The emergence of the 2009 pandemic influenza virus led to the development and manufacture of a new generation of monovalent influenza vaccines – some of which employed new adjuvants. Many countries implemented pandemic vaccine programmes targeted at new risk groups, in particular children [Reference Wichmann1–Reference Hardelid3]. A number of observational studies have since shown the pandemic influenza vaccines (PIV) administered during the pandemic had high levels of vaccine effectiveness (VE) against various end-points [Reference Andrews4]. Early mid-season analyses from 2010/11, including from the UK, have suggested some residual protection from the adjuvanted pandemic vaccine 12 months later in a number of settings [Reference Pebody5] as have some end-of-season analyses [Reference Jimenez-Jorge6].
A number of countries are considering the potential introduction of routine childhood seasonal influenza vaccine programmes. Uncertainties continue regarding the potential effectiveness of seasonal influenza vaccines in this particular age group – such information will be critical to inform estimates of the potential future impact and cost-effectiveness of implementing such vaccine programmes.
The 2010/11 influenza season was characterized by the re-introduction of the normal, unadjuvanted trivalent seasonal influenza vaccine (TIV) after the 2009 pandemic. The 2010/11 unadjuvanted TIV, as recommended by WHO, included the influenza A(H1N1)pdm 2009 strain , with the UK 2010/11 vaccination programme starting in autumn 2010 and reaching a final uptake of 50·4% in clinical at-risk groups aged 6 months to 65 years.
The UK experienced intense influenza A(H1N1)pdm 2009 transmission during the 2010/11 season, with later co-circulation of influenza B. This provided the opportunity to undertake mid-season estimate for VE using the established swab-negative case-control approach [Reference Skowronski8, Reference Fleming9]. These mid-season estimates against influenza A (H1N1)pdm 2009 infection up to January 2011 have been published [Reference Pebody5], demonstrating an effectiveness of 34% for monovalent vaccine only given in 2009/10, 46% for trivalent 2010/11 vaccine only and 63% if vaccinated with both vaccines.
This present study presents the end-of-season VE for the 2010/11 TIV in preventing confirmed influenza A(H1N1)pdm 2009 and influenza B infection in both children and adults. It also examines in further detail the potential protection from vaccination with monovalent A(H1N1)2009 vaccine administered the previous season and finally the potential accuracy of mid-season VE estimates.
Study population and period
Data was derived from five primary-care influenza sentinel surveillance schemes in England (two schemes), Northern Ireland, Scotland and Wales. Details of the Royal College of General Practitioners (RCGP), Health Protection Agency (HPA) Regional Microbiology Network (RMN), Public Health Wales and Health Protection Scotland (HPS) swabbing schemes have been presented previously [Reference Hardelid3]. The Public Health Agency of Northern Ireland operated a scheme with 37 practices in 2010/11, covering 11·6% of the registered population.
The study period ran from 1 September 2010 to 17 March 2011. Cases were defined as persons presenting during the study period in a participating practice with an acute influenza-like illness (ILI) who were swabbed and then tested positive for influenza A(H1N1)pdm 2009 or influenza B. ILI was defined as an acute respiratory illness with fever or complaint of feverishness. Controls were individuals presenting with ILI in the same period that were swabbed and tested negative for influenza. Individuals testing positive for other influenza A types were dropped.
A standardized questionnaire collected demographic, clinical and epidemiological information from cases and controls including date of birth, sex, underlying clinical risk group, date of onset of respiratory illness, date of specimen and influenza vaccination status for 2010/11 and previous season with vaccination dates was completed by the GP during the patient consultation for their respiratory illness. Vaccination data were derived from the patient's medical record.
Laboratory confirmation was undertaken using real-time polymerase chain reaction (RT–PCR) assays for circulating influenza A viruses, influenza B viruses and other respiratory viruses [Reference Ellis10, Reference Gunson11]. Samples in England were sent to the HPA Microbiology Services, Colindale (RCGP scheme) or one of the regional HPA microbiology laboratories (RMN scheme). Samples in Wales were sent to the Public Health Wales Specialist Virology Centre and in Scotland to the West of Scotland Specialist Virology Centre (HPS scheme) for molecular testing. In Northern Ireland samples were sent to the Regional Virus Laboratory, Belfast.
