Field seasonal influenza vaccine effectiveness

Abstract

Thanks to the screening method, we estimated among target groups the 2010/2011 field vaccine effectiveness (FVE) against laboratory confirmed influenza cases seen in general practice. We also compared the values of FVE estimations obtained by using three sources of the population vaccination coverage (VC) based on three different methodologies: (1) administrative data from the main social security scheme (Caisse Nationale d’Assurance Maladie des Travailleurs Salariés—CNAMTS) covering about 85% of the French population, (2) a cross-sectional national telephone survey in the general population, and (3) a declarative survey in the population seen in a one-day general practitioner (GP) consultations.

 

The FVE estimates among target groups were stratified by age (<65 y old with reported chronic illness; ≥65 y old and overall). Using the VC of the CNAMTS, the FVE of the 2010/2011 seasonal trivalent vaccine against laboratory confirmed infection with any influenza virus was 59% (95% Confidence Interval, 17 to 81). It was 85% (17 to 99) and 50% (–16 to 80) for A(H1N1)pdm09 and B influenza infections, respectively. The values of FVE using the influenza VC obtained in a sample of the general population and of the population of GPs’ patients were 73% (45 to 87) and 82% (63 to 92), respectively.

We estimated a moderate influenza FVE in preventing confirmed influenza viruses in target groups by using the VC of the CNAMTS. We also observed that the screening method generates FVE values dependent on the choice of the source of VC and thus should be used cautiously.

Keywords: :

• vaccine effectiveness
• general practitioners
• influenza
• vaccination coverage
• target groups

Introduction

Each autumn, influenza vaccine vouchers are sent by the French Social Security to all beneficiaries (approximately 10–12 million people) who are targeted according to annual criteria defining the population groups at risk.^1-3 The voucher allows the recipients to get the vaccine from the pharmacist. In recent years, thanks to this system, target population vaccination coverage (VC) has been annually estimated by the National Health Insurance Scheme for Employees (Caisse Nationale d’Assurance Maladie des Travailleurs Salariés – CNAMTS), the main social security scheme, covering about 85% of the French population.^1-3

For the 2010/2011 season, the recommended trivalent influenza vaccine included the influenza pandemic strain A/California/07/2009, in addition to the H3N2 A/Perth/16/2009 and B/Brisbane/68/2008 strains.⁴ The French 2010/2011 influenza vaccination campaign lasted from September 2010 to January 2011 with 12 million of people receiving influenza vaccine vouchers.^1,2 The influenza vaccination strategy targets three main at risk groups: persons aged 65 year old (y.o.) or above, health professionals and persons below 65 y.o. with certain underlying chronic conditions.^1-3

In France, in spite of the availability of the VC through the CNAMTS data, influenza vaccine effectiveness (VE) estimates for the target groups are poor.^5,6 Since 2003, the French Sentinelles Network, an internet surveillance system based on the declaration made by voluntary general practitioners (GPs),⁷ estimates imperfect VE. These estimates were based on the screening method,^8,9 and used an estimator stratified by age-group requiring two imperfect samples: a sample of influenza like-illness (ILI) cases (a non specific outcome) and a sample drawn from a poll on the influenza vaccine conducted each year after the epidemic season by a private independent organization.¹⁰

The aim of the present study was to measure the VE of the 2010/2011 influenza vaccine among target groups, by using two major sources of available data in France during the influenza outbreak: the PCR laboratory confirmed influenza cases and the VC from the CNAMTS administrative data.² Moreover in order to evaluate the dependence of the FVE values to the use of different VC sources, we estimated the VE using two supplementary databases of VC available during the 2010/2011 season: a cross-sectional national telephone survey in general population and a survey in the population consulting a GP. We also asses the bias caused by using non-specific endpoints, by comparing FVE against laboratory confirmed influenza vs. ILI.

