 2009b) data collection on harms began at the receipt of the vaccine or placebo and continued until the end of the study. However, harms data were solicited from a subset of subjects and no mention of method used to select them and no justification for not collecting harms data from all participants were reported.
Other potential sources of bias
It is now known that industry funding of influenza vaccines studies determines publication in high prestige journals and higher citation rates than other types of funding. In addition industry funding is associated with optimistic conclusions, but the quality of the majority of influenza vaccines studies is low, irrespective of funding. A previously cited review showed a complex web of interrelationships between these variables (Jefferson 2009a), but how this impacts on policy making is unknown.
Effects of interventions
Inactivated parenteral vaccines (Analysis 01)
Inactivated parenteral vaccines were 30% effective (95% CI 17% to 41%) against the symptoms of ILI if content matched WHO recommendations and circulating strain, but were not effective (RR 0.93, 95% CI 0.79 to 1.09) when these were unknown (Analysis 1.1.2)
Against influenza symptoms vaccines were 73% efficacious (54% to 84%) when content matched WHO recommendations and circulating strain but decreased to 44% (95% CI 23% to 59%) when it did not (Analysis 1.2).
An alternative to the use of risk ratio based formula 1-RR expressed as percentage is the use of risk difference (RD). In this case 30% of unvaccinated people versus 24% of people vaccinated with inactivated parenteral vaccines developed symptoms of ILI. This is the equivalent to saying that 70% of the unvaccinated study participants did not get ILI symptoms compared to 76% of the vaccinated study participants who did not get ILI symptoms (effectiveness). When the vaccine matched the viral circulating strain and circulation was high, 4% (2% to 5%) of unvaccinated people versus 1% of vaccinated people developed influenza symptoms (efficacy). These differences were not likely to be due to chance. When the vaccine content did not match the circulating influenza viruses 1% of vaccinated people developed symptoms compared to 2% of unvaccinated people.
Efficacy was lower (74%, 95% CI 45% to 87%) when the studies carried out during the 1968 to 1969 pandemic were excluded. Based on one study, 42% less (95% CI 9% to 63%) physician visits are carried out in those vaccinated with WHO recommended vaccines matching circulating viruses, but not in those not matching (RR 1.28, 95% CI 0.90 to 1.83) (Analysis 1.3.2). A similar result is seen in the effect on days of illness (Analysis 1.4), but there seems to be no effect on times an antibiotic or a drug were prescribed (Analysis 1.5 and 1.6). Five trials evaluated time off work, estimating that vaccination saved on average around 0.13 working days. This result was not statistically significant. Hospital admissions (evaluated in four trials) were also lower in vaccinated arms, but the difference was not statistically significant. There was little difference in complication rates between vaccinated and unvaccinated groups (Analyses 1.7 to 1.10). The conclusions of this comparison were unaffected by analysis using either random- or fixed-effect models
Local tenderness and soreness was more than three times as common among parenteral vaccine recipients than those in the placebo group (RR 3.11, 95% CI 2.08 to 4.66) (Analysis 1.11.1). There were also increases in erythema (RR 4.01, 95% CI 1.91 to 8.41) (Analysis 1.11.2), but not induration or arm stiffness. The combined local effects endpoint was significantly higher for those receiving the vaccine (RR 2.87, 95% CI 2.02 to 4.06) (Analysis 1.11.5). Myalgia was significantly associated with vaccination (RR 1.54, 95% CI 1.12 to 2.11) (Analysis 1.12.1). None other of the systemic effects were individually more common in parenteral vaccine recipients than in placebo recipients. However, the combined endpoint was increased (RR 1.29, 95% CI 1.01 to 1.64) (Analysis 1.12.6).
Live aerosol vaccines (Analysis 02)
Live aerosol vaccines have an effectiveness of 10% (95% CI 4% to 16%) and content and matching appear not to affect their performance significantly. However, overall their efficacy is 62% (95% CI 45% to 73%). Again, neither content nor matching appear to affect their performance significantly. The effectiveness of the aerosol vaccines against ILI (with no clear definition) was significant only for WHO recommended vaccine matching absent or unknown (11%, 95% CI 3% to 18%). The conclusions of this comparison were unaffected by analysis using either random- or fixed-effect models.