To assess VE based on monovalent H1N1 2009 vaccine and 2010/11 TIV status, a four-level variable was defined as previously [Reference Pebody5] with the following categories:
(1) unvaccinated in both years;
(2) receipt of monovalent 2009 PIV in 2009/10 but not in receipt of 2010/11 TIV;
(3) receipt of either PIV in 2010/11 (provided to certain risk groups) or TIV in 2010/11 or both, but not vaccinated in 2009/10;
(4) receipt of PIV in 2009/10 and TIV in 2010/11, or received first dose of PIV in 2009/10 and second dose of PIV in 2010/11.
As numbers were low (five persons had two doses), those that had received two doses of PIV in 2009/10 were not analysed separately from those who received only one dose. Persons were defined as vaccinated if date of vaccination with 2010/11 TIV or PIV was ⩾14 days before onset of illness. Those in whom the period between vaccination and onset of illness was <14 days were excluded, as immunity is unknown. When assessing VE of PIV or combination of PIV and 2010/11 TIV, only those with known vaccination status were included. To assess VE of just 2010/11 TIV, individuals could still be included if PIV status was not known. If the date of vaccination was missing, as the 2010/11 campaign occurred before influenza circulation, it was assumed that 2010/11 TIV vaccination was ⩾14 days before onset date. For PIV, if date of vaccination was missing, it was assumed the person was vaccinated in 2009/10. If date of onset of symptoms was missing then the date was assumed to have been 4 days prior to the date the swab was taken (the median interval based on the observed data). Respiratory samples with a delay of >29 days between onset of illness and sample collection were excluded as the sensitivity of the PCR test is less effective for long intervals between onset and sampling. A sensitivity analysis was also undertaken censoring at 7 days between onset of illness and sample collection.
VE was estimated as 1 – (odds ratio) using multivariable logistic regression models with influenza A(H1N1)pdm 2009 or influenza B PCR results as outcomes and seasonal or pandemic vaccination status as the linear predictor. In the analyses evaluating VE in preventing influenza A(H1N1)pdm 2009 infection, samples positive for influenza B were excluded and vice versa. Age (coded into five standard age groups: <5, 5–14, 15–44, 45–64, ⩾65 years), gender, clinical risk group, surveillance scheme (HPS, RCGP, RMN, NI, Wales) and date of sample collection (month) were investigated as potential confounding variables.
To assess the impact of missing vaccination history (or date of vaccination) and to allow inclusion of the risk group variable without dropping many individuals, multiple imputation was used in a sensitivity analysis. The analysis, which also imputed missing data for gender and age, was performed using the multiple imputations chained equation (MICE) package in R (R version 2.13.0, R Foundation for Statistical Computing, 2011). Additional variables, which had no missing data, used in the imputation were flu virology status, month swab was collected and surveillance scheme. All variables were included as possible predictors of the missing data and logistic regression was the model for gender and risk group, proportional odds regression for age group, and polytomous logistic regression for the combined seasonal/pandemic vaccine status. Five separate imputation datasets were imputed and the model estates combined [Reference van Buuren and Groothuis-Oudshoorn12].
All other statistical analyses were performed in Stata version 10 (StataCorp, USA).
A total of 7797 individuals were swabbed during the study period. Thirty-nine were dropped as they were positive for influenza H3 or influenza A but not influenza A(H1N1)pdm 2009. Of the rest, 4418 (56·9%) were collected from the RCGP scheme, 1902 (24·5%) from the HPS scheme, 884 (11·4%) from the RMN scheme, 190 (2·4%) from the Public Health Wales Scheme and 364 (4·7%) from the PHA Northern Ireland Scheme. Table 1 summarizes those individuals excluded because of missing information. Vaccine date was unknown for 25 individuals given TIV and 45 given PIV. Although date of onset was missing for 801 (10·3%) individuals, these were included with onset date defined as swab date minus 4 days.
TIV, Trivalent seasonal influenza vaccine.
The demographic and epidemiological characteristics of cases and controls are summarized in Table 2. A total of 848 individuals had received 2010/11 TIV and 616 monovalent PIV. Whereas very few children received 2010/11 TIV (14 aged <5 years and 27 aged 5–14 years).
