Effects of previous episodes of influenza and vaccination in preventing laboratory-confirmed influenza in Navarre, Spain, 2013/14 season

J Castilla 1 2 , A Navascués 3 , M Fernández-Alonso 4 , G Reina 4 , E Albéniz 5 , F Pozo 6 , N Álvarez 5 7 , I Martínez-Baz 1 2 , M Guevara 1 2 , M García-Cenoz 1 2 , F Irisarri 1 2 , I Casado 1 2 , C Ezpeleta 3 , Primary Health Care Sentinel Network and Network for Influenza Surveillance in Hospitals of Navarra 8 1. Instituto de Salud Pública de Navarra, IdiSNA Navarra Institute for Health Research, Pamplona, Spain 2. CIBER Epidemiología y Salud Pública (CIBERESP), Madrid, Spain 3. Complejo Hospitalario de Navarra, IdiSNA Navarra Institute for Health Research, Pamplona, Spain 4. Clínica Universidad de Navarra, IdiSNA Navarra Institute for Health Research, Pamplona, Spain 5. Servicio Navarro de Salud, IdiSNA Navarra Institute for Health Research, Pamplona, Spain 6. Centro Nacional de Microbiologia (World Health Organization National Influenza Centre Madrid), Instituto de Salud Carlos III, Majadahonda, Spain 7. Red de Investigación en Servicios de Salud en Enfermedades Crónicas (REDISSEC), Pamplona, Spain 8. The members of these networks are listed at the end of the article


Introduction
Influenza produces annual epidemics that spread widely in the susceptible population.About 20% of children and 5% of adults worldwide develop symptomatic influenza each year [1].This exposure could confer immunity that would protect against the same virus type and subtype in subsequent seasons.Since the 2009 pandemic, influenza virus A(H1N1)pdm09, A(H3N2) and B have been alternating, thus part of the population may have acquired natural immunity after exposure to these viruses [2].
In serological surveys, nearly all children aged nine years or older had antibodies against influenza A [3].However, this does not mean that they are totally protected against this virus type, since antigenic drift of the influenza virus allows it to escape immune control.Differences in protection could not be accounted for by differences in serum haemagglutination inhibition titres, demonstrating that multiple immune mechanisms induced by natural infection confer resistance to influenza [4,5].
Annual influenza vaccination is the primary measure to prevent influenza and its consequences [1].Trivalent seasonal influenza vaccines include strains of influenza A(H1N1), A(H3N2) and B. In the 2013/14 season, the influenza vaccine composition recommended in the northern hemisphere included an A/California/7/2009(H1N1)pdm09-like virus, an A(H3N2) virus antigenically similar to the cell-propagated prototype virus A/Victoria/361/2011, and a B/ Massachusetts/2/2011-like virus [6].
Although both natural infection and vaccination with inactivated vaccine stimulate serum haemagglutination inhibition antibodies and provide protection against homologous wild-type influenza strains, the protection associated with natural infection lasts longer and is broader than that induced by inactivated vaccine [10,11].However, the effect of natural immunity and its practical relevance are not generally evaluated.The aim of this study was to estimate the effects of previous influenza episodes and of the trivalent vaccine in preventing inpatient and outpatient cases with laboratory-confirmed influenza in Navarre, Spain, in the 2013/14 season.