Results

Description of the 2010–2011 influenza season in France

During the winter of 2010/2011, ILI incidence crossed the epidemic threshold during week from 20th to 26th December 2010, peaked on week from 03th to 09th January 2011 and fell below the epidemic threshold during week from 14th to 20th February 2011.¹¹ In 2010/2011, the influenza epidemic was characterized by a co-circulation of influenza A (H1N1)pdm09 and influenza B viruses with a good antigenic match to vaccine strains.^4,12

Laboratory-confirmed influenza cases

Over the 2010/2011 season, 532 confirmed influenza cases were reported by Sentinelles and Grog GPs between the epidemic period. Among them, 517 (97%) were reported with information about vaccination, risk factors, and age. Of this 517 cases, 37 (7%) had at least one risk factor for influenza; of which 10 (27%) aged less than 65 y.o. with chronic illness and 27 (73%) aged ≥65 y.o. (Table 1). Among these 37 patients with at least on e risk factor for severe influenza, 10 (27%) were infected by influenza virus A(H1N1)pdm09, 24 (65%) were infected by influenza viruses B, and 3 (8%) were infected by A(H3N2) virus. Ten patients (27%) were vaccinated with the trivalent 2010/2011 seasonal vaccine (Table 1). The median age was of 67 y.o. (min = 5; max = 82) and a sex ratio of 0.50.

Table 1. Description of laboratory-confirmed influenza and ILI cases included in the study

Groups, age (years)	Total described No.	Total vaccinated No. (%)
ILI patients
6m-64 y.o. with chronic illness	211	24 (11)
≥ 65 y.o.	214	78 (36)
All target groups	425	102 (24)
Laboratory-confirmed influenza patients
All influenza viruses
6m-64 y.o. with chronic illness	10	0 (0)
≥ 65 y.o.	27	10 (37)
All target groups	37	10 (27)
A(H1N1)pdm09
6m-64 y.o. with chronic illness	5	0 (0)
≥ 65 y.o.	5	1 (20)
All target groups	10	1 (10)
A(H3N2)
6m-64 y.o. with chronic illness	1	0 (0)
≥ 65 y.o.	2	1 (50)
All target groups	3	1 (33)
B
6m-64 y.o. with chronic illness	4	0 (0)
≥ 65 y.o.	20	8 (40)
All target groups	24	8 (33)

ILI cases

Over the 2010/2011 season, 7834 ILI cases were reported by Sentinelles GPs during the epidemic period. We considered a sample of 7524 ILI cases (96%) aged over six months who did not have missing information concerning age or risk factors and for whom vaccination status data had been completed for analysis. Of the 7524 ILI cases, 425 (5.6%) belonged to a target group for vaccination (<65 y.o. with reported chronic illness; ≥65 y.o. and overall). One hundred-two patients (24%) were vaccinated with the 2010/2011 seasonal vaccine (Table 1). The median age was of 64 y.o. (min = 0.75; max = 93) and a sex ratio of 0.50.

Field vaccine effectiveness (FVE)

The FVE according age, risk factors and influenza virus computed by using data of CNAMTS are detailed in Table 2. The FVE of the 2010/2011 seasonal trivalent vaccine against any influenza virus (A and B) was 59% (95% Confidence Interval [CI], 17 to 81). The FVE of the 2010/2011 seasonal trivalent vaccine against A(H1N1)pdm09 and B was 85% (17 to 99) and 50% (–16 to 80) respectively (Table 2). Due to an insufficient number of cases, the FVE of the 2010/2011 seasonal trivalent vaccine against A(H3N2) could not be estimated. The FVE of the 2010/2011 seasonal trivalent vaccine in preventing ILI was 56% (45 to 65).

Table 2. Estimates of field vaccine effectiveness (FVE) of trivalent influenza vaccine against laboratory confirmed influenza and ILI infections during periods of influenza circulation in France, 2010/2011 using administrative data of CNAMTS

Groups, age (years)	VE (%)	95% IC
ILI patients
6m-64 y.o. with chronic illness	65	48 to 78
≥ 65 y.o.	51	35 to 63
All target groups	56	45 to 65
Laboratory-confirmed influenza patients
All influenza viruses
6m-64 y.o. with chronic illness	100	-
≥ 65 y.o.	49	-9 to 78
All target groups	59	17 to 81
A(H1N1)pdm09
6m-64 y.o. with chronic illness	100	-
≥ 65 y.o.	79	-45 to 99
All target groups	85	17 to 99
B
6m-64 y.o. with chronic illness	100	-
≥ 65 y.o.	43	-38 to 78
All target groups	50	-16 to 80