Significantly more recipients experienced symptoms of upper respiratory infection, sore throats and coryza after vaccine administration than placebo administration (upper respiratory infection RR 1.66, 95% CI 1.22 to 2.27; coryza RR 1.56, 95% CI 1.26 to 1.94; sore throat 1.73, 95% CI 1.44 to 2.08)). There was no significant increase in systemic harms, although rates of fever fatigue and myalgia were higher in vaccine than placebo groups.
Inactivated aerosol vaccines (Analysis 03)
Inactivated aerosol vaccines had effectiveness of 42% (95% CI 17% to 60%) although this observations is based on four data sets from two studies. The conclusions of this comparison were substantially unaffected by analysis using either random- or fixed-effect models although effectiveness against ILI - WHO recommended content and matching vaccine went from a fixed-effect RR 0.59 (95% CI 0.43 to 0.81) to a random-effects RR of 0.47 (95% CI 0.19 to 1.13) (Analysis 3.1.1) and the subcomparison ILI - WHO recommended but with content and matching unknown went from a fixed-effect RR 0.69 (95% CI 0.51 to 0.93) to a random-effects RR 0.63 (95% CI 0.37 to 1.07) (Analysis 3.1.2).
We conclude that the presence of heterogeneity does not materially alter our conclusions. Sensitivity analysis by methodological study quality did not affect our findings.
None of the trials on inactivated aerosol vaccines reported significant harms.
Serious and rare harms
Oculo-respiratory syndrome (ORS)
On the basis of one randomised trial (Scheifele 2003) on 651 healthy adults aged around 45, trivalent split inactivated vaccine (TIV) causes mild oculo-respiratory syndrome in people with no previous history of ORS. ORS was defined as bilateral conjunctivitis, facial swelling (lip, lid or mouth), difficulty in breathing and chest discomfort (including cough, wheeze, dysphagia or sore throat). ORS (attributable risk 2.9%, 95% CI 0.6 to 5.2), hoarseness (1.3%, 95% CI 0.3 to 1.3) and coughing (1.2%, 95% CI 0.2 to 1.6) occurred within six days of vaccination. The association did not appear to be specific for any type of TIV.
Guillain-Barré Syndrome (GBS)
Three studies assessed the association between influenza vaccination and Guillain-Barré Syndrome (GBS) (rapidly progressing symmetric paralysis with usually spontaneous resolution). The first study compared GBS cases by vaccination status and the national incidence in vaccinated and unvaccinated national cohorts. The attributable risk from vaccination was just below 1 case of GBS every 100,000 vaccinations (Shoenberger 1979). The rise in GBS following rapid immunisation of millions of Americans in 1976 to 1977 led to the halting of the campaign. The second study (Kaplan 1982) was a retrospective cohort model comparing incidence of GBS in vaccinated and unvaccinated adults in the USA (minus the state of Maryland) within eight weeks from vaccination. The study reported a lack of evidence of association (RR of 0.6 and 1.4 for the two seasons included in the study; described as non-significant but with no confidence intervals reported). The study is a poor quality model with poor case ascertainment, no case definition and assumptions of the size of the exposed and non-exposed denominators. A similar design but with more sophistication was used in the Lasky et al study for the 1992 to 1993 and 1993 to 1994 seasons (Lasky 1998). Lasky et al. assessed the risk of GBS within six weeks from vaccination. Assessment of exposure was based on a random digit phone sample validated through state data on vaccine coverage and provider-sources data on vaccination timings. Two hundred and seventy three cases of GBS were identified through the CDC VAERS surveillance database and histories validated using hospital documentation. Only 180 cases were available for interview. Nineteen cases were assessed by the authors as being vaccine-associated (received vaccine in the previous six weeks (RR 1.8, 95% CI 1.0 to 3.5) adjusted for age, sex and season). The cases had a mean age of 66 years. The authors estimated the incidence of vaccine-induced GBS as 0.145 cases per million persons per week or 1.6 extra cases per million vaccinations. Despite its many limitations (mainly due to case attrition and variable reliability of exposure data) the study is well conducted and its conclusions credible, if conservative. We conclude that there may be a small additional risk of GBS. The studies demonstrate the danger of commencing a large vaccination campaign without adequate harms assessment.
