Skip Navigation

Component-Specific Effectiveness of Trivalent Influenza Vaccine as Monitored through a Sentinel Surveillance Network in Canada, 2006–2007

  1. Danuta M. Skowronski1,
  2. Gaston De Serres2,
  3. Jim Dickinson3,
  4. Martin Petric1,
  5. Annie Mak1,
  6. Kevin Fonseca4,
  7. Trijntje L. Kwindt4,
  8. Tracy Chan1,
  9. Tracy Chan1,
  10. Nathalie Bastien5,
  11. Hugues Charest2 and
  12. Yan Li5
  1. 1 British Columbia Centre for Disease Control, Vancouver, British Columbia
  2. 2 Institut national de santé publique du Québec, Québec
  3. 3 University of Calgary, Alberta
  4. 4 Alberta Provincial Laboratory, Alberta
  5. 5 National Microbiology Laboratory, Public Health Agency of Canada, Manitoba, Canada
  1. Reprints or correspondence: Danuta M. Skowronski, 655 W. 12th Ave., Vancouver, British Columbia, Canada V5Z 4R4 (danuta.skowronski{at}bccdc.ca).

  1. Presented in part: Eleventh Annual Conference on Vaccine Research, Baltimore, Maryland, 6 May 2008 (abstract S13).

 
Next Section

Abstract

Background. Trivalent inactivated influenza vaccine (TIV) is reformulated annually to contain representative strains of 2 influenza A subtypes (H1N1 and H3N2) and 1 B lineage (Yamagata or Victoria). We describe a sentinel surveillance approach to link influenza variant detection with component-specific vaccine effectiveness (VE) estimation.

Methods. The 2006–2007 TIV included A/NewCaledonia/20/1999(H1N1)—like, A/Wisconsin/67/2005(H3N2)— like, and B/Malaysia/2506/2004(Victoria)—like components. Included participants were individuals ⩾9 years of age who presented within 1 week after influenzalike illness onset to a sentinel physician between November 2006 and April 2007. Influenza was identified by real-time reverse-transcriptase polymerase chain reaction and/or culture. Isolates were characterized by hemagglutination inhibition assay (HI) and HA1 gene sequence. VE was estimated as 1-[odds ratio for influenza in vaccinated versus nonvaccinated persons].

Results. A total of 841 participants contributed: 69 (8%) were ⩾65 years of age; 166 (20%) received the 2006–2007 TIV. Influenza was detected in 337 subjects (40%), distributed as follows: A/H3N2, 242 (72%); A/H1N1, 55 (16%); and B, 36 (11%). All but 1 of the A/H1N1 isolates were well matched, half of A/H3N2 isolates were strain mismatched, and all B isolates were lineage-level mismatched to vaccine. Age-adjusted estimated VE for A/H1N1, A/H3N2, and B components was 92% (95% CI, 40%–91%), 41% (95% CI, 6%–63%), and 19% (95% CI, –112% to 69%), respectively, with an overall VE estimate of 47% (95%CI,18%–65%). Restriction of the analysis to include only working-age adults resulted in lower VE estimates with wide confidence intervals but similar component-specific trends.

Conclusions. Sentinel surveillance provides a broad platform to link new variant detection and the composite of circulating viruses to annual monitoring of component-specific VE.

Since the mid-1970s, trivalent inactivated influenza vaccine (TIV) has been reformulated annually to include standardized amounts of hemagglutinin (HA) from representative seed strains of 2 antigenically-distinct influenza A subtypes (H3N2 and H1N1) and 1 of 2 antigenically-distinct influenza B lineages (Yamagata or Victoria) [1]. Specific vaccine components are designated by the World Health Organization in February for the following influenza season, which typically spans November to April in the northern hemisphere [1]. The proportionate mix of circulating influenza types, subtypes, strains, and new variants reflects complex interaction between error-prone virus replication, random mutation, and population immunity over space and time. A linked laboratory network globally monitors the emergence and dominance of new mutations, as well as the transmission, growth characteristics, and impact of new variants to provide annual updates for recommended vaccine components [1].

Epidemiologic proof of vaccine effectiveness (VE) is not a requirement for distribution of the TIV formulation, which changes annually. Periodic studies to estimate TIV protection are generally powered to assess expected reduction in overall influenza illness in a circumscribed cohort and typically report only a single estimate of protection against any influenza. This approach minimizes sample size and cost and addresses a clinically relevant outcome, but it provides limited virologic diversity or detail to reflect the trivalent composition of the vaccine. Recent meta-analysis found that VE against laboratory-confirmed influenza in young adults was 80% (95% CI, 56%–91%) when measured during select seasons of vaccine match and 50% (95% CI, 27%–65%) during select seasons of vaccine mismatch [2]. Vaccine match or mismatch, however, are relative terms, dependent on antigenic distance between the specific vaccine component and its circulating counterpart. Impact at the population level further depends on the proportionate mix of circulating A and B subtypes and strains. Although some HA-based crossprotection exists between related influenza strains, HA-based serum antibody is not expected to give protection across different A subtypes or B lineages [2–10].