Assessment of bias due to the reference population samples

The proportions of the population vaccinated (PV) in the two categories selected (<65 y.o. with reported chronic illness; ≥65 y.o.) with respect to the methodology and the origin of data were described in Table 3. Overall, the influenza VC was lower in people aged less than 65 y.o. with reported chronic illness compared with the ≥65 y.o. independently of the origin of data and the methodology used. People aged ≥65 y.o. belonging to the population of GPs patients had the highest VC (72%) with respect to those observed in the general population ranging from 53.8% (administrative data) to 61% (national telephone survey). Among people aged less than 65 y.o. with reported chronic illness, VC ranged from 26.8% (administrative data) to 46.6% (general population).

Table 3. Vaccination coverage available among target groups in France, 2010/2011, according three sources of data

Methodology	Origin of data	VC by age groups (95% IC)
		6 m-64 y.o. with chronic illness	≥ 65 y.o.
Administrative data	CNAMTS administrative data (general population)	26.8%	53.8%
Cross-sectional national telephone survey	French Institute for public health surveillance's survey (general population)	46.6% (39.7–53.6)	61.0% (56.7–65.0)
Standardized survey on a randomly assigned single day in general practice	GPs patients of the French Sentinelles Network (GPs’ patients population)	46.0% (40.0–51.0)	72.0% (70.0–75.0)

The values of FVE against laboratory confirmed infection, using the influenza VC obtained in (1) a sample of the general population and (2) of the population of GPs’ patients were 73% (45 to 87) and 82% (63 to 92), respectively (Table 4). The FVE according age, risk factors and source of VC are detailed in Table 4.

Table 4. Estimates of field vaccine effectiveness (FVE) of trivalent influenza vaccine against laboratory confirmed influenza during periods of circulation in France by using three sources of vaccination coverages, 2010/2011

Origin of data	FVE(95% CI)
	6 m-64 y.o. with chronic illness	≥ 65 y.o.	All target groups
CNAMTS administrative data (general population)	100	49 (-9 to 78)	59 (17 to 81)
Cross-sectional national telephone survey (general population)	100	62 (19 to 83)	73 (45 to 87)
Standardized survey on a randomly assigned single day in general practice (GPs’ patients population)	100	77 (51 to 90)	82 (63 to 92)

Discussion

In this study conducted during the 2010/2011 French influenza season, we estimated a moderate influenza FVE in preventing laboratory confirmed influenza infections in target groups (< 65 y.o. with reported chronic illness; ≥ 65 y.o. and overall). By using the screening method, we estimated the FVE by comparing the VC among the influenza positive patients to the VC of the same risk groups in the general population estimated by using administrative data of the CNAMTS, covering about 85% of the French population.

We also obtained the FVE by using different sources of VC estimates. Results showed important variability with a tendency to overestimation among VE estimated using VC issued from the declarative survey in the population seen in a one-day GP consultations and from the cross-sectional national telephone survey in the general population. As previously explained,^13-15 misestimation could be caused by the large temporal difference between VC of the referent population and cases. The VC of the population consulting GPs of the Sentinelles Network has not been estimated in the period of occurrence of cases but in post-epidemic period. Thus the data here reported represent the population attended by the GPs but not the populations attended by the GPs during the influenza epidemic period and were not representative of the cases population here analyzed, leading to a danger of VE overestimation. The coverage data observed among the population consulting GPs of the Sentinelles Network, were 10 to 20% higher than those collected by CNAMTS administrative data (Table 3).