One case-control study and case-series based in the German-speaking regions of Switzerland assessed association between an intranasal inactivated virosomal influenza vaccine and Bell's palsy (Mutsch 2004). Two hundred and fifty cases that could be evaluated (from an original 773 cases identified) were matched to 722 controls. All were aged around 50. The study reports a massive increase in risk (adjusted OR 84, 95% CI 20.1 to 351.9) within 1 to 91 days since vaccination. Despite its many limitations (case attrition - 187 cases could not be identified - and ascertainment bias - physicians picked controls for their own cases - confounding by indication - different vaccine exposure rate between controls and the reference population) it is unlikely that such a large OR could have been affected significantly by systematic error. The authors called for larger pre-licence harms trials, given the rarity of Bell's palsy. On the basis of this study the vaccine was withdrawn from commerce.
The association between influenza vaccines and cutaneous melanoma was assessed by a case-control study on 99 cases and 104 controls (Mastrangelo 2000). The authors report a protective effect of repeated influenza vaccination on the risk cutaneous melanoma (OR 0.43, 95% CI 0.19 to 1.00). The study is at high risk of bias because of the selective nature of cases (all patients in the authors' hospital), attrition bias (four cases and four controls eliminated because of "failure to collaborate", recall bias (up to five years exposure data were based on patients' recollection) and ascertainment bias (non-blinded exposure survey).
Primary cardiac arrest
The association between influenza vaccination the previous year and the risk of primary (i.e. occurring in people with no previous history of cardiac disease) cardiac arrest was assessed by a case-control study on 360 cases and 418 controls (Siscovick 2000). The authors concluded that vaccination is protective against primary cardiac arrest (OR 0.51, 95% CI 0.33 to 0.79). The difficulty of case ascertainment (77% of potential cases had no medical examiner report and/or autopsy), recall bias (spouses provided exposure data for 304 cases, while 56 survivor cases provided data jointly with their spouses) make the conclusions of this study unreliable. It is impossible to judge the reliability of this study because of a lack of details on the circulation of influenza in the study areas in the 12 months preceding cardiac arrest (the causal hypothesis is based on the effects of influenza infection on the oxygen supply to the myocardium through lung infection and inflammation).
The effects of different types of live attenuated cold recombinant influenza vaccination on pulmonary function were assessed by a double-blind placebo-controlled randomised trial on 72 healthy volunteers aged around 26 (Atmar 1990) (data on 17 asthmatics were not extracted). The authors report several non-significant drops in lung function up to seven days post-inoculation and a higher incidence of influenza like illness (17/46 versus 4/26) in the vaccinated arms.
Vaccines for the 1968 to 1969 (H3N2) influenza pandemic (Comparisons 04 to 08)
Five studies yielded 12 data sets (Eddy 1970; Mogabgab 1970a; Mogabgab 1970b; Sumarokow 1971; Waldman 1969a; Waldman 1969b; Waldman 1969c; Waldman 1969d; Waldman 1972a; Waldman 1972b; Waldman 1972c; Waldman 1972d). As one would expect, vaccine performance was poor when content did not match the pandemic strain (Analysis 4). However, one-dose or two-dose monovalent whole-virion (i.e. containing dead complete viruses) vaccines achieved 65% (95% CI 52% to 75%) protection against ILI and 93% (95% CI 69% to 98%) protection against influenza, and 65% (95% CI 6% to 87%) against hospitalisations (Analysis 5). Approximately half a working day lost and half a day of illness were saved but no effect was observed against pneumonia. All comparisons except for influenza-like illness are based on a single study (Analysis 5). The large effect on ILI is coherent with the high proportion of these illnesses caused by influenza viruses in a pandemic (i.e. the gap between efficacy and effectiveness of the vaccines is narrow). Aerosol polyvalent or monovalent vaccines had modest performance (Analyses 6 to 8).