Study population
This study was performed in the region of Navarre, Spain.The Regional Health Service provides healthcare, free at point of service, to 97% of the population.
The Navarre Ethical Committee for Medical Research approved the study protocol.
The seasonal vaccination campaign took place from 14 October to 30 November 2013.The trivalent inactivated split non-adjuvanted vaccine was recommended and offered free of charge to people aged 60 years or older and to those with risk factors or major chronic conditions [12].Other people were also vaccinated if they paid for the vaccine.
In the 2013/14 season and the preceding seasons, influenza surveillance was based on automatic reporting of cases of medically-attended influenza-like illness (MA-ILI) from all primary healthcare centres and hospitals.ILI was considered to be the sudden onset of any general symptom (fever or feverishness, malaise, headache or myalgia) and any respiratory symptom (cough, sore throat or shortness of breath).In addition, a sentinel network composed of a representative sample of primary healthcare physicians, covering 16% of the Navarre population, was asked to take double swabs, nasopharyngeal and pharyngeal, after obtaining verbal informed consent, from all their patients diagnosed with ILI whose symptoms had begun less than five days before the consultation.The protocol for influenza cases in hospitals foresees nasopharyngeal and pharyngeal swabbing of all hospitalised patients with ILI.
Swabs were analysed by real-time RT-PCR, using either of two commercial real-time RT-PCR assays: RealCycler FLURSV (Progenie Molecular, Spain) and Real Time Ready Influenza A(H1N1) Detection Set (Roche Diagnostics, Switzerland).Detection of influenza A and B was based on the matrix protein gene and subtyping was based on the haemagglutinin (HA) gene.The internal amplification control was positive in all influenza-negative samples, indicating that failure to detect influenza virus was not due to inhibition.
Strains systematically selected among culture-positive samples by week and virus type/subtype were sent to the National Influenza Centre laboratory in Madrid for genetic characterisation based on partial sequencing of the HA gene (subunit HA1).

Study design and statistical analysis
We carried out a test-negative case-control study in the population covered by the Navarre Health Service.Healthcare workers, persons living in nursing homes and children under six months of age were excluded.
The study included the consecutive weeks in which influenza virus was detected, i.e. the period from 9 December 2013 (week 50) to 23 March 2014 (week 12).All information related to patients was linked using a unique identification number.
The cases were MA-ILI patients in primary healthcare or in hospitals for whom influenza virus infection was confirmed by RT-PCR, and the controls were MA-ILI patients who tested negative for influenza virus.Their vaccination status for the trivalent seasonal influenza vaccine was obtained from the regional vaccination register [13].Subjects were considered to be protected starting 14 days after vaccine administration.
From the electronic records of epidemiological and virological surveillance we obtained information on MA-ILI diagnosis and RT-PCR-positive patients in previous seasons for the study subjects.We defined previous MA-ILI related to a specific virus subtype as a laboratory-confirmed influenza infection with this virus subtype (virological criterion) that had occurred in the seasons from 2008/09 through 2012/13 or as MA-ILI that occurred in these seasons in weeks where more than 25% of swabs were confirmed for this influenza virus subtype (epidemiological criterion).Five previous seasons were considered given the long-lived protection associated with natural infection [10,11] and because no major shift had affected the circulating viruses involved in the analysis.Table 1 shows the periods when more than 25% of patients tested positive to the predominant virus type/subtype and the average percentage of swabbed patients who tested positive for the predominant circulating influenza virus by season.Finally, previous MA-ILI related to any influenza virus included all laboratory-confirmed influenza cases or MA-ILI patients that had occurred in the seasons 2008/09 through 2012/13 in weeks with more than 25% of swabs confirmed for any influenza virus, although on average 64% of swabbed patients tested positive for any influenza virus during these periods.
Percentages were compared by chi-square test.The odds of influenza vaccination and the odds of MA-ILI in the previous five seasons were compared between cases and controls.Logistic regression was used to calculate the odds ratios (OR) with their 95% confidence intervals (CI), adjusting for sex, age group (< 5, 5-14, 15-44, 45-64 and ≥ 65 years), major chronic conditions (heart disease, respiratory disease, renal disease,  a Includes seven cases of not subtyped influenza A. Two patients had simultaneous positive test results for influenza A(H1N1)pdm09 and influenza A(H3N2).b Two patients were attended in primary healthcare and referred to hospital.c Medically-attended influenza-like illness virologically or epidemiologically related to influenza in the previous five seasons.cancer, diabetes mellitus, liver cirrhosis, dementia, stroke, immunodeficiency, rheumatic disease and body mass index ≥ 40 kg/m 2 ), month of sample collection and healthcare setting (primary healthcare and hospital).Separate analyses were done by type/subtype of influenza, age group and healthcare setting.The fraction of prevented disease in exposed individuals or vaccine effectiveness (VE) was estimated as (1 − OR) x 100.

