key: cord-0712697-3d8euosy
authors: Janjua, Naveed Z.; Skowronski, Danuta M.; Hottes, Travis S.; Osei, William; Adams, Evan; Petric, Martin; Sabaiduc, Suzana; Chan, Tracy; Mak, Annie; Lem, Marcus; Tang, Patrick; Patrick, David M.; De Serres, Gaston; Bowering, David
title: Seasonal Influenza Vaccine and Increased Risk of Pandemic A/H1N1-Related Illness: First Detection of the Association in British Columbia, Canada
date: 2010-11-01
journal: Clin Infect Dis
DOI: 10.1086/656586
sha: da70820b39071e92858b5a5db50ae98eb581abe8
doc_id: 712697
cord_uid: 3d8euosy

Background. In April 2009, an elementary school outbreak of pandemic H1N1 (pH1N1) influenza was reported in a community in northern British Columbia, Canada-an area that includes both non-Aboriginal and Aboriginal residents living on or off a reserve. During the outbreak investigation, we explored the relationship between prior receipt of trivalent inactivated influenza vaccine (TIV) and pH1N1-related illness. Methods. A telephone survey was conducted from 15 May through 5 June 2009 among households of children attending any school in the affected community. Members of participating households where influenza-like illness (ILI) was described were then invited to submit blood samples for confirmation of pH1N1 infection by hemagglutination inhibition and microneutralization assays. Circulation of pH1N1 was concentrated among households of the elementary school and elsewhere on-reserve to which analyses of TIV effect were thus restricted. Odds ratios (ORs) for the TIV effect on ILI were computed through logistic regression, with adjustment for age, comorbidity, household density, and Aboriginal status. The influence of within-household clustering was assessed through generalized-linear-mixed models. Results. Of 408 participants, 92 (23%) met ILI criteria: 29 (32%) of 92 persons with ILI, compared with 61 (19%) 316 persons without ILI, had received the 2008–2009 formulation of TIV. Fully adjusted ORs for 2008- 2009 TIV effect on ILI were 2.45 (95% confidence interval, 1.34–4.48) by logistic regression and 2.68 [95% confidence interval, 1.37–5.25) by generalized-linear-mixed model. Conclusions. An outbreak investigation in British Columbia during the late spring of 2009 provided the first indication of an unexpected association between receipt of TIV and pH1N1 illness. This led to 5 additional studies through the summer 2009 in Canada, each of which corroborated these initial findings.

Telephone survey. After identification of the elementary school outbreak, the sampling frame for investigation was broadly inclusive of all households of students enrolled in the 6 local community schools. Approximately one-half of all households in the community included at least 1 child [1] .

Before initiating the telephone survey, letters were distributed via students to households to explain the purpose of the investigation. Trained interviewers then contacted households, obtained consent, and conducted telephone interviews with 1 adult per household, who provided information for all household members. The telephone survey was conducted between 15 May and 5 June 2009. Household information included the number of household members and sleeping rooms, Aboriginal versus non-Aboriginal status, and whether residency was on or off a reserve. Individual-level information included age, whether flulike symptoms were experienced and the related dates of onset, duration of time "in bed" (in days), and data on health care visits, comorbidity (corresponding to high-risk categories specified by the National Advisory Committee on Immunization Laboratory surveillance for respiratory viruses. All respiratory virus testing for this community was provided by the BC Centre for Disease Control (Vancouver, BC) Public Health Microbiology and Reference Laboratory (Appendix, which appears only in the electronic version of the journal). During initial outbreak investigation, nasal or nasopharyngeal specimens were collected on 29 April by public health staff from a sample of students attending the affected elementary school. Thereafter, specimens were collected during the outbreak period at the clinician's or public health official's discretion and were evaluated as part of routine surveillance.

Sero-survey of affected households. To validate the clinical case definition, community households with at least 1 member reporting ILI were subsequently invited to provide serum samples from both symptomatic and asymptomatic household members. An on-site clinic was arranged during 6-8 August 2009 . Antibody response to pH1N1 was quantified using the hemagglutination inhibition (HI) and microneutralization (MN) assays (Appendix). An HI threshold of у40 was used to designate seropositive versus seronegative participants, and status was confirmed using the MN assay [5] .

The primary focus of analyses presented here is estimation of the effect of receipt of the 2008-2009 TIV on risk of developing ILI during a documented pH1N1 outbreak. Symptom experience since 1 April 2009 was elicited. ILI was defined at the analysis stage as a report of fever and cough plus у1 of the following symptoms during that period: headache, general aches, sore throat, or prostration. Corresponding control sub-jects were participants who had been symptomatic in the period since 1 April 2009 who did not meet the ILI case definition or who were asymptomatic. Participants aged р6 months as of 31 December 2008 were excluded at the analysis stage, because they would not have been eligible to receive the vaccine.