Heterogeneity in the proportionate mix of influenza viruses in Canada and their match to annually selected vaccine components is illustrated in table 1 for strains characterized each season between 2000–2007 [11–18]. During that period, the A/H1N1 virus showed no HA antigenic change, and the vaccine component was well matched each year. Conversely both virus drift and relative vaccine mismatch at the strain level were observed for the A/H3N2 component from 2003–2007. More problematic was vaccine match for the B component. Viruses of the B/Victoria lineage predominated during the 1980s, while B/Yamagata— lineage viruses accounted for the majority of isolates in most countries from 1990–2001. In 2001, B/Victoria viruses reemerged in North America for the first time in more than a decade [19], 20. Since then, viruses belonging to the B/Yamagata and B/Victoria lineages have varied in terms of their contribution, resulting in lineage-level mismatch for the B component of the influenza vaccine during 4 of 6 subsequent seasons.

Complexity in virus evolution and vaccine match has led to increased interest in field methods for monitoring vaccine performance each year [4, 6, 21–24]. For ethical reasons and practical reasons, including cost, randomized controlled trials (RCTs) cannot be routinely conducted every year. Observational designs offer an efficient alternative for estimating VE annually, provided that cases and controls can be reliably recruited from the same source population and the influence of confounders can be adequately addressed [25]. Regardless of study design, laboratory-confirmed outcomes are needed [26–28]. Component-specific estimates of vaccine protection would enhance understanding of the impact of virus variation and improve the generalizability of results to areas experiencing a different mix of viruses. We describe a sentinel surveillance platform to link detection of new variants and the composite of circulating influenza viruses to the annual estimation of component-specific VE.

Previous SectionNext Section

Methods

Sentinel influenza surveillance system. Sentinel physician networks in British Columbia, Alberta, and Quebec, Canada, contributed to the study. Participating sentinel physicians were provided with instruction manuals, data capture materials, and swab sample kits for collecting specimens from individuals who presented with influenzalike illness (ILI). ILI was defined by national surveillance criteria as the acute onset of fever and cough and one or more of the following: sore throat, arthralgia, myalgia, or prostration [29]. Sentinel physicians submitted respiratory specimens (nasal swab samples were preferred, but throat swab samples were accepted) for influenza testing along with epidemiologic information collected from consenting individuals who presented with ILI within 7 days after onset. Provincial and national ethics boards approved the study.

Vaccination. In participating provinces, influenza vaccination is provided free to elderly persons ⩾65 years of age and persons with designated high-risk conditions, as well as their household contacts and care providers; others are encouraged to receive vaccination, but must purchase it [30]. Split TIV containing 15 µg HA for each strain is recommended as a single 0.5 mL dose administered intramuscularly to persons ⩾9 years of age. The 2006–2007 TIV components included A/New Caledonia/ 20/1999(H1N1)—like, A/Wisconsin/67/2005(H3N2)—like, and B/Malaysia/2506/2004(Victoria)–like strains [30]. During 2006–2007, national vaccine allotment was such that >80% of TIV distributed in participating provinces was supplied by GlaxoSmith-Kline (Fluviral). Study participants received the commercially available TIV in accordance with usual vaccination practice during the regular fall vaccination campaign. Vaccination status and timing were then determined on the basis of history elicited from the participant. Participants were considered to have been immunized if vaccine was given ⩾2 weeks prior to ILI onset [31].

Laboratory testing. All submitted specimens were tested for influenza A/B at the respective provincial laboratories by real-time reverse-transcriptase polymerase chain reaction (RT-PCR) [32, 33]. Specimens identified as positive for influenza A were further tested by real-time RT-PCR to determine the subtype (H3 or H1). All respiratory specimens obtained in British Columbia, all specimens from Quebec that were positive for influenza by real-time RT-PCR, and a proportion of specimens from Alberta that were positive for influenza by RT-PCR were inoculated into conventional cell culture for virus isolation. Virus isolates were submitted to the National 7/22/2010 12:12:45 PM Microbiology Laboratory, Manitoba, for strain characterization by hemagglutination inhibition (HI) assay. HI testing was performed by using postinfection fowl sera against the specific 2006–2007 vaccine strains [30]. An isolate was identified as antigenically most similar to a prototype strain in accordance with the reciprocal of the highest HI titer [34]. If specimens showed 4-fold reduced titers in response to the vaccine strain, HI testing was also performed against specific variants including A/SolomonIslands/3/2006(H1N1)—like, A/Brisbane/ 10/2006(H3N2)—like, and B/Shanghai/361/2002 (Yamagata)—like strains [35].

Viral RNA was extracted from H3N2 cell culture lysates and the HA1 gene was amplified by real-time RT-PCR. Two sets of primers were used, resulting in 2 amplicons that overlap by ∼300 bases. For each isolate, the amplicons were sequenced in both the 3' and 5' directions [36]. Sequences were annotated and compared with recently circulating H3N2 viruses.

Inclusion criteria. The residents of participating provinces were eligible if they presented to a sentinel physician within 7 days after ILI onset, between 20 November 2006 (week 47) and 30 April 2007 (week 17) (table 2). In Quebec, participation began at the end of January (week 5). Respiratory specimens that were accompanied by complete questionnaire information were included. Both the vaccination schedule and the vaccine dose are different for children <9 years of age, and for this reason, their specimens were not included in VE analysis [30]. Duplicate visits withi positiven a 2-week period were consolidated into a single record, and test results were considered for influenza if either result was positive.

