Case control study for measuring influenza vaccine effectiveness in Portugal - Season 2010-11- Final report

• Autor(a) principal: Nunes, Baltazar
• Data de Publicação: 2011
• Outros Autores: Pechirra, Pedro, Machado, Ausenda, Falcão, Isabel, Gonçalves, Paulo, Conde, Patricia, Batista, Inês, Guiomar, Raquel, Marinho Falcão, José
• Tipo de documento: Relatório
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.18/448
• topic: Cuidados de Saúde; Estados de Saúde e de Doença; Influenza; Vaccine Effectiveness; Test Negative Design; Screening Method

Every year, influenza virus is responsible for epidemics that affect human health causing respiratory infections that could lead to serious health complications of individuals belonging to risk groups, as well as on the functioning of health services. In order to mitigate influenza impacts, vaccination has been one of the main measures, being recognized its role in reducing the risk of developing the disease and the occurrence of their complications. Thus, since the vaccine is reformulated every season estimating the influenza vaccine effectiveness (VE) every season and in an early stage is of major importance to support public health decisions. Since 2008-2009, Portugal has been participating in I-MOVE project that aims to estimate seasonal and pandemic vaccine effectiveness during and after the influenza season. Last season, 2010-2011, Portugal once again joined the I-MOVE multi-center case control study (with the national VE study- Euroeva) together with Spain, Ireland, France, Italy, Romania, Hungary and Poland, using a common protocol and with the objective of estimate the 2010-11 seasonal influenza vaccine effectiveness respectively in the elderly (65+) and in all age groups. Additionally, using information on 2010-11 seasonal vaccine coverage in the population, it has been also proposed to estimate 2010-11 seasonal vaccine effectiveness using the screening method. Material and Methods Two different approaches were used so as to estimate vaccine effectiveness: a) For the test negative design (TND), a case-control approach was used, where laboratory confirmed influenza cases (ILI+) were compared to laboratory negative influenza ILI patients (ILI-). On a weekly basis, each GP selected systematically ILI patients (two per week from all ages and all ILI patients with 65 years and more) using the EU ILI case definition. Data on confounding factors and effect modifiers was collected using a standardized questionnaire. VE was estimated as one minus the odds ratio of being vaccinated in cases versus controls adjusted for confounders by logistic regression. b) For the screening method, the 2010-11 seasonal vaccine coverage was compared between a sample of ILI cases and a sample of ILI cases laboratory confirmed for influenza with the vaccine coverage estimated in the general population. ILI cases and ILI Positive cases were the same as the one used in the TND. Vaccine coverage in the population was obtained from a sample of 1074 households stratified by region (homogeneous allocation) selected from a dual sample frame – random digit dialing mobile and landline phones (ECOS sample). The relevant information was collected by CATI (Computer Assisted Telephone interview) with the same questionnaire – one respondent by household (proxy for the rest of the household members). VE was estimated by comparing the proportion of cases vaccinated to the vaccine coverage in the source population using the Orenstein formula and the Farrington method to adjust for confounders. For both methods an ILI patient was considered vaccinated if he/she had received one dose of the vaccine at least 14 days prior onset of symptoms. Data analysis comprised ILI positive cases selected between week 45 of 2010 and week 11 of 2011. Results In Portugal, the 2010-2011 influenza season was characterized by a mixed circulation of influenza virus. In the early beginning of the season, B type virus dominated the season, until week 1 where A(H1N1)2009 virus started to dominate. Also in circulation, but in a minor proportion was the A(H3) virus. a) Test negative design results: Considering TND results, 58 GP’s accepted to participate in the study, with 60% participation rate (35 GP’s effectively participated in the study by selecting patients). Excluding 33 ILI cases (for not meeting the inclusion criteria) the final sample for analyses consisted on 253 ILI patients with a high positive rate (57%). Among cases, 73 were positive for B virus, 69 for A(H1N1)2009 and 2 for A(H3). For analysis purposes three groups of cases were defined: All influenza, Influenza B and Influenza A(H1N1)2009 that were compared to ILI cases that tested negative for influenza virus (109 Controls). After adjustment for age group, pandemic and seasonal vaccine 2009-10 season, any chronic disease, target group, GP visits and month of onset, VE point estimates were: • 58% (CI95% -61 – 89) for All influenza • 34% (CI95% -98 – 97) for Influenza A(H1N1)2009 and • 75% (CI95% -255 – 88) for Influenza B. Nevertheless, no statistical significance was obtained for either the analysis. b) Screening method results In season 2010-2011, 903 households of the ECOS-sample were interviewed (which corresponds to 2684 individuals). The final estimated vaccine coverage was 17.5% (CI95% 15.1-20.3), with a gradient evolution since September 2010 (3.5%) till December 2010 (16.6%). Crude and adjusted VE (using the Farrington method) estimates were computed for medically attended ILI cases and ILI influenza positive cases. Overall results indicate that after adjustment for confounding (age group and presence of chronic diseases), VE point estimates decreased from 47.0% to 33.3% in ILI cases and from 70.1% to 63.7% in ILI positive cases. Due to the small sample size, no VE estimates were computed for the individuals of the vaccination target group, i.e., individuals with 65 years and more and the ones with at least one chronic disease. Conclusions Overall results obtained by the Euroeva study indicate that crude 2010-11 seasonal VE estimate against medically attended influenza was 79% (CI95% 43-94) and 70% (CI95% 32-87), respectively for the TND and screening method. After adjustment the respective VE estimates decreased: 58 (CI95% -61-89) and 64% (CI95% 17-84). These results were in accordance to the up to now published results (42-65%). The TND study was also able to provide strain specific 2010-11 seasonal VE estimates: influenza B, crude VE=87% (CI95% 41-99) and adjusted VE=75% (CI95% -98-97) and for influenza A(H1N1)2009, crude VE=74% (CI95% 14-94) and adjusted VE=34% (CI95% -254-88). These results suggest that the 2010-11 seasonal VE was lower than the monovalent A(H1N1)2009 VE estimated by the IMOVE multicenter study in the season 2009-10: 72% (CI95% 46-86). Our study was unable to estimate VE for specific seasonal vaccine target groups. This result enhances, as in previous studies, the unavoidable need for pooling data from network of VE studies with common protocol as IMOVE. Recommendations The main recommendations focused on: • To calculate sample size taking into consideration: o the context of the multicentre case-control study: minimum sample size per site in order to assure a minimum homogeneity for pooled analysis; o different expected VE point estimates, i.e. for low, medium and high VE; o minimum set of factors for stratified analysis; o the adjustment for confounders. • To increase sample size, mainly in the elderly population (aged 65 years or more); • To increase the total number of participating GP’s in the study by exploring other sources of GP’s recruitment; • To study the inclusion of the population based vaccine coverage uncertainty in the screening method; • To explore with participating GP’s the best way to obtain estimates of the Euroeva ILI sample fraction. Finally we also recommend continuing the harmonization of the study designs between participating countries with the multi-centre study objective.