Although this review presents a large number of comparisons and outcomes based on a number of different groupings of studies and trials, most of the discussion was based on the results of the analysis of a WHO recommended vaccine against placebo. Parenterally administered influenza vaccines appear significantly better than their comparators and can reduce the risk of developing influenza symptoms by around 4%, if the WHO recommendations are adhered to and the match is right. However, whilst the vaccines do prevent influenza symptoms, this is only one part of the spectrum of "clinical effectiveness" as they reduce the risk of total "clinical" seasonal influenza (i.e. influenza-like illness) symptoms by around 1%. When the results of our analysis are expressed as RD the effect appears minimal. This is remarkable as healthy adults are the population in which inactivated vaccines perform best. We found no evidence that vaccines prevent viral transmission or complications.
It is not possible to give a definite indication on the practical use of live aerosol vaccines, because the assessment of their effectiveness is based on a limited number of studies presenting conflicting results. The effectiveness, according to WHO criteria, appears relatively low. Results regarding inactivated aerosol vaccine are based on the analysis of a few trials reporting only clinical outcomes not directly comparable, owing to non-homogeneous definitions. It does not seem wise to draw conclusions from these data. Rates of complications caused by influenza in these trials were very low and analysis of the few trials which contained this outcome, did not reveal a significant reduction with the influenza vaccine. This result appears to contrast with assertions of policy makers (ACIP 2006) and may be due to the general rarity of complications caused by respiratory infection in healthy adults. Hospitalisation was assessed in four trials and did not show a significant benefit from vaccination. Working days lost in placebo recipient and vaccine recipients were significantly reduced in the vaccinated group, but by less than half a day on average.
Inactivated vaccines cause local (redness, induration) and systemic harms (myalgia, possibly fatigue). In rare cases there may be an increased risk of GBS, of ORS and Bell's palsy but this may be product-specific. Given the low effectiveness of the aerosol vaccines, the effects classified as harms (sore throat and cough) may be caused by influenza. Although the possibility of causing serious harm may be rare, it must be born in mind when proposing the inception of a mass campaign of immunisation to a whole population, i.e. when exposure to the vaccines is increased.
While the parenteral vaccine efficacy against seasonal (i.e. non-pandemic) influenza is around 75% for the WHO recommended and matched strain, its impact on the global incidence of clinical cases of influenza (i.e. ILI) is limited (around 16% in best case scenario). The universal immunisation of healthy adults should achieve a number of specific goals: reducing the spread of the disease, reducing the economic loss due to working days lost and reducing morbidity and hospitalization. None of the studies included in the review presented results evaluating the ability of this vaccination to interrupt the spread of the disease. Some studies presented data on reduction of working days lost and showed a very limited effect. Similarly a very limited effect was found on morbidity and no effect was found on hospitalization. Given the limited availability of resources for mass immunisation, the use of influenza vaccines should be primarily directed where there is clear evidence of benefit.
Whole-virion monovalent inactivated vaccines may help control a pandemic, if the antigenic match between virus and vaccine is right. Although this observation is based on a limited number of old trials, the high effectiveness of the vaccine (i.e. against ILI) would seem to confirm its potential for use. Efforts to update and enhance these vaccines should have priority.
A number of problems should be taken into consideration when interpreting the results of this review.
- None of the live aerosol vaccines included in the review were registered.
- Methods of vaccine standardisation have changed significantly.
- Recent vaccines present significant differences in purity when compared with older ones.
