Influenza vaccine effectiveness assessment in the UK Nick Andrews, Statistics Unit, Health Protection Agency 1
Outline Introduction The UK swabbing schemes Assessment by the test negative case control design 2010/11 example Results 2004/06 to 2011/12 Cohort studies in England and Scotland Comments 2
Introduction In the UK trivalent influenza vaccine (TIV) has been used for many years in those aged 65 years and over, in clinical risk groups and health care workers In 2009/10 the H1N1 adjuvanted monovalent influenza vaccine (MIV) (Pandemrix) was used targeted at risk-groups and all children <5 years. In 2014/15 introduction of live attenuated nasal vaccine programme for all 2-16 year olds is planned. Vaccine uptake across the UK for 2011/12 was 68-77% in those aged 65+ and 50-70% in risk groups. 3
Swabbing Schemes Swabbing is performed by GPs in five schemes across the UK Royal College of General Practitioners England (~100 GPs) HPA Region Microbiology Network England (45 GPs) Health Protection Scotland (90 GPs) Public Health Wales (44 GPs) Public Health Agency of Northern Ireland (37 GPs) Total population covered by the GPs is about 2.5 million 2004/05 to 08/09 RCGP data used to estimate VE by test negative case control (TNCC) Since 2009/10: Data pooled across the UK schemes for TNCC Mid season estimates produced RCGP /Scotland have done cohort studies. RCGP also do within season screening assessment 4
Test-negative case-control design (TNCC) Started to be used for Influenza VE in about 2005 by various groups (UK, Canada) Individuals are tested for infection with a certain vaccine preventable disease. Vaccination history is also ascertained. Sample Sample Tested + Tested Vaccinated a b Unvaccinated c d VE = 1 (a/c)/(b/d) Analysis is as with a classic case control study (logistic regression) 5
Influenza VE using the TNCC in the UK Individual presents to GP with ILI Results and epidata go into a database. For the RCGP scheme lab results coded back into GP database by GPs GP is part of a network of swabbing GPs Sample tested by PCR for flu (and type of flu). Also testing for other viruses (RSV) Oct-March: GP swabs all or a subset of ILI patients Sample sent to National Laboratory along with a form with vaccination history, date of onset, age, gender and risk group. 6
Swab and instructions 7
Key Potential Confounding variables Age (certainly a confounder) Period (certainly a confounder) Region/country (possibly) Being in a risk group for vaccination (possibly) 8
Example of swabs taken in 2010/11 3500 3000 2500 timing of vaccine uptake in the general population negative B H1N1(2009) swabs 2000 1500 1000 500 0 sept oct nov dec jan feb mar month 9
Description of swabs taken, 2010/11 10
Results for 2010/11 (Pebody 2012*) H1N1 B Negative Vaccinated (>14d) 81 58 604 Unvaccinated 1626 1064 3693 Crude TIV VE for H1N1pdm(09) = 70% Crude TIV VE for B = 67% Crude estimates not too meaningful adjustment for age and period makes the most difference Adjusted TIV VE against H1N1pdm(09) = 56%, 95% CI (42-66%) Adjusted TIV VE against B = 57% (42-68%) Adjustment for risk group / scheme made very little difference to the estimates We also looked at VE of monovalent pandemic vaccine in 2009/10 which was 72% (15-91%) in under 5 year olds and 10% (-36 to 41%) in over 5 year olds. * Pebody et al. Flu vaccine effectiveness in UK, 2010/11, Eurosurveillance 2012 11
VE estimates against H1N1 by age group, 2010/11 wide 95% CIs * Pebody et al. Flu vaccine effectiveness in UK, 2010/11, Eurosurveillance 2012 12
Results since 2004/05 (except 2009/10 when TIV VE was 0% and MIV approx 70%) No data No data 13
Methodological Issues Bias from lack of sensitivity - Orenstein (IJE 2007) considered this and gave formulas. GP selection of ILI cases to swab. We ask for swabs irrespective of vaccination history. No reason to suspect biased selection and GP does not know test result when selecting who to swab. Should be good matching on propensity to consult. 14
Part of i-move Cohort studies RCGP and Scotland using Primary Care data extracts ILI, LRTI, ARTI end points examined RCGP swabbing results could also be used as laboratory results are written back into the GP record. Results show sensible VE against ILI but ARTI/LRTI not specific enough (VE 0 or negative). Issues of bias due to propensity to consult need to look at consulting frequency. 15
16
2010/11: If true VE is 55% and 50% of ILI swabs are positive then expect to observe 27% VE 17
Comments Using surveillance data and the TNCC design has consistently produced results that are sensible and in line with other studies. Pooling UK data form similar schemes has increased power with no evidence of estimates differing by scheme. Only adjustment for age and period seems necessary. For the cohort analysis more issues with bias from propensity to consult but ILI outcome seems to give sensible results 18
Acknowledgements HPA: Richard Pebody, George Kafatos, Pia Hardelid, John Watson, Liz Miller, Maria Zambon, Praveen Sebastian Pillai, Joanna Ellis. RCGP: Douglas Fleming, Hayley Durnall, Michele Barley Scotland: Jim McMenamin, Chris Robertson, William Carman, Kim Kavanagh Wales: Simon Cottrell, Catherine Moore, Daniel Rhys Thomas Northern Ireland: Brian Smyth, Catriona Kearns i-move collaborators London School of Hygiene and Tropical Medicine: Punam Mangtani All participating general practices in the schemes for provision of the data. 19
References for UK studies Pebody, R. G., et al. (2012). Age-specific vaccine effectiveness of seasonal 2010/2011 and pandemic influenza A(H1N1) 2009 vaccines in preventing influenza in the United Kingdom. Epidemiology and infection 2012. Pebody, R., et al. (2011). Effectiveness of seasonal 2010/11 and pandemic influenza A(H1N1)2009 vaccines in preventing influenza infection in the United Kingdom: mid-season analysis 2010/11. Euro surveillance, 16(6), id 19791. Simpson CR, et al. Effectiveness of H1N1 vaccine for the prevention of pandemic influenza in Scotland, UK: a retrospective observational cohort study. Lancet Infect Dis. 2012 Sep;12(9):696-702 Kavanagh K, et al Assessment of the variability in influenza A(H1N1) vaccine effectiveness estimates dependent on outcome and methodological approach. PLoS One. 2011;6(12):e28743. Hardelid, P et al. (2012). Effectiveness of trivalent and pandemic influenza vaccines in England and Wales 2008-2010: Results from a cohort study in general practice. Vaccine, 30(7), 1371 8. Andrews, N.,et al. (2011). Age-specific effectiveness of an oil-in-water adjuvanted pandemic (H1N1) 2009 vaccine against confirmed infection in high risk groups in England. The Journal of infectious diseases, 203(1), 32 9. Hardelid, P, et al. (2011). Effectiveness of pandemic and seasonal influenza vaccine in preventing pandemic influenza A(H1N1)2009 infection in England and Scotland 2009-2010. Euro surveillance, 16(2), id19763. Fleming, D. M, et al. (2010). Estimating influenza vaccine effectiveness using routinely collected laboratory data. Journal of epidemiology and community health, 64(12), 1062 7. 20