Effect of previous and current vaccination against influenza A(H1N1)pdm09, A(H3N2), and B during the post-pandemic period 2010-2016 in Spain

Background Recent studies suggest that the protective effect of the current influenza vaccine could be influenced by vaccination in previous seasons. We estimated the combined effect of the previous and current influenza vaccines from the 2010–2011 season to the 2015–2016 season in Spain. Methods We performed a test-negative case-control study in patients ≥9 years old. We estimated the influenza vaccine effectiveness (IVE) against influenza A(H1N1)pdm09, A(H3N2), and B virus. Results We included 1206 influenza A(H1N1)pdm09 cases, 1358 A(H3N2) cases and 1079 B cases. IVE against A(H1N1)pdm09 virus in the pooled-season analysis was 53% (95% Confidence Interval (CI): 21% to 72%) for those vaccinated only in the current season and 50% (95%CI: 23% to 68%) for those vaccinated in the both current and previous seasons. Against the influenza A(H3N2) virus, IVE was 17% (95%CI: -43% to 52%) for those vaccinated only in the current season and 3% (95%CI: -33% to 28%) for those vaccinated in both seasons. Regarding influenza B, we obtained similar IVEs for those vaccinated only in the current and those vaccinated in both seasons: 57% (95%CI: 12% to 79%) and 56% (95%CI: 36% to 70%), respectively. Conclusion Our results suggested no interference between the previous and current influenza vaccines against A(H1N1)pdm09 and B viruses, but a possible negative interference against A(H3N2) virus.


Introduction
Influenza represents a public health problem, therefore it is generally recommended that population groups at risk of severe complications or death are vaccinated: the elderly (above 64 years old), those younger than 65 with chronic conditions, pregnant women or persons at risk due to their profession [1; 2].The influenza vaccine is the main preventive measure available, but as influenza viruses undergo frequent changes in their surface antigens, it needs re-formulation each year, with the aim of antigenically matching the circulating strains [3].
Since the 2008-2009 season, have estimated the influenza vaccine effectiveness (IVE) in Spain using the cycEVA study (case-control study for the influenza vaccine effectiveness in Spain), which is the Spanish component of the I-MOVE project for monitoring the IVE in Europe. Ever since, within the cycEVA study we have continuously employed a case-control test-negative design. This is a method that is used worldwide as one of the most appropriate to estimate IVE, because it minimizes the habitual biases of observational studies [4][5][6].
A series of factors influenced the yearly-observed IVE estimates. Apart from the antigenic match with the vaccine-circulating strain, recent studies have suggested that the protective effect of the current season´s influenza vaccinecould be also influenced by influenza vaccination in previous seasons (N.B. This article refers to "current" in the sense of the season under analysis rather than the present moment). While several papers did not detect any evidence for decreasing protection with repeated vaccination [7;8], other authors described an effect between the previous season´s vaccination and the current one. This was observed in a single season or across various seasons, with some studies suggesting a negative interference between the previous and the current season´s influenza vaccine [9][10][11][12].
We present the IVE estimates for the current season´s influenza vaccine, as well as the combined effect of the previous and current vaccine, as obtained during six post-pandemic seasons in Spain, 2010-2011 to 2015-2016, with the cycEVA study. We also interpret these results in the context of the similarities between the circulating strains and vaccine strains, both in current and previous season.