HPS, Health Protection Scotland; RCGP, Royal College of General Practitioners' surveillance scheme; RMN, Health Protection Agency (HPA) Regional Microbiology Network; TIV, trivalent seasonal influenza vaccine.
* Five cases positive by both H1N1 (2009) and influenza B are shown in this column.
† Missing onset date was calculated as 4 days prior to sample date.
Model fitting for VE estimation
When estimating vaccine effects, age group, gender, time period and surveillance scheme were adjusted for in a multivariable logistic regression model. Although all these variables were significantly associated with having a positive swab, only age group and time period were confounders for the vaccine effects.
Risk group was missing for 2270 (29%) out of 7758 samples, and was therefore not included in the final model. If risk group was included, or if multiple imputation was used, the VE estimates remained similar. Tables 3 and 4, as well as Figures 1 –3 show VE estimates against influenza A(H1N1)pdm 2009 and influenza B according to vaccination status and, in the figures, by age group and scheme.
CI, Confidence interval; TIV, trivalent seasonal influenza vaccine.
* Adjusted for age group, gender, time period and surveillance scheme.
CI, Confidence interval; TIV, trivalent seasonal influenza vaccine.
* Adjusted for age group, gender, time period and surveillance scheme.
VE against influenza A(H1N1)pdm 2009 infection
The adjusted VE estimates against influenza A(H1N1)pdm 2009 increased from 28% [95% confidence interval (CI) −6 to 51] for PIV only in 2009/10; to 55% (95% CI 31–71) for vaccination only in 2010/11 mainly with 2010/11 TIV; to 60% (95% CI 39–73) if vaccinated in both seasons with PIV then mainly with 2010/11 TIV (Table 3). Persons who had received vaccination in both 2009/10 and 2010/11 or just in the 2010/11 season had a significantly higher VE compared to persons who received PIV only in 2009/10 (Wald test P = 0·012, 0·002 for just 2010/11, both respectively). There was no difference in VE between those vaccinated in both seasons (PIV in 2009/10 and mainly TIV in 2010/11) and those just vaccinated (mainly with TIV) in 2010/11 (P = 0·58). The VE for 2010/11 TIV, irrespective of previous PIV status, was 56% (95% CI 42–66) (Table 4).
For 2010/11 TIV, there was no evidence VE differed significantly by age group (P = 0·16). VE was 87% (95% CI 45–97) in the <5 years age group and 84% (27–97) in the 5–14 years age group (Fig. 1a). The point estimate was lower in older age groups (Fig. 1a). For PIV there was no significant difference in VE estimates by age when using the five age groups (P = 0·12) but a difference was found when splitting age as <5 and ⩾5 years (P = 0·04) (Fig. 2). This age division was chosen due to the different strategy for the <5 years age group who were all recommended the monovalent vaccine during 2009/10. This gave VE estimates of 72% (95% CI 15–91) for the <5 years age group and 10% (95% CI −36 to 41) for the ⩾5 years age group. No significant difference in VE was observed by surveillance scheme (Fig. 3).
In a sensitivity analyses of VE against influenza A(H1N1)pdm 2009, censoring samples taken >7 days after symptom onset gave slightly higher VE estimates with broader confidence intervals: the adjusted VE for those vaccinated in the previous season (2009/10) was 39% (95% CI 5–61); for those vaccinated only in the current season (2010/11) VE was 61% (95% CI 35–77) and for those vaccinated both seasons VE was 65% (95% CI 43–79). The VE for 2010/11 TIV irrespective of PIV status increased to 63% (95% CI 48–73). With multiple imputation, the results were similar to those found without imputation, but did allow adjustment for risk group: the imputed VE for vaccination in both seasons was 59·0% (95% CI 36·4–73·6), which is similar to the VE estimate given in Table 3.
Adjusting for month had a large effect on VE for those vaccinated in 2009/10, decreasing it from 53% (crude) to 36% after adjustment.
VE against influenza B infection
There was no evidence of any VE of PIV against influenza B (Table 3). The VE analysis therefore only considers 2010/11 vaccination status. The adjusted VE was 57% (95% CI 42–68). Censoring samples taken >7 days after symptom onset increased VE to 61% (95% CI 45–72).