- Different doses and schedules were pooled in the analysis
The content and results of previous versions of this review have been extensively misquoted especially in public policy documents (Jefferson 2009c). Two types of common misquotes are the generalisation of evidence from this review to all age and risk groups and the generalisation of estimates of effect to all outcomes (especially complications and deaths). The misquotes then assume that the performance of influenza vaccines is uniform across all age groups and from symptom prevention to all outcomes. Both generalisations are not supported by any evidence and seem to originate from the desire to use our review to support decisions already taken. The misquotes appear to be based on both the abstract and Plain language summary (which is what you would expect from a superficial reading of the review by people with a specific agenda). It is for these reasons that in this 2010 update we have tried to minimise the risk of being misquoted by presenting effects on major outcomes both in RR and RD format and have inserted a general warning on the quality of evidence in the field of influenza vaccines. Recent examples of misquotes of this review come from page 11 of the 2009 ACIP document (ACIP 2009). The 2007 version of the review is indicated as reference 121: “When the vaccine and circulating viruses are antigenically similar, TIV prevents laboratory-confirmed influenza illness among approximately 70% to 90% of healthy adults aged < 65 years in randomised controlled trials (121, 124). Vaccination of healthy adults also has resulted in decreased work absenteeism and decreased use of health-care resources, including use of antibiotics, when the vaccine and circulating viruses are well-matched (121, 123). Efficacy or effectiveness against laboratory-confirmed influenza illness was 47% - 77% in studies conducted during different influenza seasons when the vaccine strains were antigenically dissimilar to the majority of circulating strains (117,119,121,124). However, effectiveness among healthy adults against influenza-related hospitalization, measured in the most recent of these studies, was 90% (125)”. There are three subtle manipulations in the text. First, the review is cited with single study references. Second, the impression reading the text is that vaccines have effect against all outcomes when the evidence quoted refers to cases (or symptoms as we call them in this latest update of the review). Third, our review (which only includes RCT evidence of effectiveness) shows no effect on hospitalisations, CDC quote reference 125 which is a 2007 observational study. The CDC authors clearly do not weight interpretation by quality of the evidence, but quote anything that supports their theory.
Summary of main results
Inactivated influenza vaccines decrease the risk of symptoms of influenza and time off work, but their effects are minimal, especially if the vaccines and the circulating viruses are mismatched. There is no evidence that they affect complications or transmission.
Overall completeness and applicability of evidence
Taken alone, the review shows that according to randomised evidence, inactivated vaccines have a small effect in preventing symptoms of influenza and getting workers back to work quicker.
Quality of the evidence
We found evidence from more than 80,000 people in 50 randomised studies. Regardless of quality, all studies fail to report any evidence of effect on complications. The safety evidence base from randomised trials of inactivated vaccines is very small, probably indicating less concern with harms. Inactivated vaccines cause rare major harms which appear to be mostly linked to specific products or lots.
Potential biases in the review process
The review conclusions are uncertain about the safety profile of inactivated vaccines which is a reflection of the size of the evidence base.
Agreements and disagreements with other studies or reviews
We are not aware of other systematic reviews on this topic.
Implications for practice
The results of this review seem to discourage the utilisation of vaccination against influenza in healthy adults as a routine public health measure. As healthy adults have a low risk of complications due to respiratory disease, the use of the vaccine may be only advised as an individual protection measure against symptoms in specific cases.
Implications for research
The major differences in effect size between outcomes highlight the need for careful consideration of the best study design to assess the effects of public health measures such as vaccines. Large studies encompassing several influenza seasons are required to allow assessment of the effect of the vaccines on seemingly rare outcomes such as complications and death.