Study setting, design and data collection
We have used data obtained from the cycEVA study, from the 2010-2011 to 2015-2016 influenza seasons. The cycEVA study is an observational case-control study to monitor influenza vaccine effectiveness (IVE) in Spain, which is conducted within the framework of the Spanish Influenza Sentinel Surveillance System (SISSS) [13]. During the study period, between five and eight out of the 17 regional sentinel networks participated each season.
The methodology of the cycEVA study has been described previously in the literature [14;15]. Briefly, following a common European protocol [16], sentinel practitioners (SPs) reported cases of influenza-like illness (ILI) on a weekly basis according to a definition that is based on the European Commission (EU) ILI case definition [17]. They systematically National Centre of Epidemiology, has the mandate to collect, analyse and disseminate surveillance data on infectious diseases in Spain. There is not direct access to the RENAVE database, but requests from third parties not belonging to the RENAVE with research purposes are solved by the National Centre of Epidemiology. In this case, the person responsible for the database of this study is responsible for the National Influenza Surveillance Program in Spain, who is also the corresponding author of the manuscript: Amparo Larrauri. Data are available upon reasonable request to Amparo Larrauri, responsible for the National Influenza Surveillance Program in Spain.
swabbed patients below 65 years old (the first two patients with influenza-like illness (ILI) who had consulted a sentinel physician each week) and all patients above 64 years old. For each recruited patient, the SPs collected information on demographic data (age, sex, sentinel network), previous and current vaccination status (including date of vaccination) and presence of a chronic condition (i.e., chronic cardiovascular, pulmonary, hepatic or renal diseases, congenital or acquired immunodeficiency and diabetes mellitus), as well as pregnancy status, and obesity (defined as a body mass index BMI !40 kg/m 2 ).
For all six seasons, we used a case-control test-negative design. We defined cases as ILI patients who had a reverse transcription polymerase chain-reaction (RT-PCR) and/or cell culture test, positive for influenza (A(H1N1)pdm09, A(H3N2) or B); whereas controls tested negative for all influenza viruses. In each season, the study started when a sporadic circulation of influenza viruses was detected in the participant surveillance networks. We considered a patient vaccinated if he/she had received the trivalent seasonal influenza vaccine at least 14 days before the onset of ILI symptoms; patients receiving the influenza vaccine less than 14 days before symptoms' onset were considered unvaccinated, whereas those without a vaccination date were excluded from the analysis.

Data analysis
Since the objectives of this work are different from those of previous cycEVA studies [18][19][20][21], we reanalyzed data within cycEVA using the same analysis approach across all six seasons studied. We have considered only patients of nine years and above, since below this age complete vaccination status might include vaccination in the previous season; we also restricted the analysis to the ILI patients swabbed less than eight days since the onset of symptoms. For each season, we estimated the IVE against the predominant influenza virus (!60% of the total type/subtype of influenza virus circulation), or against viruses with a circulation of at least 25% of the total influenza viruses, during the period with continuous circulation of that specific type/subtype. We estimated the IVE for all those aged nine and above, the target groups for influenza vaccination; and, in a further sensitivity analysis, we restricted the analysis to those swabbed four days or less since onset of symptoms. We considered target groups as all those patients above 64 years old (60 in some sentinel networks), those with a chronic condition or other risk factor for influenza (obesity or pregnancy), or who met any other criteria to be vaccinated in Spain (such as healthcare workers or caregivers). We explored the distribution of the covariates between cases and controls using a χ 2 test. We used logistic regression models to evaluate the odds ratios (OR) with 95% confidence intervals (95%CI), adjusting for age-groups (9-14, 15-44, 45-64, !65 years), sex, chronic condition, sentinel network and week of swabbing, as well as influenza season in the pooled analysis. We estimated the IVE with the formula (1-OR)x100 for vaccination.
In the main study analysis, we estimated the effect of previous and current seasons' vaccination, against considered virus type/subtype, in pooled analysis and by each season for all study subjects. In a further refinement, we restricted the analysis to those belonging to the target groups. In the 2010-2011 influenza season, the pandemic vaccine was considered as a previous vaccination against A(H1N1)pdm09 for the purposes of the IVE. To determine current vaccination status we used the available vaccination date, while for patients vaccinated in the previous season we had only yes/no information. The effects of previous and current vaccinations were evaluated using four categories: vaccinated only in the previous season, vaccinated only in the current season, receiving both vaccines, and unvaccinated. We estimated the adjusted IVE with the 95%CI for each category, adjusting for the same variables as in the primary analysis and using the unvaccinated as a reference. Differences in the IVE between those vaccinated only in the "current" season and those receiving both vaccines were explored, using those vaccinated only in the current season as reference.
For each season we have provided the results on the genetic characterization of influenza strains, from cases notified at the national level in Spain. Strains were genetically characterized at the World Health Organization (WHO) National Influenza Centre in Madrid (Spain), by sequencing the HA1 fragment of the viral hemagglutinin gene.  (Fig 1).

Influenza seasons and characteristics of cases and controls
A total of 507, influenza A(H1N1)pdm09 cases were included for the 2010-2011 season, 303 for 2013-2014 and 396 for 2015-2016, respectively. There were less cases than controls belonging to target groups for vaccination in all A(H1N1)pdm09 influenza seasons. Compared to the controls, cases were less vaccinated in the current, as well as in the previous season (Table 1).  (Table 3).