VE analysis. VE was estimated by use of case-control “testnegative” design [27]. Case subjects were individuals who provided samples in which influenza was identified; control subjects were individuals who provided samples in which neither influenza A nor B was identified. For estimates of VE against influenza B, records positive for influenza A were excluded, and vice versa; a similar approach was followed for subtype analysis.

Proportions were compared by use of the ϰ2 test. Medically attended, laboratory-confirmed influenza is uncommon, and for this reason, case subjects are a small fraction of the total source population. In addition, control subjects are a proportion of the total population that reflects vaccine coverage and is anticipated to be stable over the study period. In that context, the odds ratio (OR) is appropriate to estimate relative risk [37]. Logistic regression was thus used to estimate the odds ratio for laboratory-confirmed influenza in vaccinated versus unvaccinated participants, and VE was estimated as 1-[odds ratio for influenza in vaccinated versus nonvaccinated persons]] [21]. Covariate adjustment included the following 4 age groups: 9–19 years (school age), 20–49 years (young working age), 50–64 years (older working age), and ⩾65 years (elderly). Other covariates included the presence of chronic conditions (binary for vaccine-eligible condition [30]), interval from ILI onset to specimen collection (0–4 days, 5–6 days, or 7 days), province, month, and swab sample site.

Previous SectionNext Section

Participation Profile

There were 64 contributing sentinel physicians in British Columbia, 53 in Alberta, and 30 in Quebec. Over the study period, 1127 respiratory specimens were submitted for influenza testing. After exclusion criteria were applied, 841 participants were analyzed: 486 were from British Columbia, 235 from Alberta, and 120 from Quebec (table 2).

Of the 829 participants whose sex was known, 449 (54%) were women, and the median age was 36 years (range, 9–97 years). One hundred fifty-nine participants (19%) were school age, 482 (57%) were young working age, 131 (16%) were older working age, and 69 (8%) were elderly. Overall, 115 (14%) had a chronic condition, though the percentage of participants with a chronic condition varied significantly by age group (P >.001); it was higher among elderly indiviudals (32 [46%] of 69), compared with older (30 [23%] of 131) or younger (46 [10%] of 482) working-age adults and school-aged children (7 [4%] of 159).

Of 841 participants, 193 (23%) received the 2006–2007 TIV and 166 (20%) received vaccine ⩾2 weeks prior to ILI onset; for the purpose of VE analysis, only the latter were considered to have been immunized. Of the 504 participants who tested negative for influenza, 151 (30%) received the 2006–2007 TIV– comparable to the vaccination coverage in participating provinces estimated through a separate survey [38]. The proportion of participants who had been vaccinated did not vary significantly by month or between provinces. The percentage of immunized individuals varied significantly by age (P <.001); it was higher among elderly individuals (51 [74%] of 69), compared with older (39 [30%] of 131) or younger working-age adults (61 [13%] of 482) or school-age children (15 [9%] of 159). The OR of being vaccinated for persons with chronic conditions compared to those without chronic conditions decreased with age. The OR was 8.8-fold (43% vs. 8%; P = .02) in schoolage children, 3.2-fold (28% vs. 11%; P < .001) in younger working-age adults, 2.7-fold (47% vs. 25%; P = .02) in older working-age adults, and 1.1-fold (75% vs. 73%; P > .05) in elderly adults.

Of 166 participants immunized, 134 (81%) presented with ILI within 0–4 days after onset, 21 (13%) presented within 5–6 days, and 11 (7%) presented at 7 days, compared with 597 (88%) of 675 unvaccinated individuals who presented with ILI within 0–4 days after onset, 65 (10%) who presented within 5–6 days, and 13 (2%) who presented at 7 days (P = .02). There was no significant difference in the timeliness of medical visit by age group or presence of chronic condition.

Previous SectionNext Section

Influenza Identification

Provincial profiles. The weekly incidence of influenza detection by province, study week, and subtype is shown in figure 1A–1C; overall influenza activity mirrored patterns described elsewhere during the 2006–2007 season (table 1) [18, 34]. British Columbia experienced mostly influenza A activity (203 [95%] of 214 PCR-positive specimens): H3N2 was recovered from 172 (85%) of 203 specimens and H1N1 from 31 (15%) specimens, with limited B activity throughout (figure 1A and table 3). Alberta also experienced primarily influenza A (66 [93%] of 71 PCR-positive specimens), but this comprised ∼60% H1N1 detected mostly before January and ∼40% H3N2 thereafter. Minimal B activity occurred in Alberta at season's end (figure 1B and table 3). In Quebec, influenza activity after February 1 was a mix of 60% A/H3N2 and 40% B (figure 1C and table 3).

Figure 1.
View larger version:
Figure 1.

Influenza detection by subtype and week, 2006–2007. A, British Columbia; B, Alberta (4 influenza A isolates of unknown subtype collected during weeks 5, 8, 10, and 13 were not included); C, Quebec (participation did not begin until week 5). The study period extended from 20 November 2006 (week 47) to 30 April 2007 (week 17).

Figure 2.
View larger version:
Figure 2.