The results of the national telephone survey (Table 3), were between 5 and 10% higher than figures produced by the CNAMTS for the same influenza season.¹ This discrepancy could be explained by the biases observed in telephone surveys (participation bias in favor of better vaccinated individuals, recall bias especially in the elderly and social desirability bias in the answers provided), which could lead to some over-estimation of coverage in the telephone survey.¹ The VC drawn from administrative data (CNAMTS), covering about 85% of the French population, was the most adequate to compute estimations of the FVE among target groups, with respect to the two VC data here analyzed and issued both from surveys. On the other hand, administrative data could be affect by under-estimations of VC data, calculated as the number of vouchers receipts divided par the number of the vouchers sent by the CNAMTS, because of people may have access to vaccination without using vouchers (example in nursing homes).^2,3

Our estimates (VE = 59%; 95% CI, 17 to 81), measured using PCR virological data and administrative data, were higher to adjusted VE against all influenza viruses reported by a case-control Spanish study in general practice among patients targeted for vaccination (36%; −39 to 71),¹⁶ and closed to a European multi-center case control study among target groups (56.2%; 34.3 to 70.7).¹⁷ It was below the estimates of a case-control German study among people with chronic illness (80%; 32.7 to 94.4).¹⁸ As explained by the authors of the German study, because their points estimates were less specific than those of the other published studies on influenza VE, their results could be considered similar.¹⁸ As expected, our estimates of FVE among target groups were slightly lower than values of VE estimated in all patients (with and without risk factors of severe influenza) ranging from 52% to 89.4%.^16-20

According to virus type, our overall VE estimates of the influenza 2010/2011 vaccine showed a good protection against A(H1N1)pdm09 infection (85%; 17 to 99), that were above the results reported among target groups in a European multicenter case-control study (VE A[H1N1]pdm09 = 58.9%; 32.0 to 75.1)¹⁷ and a Spanish case-control study among patients eligible for vaccination (VE A[H1N1]pdm09 = 47%; 0 to 74).¹⁶ Our estimates were also above than those reported for the pandemic vaccine during 2009/2010 among target groups for influenza vaccination, ranging from 49% to 71%.²¹ The moderate protective effect of the 2010/2011 vaccine against influenza B infection reported in this study (50%; −16 to 80), not explained by antigenic mismatch,^4,12 was due to a relative important number of vaccine failures in person aged of ≥65 y.o. The FVE reported for this season in other European studies was variable and ranged from 19% to 63% among target groups.^16-20

As ILI is a nonspecific outcome, it would normally be expected that FVE estimations using ILI would under-estimate the real VE,²² because of only a portion of ILI cases may be due to influenza virus infection. In our study, FVE estimation using ILI cases (56%; 45 to 65) was closed to FVE observed using laboratory confirmed influenza cases (59%; 17 to 81). If we analyze our results according the influenza types/sub-types, we observed a lower VE in preventing ILI (VE = 56%; 45 to 65) than in preventing confirmed influenza A(H1N1)pdm09 (85%; 17 to 99). Similarly, during the (2009/2010), through a validation study on a sample of laboratory-confirmed influenza enrolled by French GPs, we estimated the same bias in VE inherent to the use of such a non-specific outcome.²² On the other hand, when we compared the VE in preventing ILI vs. the VE in preventing confirmed influenza virus B, the estimates were highest for ILI. This result could be explained by the important age differences observed between influenza B cases with patients aged of >65 y.o. three times more enrolled than patients aged of less 65 y.o. Stochastic variations due to little sample size are another possible explanation.

The present study has several advantages. The estimations of FVE have been obtained by using as endpoint PCR confirmed influenza, the most specific outcome, as viral culture will miss cases and serology will overestimate VE for trivalent inactivated vaccines.^13,23 Furthermore, the completeness of the VC of the reference population estimated by using administrative data of the CNAMTS, is high because covering about 85% of the French population.^1-3 By restricting analyses to the people with chronic illness at risk of severe illness and death after influenza infection, targeted by the French vaccination recommendation, allowed us to measure the impact of the French health policy.