There was no evidence VE varied by age group (no significant age–vaccine interaction, likelihood ratio test, P = 0·46). VE in the <5 years age group was 47% (95% CI −337 to 93) and in 5–14 years age group it was 75% (95% CI 32–91) (Fig. 1b). There was, however, evidence that VE against B infection varied for both surveillance scheme (Fig. 3) and strain (B/Victoria and B/Yamagata). The majority (93%) of B viruses circulating in England and Scotland in 2010/11 were of B/Victoria lineage. Of these B/Victoria isolates, 7/268 cases had been vaccinated with 2010/11 TIV giving an adjusted VE of 78% (95% CI 51–91). This compared to 3/15 B/Yamagata isolates, giving an adjusted VE of −34% (95% CI −448 to 68).
This observational study of influenza VE in the UK has several key findings: first, vaccination with the 2010/11 TIV provided significant protection against laboratory-confirmed infection for both influenza A(H1N1)pdm 2009 and influenza B; second, there was evidence of significant protection with 2010/11 TIV for school children for both influenza A(H1N1)pdm 2009 and B infections; third, immunization with A(H1N1)2009 vaccine in 2009/10 followed by TIV in the 2010/11 season provided similar protection against confirmed influenza A(H1N1)pdm 2009 infection to just receiving TIV in 2010/11; fourth, protection against influenza A(H1N1)pdm 2009 infection in 2010/11 following vaccination with PIV 1 year previously in 2009/10 was reduced, although for children aged <5 years VE was maintained; fifth, there was evidence of strain-specific variation in VE for confirmed influenza B infection and finally the findings reinforce earlier published mid-season estimates of VE [Reference Pebody5].
The swab-negative case-control study design is now a well established approach to estimate influenza VE, with several studies published on the methodology [Reference Ferdinands13, Reference Orenstein14]. The potential limitations presented in this paper have been outlined previously and relate to convenience sampling of biological specimens resulting in the potential for selection bias; missing data items and lack of information on risk status [Reference Hardelid3]. Furthermore, for children aged <13 years the analysis is based on having received one or more doses of vaccines, as we were not able to disentangle those who had received the recommended two doses from those who only received one dose. This could potentially underestimate the VE of the recommended two-dose schedule in this age group. We have presented various sensitivity analyses to attempt to address the potential impact on VE of missing data items. In addition, we have applied multiple imputation methods, which have allowed us to adjust for risk status and demonstrate that this was not an important confounding variable in this particular analysis.
This study confirms that 2010/11 TIV was effective in protecting against both confirmed influenza A(H1N1)pdm 2009 and influenza B infection in people consulting their GP with an acute respiratory illness. This also confirms published findings from several settings including an earlier mid-season UK analysis. These all demonstrated that the 2010/11 seasonal influenza vaccine provided moderate protection against influenza A(H1N1)pdm 2009 infection [Reference Pebody5, Reference Kissling and Valenciano15–Reference Skowronski18]. This VE is consistent with studies on the effectiveness of TIV in the pre-pandemic era, which suggests that the vaccine is protective against the circulating influenza strains [Reference Skowronski8, Reference Fleming9].
Among children we found that 2010/11 TIV was significantly effective in both the <5 and 5–14 years age groups in preventing influenza A(H1N1)pdm 2009 infection and in the 5–14 years age group only for protecting against confirmed influenza B infection. Only a small number of studies have been published on the effectiveness of unadjuvanted seasonal influenza vaccines mainly in pre-school children [Reference Szilagyi19–Reference Eisenberg21] and this current study provides further useful evidence of the effectiveness of unadjuvanted seasonal TIV in children of all ages. This will be helpful in informing decisions around possible introduction of seasonal influenza vaccines into childhood – particularly for school-aged children.