Results
During the 2013/14 season in Navarre, the incidence of MA-ILI, the number of swabbed patients and the number of influenza-positive cases followed similar trends, peaking in week 3 of 2014 (Figure).
In the study period, a total of 1,170 MA-ILI patients were swabbed, of whom 525 were attended in primary healthcare and 645 were hospitalised.A total of 589 (50%) were confirmed for influenza virus, all of them for influenza A. Influenza A(H3N2) virus was detected in 349 cases, influenza A(H1N1)pdm09 in 235, and seven remained non-subtyped.Two patients had a simultaneous positive test result for influenza A(H1N1)pdm09 and A(H3N2).Sequence analysis of the amplification product (the HA1 fragment of the haemagglutinin gene) was available for 114 influenza viruses.All 42 A(H1N1) pdm09 viruses were A/StPetersburg/27/2011-like and all 72 A(H3N2) viruses were A/Texas/50/2012-like.
A similar proportion of laboratory-confirmed cases and influenza-negative controls had had MA-ILI in the previous five seasons (11% vs 10%; p = 0.759), but only 17% of them (20/120) had been laboratory-confirmed for influenza virus in the previous episode.Of the 120 patients who had had any MA-ILI episode in the previous five years, 18 had had more than one episode and only one had had two episodes related to the same virus subtype.Among the 589 cases, 144 (24%) had  a Results obtained from a logistic regression model adjusted for sex, age group (< 5, 5-14, 15-44, 45-64 and ≥ 65 years), month of sample collection, major chronic conditions, healthcare setting (primary healthcare and hospital), medically-attended influenza-like illness virologically or epidemiologically related to the analysed influenza virus in the previous five seasons, and 2013/14 influenza vaccine.
b Medically-attended influenza-like illness virologically or epidemiologically related to influenza in the previous five seasons.
c Patients attended in primary healthcare and referred to hospital were included in both subanalyses.
The proportion of patients vaccinated in the current season was lower among those with previous MA-ILI than in those without a history of MA-ILI (17% vs 31%; p = 0.001).While previous MA-ILI was more frequent in patients between five and 44 years-old, in those without major chronic conditions and in those attended in primary healthcare, vaccination in the current season was more frequent in patients 65 years and older, in those with major chronic conditions and in patients attended in hospitals (Table 3).
In the analysis adjusted by influenza vaccination and other potential confounders, previous MA-ILI related to any influenza virus showed a 30% (95% CI: -7 to 54) protection against a new episode of laboratory-confirmed influenza, although this did not reach statistical significance.The overall adjusted estimate of the influenza VE was 31% (95% CI: 5-50).The estimate of the VE was 21% (95% CI: -45 to 57) in the analysis restricted to primary healthcare patients, and 35% (95% CI: 4-56) in hospitalised patients (Table 4).
In the comparison between influenza A(H1N1)pdm09 cases and controls, previous episodes of MA-ILI related to A(H1N1)pdm09 virus were 63% (95% CI: 16-84) protective against laboratory-confirmed A(H1N1)pdm09 influenza, even though the natural exposure had in most cases occurred more than two years before.The protective effect was similar in the analysis restricted to patients attended in primary healthcare and to those younger than 65 years.One case without comorbidity that had been confirmed with influenza A(H1N1)pdm09 in the 2009/10 season was again confirmed with influenza from the same virus subtype in the 2013/14 season.In the same models, the overall adjusted VE was