In interpreting our results some limitations could be considered. As we estimated VE among target groups, the sample size was limited and it was not adequate for some subgroup analyses as influenza VE estimates against influenza B infection. Moreover, even though the VC was calculated in the population with chronic disease, it was not possible to investigate the type and severity of the underlying conditions. As we analyzed two large age groups, a high variability inside would be considered. Moreover, we could not adjust or stratified by confounding factors such as health care utilization, sex, hospitalization, previous vaccinations, smoking, etc.¹⁵ Finally, including patients vaccinated with a monovalent vaccine in the previous season could also have affected the VE estimated leading to overestimation of VE against A(H1N1)pdm09 even if the residual protective effect 1 y after the pandemic was limited.^16,19

Despite its limitations these results show that the screening method could provide early estimate of FVE by using laboratory confirmed endpoints and administrative data among target groups seen in general practice, and could be useful to detect in real time low VE and alert public decision makers. We also observed that the screening method generates FVE values dependent on the choice of the source of VC and thus should be used cautiously. These observations are important for future studies using the screening methodology.

Moreover, this method could be used as a complement to case-control study where adjustment for confounding factors is possible.

Material and Methods

Ethical statement

The protocol was conducted in agreement with the Helsinki declaration. We obtained authorization from the French Data Protection Agency (CNIL, registration number #471393). Specific written informed patient consent was not needed for this observational study.

The screening method

We estimated VE through the use of the screening method.^8,9 The screening method is a “case-cohort” or “case-base” design. Its principle is to calculate VE (or FVE for field vaccine effectiveness) using the following equation:

$FVE=\frac{PV-PCV}{PV(1-PCV)}$

Where PCV is the proportion of vaccinated among influenza cases and PV is the proportion of vaccinated among the population.^8,9 Two influenza data sets were used as PCV: one of laboratory-confirmed influenza cases, and one of clinically defined ILI cases to estimate the bias due to the use of a nonspecific influenza outcome. PV was obtained from administrative sources (CNAMTS).² Because influenza vaccines were not given to children under six months old, they were excluded from the study. Individuals with missing age, and/or risk factors for severe influenza and/or vaccination status were also excluded. Vaccination status was reported by GPs, based on GPs records or patient report. We did not collect data on vaccine manufacturer and assumed all vaccines were equally effective. Vaccines were considered as potentially effective if administrated at least 3 weeks prior to symptom onset. Patients whose vaccination status occurred <3 weeks prior to symptom onset were considered as not vaccinated.

Study period

For the acquisition of laboratory confirmed influenza cases and ILI cases, the epidemic study period began during week from 20th to 26th December 2010, which was the beginning of the influenza epidemic as declared by the French Sentinelles Network in France, and ended during week from 14th to 20th February 2011.¹¹

Case recruitment for determining PCV

Influenza like illness (ILI) cases were reported by Sentinelles GPs in France excluding overseas territories but including Corsica, as part of a surveillance routine using the following definition: “sudden onset of fever >39 °C (102°F) with respiratory signs and myalgia”⁷. Laboratory-confirmed influenza cases in France excluding Corsica and overseas territories were reported by GPs from the Regional Groups of Influenza Observation (GROG) as part of a surveillance routine. GROG is a network of private-practice GPs dedicated to the virological surveillance of influenza.²⁴ Corsican laboratory-confirmed influenza cases were reported by Sentinelles GPs.^22,25 Nasopharyngeal swabs were collected through a randomized selection routine. GPs included the first patient of each week, of any age (Sentinelles protocol) or of a personally assigned age group among 0–4, 5–14, 15–64, and >65 y-old (GROG protocol). Only patients presenting between 0 and 7 d after symptom onset were swabbed. GROG GPs swabbed patients presenting with acute respiratory infection, defined as: sudden onset of a respiratory sign and a systemic sign evoking an infection (fever, asthenia, headache, myalgia, faintness). Corsican Sentinelles GPs swabbed patients presenting with ILI according to the Sentinelles definition.⁷ Swabs were analyzed in 11 laboratories, depending on the region of swabbing (two national reference centers and nine laboratories), by real-time polymerase chain reaction.

The following information was collected for both laboratory-confirmed influenza cases and ILI: date of consultation, age, sex, vaccine status for 2010/2011seasonal trivalent vaccine, time since vaccination (more or less than 3 weeks) and presence of risk factors (chronic illness). For laboratory confirmed influenza cases, the viral strain was determined.

As previously stated,²² we used a three weeks period between vaccination and consultation to identify vaccinated influenza cases fully immunized from those who were still unprotected.