. Characteristics of influenza A(H1N1)pdm09 cases and test negative controls in Spain
When restricting the analysis to the patients swabbed 4 days or less after the onset of symptoms, we obtained similar results compared to all population analysis (Table 4).

Previous and current vaccination analysis
IVE against A(H1N1)pdm09 showed a null or low adjusted estimate for patients receiving only the previous vaccination, with the exception of the 2010-2011 season where patients had received the previous monovalent pandemic vaccine: 65% (95%CI: -13% to 89%) ( Table 5).
The IVE point estimates for patients receiving both current and previous vaccine were higher compared to those vaccinated only in the current season in the 2010-2011 and 2013-2014 seasons (p comparison = .308 and .352, respectively), but lower in the 2015-2016 season (p comparison = .321), although the differences were not significant. The pooled analysis revealed an IVE of 26% (95%CI: -39% to 61%) for those having received only the previous vaccination, and similar moderate values around 50% for those receiving either vaccines or only the current one (Table 5).
Regarding influenza B, we found negative IVE point estimates for patients receiving only the previous season's vaccine. Having received both the previous and current season's vaccines resulted in an IVE similar to those receiving only the current one, with point estimates slightly Combined effect of previous and current influenza vaccination on influenza vaccine effectiveness in Spain   (Table 5).
When estimating the IVE against influenza A(H3N2), A(H1N1)pdm09 and B and restricting the analysis to the population group targeted for vaccination, we obtained similar results ("S1 Table").

Vaccine and circulating strains
The A(H1N1)pdm09 circulating strains for the three seasons included in the analysis matched the vaccine strain (Table 5).
In In all four seasons considered for IVE against influenza B, the previous and the current vaccine lineages were the same. Moreover they matched the current season's circulating lineage. The only exception was the 2015-2016 season, when there was a mismatch between the current vaccine (Yamagata lineage) and the circulating strains (Victoria lineage) ( Table 5).