Although recently published work has demonstrated that the pandemic influenza A(H1N1) 2009 vaccine had good effectiveness in 2009/10 in preventing confirmed influenza A(H1N1)pdm 2009 infection during the pandemic period in a wide range of geographical settings [Reference Valenciano2, Reference Hardelid3], this present study indicates that overall adjuvanted pandemic vaccine protection does not last across the season to the following year. This corroborates findings from our earlier mid-season analysis in the UK [Reference Pebody5] and Spain [Reference Castilla17]. There was, however, evidence in the current study that PIV protection was maintained in children aged <5 years. It is important to note that this particular target group was broader compared to the older age groups. In the UK all healthy children up to age 5 years, not just those in a clinical risk group, were targeted with PIV in 2009/10 reaching an uptake of 23·6%. Furthermore, this part of the pandemic vaccine programme was delivered later in the 2009/10 season with most vaccine given in the spring of 2010 compared to those with an underlying clinical risk factor aged <65 years, where the programme was started in autumn 2009 and reached an uptake of 35·4%. Both these factors may explain why protection is maintained in this age group compared to older age groups. Our study is also congruent with other studies that have suggested pandemic influenza vaccine protection is lower in older children and adults in the 2009/10 season [Reference Griffin22], with others suggesting that antibodies persist after 1 year in children after vaccination with adjuvanted PIV [Reference Walker23].
The lower VE estimate against influenza B for the Yamagata lineage strain is also of some interest. In any season there may be several B variants within the main influenza types/subtypes co-circulating. It would be expected that the calculated VE would vary dependent on how well the circulating strains match the vaccine components. The trivalent seasonal vaccine, however, contained only a single influenza B strain in 2010/11; a B/Brisbane/60/2008-like virus. During 2010/11, both Victoria lineage (B/Brisbane/60/2008-like viruses) and Yamagata lineage (B/Bangladesh/3333/2007-like viruses) influenza B viruses were identified as circulating by molecular analysis or antigenic typing. There is evidence of some variation within the UK in the relative contribution of each of these influenza B lineage strains; in England 7% of influenza B viruses from any source were characterized as B/Bangladesh/3333/2007-like, while in Scotland although this figure was higher (21%), it is based on a much smaller number of samples submitted for molecular typing rather than antigenic typing.
Although the earlier published mid-2010/11 season UK estimates were reasonably accurate, one of the main changes [Reference Pebody5] in this end-of-season analysis is that the possible dose–response relationship that persons who received vaccination in both 2009/10 and 2010/11 seasons had a non-significantly higher VE compared to persons who received vaccine only 2009/10 is now significant in the current study. This highlights the importance of annual re-vaccination with influenza vaccine.
The current estimates are also now more precise than the mid-season analysis, in particular for VE against influenza B. Thus, although the mid-season estimate provided some important early findings, this end-of-season study provides the definitive results and also allows an age-specific analysis.
In conclusion, this end-of-season study provides important evidence that the 2010/11 season's TIV provided protection against infection to both strains of influenza circulating in the 2010/11 season [influenza A(H1N1)pdm 2009 and influenza B] in the UK. In particular, the study provides evidence of TIV effectiveness in school-aged children, which will be an important finding in consideration of potential extension of the national programme to children. The findings also provide evidence that PIV protection wanes after 1 year except for those aged <5 years. The study reinforces the recommendation that annual re-immunization of target groups is required regardless of vaccination the previous season (including those vaccinated with an adjuvanted vaccine). Furthermore, the study confirms the potential value of undertaking a mid-season analysis to provide an early estimate of protection. This can provide key information to assist decision-making, e.g. the WHO vaccine strain selection for the following season.
We are grateful to the many primary-care physicians in England, Wales, Northern Ireland and Scotland who supplied clinical information on their patients; to the staff of the HPA Respiratory Virus Unit, the HPA regional microbiology laboratories, Public Health Wales Specialist Virology Centre, the West of Scotland Specialist Virology Centre and the Regional Virus Laboratory, Belfast who undertook analysis of specimens. We thank the staff of the Health Protection Agency, RCGP, Public Health Wales, Public Health Agency Northern Ireland and Health Protection Scotland teams who coordinate the GP schemes, in particular Joy Field from the HPA; Catherine Frew and Rory Gunson from WoSSVC and Arlene Reynolds, Rachael Darroch and Drew Palmer from HPS for overseeing data collection.
This work has been undertaken as part of a European Centre for Disease Control and Prevention (ECDC)-funded project ‘I-MOVE’, which is coordinated by Epi-Concept (http://www.epiconcept.fr/).
DECLARATION OF INTEREST
D. M. Fleming has received funding to attend influenza-related meetings and has received consultancy fees from influenza vaccine manufacturers who might have an interest in the submitted work in the previous 3 years. In addition, The Virus Reference Department of the Health Protection Agency receives funding from a variety of vaccine manufacturers who might have an interest in the submitted work.