The authors gratefully acknowledge the help received from Drs Brian Hutchison, Alan Hampson, James Irlam, Andy Oxman, Barbara Treacy, Gabriella Morandi, Kathie Clark, Hans van der Wouden, Nelcy Rodriguez, Leonard Leibovici, Mark Jones, Jeanne Lenzer, Janet Wale, Clare Jeffrey, Robert Ware, Roger Damoiseaux, and Maryann Napoli. The original review was funded by the UK Ministry of Defence, the 2004 update was supported by the two Italian Local Health Authorities in which two of the review authors were employed, the 2007 update was funded by the same Local Health Authorities and the UK's Department of Health Cochrane Incentive Scheme. The 2010 update was not funded. Professor Jon Deeks designed and carried out statistical analyses in earlier versions of the review. Finally, the review authors wish to acknowledge Daniela Rivetti and Vittorio Demicheli as previous authors.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. MEDLINE search strategy for 2004 update
#1 ("Influenza Vaccine/administration and dosage"[MeSH] OR "Influenza Vaccine/adverse effects"[MeSH] OR "Influenza Vaccine/contraindications"[MeSH] OR "Influenza Vaccine/immunology"[MeSH] OR "Influenza Vaccine/metabolism"[MeSH] OR "Influenza Vaccine/radiation effects"[MeSH] OR "Influenza Vaccine/therapeutic use"[MeSH] OR "Influenza Vaccine/toxicity"[MeSH]) OR ("Influenza/epidemiology"[MeSH] OR "Influenza/immunology"[MeSH] OR "Influenza/mortality"[MeSH] OR "Influenza/prevention and control"[MeSH] OR "Influenza/transmission"[MeSH])
#2 (influenza vaccin*[Title/Abstract]) OR ((influenza [Title/Abstract] OR flu[Title/Abstract]) AND (vaccin*[Title/Abstract] OR immuni*[Title/Abstract] OR inoculati*[Title/Abstract] OR efficacy[Title/Abstract] OR effectiveness[Title/Abstract])
#3 #1 OR #2
# 4 "Randomized Controlled Trial"[Publication Type] OR "Randomized Controlled Trials"[MeSH] OR "Controlled Clinical Trial"[Publication Type] OR "Controlled Clinical Trials"[MeSH] OR "Random Allocation"[MeSH] OR "Double-Blind Method"[MeSH] OR "Single-Blind Method"[MeSH]
#5 controlled clinical trial*[Title/Abstract] OR randomised controlled trial*[Title/Abstract] OR clinical trial*[Title/Abstract] OR random allocation[Title/Abstract] OR random*[Title/Abstract] OR placebo[Title/Abstract] OR double - blind[Title/Abstract] OR single - blind[Title/Abstract] OR RCT[Title/Abstract] OR CCT[Title/Abstract] OR allocation[Title/Abstract] OR follow - up[Title/Abstract]
#6 #4 OR #5
#7 #3 AND #6
Appendix 2. MEDLINE (PubMed) search strategies for 2010 update
#1 "Influenza Vaccines"[MeSH] OR ("Influenza, Human/complications"[MeSH] OR "Influenza, Human/epidemiology"[MeSH] OR "Influenza, Human/immunology"[MeSH] OR "Influenza, Human/mortality"[MeSH] OR "Influenza, Human/prevention and control"[MeSH] OR "Influenza, Human/transmission"[MeSH])
#2 ((influenza vaccin*[Text Word]) OR ((influenza [Text Word] OR flu[Text Word]) AND (vaccin*[Text Word] OR immuni*[Text Word] OR inoculation*[Text Word] OR efficacy[Text Word] OR effectiveness[Text Word])))
#3 #1 OR #2
#4 randomized controlled trial [pt] OR controlled clinical trial [pt] OR randomized [tiab] OR placebo [tiab] OR drug therapy [sh] OR randomly [tiab] OR trial [tiab] OR groups [tiab]) AND humans [mh]
#5 ("cross over" OR "crossover" OR "Follow Up") OR ("Cross-Over Studies"[MeSH] OR "Follow-Up Studies"[MeSH] OR "Prospective Studies"[MeSH]) OR ("time series" OR "interrupted time series") OR (placebo* OR random* OR "double blind" OR "single blind" OR clinical trial* OR trial design) OR ("Case-Control Studies"[MeSH] OR (cases[Title/Abstract] AND controls[Title/Abstract])) OR ("Cohort Studies"[MeSH] OR cohort*) OR ("Comparative Study"[Publication Type]) OR ("before after"[Title/Abstract] OR "before-after"[Title/Abstract] OR "before/after"[Title/Abstract] OR "before and after"[Title/Abstract]) OR (volunteer*[Title/Abstract]) OR (control*[Text Word] AND evaluation[Text Word])
#6 #4 OR #5