Amino-acid sequences of the HA1 gene from H3N2 isolates. A total of 111 isolates were included: isolates with the same sequence are grouped together, and the number of isolates in each group is indicated in parentheses. All sequences were trimmed to 347 aa, encoded by nucleotides 49 to 1089. Sequence alignment was performed with the ClustalW algorithm, and the phylogenetic tree was generated by using MegAlign (version 5.05; DNAStar)

Table 1.
View larger version:
Table 1.

Influenza viruses characterized and percentage similar to vaccine components, 2000–2007, Canada.

Table 2.
View larger version:
Table 2.

Study exclusion and inclusion criteria.

Table 3.
View larger version:
Table 3.

Laboratory profile of respiratory specimens included in analysis, according to location.

Virus contribution. Among the 29 H1N1 isolates available for strain characterization, all but 1 was well matched to the vaccine component (table 3). Among 8 influenza B isolates, all were lineage-level mismatched to the vaccine component (titers <1:40) and characterized as B/Shanghai/361/2002(Yamagata)—like. All 110 H3N2 isolates were initially characterized by HI as similar to the vaccine component (A/Wisconsin/67/2005 [H3N2]—like). However, 39 (44%) of 89 H3N2 isolates in British Columbia demonstrated 4-fold reduced titers to the vaccine component: compared to a reference titer of 1:640, titers among these isolates ranged between 1:80 (3 of 39) and 1:160 (36 of 39) for A/Wisconsin. Similarly, in Alberta, 2 (17%) of 12 H3N2 isolates had HI titers of 1:320 to A/Wisconsin, compared with a reference titer of 1:1280. None of the 9 H3N2 isolates recovered in Quebec showed reduced HI titers to A/Wisconsin. H3N2 isolates were also analyzed through sequencing of the HA1 gene, including 94 from British Columbia, 12 from Alberta, and 5 from Quebec (table 3 and figure 2). These isolates were collected between December and April, from participants ranging in age from 9 to 73 years. With 1 exception, sequence analysis revealed equal clustering of all H3N2 isolates as A/Brisbane/10/ 2006(H3N2)—like and A/Nepal/921/2006(H3N2)—like variants (figure 2). This included 64 isolates (27 A/Brisbane and 37 A/Nepal) that had not initially shown reduced HI titers to A/Wisconsin. When ferret antisera for A/Brisbane/10/2006(H3N2)—like virus became available much later, repeated HI assays that used antisera for both A/Wisconsin/67/2005(H3N2)—like and A/Brisbane/ 10/2006(H3N2)—like viruses identified >50% of the remaining available H3N2 isolates to have been the latter drift variant, including specimens collected as early as December 2006 (table 1).

Epidemiologic covariates. Influenza detection varied significantly by age category (P <.001), including variation with respect to subtype distribution (P = .02) (table 4). Of the 55 H1N1 isolates recovered, 40% were recovered from school-age children, 54% from young working-age adults, 4% from older working-age adults, and 2% from elderly individuals. Comparable percentages of distribution for the 242 H3N2 isolates recovered were 24%, 60%, 12%, and 4%, respectively. The 36 influenza B isolates detected were distributed among the age groups as follows: 8%, 69%, 19%, and 3%, respectively. The rate of influenza detection was higher among those without chronic conditions (P = .01), compared to those with such conditions. The rate of detection for influenza A/H3N2 varied with the timeliness of medical visit ; 224 (31%) 727 specimens obtained at 0–4 days after onset were positive, 15 (17%) of 86 specimens obtained at 5–6 days were positive, and 3 (12%) of 24 specimens obtained at 7 days (P = .003). Detection of A/H1N1 and B did not vary with timeliness of medical visit.

Table 4.
View larger version:
Table 4.

Influenza detection by subtype, age, chronic conditions and vaccination status.

VE estimation. Component-specific and overall VE estimates are shown in table 5. Adjustment for age reduced VE estimates, whereas other covariates, including chronic conditions, had little influence. Age-adjusted VE for the A/H1N1, A/H3N2, and B components was 92% (95% CI, 40%–99%), 41% (95% CI, 6%–63%), and 19%(95%CI, –112% to 69%), respectively, with an overall VE of 47% (95% CI, 18%–65%). Analysis restricted to include only working-age adults gave lower VE estimates with wide confidence intervals but similar component-specific trends.

Table 5.
View larger version:
Table 5.

Overall and component-specific vaccine effectiveness estimates.

Previous SectionNext Section

Discussion

To our knowledge, this is the first public health initiative in North America to use an existing sentinel surveillance system to assess VE against laboratory-confirmed influenza. With the contribution of respiratory specimens and virus isolates from a broad geographic area, this study is also the first to report separate but simultaneous VE estimates for each component of the trivalent vaccine, as far as we know. Through this combined surveillance approach, we were able to detect and characterize new influenza variants, assess their prominence in a well-defined and representative group, and consider their impact on vaccine protection in a real-world context.

We demonstrated very different estimates of VE for the H1N1, H3N2, and B components of the 2006–2007 TIV formulation in Canada. For the seventh consecutive year in Canada, the HA of the H1N1 component was well matched to circulating virus; consistent with this, VE for that component was high. Conversely, strain mismatch to the vaccine component was detected among half of the H3N2 isolates assessed, with evidence of reduced protection. Finally, the B component was substantially mismatched to circulating virus at the lineage level, with further evidence of reduced protection. Although sample size was insufficient for precise estimation, the good concordance we observed between separate degrees of component-specific mismatch and separate estimates of component-specific VE is reassuring with respect to internal validity. Using the same methodology to estimate VE for each vaccine component, we may expect that underlying biases would apply to each estimate. The same biases would then have to pull in dramatically opposite directions to explain the divergent component-specific VE estimates we found.