Discussion
Clarifying the combined effect of previous and current influenza vaccination remains a complex challenge; we have tried to disentangle this effect by using data across six influenza seasons in Spain for the firs time. To better understand the role of the previous year's vaccine, we discuss our results taking also into account the match between the circulating and vaccine influenza strains.
Overall, we have found better vaccine effectiveness against confirmed A(H1N1)pdm09 and B influenza than against A(H3N2), similar results were obtained both for all patients included in our study as well as for the target groups for vaccination (Table 1).This was suggested both by the pooled analysis and for each influenza season under consideration. These results appear to be in line with previous studies performed in Spain and elsewhere, and confirm the usual poor performance of influenza vaccines against A(H3N2) [22][23][24][25].
When taking into account the previous vaccination, we have found some residual protective effect of the pandemic vaccine against 2010-2011 A(H1N1)pdm09 infection, that could be related with an adjuvant vaccine administered universally in Spain during the pandemic. Also we have seen that having received both the pandemic and the current vaccine, protected against A(H1N1)pdm09 infection in 2010-2011, an effect also described in previous studies [14;26;27]. Within each studied season, we did not find any interference between the previous season's vaccine and the current vaccine's performance, findings also confirmed by the pooled analysis. Along the same lines, a recently published study suggested an optimal vaccine protection against A(H1N1)pdm09 as individuals having received the current season's vaccine and 1-2 prior doses [28]. These results could be explained by the relative genetic stability of the A (H1N1)pdm09 influenza virus since the pandemic, leading to the inclusion of the same vaccine strain A/California/7/2009 and a continued good match with strains circulating within in each of the three studied seasons [14;18;29].
The low or null IVE estimates obtained against A(H3N2) do not always appear to be in line with the degree of matching between the vaccine and circulating strains. In 2011-2012 and 2014-2015 seasons, a mismatch was described, whereas in 2013-2014 the vaccine did match the circulating strain; however the overall IVE remained suboptimal in all three seasons (Table 5). We registered the highest protective effect against A(H3N2) (34%) in the 2011-2012 season, the first post-pandemic influenza season dominated by A(H3N2) which had Combined effect of previous and current influenza vaccination on influenza vaccine effectiveness in Spain circulating strains discordant with the vaccine strain. We also found some residual protective effect of the previous season's vaccine against A(H3N2) infection in the 2011-2012 and 2013-2014 seasons, but not in the 2014-2015 season. The current season's vaccines showed suboptimal but slightly better protection against A(H3N2) for the vaccination-targeted population than in the general population (S1 Table), findings which are in line with a previous study [30]. While previous research described a poor correlation between the IVE and the circulating-vaccine A(H3N2) strain match [31], others studies have shown a correlation of the IVE with the match between circulating and vaccine strains [32][33][34]. When taking into account the previous vaccination, our results suggest a negative interference between the previous season's vaccine and the current one for the 2014-2015 season. This effect was observed and described in the same season in Canada [35;36] and is consistent with the antigenic distance theory proposed by Smith & al. [37]. Regarding the protection against influenza B, the pooled analysis suggested no protective effect from the previous season's vaccine, consistent with another similar study [39], and no negative interference between the previous season's vaccine and current vaccine protection for all four studied seasons. By contrast, another study taking into account repeated vaccination over eight seasons in the United States found evidence of residual protection against influenza B [12]. For the 2010-2011 and 2012-2013 seasons, we have found no suggestion of negative interference between the previous and the current season's vaccination; having received both vaccines resulted in moderate protective effect (around 65%) against confirmed Influenza B, significant for the 2012-2013 season and comparable to results described elsewhere [40]. For 2014-2015 and 2015-2016 seasons, a slightly decrease in IVE point estimates was registered for those receiving both vaccines compared to those having only the current, which was more pronounced for the 2014-2015 season. The small sample size and overlapping 95%CI however, do not allow strong conclusions to be drawn with respect to negative interference from the previous vaccination. In all but one influenza season, the circulating Influenza B strain matched the vaccine strain, partially explaining the moderate effectiveness of the vaccine. The 2015-2016 season was characterized by the discordance of circulating and vaccine lineages (Victoria and Yamagata, respectively), but the IVE was still moderate, suggesting possible cross-lineage protection, previously described in other settings [10;40].
In target groups for vaccination, the higher vaccine protection against A(H1N1)pdm09 and B was provided by vaccination in both previous and current seasons, whereas a possible interference between the previous and current vaccine was also observed against A(H3N2) ( Table 5).
Our study has several limitations. The most important of which are the low vaccine coverage and small sample size, especially for the categories used when evaluating the previous vaccination effect, rendering imprecise estimations in some analysis, and reflected by wide confidence intervals. Moreover, for this analysis we could take into account the vaccination in the previous season, but we could not evaluate the repeated vaccination effect, since this information was not available. We discussed our results taking into account the match between circulating and vaccine strains, which was available for genetic characterization at national level but not in the cycEVA sentinel regions. Although differences in the distribution of circulating strains might exist, we consider these differences too small to significantly influence our results. We agree however with other researchers, however, that studies estimating the IVE against specific clade, with a randomized, and therefore unbiased selection of strains to be genetically characterized, would be of added value towards minimizing this limitation. Additionally, our study did not consider the effect that the use of several vaccine types might have on our results since it has varied by age group and Spanish sentinel network; however adjusting by these two variables might overcome this limitation.
In conclusion, our study results suggest an moderate protective effect against overall influenza A(H1N1)pdm09 and B, and a low vaccine effectiveness against A(H3N2). The pooled results of the studied seasons against predominant influenza type/subtype revealed no interference between the previous and current vaccine against A(H1N1)pdm09 and B and a possible negative interference against A(H3N2). Vaccine protection achieved against A(H1N1)pdm09 and B with the current vaccine or with both, the previous and current vaccine, is always superior to not being vaccinated. We have tried to explain these results by considering the previous vaccination effect and taking into account the match between circulating and vaccine strain. We have concluded that only the match between circulating and vaccine strains alone, cannot explain the obtained IVEs, and we underline the need of combined immunological and epidemiological studies, to better understand how these two elements are correlated. We also gratefully acknowledge the contribution of all the sentinel general practitioners and paediatricians, epidemiologists and virologists who have ever participated in the cycEVA study and in the Spanish Influenza Surveillance System.

Author Contributions
Conceptualization: AG IMB JC AL.

Data curation: AG AL.
Formal analysis: AG IMB AL.