On the basis of results of serologic tests, the sensitivity, specificity, and positive and negative predictive values for the ILI case definition were explored. Odds ratios (ORs) for seasonal influenza vaccine effect (2008-2009 and 2007-2008) on ILI were computed via logistic regression analysis, with adjustment for combinations of age, chronic conditions, Aboriginal status, and household density (calculated as the number of household members/number of sleeping rooms). We also accounted for within-household clustering while assessing vaccine effect by using generalized-linear-mixed models (GLMMs) for binary outcomes, adjusting for the same covariates [6] . Because surveillance data suggested that children experienced higher pH1N1 attack rates and that older adults were at lower risk [7, 8] , we explored vaccine effects stratified for participants aged !20 years and !50 years. We also explored the effect of TIV receipt on pH1N1 infection defined by HI and/or MN seropositive status.

The initial telephone survey was conducted as a public healthmandated outbreak investigation, with verbal consent provided at interview. The serologic component was reviewed and approved by the Research Ethics Board of the University of British Columbia, and individual written consent was obtained for blood sample collection and analysis.

Respiratory specimen surveillance. Respiratory virus testing by the BC Centre for Disease Control for the local community included 30 specimens collected during the period 29 April through 5 June 2009. pH1N1 was confirmed in 14 of these 30 specimens by reverse-transcription polymerase chain reaction (RT-PCR). Other respiratory viruses detected in the local community during that period included coronavirus (in 2 specimens) and rhinovirus or enterovirus (in 2 specimens).

Of the 30 specimens submitted, 9 were from households of the affected elementary school that also participated in the telephone survey; pH1N1 was detected in 6 (67%) of the 9 specimens from survey participants. Five (56%) of these 9 specimens were from patients living on a reserve; pH1N1 was detected in 4 (80%) of the 5 specimens. Of note, surveillance data indicated the last detection of seasonal influenza (A/H3N2) in the local health area was in February 2009. Serologic test results. In total, 135 households with at least 1 member with ILI identified during the community survey were invited to participate in the serologic study. Ultimately, 42 households contributed serologic specimens, resulting in 106 individual serum samples available for analysis, including 58 (54%) from households associated with the affected elementary school (n p 45) or on a reserve (n p 29). Details of the serosurvey participants are shown in Table A1 in the Appendix.

In total, 44 (42%) of the 106 serologic survey participants reported ILI during the study period. Of the 106 serum samples, 28 (26%) had HI titers to pH1N1 у40 (denoting seropositivity), and of these, 22 (79%) were from persons who had reported ILI. Among the 106 serologic survey participants overall, there was strong correlation between log-transformed HI and MN titers (r p 0.92). Of the 28 specimens with an HI titer у40, all but 3 had MN titers у80 and exceeding HI. Of the 3 participants whose specimens yielded MN titers less than the HI titers, none reported ILI.

Of the 45 participants from the affected elementary school households, 14 (31%) were seropositive for pH1N1 (13 of 14 persons with ILI). Of the 29 on-reserve participants, 13 (45%) were seropositive for pH1N1 (12 of 13 persons with ILI). Of the 48 serum samples from participants belonging to off-reserve households without children in the affected elementary school, 12 (25%) of 48 were seropositive for pH1N1 (8 of 12 persons with ILI).

Clinical case definitions: characteristics in relation to serologic status. To guide analyses, we assessed the sensitivity, specificity, and positive and negative predictive values of the ILI case definition in relation to HI or MN seropositive status (Table 1) . Parameters were highest in combination among participants belonging to households associated with the initial elementary school outbreak or who lived on a reserve. The ILI case definition had less value in predicting pH1N1 seropositive status among telephone survey participants who belonged to households other than those associated with the elementary school or who lived on a reserve (positive predictive value, 38%, 68%, and 75%, respectively) ( Table 1) .

Overall, 266 households and 1154 individuals contributed to the community telephone survey. Because respiratory virus surveillance and the follow-up serologic survey both indicated greatest pH1N1 circulation among the originally affected elementary school and on-reserve households, we restricted analyses of TIV effect on ILI to those households, hereafter referred to as "elementary school" or "on-reserve" participants. The epidemic curve of ILI, by date of onset for the elementary school and/or on-reserve participants, is shown in Figure 1 .