In our study, the profile of circulating influenza viruses varied substantially by participating province. Against that backdrop, and given variation in component-specific VE, vaccine protection is also likely to have varied by geographic area across the 2006–2007 winter. In British Columbia, H3N2, including a substantial proportion of drift variants, dominated throughout. In Alberta, protection is expected to have been high in association with H1N1 activity before January but may have dropped thereafter as H3N2—both vaccine and drift strains—and finally lineage-mismatched B viruses became more prominent at season's end. In Quebec, protection after February reflected a different mix, consisting of 60% H3N2 and 40% B viruses. These results illustrate the limits of reporting a single estimate of VE as representative of TIV protection in all areas, across and throughout seasons. Instead, VE against influenza requires more nuanced analysis than may be required for any other vaccine preventable disease. Routine component-specific analysis could better support regional interpretation of vaccine protection on the basis of varying profiles of virus circulation. Furthermore, it is necessary to know the profile of circulating influenza viruses and variation in VE between regions and from year to year to evaluate and compare program changes (targeted or universal) at the population level over space and time.

Antigenic drift within the influenza A/H3N2 virus occurred in western Canada as early as December 2006, but this was initially suspected only through special investigations, including the recognition of reduced HI titers to the vaccine component and gene sequence analysis. The precise identification of drift variants by HI assay is otherwise limited midseason by the availability of specific antisera to newly emerging strains. Initial sequencing and HI assay results were not consistently correlated, and final confirmation of antigenic drift was possible only after specific reference antisera for the A/Brisbane/10/2006(H3N2)—like drift variant later became available. Identification of A/Brisbane among approximately half of H3N2 isolates available was in keeping with the suboptimal estimates of VE we found for that component of the 2006–2007 vaccine. By then, however, the World Health Organization had already designated the 2007– 2008 vaccine components, choosing to retain the same H3N2 and B components from the 2006–2007 formulation [39]. During the southern hemisphere's influenza season (May—September 2007), the A/Brisbane/10/2006(H3N2)—like drift variant accounted for an increasing proportion (80%) of circulating A/H3N2 viruses, and virtually all B viruses belonged to the Yamagata lineage, with some areas reporting higher than expected rates of influenza illness [40, 41]. A high rate of H3N2 and B antigenic mismatch continued into the northern hemisphere's 2007–2008 influenza season [42–44]. Our results for 2006–2007 suggest that areas of the northern hemisphere experiencing substantial activity due to A/Brisbane/10/2006(H3N2)—like strains (as in the United States) or activity due to B/Yamagata strains (most regions) during the 2007–2008 season would have experienced suboptimal vaccine protection against these viruses—as subsequently shown in interim 2007–2008 VE estimates since published for the United States [42 –45].

Since reemergence of the B/Victoria lineage in 2001, the B component of TIV has been lineage mismatched to circulating virus during 4 of 6 subsequent seasons. Little or no HA-based serum cross-protection is anticipated between B lineages, and the relative importance of this will vary with the proportionate contribution of HA-mismatched B virus to overall influenza activity each season (table 1). As a result of reassortment in 2003, all B viruses, both Victoria and Yamagata lineages, share the same Yamagata neuraminidase (NA) protein. The amount of NA in TIV, however, is not standardized, so NA-based vaccine cross-protection, although theoretically possible, cannot be assured [6, 19]. Given frequent lineage-level vaccine mismatch, it may be tempting to consider inclusion of HA from both B lineages in a quadrivalent vaccine formulation. If feasible, however, the incremental benefit of such an approach is likely to be limited by the high cost of new vaccine development and the typically lower impact overall of influenza B [46]. During the 2006–2007 season, B virus comprised approximately 10% of all influenza detections in Canada. With influenza B contributing so little to the current study, no firm conclusions can be drawn. Better understanding of vaccine benefit against influenza B is needed, and further evaluation is therefore warranted.

Our methodology was predicated on certain assumptions and accompanied by caveats inherent to observational designs. Vaccine status was determined on the basis of participant or sentinel physician report without validation. We assumed that vaccinated persons in the community have the same likelihood of being exposed to influenza as unvaccinated persons, and that members of each group present as frequently to a physician if they develop ILI of similar severity. Because case and control subjects both emerge from the group of those presenting with ILI, we partly address the bias of health-seeking behavior. Our estimates mostly reflect the protection conferred to young persons with few elderly participants. Overestimation of VE against serious outcomes has been reported among elderly persons as a result of better general health status among the vaccinated, compared with the unvaccinated. It is unclear whether this bias may also apply to younger individuals seeking medical consultation [47, 48]. Direct age-stratified analyses and comparisons were precluded by insufficient sample size, but it is of note that adjustment for age and restriction of analysis to working-age adults resulted in a general lowering of VE estimates. This may be explained by instability due to small sample size and/or systematic differences in the distribution of vaccination and disease risk (susceptibility and severity) associated with both known and unknown factors. Current protection also likely reflects the diverse repertoire of antigenic exposure and immune response that is accumulated and sustained through boost across the life span. We did not record prior influenza or vaccination history, but repeated vaccinations over multiple years may be more strongly associated with current vaccination among elderly persons or those with comorbidities. The potential influence of crossmemory may be especially relevant to consider during seasons of significant vaccine mismatch.