Table 4b
Preventive effect of previous episodes of medically-attended influenza-like illness and of the trivalent inactivated influenza vaccine against new cases of laboratory-confirmed influenza in Navarre, Spain, 2013/14 season (n = 1,170) CI: confidence interval; MA-ILI: medically attended influenza-like illness.
a Results obtained from a logistic regression model adjusted for sex, age group (< 5, 5-14, 15-44, 45-64 and ≥ 65 years), month of sample collection, major chronic conditions, healthcare setting (primary healthcare and hospital), medically-attended influenza-like illness virologically or epidemiologically related to the analysed influenza virus in the previous five seasons, and 2013/14 influenza vaccine.
b Medically-attended influenza-like illness virologically or epidemiologically related to influenza in the previous five seasons.
c Patients attended in primary healthcare and referred to hospital were included in both subanalyses.
45% (95% CI: 12-65), and similar estimates of the VE were found in the analysis stratified by age group or healthcare setting (Table 4).
The comparison of influenza A(H3N2) cases and controls showed that previous episodes of MA-ILI related to A(H3N2) virus were 65% (95% CI: 13-86) protective against laboratory-confirmed influenza A(H3N2) and 70% (95% CI: 15-90) protective in the analysis restricted to patients younger than 65 years.On the other hand, the overall adjusted VE was 20% (95% CI: −15 to 45), and other estimates of the VE for subgroups of patients were also low and not statistically significant (Table 4).In most cases, the natural exposure had occurred more than a year before.

Discussion
In this study we estimated at the same time the protection conferred by previous episodes of MA-ILI and by influenza vaccination in a season with intense cocirculation of influenza A(H1N1)pdm09 and A(H3N2).People with a history of MA-ILI attributable to a specific virus subtype in the previous five seasons had a markedly lower risk of disease due to the same subtype.The trivalent inactivated vaccine showed moderate VE in preventing laboratory-confirmed influenza A(H1N1)pdm09 and low effectiveness against influenza A(H3N2).Even though the natural exposure had in most cases occurred more than a year before, it conferred the same or greater protection against the same virus subtype than the vaccine administered a few months previously.In accordance with McLean et al., five previous seasons were considered for natural protection [14] because the protection following natural exposure is stronger and longer-lasting and covers a greater variety of viral strains, which has been related to activation of a more complete immune response that includes mechanisms of cellular immunity [4,15,16].No major shift had affected the circulating viruses involved in the analysis.
It was possible to define the virus that most probably caused the cases of MA-ILI in the previous five seasons thanks to the fact that one virus clearly predominated in Navarre in each of those five seasons.In seasons with simultaneous co-circulation of various viruses, it would be more difficult to attribute the cases of MA-ILI with certainty to a specific virus subtype.
Since the appearance of the A(H1N1)pdm09 virus in 2009, the circulating strains of this virus have been well matched with the vaccine strain A/ California/7/2009(H1N1) [2], which could explain the protection of the vaccine and of influenza episodes in previous seasons.
Although the influenza A(H3N2) virus strains which circulated in the 2013/14 season had a good genetic match with the vaccine strain [2], the observed VE was low.However, this virus showed a high cross protection with the strains circulating in the previous seasons 2008/09 and 2011/12.This difference between natural and vaccine protection with matched strains should encourage the exploration of alternative ways of obtaining better vaccines against influenza.
In the study population, natural and vaccine immunity were distributed in a complementary manner.A history of MA-ILI was more frequent in persons aged five to 44 years, which explains why this protective mechanism was more important in population groups that do not normally get vaccinated against influenza.
Although previous diagnosis of disease from the same virus subtype was associated with high protection, previous MA-ILI related to any influenza virus but not restricted to the same virus subtype conferred only low protection against a new episode of laboratoryconfirmed influenza.This is mainly explained by the likelihood of infection by a different type or subtype of influenza virus.Therefore, in persons with risk factors for influenza complications, having had the disease in previous seasons should not be a reason not to get vaccinated.While natural exposure protects specifically against the virus subtype to which one has been exposed, the protection conferred by the trivalent vaccine, although less strong, covers all three virus types/ subtypes simultaneously.
Previous episodes of influenza are not usually taken into account as potential confounding factors in studies evaluating influenza VE.To our knowledge, only McLean et al. had adjusted for influenza diagnoses in the prior five seasons in the analysis of influenza vaccine effectiveness [14].In this and in our study, the estimated VE did not change regardless of whether the models included this history, suggesting that this variable does not act as a confounding factor that needs to be controlled.
Although our end-of-season estimate of VE was additionally adjusted for previous episodes of influenza, it was consistent with mid-season estimates obtained in Navarre and Spain for this same season [17,18], and with estimates obtained at the end of the season in a European multicentre study and in Greece [19,20]; it was less consistent, however, with estimates from other countries with different distribution of virus types, subtypes and strains detected in the same season [21][22][23].
Some limitations should be considered in interpreting the results of this study.Previous episodes of MA-ILI reflect the history of exposures to the influenza virus from the healthcare perspective and may be considered a proxy for natural immunity.Some 10% of subjects included in the study had a history of MA-ILI in the previous five seasons.However, the proportion of the population with natural immunity against influenza could be considerably higher, since it is estimated that 30-50% of influenza infections are asymptomatic [24].
In one study conducted in Navarre, 36% of symptomatic cases had not sought medical care [25].It should also be added that there is possible immunity from exposures occurring more than five years previously.This misclassification in the previous influenza infection is probably non-differential and would bias the estimates towards the null effect.In the absence of this bias, the protection due to previous episodes would have been higher.
Of the patients with a previous episode of MA-ILI, only 17% had a laboratory-confirmed diagnosis, while the rest met only one epidemiological criterion for the disease.Based on the percentage of swabs confirmed for influenza in each season (Table 1), we estimate that this criterion ensures the correct classification of 70% of cases with a history of influenza A(H3N2), of over 50% of cases with a history of influenza A(H1N1) pdm09, and of 64% of cases with a history of any influenza in the previous five years.Accordingly, we cannot totally rule out the possibility of incorrect classification that arose from considering cases that could have been due to another cause such as previous episodes related to a specific virus.If we had had laboratory confirmation of all the cases of influenza in previous years, the protective effect of this history would probably have been greater.
The results presented had limited statistical power for some analyses, mainly because of the low numbers of cases and controls with previous MA-ILI included in the study.Laboratory-confirmed cases were compared with controls recruited in the same healthcare settings before either patient or physician knew the laboratory result, a fact that reduced selection bias [26].
This study included MA-ILI patients recruited from the same population in both primary healthcare centres and hospitals.The healthcare setting could have acted as a confounding factor, therefore the analyses were adjusted for this variable.The possibility that the healthcare setting might have modified the effect or biased the results can be ruled out given the consistency of the estimates obtained in these two patient groups and in the joint analysis.The joint analysis achieved representation of the whole spectrum of patients with influenza in the population.