Elementary school household participants. There were 118 households with at least 1 child enrolled at the affected elementary school. Sixty-three of these households participated in the survey, contributing data on 271 individuals. After exclud-ing those with unknown 2008-2009 TIV status (n p 17) and unknown chronic conditions (n p 1), the analysis included 253 participants associated with the elementary school. Among these participants, 36 (15%) lived on one of the local reserves. Overall, 153 (60%) of 253 persons were aged !20 years. There were few older adults (11 [4%] ) aged у50 years.

Of the 253 elementary school household participants, 66 (26%) reported ILI, with the rate highest among young children ( Table 2 ). The secondary attack rate among elementary school households was 27% (32 of 119); after excluding on-reserve households, the rate was 20% (19 of 94). Only 1 participant reported travel to Mexico since mid-March 2009. The proportion in the fourth quartile of household density was greater among participants with ILI than among those without ILI (40% vs 20%). The proportion of participants with chronic conditions was comparable to the BC proportion overall for children and young adults (∼10%) [9] . Self-reported influenza immunization rates among non-ILI control subjects were also comparable to BC rates estimated through other surveys among children and young adults (15%-20%) in British Colombia [10] .

On-reserve household participants. Two hundred twentyfour individuals who participated in the community survey were reported as living on a surrounding reserve. After excluding individuals with unknown 2008-2009 TIV status (n p 29), unknown chronic condition (n p 1), or age р6 months on 31 December 2008 (n p 4), the on-reserve analysis included 191 survey participants, of whom 100 (52%) were aged !20 years and 26 (14%) were aged у50 years.

Of the 191 on-reserve participants, 44 (23%) reported having ILI, again with the rate highest among young children ( Table  2 ). The secondary attack rate among on-reserve households was 24% (23 of 94). None of these participants reported travel to Mexico since mid-March. The proportion of participants living in households in the fourth quartile of density was greater for on-reserve participants (39%) than for elementary school participants (25%) and among on-reserve participants with ILI (50%) than among those without ILI (35%). A higher proportion of on-reserve participants with ILI sought medical care ( 16) . No participant in either population was prescribed antivirals, and none were hospitalized.

A higher proportion of Aboriginals in British Colombia and Canada, compared with the general population, have at least 1 chronic condition [11, 12] . Accordingly, the proportion of young on-reserve participants in our study with at least 1 chronic condition (27 [14%] of 191) was higher than the proportion of elementary school participants and was within the NOTE. Seropositivity was defined as a hemagglutination inhibition assay titer у40, and seronegativity was defined as a hemagglutination inhibition titer !40. NPV, negative predictive value; NSS, insufficient sample size owing to no on-reserve participants without ILI who were seronegative; PPV, positive predictive value. a ILI was defined at the analysis stage as a report of fever and cough plus у1 of the following symptoms since 1 April 2009: headache, general aches, sore throat, or prostration.

expected range overall for young Aboriginals who lived on a reserve. Influenza vaccination rates were higher among Aboriginals who lived on a reserve in British Colombia overall (45%) than among the general population (30%) [13] . Thus, among non-ILI control subjects in our survey, young on-reserve Aboriginals had higher influenza immunization rates (42 [29%] of 147 persons) than did elementary school household participants (26 [14%] of 187 persons) ( Table 2) .

Fully adjusted overall and stratified ORs for the effect of 2008-2009 TIV on ILI risk are shown in Table 3 . Fully-adjusted OR overall was 2.45 (95% confidence interval [CI], 1.34-4.48) when estimated by logistic regression and 2.68 (95% CI, 1.37-5.25) by GLMM. ORs were higher with restriction to younger participants aged !20 years or !50 years and to on-reserve participants, especially with similar age restriction (Table 3) . Although ORs were higher among Aboriginals who lived on a reserve than among non-Aboriginal participants, an analysis for interaction did not reach statistical significance (P 1 .05).

In additional sensitivity analyses based on logistic regression and restricted to the period of peak ILI activity May 2009), the overall OR for 2008-2009 TIV effect adjusted for age, chronic conditions, Aboriginal status, and household density was higher (3.55; 95% CI, 1.70-7.34). When we used control subjects defined as fully asymptomatic persons rather than those who merely lacked ILI, the overall adjusted OR was 2.51 (95% CI, 1.3-4.82), comparable to the primary analysis.