Most countries conduct influenza surveillance, including laboratory and epidemiologic contributions, to monitor virus evolution and impact on patterns of clinical ILI. Given rapidly changing viruses, annual vaccine reformulation, and the increasing inclusiveness of vaccination programs, influenza surveillance activities should also include routine assessment of VE. It is clear that component-specific, age-stratified, or other stratified VE analysis requires larger sample size, which is most efficiently attained through observational study design. We illustrate how this could be achieved annually by using the broad platform of sentinel surveillance networks. Such a system cannot address all the uncertainties inherent in observational designs, and results will remain imperfect compared with those obtained in RCTs. On the other hand, annual RCTs are not possible. Given that this is the case, the proposed surveillance approach offers the advantage of a sustainable system for comparative trend analysis across vaccine components, seasons, and geographic areas. Given that there is so much inherent variability associated with influenza, consideration of trends may, in fact, be more relevant, reliable, and informative than the literal interpretation of individual point estimates. The annual use of this surveillance approach for VE estimation by multiple jurisdictions has the potential to simultaneously establish baseline measures for ongoing comparison, contribute to methodologic refinement, improve the selection of vaccine components, and provide important immuno-epidemiologic and program insights. We encourage its further development and expansion as an integral component of annual influenza surveillance activities.

Previous SectionNext Section

Acknowledgments

We wish to acknowledge the contribution of sentinel sites in each participating province.

Previous SectionNext Section

Footnotes

  • Potential conflicts of interest: G.D.S. has been a coinvestigator on an unrelated research project funded by GlaxoSmithKline. No other authors report relevant conflicts of interest.

  • Financial support: British Columbia Centre for Disease Control; Alberta Health and Wellness; Institut national de santé publique du Québec; Public Health Agency of Canada and the Canadian Institute for Health Research.

  • Nucleotide sequences for influenza A/H3N2 isolates have been submitted to the International Nucleotide Sequence Databases with accession numbers of EU659819–EU659851.