Population data for determining PV

The proportion of vaccinated individuals among people with <65 y.o. with reported chronic illness and ≥65 y.o. was obtained through the administrative data provided by CNAMTS.²

The French influenza vaccination strategy targets three main at risk groups: persons aged 65 y.o. or above, health professionals and persons below 65 y.o. with certain underlying chronic conditions such as diabetes, heart failure, cardiac arrhythmias, cardiovascular disease, congenital heart disease with an additional gravity criterion; invalidating cerebrovascular accident; severe chronic respiratory failure; epilepsy and serious neurological or muscular conditions; severe primary immune deficiency and human immune deficiency (HIV); chronic nephropathy and classical nephropathic syndrome; cystic fibrosis.³ Each autumn, an individual vaccination voucher is sent by the various Social Security Schemes (SSS) to those three at risk populations. The underlying chronic condition is attributed to patients on request of their GP and after validation by a physician of the SSS (doctor who determines the diseases and medical treatments to benefit from 100% social security coverage). The voucher allows the recipient to get vaccinated for free. The voucher is then returned to the SSS by the pharmacist who delivered the vaccine in order to get refunded. At the end of each season, each of the 3 main SSS estimates vaccine uptake in the three main risk groups based on the voucher return rates.^2,3 For each seasonal influenza vaccination campaign, the number of vaccinations effectuated is estimated by the number of vaccine vouchers presented at a pharmacy. Vaccination coverage is usually monitored through the data provided by the CNAMTS, the main SSS which covers about 85% of the French population. In 2011, pregnant women and obese persons were considered at risk of severe influenza but they did not receive a voucher if they had no chronic co-morbidities, thus VC were not available for these categories.

Estimation of field vaccine effectiveness

In this study, we estimated, the FVE of the 2010/2011 seasonal trivalent vaccine against any influenza infection, against the subtype A(H1N1)pdm09 and the influenza B viruses respectively, by using the screening method proposed by Farrington.²⁶ In practice, VE for “all target groups” was estimated with a logistic regression model allowing a different offset in each age strata (two strata: <65 y.o. with reported chronic illness; ≥65 y.o.).

We conducted the analyses using R v2.8.1.

Assessment of biases

Bias due to the use of a nonspecific influenza outcome (ILI)

To assess the bias of using ILI (a non specific outcome) as an endpoint to estimate VE, we compared VE estimates obtained using the ILI cases sample to VE estimates by using the confirmed influenza cases.

Bias due to the reference population samples

Estimates of effectiveness from the screening method are very sensitive to minor errors in input estimates.²⁷ Thus, we estimated FVE obtained by using influenza confirmed cases as endpoint and two others sources of influenza VC available in 2010/2011 in France among <65 y.o with reported chronic illness; ≥65 y.o and overall, in addition of data from the administrative source (CNAMTS) and describe above: (1) data from a cross-sectional national telephone survey in general population and (2) data from a survey in the population consulting a GP (GP practices). The VC are shown in Table 2.

The cross-sectional national telephone survey was undertaken in January 2011 by the French Institute for Public Health Surveillance among a representative sample of households in continental France connected to a land-line telephone, which represents more than 85% of French households. Individuals in the household belonging to a target group (persons aged 65 y.o. or above and persons below 65 y of age with chronic illness) were searched and included if oral consent was obtained.¹

To assess the VC in the population consulting a GP, we conducted a cross-sectional study²⁸ among GPs of the French Sentinelles Network. A day of participation was randomly assigned to each GP who agreed to participate (n = 262) and randomization was balanced to reflect the distribution of the usual days of consultation from Monday to Saturday, as declared by participants. From June to September 2011, 203 GPs included 4248 patients. Through this declarative survey, the following information was collected from each patient: age, sex, pregnancy at any trimester, obesity (BMI ≥ 30), chronic illness and vaccination status for seasonal trivalent vaccine 2010/11. Patients were included if oral consent was obtained.

Acknowledgments

The authors wish to thank all the participating general practitioners. We thank the French National Influenza Centre for virological data.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

Funding for this study was provided by Sanofi Pasteur. The views expressed in this paper are those of the authors and do not necessarily reflect the position of Sanofi Pasteur.