In further sensitivity analyses based on logistic regression and restricted to participants without chronic conditions, the OR for 2008-2009 TIV effect adjusted for age, Aboriginal status, and household density was 3.44 (95% CI, 1.80-6.59). In an analysis further restricted to on-reserve participants without chronic conditions, the OR for the 2008-2009 TIV effect adjusted for age and household density was 5.38 (95% CI, 2. 27-12.75 ). In analysis restricted to non-Aboriginal households of the elementary school population, the OR adjusted for age, Analyses of TIV effect on serologically confirmed pH1N1 versus seronegative control subjects are shown in Table 4 . Sample size was small and 95% CIs intervals were wide, but the same trend in point estimates for TIV effects was observed, with a fully adjusted OR of 2. 

In this article, we present the first observation of an unexpected association between prior seasonal influenza vaccination and pH1N1 illness during the spring and summer of 2009 in Canada. Specifically, outbreak investigation conducted during the early stages of the pandemic in a northern BC community identified that participants reporting pH1N1-related ILI during the period 1 April through 5 June 2009 were more than twice as likely to report having previously received seasonal influenza vaccine.

Advantages offered by the original outbreak investigation described here include active and standardized inquiry about pH1N1-related illness, TIV receipt, and relevant covariates for all household members, obviating potential selection biases associated with differential health care access or health careseeking behavior in other methods of participant recruitment or case detection. By restricting analysis to households with children in a single community and in groups among whom pH1N1 circulation was confirmed serologically, we ensured the population at risk was well circumscribed and that control subjects were drawn from the same source population as cases, further minimizing the risk of selection bias. To account for influential covariates, we used recognized analysis techniques of restriction as well as adjustment for age, comorbidity, household density, and Aboriginal status.

There are, however, several limitations to this study warranting cautious interpretation of the results. First, the study (26) 12 (27) 46 (31) 12 (21) 29 (32) 106 (34) 29 (21) 20-49 11 (17) 78 (42) 11 (12) 10 (23) 55 (37) 10 (15) 20 (22) 124 (39) 20 (14) 50-64 1 (2) 9 (5) 1 (10) 6 (14) 13 (9) 6 (32) 6 (7) 19 (6) 6 (24) 33 (25) 28 (64) 78 (53) 28 (26) 50 (56) 167 (53) 50 (23) Male 32 (49) 86 (46) 32 (27) 15 (34) 68 (46) 15 (18) 40 (44) 146 (47) 40 (22) Chronic conditions Yes 4 (6) 9 (5) 4 (31)

10 (23) 17 (12) 10 (37) 14 (15) 26 (8) 14 ( restricted analyses to households of the affected elementary school and on a reserve where pH1N1 circulation was more certain. This improved the positive predictive value of the ILI case definition in representing pH1N1-related illness (160%), although some misclassification may have still occurred. The effect of any persisting misclassification due to a nonspecific outcome would be to underestimate the association between TIV and pH1N1 illness (ie, to drive the OR toward a null effect) [14] . A second limitation is that we relied on proxy report by 1 adult for all household members. Both ILI experience and TIV history may have been less well known for other household members. Third, the collection of TIV status after any ILI experience may have introduced recall bias. If recall bias related to TIV receipt were operating, it would introduce misclassification (information bias) related to exposure. Ultimately, the impact of this misclassification would depend on whether participants believed a priori that seasonal vaccine ought to have decreased or increased the risk of ILI during the study period. The lack of an immunization registry in British Colombia precluded further confirmation of immunization status.

It is noteworthy that ORs were higher among Aboriginal onreserve participants who also had higher rates of repeat seasonal influenza immunization. Repeated vaccination has been hypothesized to block potentially cross-protective immunity otherwise afforded by heterotypic infection [15] [16] [17] . Higher ORs among on-reserve Aboriginals could also reflect greater susceptibility to the effects of TIV on pH1N1 risk, although an analysis for interaction did not yield statistically significant results. Some studies have reported genetic polymorphisms that favor Th2 skew among Canadian Aboriginals with higher expression of interleukin-6 and lower production of tumor necrosis factor-a, interferon-g, and interleukin-10 that may be relevant to enhanced immune-mediated effects in this population [18] . Other studies have suggested that Aboriginal status is an independent risk factor for more severe pH1N1 outcomes, but we did not specifically assess that hypothesis [19] . However, higher ORs may also reflect the methodological influences of bias or confounding. Aboriginal populations are known to have higher rates of both chronic conditions and influenza vaccination [11] [12] [13] . To address this, we adjusted for comorbidity and conducted sensitivity analyses restricted to on-reserve participants without chronic conditions, resulting in similar or higher ORs. Despite these reassurances, neither residual confounding nor bias can be fully ruled out. As a final limitation, the sample sizes were small and confidence intervals were wide for stratified and serologically-confirmed analyses so that these in particular should be interpreted cautiously.