  • Received April 17, 2008.
  • Accepted August 21, 2008.
Previous Section
 

References

  1. 1.↵
    1. Nicholson KG,
    2. Webster RG,
    3. Hay AJ
    1. Wood JM
    . Standardization of inactivated influenza vaccines. In: Nicholson KG, Webster RG, Hay AJ, editors. Textbook of influenza. Malden, MA: Blackwell Science Ltd.,; 1998. p. 333-45.
  2. 2.↵
    1. Jefferson TO,
    2. Rivetti D,
    3. Di Pietrantonj C,
    4. Demicheli V.
    Vaccines for preventing influenza in healthy adults. Cochrane Database Syst Rev 2007;CD001269.
  3. 3.
    1. Bridges CB,
    2. Thompson WW,
    3. Meltzer MI,
    4. et al
    . Effectiveness and cost-benefit of influenza vaccination of healthy working adults: a randomized controlled trial. JAMA 2000;284:1655-63.
    Abstract/FREE Full Text
  4. 4.↵
    1. Skowronski DM,
    2. Gilbert M,
    3. Tweed SA,
    4. et al
    . Effectiveness of vaccine against medical consultation due to laboratory-confirmed influenza: results from a sentinel physician pilot project in British Columbia, 2004–05. Can Commun Dis Rep 2005;31:181-92.
    Medline
  5. 5.
    1. Russell KL,
    2. Ryan MA,
    3. Hawksworth A,
    4. et al
    . Effectiveness of the 2003–2004 influenza vaccine among U.S. military basic trainees: a year of suboptimal match between the vaccine and circulating strain. Vaccine 2005;23:1981-5.
    CrossRefMedlineWeb of Science
  6. 6.↵
    1. Skowronski DM,
    2. Masaro C,
    3. Kwindt TL,
    4. et al
    . Estimating vaccine effectiveness against laboratory-confirmed influenza using a sentinel physician network: results from the 2005–2006 season of dual A and B vaccine mismatch in Canada. Vaccine 2007;25:2842-51.
    CrossRefMedlineWeb of Science
  7. 7.
    1. Herrera GA,
    2. Iwane MK,
    3. Cortese M,
    4. et al
    . Influenza vaccine effectiveness among 50–64 year-old persons during a season of poor antigenic match between vaccine and circulating influenza virus strains: Colorado, United States, 2003–2004. Vaccine 2007;25:154-60.
    CrossRefMedlineWeb of Science
  8. 8.
    1. Shuler CM,
    2. Iwamoto M,
    3. Bridges CB,
    4. et al
    . Vaccine effectiveness against medically attended, laboratory-confirmed influenza among children aged 6 to 59 months, 2003–2004. Pediatrics 2007;119:e587-e595.
    Abstract/FREE Full Text
  9. 9.
    1. Allison MA,
    2. Daley MF,
    3. Crane LA,
    4. et al
    . Influenza vaccine effectiveness in healthy 6- to 21-month-old children during the 2003–2004 season. J Pediatr 2006;149:755-62.
    CrossRefMedlineWeb of Science
  10. 10.↵
    1. Nichol KL,
    2. Nordin JD,
    3. Nelson DB,
    4. et al
    . Effectiveness of influenza vaccine in the community-dwelling elderly. New Engl J Med 2007;357:1373-81.
    CrossRefMedline
  11. 11.↵
    1. Macey JF,
    2. Zabchuk P,
    3. Winchester B,
    4. Tam TWS
    . Influenza in Canada: 2000–2001 season. Can Commun Dis Rep 2002;28:17-28.
    Medline
  12. 12.
    1. Macey JF,
    2. Tam TWS,
    3. Li Y,
    4. Winchester B,
    5. Zabchuk P
    . Influenza in Canada: 2001–2002 season. Can Commun Dis Rep 2003;29:45-60.
    Medline
  13. 13.
    1. Aziz S,
    2. Tam TWS,
    3. Macey JF,
    4. et al
    . Appendix A: summary of 2002–2003 influenza season. Can Commun Dis Rep 2005;31:17-20.
  14. 14.
    National Advisory Committee on Immunization. Statement on influenza vaccination for the 2003–04 season. Can Commun Dis Rep ACS-4 2003;29:1-20.
  15. 15.
    1. Aziz S,
    2. Tam TWS,
    3. Macey JF,
    4. et al
    . Influenza in Canada: 2003–04 season. Can Commun Dis Rep 2005;31:1-17.
    Medline
  16. 16.
    1. Xie L,
    2. Squires SG,
    3. Macey JF,
    4. et al
    . Influenza in Canada: 2004–2005 season. Can Commun Dis Rep 2006;32:57-74.
    Medline
  17. 17.
    1. Reyes F,
    2. Macey JF,
    3. Aziz S,
    4. et al
    . Influenza in Canada: 2005–2006 season. Can Commun Dis Rep 2007;33:21-41.
    Medline
  18. 18.↵
    1. Reyes F,
    2. Aziz S,
    3. Li Y,
    4. et al
    . Influenza in Canada: 2006–2007 season. Can Commun Dis Rep 2008;34:1-25.
    Medline
  19. 19.↵
    1. Shaw MW,
    2. Xu X,
    3. Li Y,
    4. et al
    . Reappearance and global spread of variants of influenza B/Victoria/2/87 lineage viruses in the 2000–2001 and 2001–2002 seasons. Virology 2002;303:1-8.
    CrossRefMedlineWeb of Science
  20. 20.↵
    1. Lin YP,
    2. Gregory VP,
    3. Bennett M,
    4. Hay A
    . Recent changes among influenza viruses. Virus Res 2004;103:47-52.
    CrossRefMedlineWeb of Science
  21. 21.↵
    1. Orenstein WA,
    2. Bernier RH,
    3. Dondero TJ,
    4. et al
    . Field evaluation of vaccine efficacy. Bull World Health Organ 1985;63:1055-68.
    MedlineWeb of Science
  22. 22.
    1. Carrat F,
    2. Flahault A,
    3. Boussard E,
    4. Farran N,
    5. Dangoumau L,
    6. Valleron AF
    . Surveillance of influenza-like illness in France: the example of the 1995/1996 epidemic. J Epidemiol Community Health 1998;52:32-38.
  23. 23.
    1. Castilla J,
    2. Arregui L,
    3. Baleztena J,
    4. et al
    . Incidence of influenza and influenza vaccine effectiveness in the 2004–2005 season [in Spanish]. An Sist Sanit Navar 2006;29:97-106.
    Medline
  24. 24.↵
    1. Uphoff H,
    2. Hauri AM,
    3. Schweiger B,
    4. et al
    . Estimation of influenza vaccine effectiveness using routine surveillance data. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2006;49:287-95.
    CrossRefMedline
  25. 25.↵
    1. Hak E,
    2. Verheij TJ,
    3. Grobbee DE,
    4. Nichol KL,
    5. Hoes AW