#7 #3 AND #6
Appendix 3. EMBASE.com search strategy for 2010 update
#1 'influenza vaccine' /exp OR 'influenza vaccine' OR ( influenza OR flu AND ( vaccin* OR immuni* OR inoculat* )) OR 'influenza vaccine' /syn OR ( 'influenza' /exp AND 'vaccine' /exp)
#2 'case control study' /syn OR 'case control' :de,ab,ti OR ( cases :ab,ti AND controls :ab,ti) OR 'cohort analysis' /syn OR 'cohort study' :de,ab,ti OR 'study cohort' :de,ab,ti OR prospectiv* :ab,ti OR volunteer* :ab,ti OR observational :ab,ti OR 'clinical trial' :it OR 'randomized controlled trial' :it OR 'drug therapy' /exp OR 'drug therapy' :de OR randomized :ab,ti OR randomised :ab,ti OR placebo :ab,ti OR randomly :ab,ti OR trial :ab,ti OR groups :ab,ti
#3 'clinical trial' :it OR 'randomized controlled trial' :it OR 'drug therapy' /exp OR 'drug therapy' :de OR randomized :ab,ti OR randomised :ab,ti OR placebo :ab,ti OR randomly :ab,ti OR trial :ab,ti OR groups :ab,ti
#4 'clinical trial' :it OR 'randomized controlled trial' :it OR 'randomized controlled trial' /exp OR 'randomization' /exp OR 'single blind procedure' /exp OR 'double blind procedure' /exp OR 'clinical trial' /exp OR 'clinical' NEAR/0 'trial' OR 'clinical trial' OR ( singl* OR doubl* OR trebl* OR tripl* AND ( mask* OR blind* )) OR 'placebo' /exp OR placebo* OR random* OR 'control group' /exp OR 'experimental design' /exp OR 'comparative study' /exp OR 'evaluation study' OR 'evaluation studies' /exp OR 'follow up' /exp OR 'prospective study' /exp OR control* OR prospectiv* OR volunteer* AND [humans]/lim
#5 #2 OR #3 OR #4
#6 #1 AND #5
#7 #1 AND #5 AND [humans]/lim AND [embase]/lim
Appendix 4. Glossary
the impact of an intervention (drug, vaccines etc) on a problem or disease in ideal conditions - in this case the capacity of vaccines to prevent or treat influenza and its complications.
the impact of an intervention (drug, vaccines etc) on a problem or disease in field conditions - in this case the capacity of vaccines to prevent or treat ILI and its complications.
an acute respiratory infection caused by a virus of the Orthomyxoviridae family. Three serotypes are known (A, B and C). Influenza causes an acute febrile illness with myalgia, headache and cough. Although the median duration of the acute illness is three days, cough and malaise can persist for weeks. Complications of influenza include otitis media, pneumonia, secondary bacterial pneumonia, exacerbations of chronic respiratory disease and bronchiolitis in children. These illnesses may require treatment in a hospital and can be life-threatening especially in 'high-risk' people e.g. the elderly and people suffering from chronic heart disease. Additionally, influenza can cause a range of non-respiratory complications including febrile convulsions, Reye's syndrome and myocarditis. The influenza virus is composed of a protein envelope around an RNA core. On the envelope are two antigens: neuraminidase (N antigen) and hemagglutinin (H antigen). Hemagglutinin is an enzyme that facilitates the entry of the virus into cells of the respiratory epithelium, while neuraminidase facilitates the release of newly produced viral particles from infected cells. The influenza virus has a marked propensity to mutate its external antigenic composition to escape the hosts' immune defences. Given this extreme mutability, a classification of viral subtype A based on H and N typing has been introduced. Additionally, strains are classified on the basis of antigenic type of the nucleoprotein core (A, B ), geographical location of first isolation, strain serial number and year of isolation. Every item is separated by a slash mark (e.g. A/Wuhan/359/95 (H3N2)). Unless otherwise specified such strains are of human origin. The production of antibodies against influenza beyond a conventional quantitative threshold is called seroconversion. Seroconversion in the absence of symptoms is called asymptomatic influenza.