Conclusion
Our results suggest low to moderate influenza VE in the 2013/14 season, which prevented almost a third of the influenza cases and hospitalisations in the vaccinated population; while not entirely satisfactory, this result is important in terms of individual and public health.Previous influenza episodes were highly effective against new influenza illness by the same virus subtype, and this effect seemed to persist over various seasons, which may point to possible avenues of obtaining better vaccines against influenza.In any case, annual influenza vaccination remains the principal preventive option in persons at high risk of developing complications if they contract influenza.

Figure
FigureWeekly incidence of patients with medically attended influenza-like illness and number of swabbed patients by test result, Navarre, Spain, influenza season 2013/14 (n = 1,170 in the study period )

Table 2
Characteristics of patients with medically-attended influenza-like illness included in the test negative case-control analysis, by test result, Navarre, Spain, 2013/14 season (n = 1,170)

Table 3
Characteristics of patients with medically-attended influenza-like illness, by previous influenza diagnosis and influenza vaccination status, Navarre, Spain, 2013/14 season (n = 1,170) MA-ILI: medically attended influenza-like illness; NA: not applicable.a Two patients were attended in primary healthcare and referred to hospital.b Medically-attended influenza-like illness virologically or epidemiologically related to any influenza virus in the previous five seasons.

Table 4a
Preventive effect of previous episodes of medically-attended influenza-like illness and of the trivalent inactivated influenza vaccine against new cases of laboratory-confirmed influenza in Navarre, Spain, 2013/14 season (n = 1,170)