    . Confounding by indication in non-experimental evaluation of vaccine effectiveness: the example of prevention of influenza complications. J Epidemiol Community Health 2002;56:951-55.
    Abstract/FREE Full Text
  26. 26.↵
    1. Halloran ME,
    2. Longini IM Jr.
    . Using validation sets for outcomes and exposure to infection in vaccine field studies. Am J Epidemiol 2001;154:391-8.
    Abstract/FREE Full Text
  27. 27.↵
    1. Orenstein EW,
    2. De Serres G,
    3. Haber MJ,
    4. et al
    . Methodologic issues regarding the use of three observational study designs to assess influenza vaccine effectiveness. Int J Epidemiol 2007;36:623-31.
    Abstract/FREE Full Text
  28. 28.↵
    1. Simonsen L
    . Commentary: Observational studies and the art of accurately measuring influenza vaccine benefits. Int J Epidemiol 2007;36:631-2.
    FREE Full Text
  29. 29.↵
    Public Health Agency of Canada. FluWatch. Definitions for the 2006–2007 season. [3 December 2008]. Available at: http://www.phac-aspc.gc.ca/fluwatch/06-07/def06-07_e.html.
  30. 30.↵
    National Advisory Committee on Immunization. Statement on influenza vaccination for the 2006–07 season. Can Commun Dis Rep ACS-7 2006;32:1-28.
  31. 31.↵
    1. Nicholson KG,
    2. Webster RG,
    3. Hay AJ
    1. Nichol KL
    . Efficacy/clinical effectiveness of inactivated influenza virus vaccines in adults. In: Nicholson KG, Webster RG, Hay AJ, editors. Textbook of influenza. Malden, MA: Blackwell Science Ltd.; 1998. p. 358-72.
  32. 32.↵
    1. Schweiger B,
    2. Zadow I,
    3. Heckler R,
    4. Timm H,
    5. Pauli G
    . Application of a fluorogenic PCR assay for typing and subtyping of influenza viruses in respiratory samples. J Clin Microbiol 2000;38:1552-8.
    Abstract/FREE Full Text
  33. 33.↵
    1. van Elden LJ,
    2. Nijhuis M,
    3. Schipper P,
    4. Schuurman R,
    5. van Loon AM
    . Simultaneous detection of influenza viruses A and B using real-time quantitative PCR. J Clin Microbiol 2001;39:196-200.
    Abstract/FREE Full Text
  34. 34.↵
    1. Kendal AP,
    2. Pereira MS,
    3. Skethel JJ
    . Concepts and procedures for laboratory-based influenza suveillence. Atlanta, GA: Centers for Disease Control and Prevention; 1982. US Department of Health and Human Services, Public Health Service.
  35. 35.↵
    National Advisory Committee on Immunization. Statement on influenza vaccination for the 2007–08 season. Can Commun Dis Rep ACS-7 2007;33:1-38.
  36. 36.↵
    1. Mak A,
    2. Rahmanian R,
    3. Lei V,
    4. et al
    . Longitudinal analysis of genotype distribution of influenza a virus from 2003 to 2005. J Clin Microbiol 2006;44:3583-8.
    Abstract/FREE Full Text
  37. 37.↵
    1. Rodrigues L,
    2. Kirkwood BR
    . Case-control designs in the study of common diseases: updates on the demise of the rare disease assumption and the choice of sampling scheme for controls. Int J Epidemiol 1990;19:205-13.
    Abstract/FREE Full Text
  38. 38.↵
    1. Kwong JC,
    2. Rosella LC,
    3. Johansen H
    . Trends in influenza vaccination in Canada, 1996/1997 to 2005. Health Reports (Statistics Canada Catalogue 82–003) 2007;18:1-11.
  39. 39.↵
    World Health Organization. Recommended composition of influenza virus vaccines for use in the 2007–08 influenza season. Wkly Epidemiol Rec. 2007. Available at: http://www.who.int/wer/2007/wer8209.pdf.
  40. 40.↵
    Centers for Disease Control. Update: influenza activity—United States and worldwide, May 20–September 15, 2007. MMWR Morb Mortal Wkly Rep. Vol. 56. 2007. p. 1001-4. Available at: http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5638a4.htm?s_cid-mm5638a4_e.
    Medline
  41. 41.↵
    1. Miller ER,
    2. Fielding JE,
    3. Grant KA,
    4. Barr IG,
    5. Papadakis G,
    6. Kelly H
    . Higher than expected seasonal influenza activity in Victoria, 2007. Commun Dis Intell 2008;32:63-70.
    Medline
  42. 42.↵
    Public Health Agency of Canada. FluWatch (2006–07). Available at: http://www.phac-aspc.gc.ca/fluwatch/index-eng.php.
  43. 43.
    Centers for Disease Control. MMWR Morb Mortal Wkly Rep. Vol. 57. 2008. Update: influenza activity—United States, September 30, 2007–April 5, 2008 and composition of the 2008–09 influenza vaccine; p. 404-8.
    Medline
  44. 44.↵
    Centers for Disease Control and Prevention. Flu activity and surveillance: FluView—a weekly influenza surveillance report prepared by the influenza division. Available at: http://www.cdc.gov/flu/weekly/fluactivity.htm.
  45. 45.↵
    Centers for Disease Control. MMWR Morb Mortal Wkly Rep. Vol. 57. 2008. Interim within-season estimate of the effectiveness of trivalent inactivated influenza vaccine—Marshfield, Wisconsin, 2007–08 influenza season; p. 393-8.
    Medline
  46. 46.↵
    1. Nicholson KG,
    2. Webster RG,
    3. Hay AJ
    1. Nicholson KG
    . Human influenza. In: Nicholson KG, Webster RG, Hay AJ, editors. Textbook of influenza. Malden, MA: Blackwell Science Ltd.; 1998. p. 219-64.
  47. 47.↵
    1. Jackson LA,
    2. Jackson ML,
    3. Nelson JC,
    4. Neuzil KM,
    5. Weiss NS
    . Evidence of bias in estimates of influenza vaccine effectiveness in seniors. Int J Epidemiol 2006;35:337-44.
    Abstract/FREE Full Text
  48. 48.↵
    1. Jackson LA,
    2. Nelson JC,
    3. Benson P,
    4. et al
    . Functional status is a confounder of the association of influenza vaccine and risk of all cause mortality in seniors. Int J Epidemiol 2006;35:345-52.
    Abstract/FREE Full Text
| Table of Contents

This Article

  1. J Infect Dis. (2009) 199 (2): 168-179. doi: 10.1086/595862
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 205 (5)
  1. Journal of Infectious Diseases
  1. Alert me to new issues

Most