Influenza-like illness (ILI)
an acute respiratory illness caused by scores of different viruses (including influenza A and B) presenting with symptoms and signs which are not distinguishable from those of influenza. ILI does not have documented laboratory isolation of the causative agent and is what commonly presents to physicians and patients (also known as the flu").
Inconsistency between results and abstract
We feel there is some inconsistency between results and abstract of this review regarding off work time.
In the results it states that 0.4 days are saved, but that this result is not statistically significant. In the abstract, however, this difference is labelled significant. Can you help us in understanding this?
I certify that I have no affiliations with or involvement in any organisation or entity with a direct financial interest in the subject matter of my criticisms.
JC van der Wouden
Feedback added 16/04/07
The difference is statistically significat as it says in the abstract. In the results the word "statistical" has been used instead of "clinical". Indeed the meaning of the comment was to underline that, although statistically significant, a difference of 0.4 day is clinically inconsistent.
Comments regarding the conclusion
Your conclusion is confusing. You write: "Universal immunization of healthy adults is not supported by the results of this review." If so, why the first sentence? You wrote in the Discussion that "serologically confirmed cases of influenza are only part of the spectum of clinical effectiveness." Furthermore, it would be helpful if you had explained the difference between influenza and influenza-like illness in the abstract. Also, the title of the synopsis is inaccurate. Why say "not enough evidence" when there are so many trials in your review? It should read: Clinical trials do not support the universal recommendation, etc. And "by a quarter" is not going to be understood by the general public. Please put in absolute terms.
I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.
Feedback added 05/04/06
This comment has been superseded and addressed by the 2006 latest update.
The review authors
Vaccines for preventing influenza in healthy adults, 13 May 2013
There seems to be an inconsistency in the presentation of the Cochrane Summary: "Vaccines to prevent influenza in healthy adults". The Plain language summary states that "Vaccine use did not affect the number of people hospitalised or working days lost", but under Main Results we read that "Vaccination had a modest effect on time off work and had no effect on hospital admissions". These two claims seem to be at odds regarding working days/time lost.
I agree with the conflict of interest statement below:
I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.
Last assessed as up-to-date: 3 June 2010.
Protocol first published: Issue 4, 1998
Review first published: Issue 4, 1999
Contributions of authors
For the 2010 update Tom Jefferson (TJ) and Eliana Ferroni (EF) designed the update.
Alessandro Rivetti (AR) carried out the searches and preliminary screening of references.
TJ and LAA applied inclusion criteria.
TJ, LAA, EF and GB extracted data.
Carlo Di Pietrantonj (CDP) arbitrated and checked the data extraction.
CDP performed the meta-analysis and carried out statistical testing.
TJ and AR wrote the final report.
All review authors contributed to both the protocol and final report.
Statistical support to previous review versions was provided by JJ Deeks.
Declarations of interest
TJ owned shares in GlaxoSmithKline and has received consultancy fees from Sanofi-Synthelabo (2002) and Roche (1997 to 1999). All other review authors have no conflicts to declare.
Sources of support
- ASL 19 and 20, Piemonte, Italy.
- Ministry of Defence, UK.
- NHS Dept of Health Cochrane Incentive Scheme, UK.
Medical Subject Headings (MeSH)
Drug Industry; Influenza A virus; Influenza B virus; Influenza Vaccines [adverse effects; *therapeutic use]; Influenza, Human [*prevention & control; virology]; Publication Bias; Research Support as Topic
MeSH check words