Because of limitations in study design and because they represented unexpected findings, we interpreted the results of this outbreak investigation as a paradoxical signal of possible concern-thought-provoking but inconclusive and warranting fur-ther evaluation. Canadian investigators thus embarked on a series of confirmatory studies using more rigorous methods and laboratory-confirmed outcomes through the summer of 2009, each of which corroborated findings from this initial outbreak investigation. In combination, these showed 1.4-2.5fold increased risk of medically attended, laboratory-confirmed pH1N1 illness among prior 2008-2009 TIV recipients [17] . An additional Canadian study using the linked Manitoba immunization registry and administrative databases has also shown similar findings of increased risk [4] (Dr Carole Beaudoin, Public Health Agency of Canada, personal communication). Thus, in Canada, 6 observational studies based on different methods and settings, including the current outbreak investigation, consistently showed increased risk of pH1N1 illness during the spring and summer of 2009 associated with prior receipt of the 2008-2009 TIV [4, 17] . Conversely, studies conducted outside of Canada have provided inconsistent results: 3 studies (from the United States and Australia) reported null effects [20] [21] [22] , 4 (from the United States and Mexico) reported protective effects [23] [24] [25] [26] , and 1 other outbreak investigation (from the Untied States) reported increased risk [27] .

Findings of pH1N1 risk associated with TIV-consistent in Canada but conflicting elsewhere-may have been due to methodological differences and/or unrecognized flaws, differences in immunization programs or population immunity, or a specific mechanistic effect of Canadian TIV. High rates of immunization and the use of a single domestic manufacturer to supply 175% of the TIV in Canada may have enhanced the power within Canada to detect a vaccine-specific effect. Given the changed immunologic landscape following the first springsummer pandemic wave and the mass pH1N1 vaccination campaign during the fall 2009, it may not be possible to further resolve this issue epidemiologically. Studies using animal models, banked serum samples, or other in vitro experiments are needed to further assess this association.

Socio-economic profiles local health areas: maps. Victoria, British Colombia, Canada: BC Stats

British Columbia Indian reserve census figures: 2006 census total population results. Victoria, British Colombia, Canada: BC Stats

Novel Swine-Origin Influenza A (H1N1) Virus Investigation Team. Emergence of a novel swine-origin influenza A (H1N1) virus in humans

Statement on seasonal trivalent inactivated influenza vaccine (TIV) for the 2009-2010 season

Note for guidance on harmonization of requirements for influenza vaccines. London, UK: European Agency for the Evaluation of Medicinal Products

Models for discrete longitudinal data

Mortality from pandemic A/H1N1 2009 influenza in England: public health surveillance study

Pandemic influenza A(H1N1)v in New Zealand: the experience from

Health indicator profile, annual estimates, by age group and sex, Canada, provinces, territories, health regions (2007 boundaries) and peer groups, occasional (1209600 series). Ottawa, Ontario

British Columbia, Provincial Health Officer. Pathways to health and healing: 2nd report on the health and well-being of Aboriginal people in British Columbia. Provincial Health Officer's Annual Report

British Columbia First Nations Health Council. Healthy children, healthy families, healthy communities: the road to wellness. BC First Nations Regional Longitudinal Health Survey

First Nations and Inuit Adult Immunization Coverage Survey (FNIICS). First Nations and Inuit Health Branch, Health Canada (contract #H1011-060001/001/CY)

Methodologic issues regarding the use of three observational study designs to assess influenza vaccine effectiveness

Vaccination against human influenza A/H3N2 virus prevents the induction of heterosubtypic immunity against lethal infection with avian influenza A/H5N1 virus

Yearly influenza vaccinations: a double-edged sword?

Association between the 2008-09 seasonal influenza vaccine and pandemic H1N1 illness during spring-summer 2009: four observational studies from Canada

Nickerson P. Differential cytokine genotype frequencies among Canadian Aboriginal and Caucasian populations

Correlates of severe disease in patients with 2009 pandemic influenza (H1N1) virus infection

Effectiveness of 2008-09 trivalent influenza vaccine against 2009 pandemic influenza A (H1N1)-United Seasonal Vaccine Effect on pH1N1 Risk • CID

Notes from the field: outbreak of 2009 pandemic influenza A (H1N1) virus at a large public university in Delaware

Interim analysis of pandemic influenza (H1N1) 2009 in Australia: surveillance trends, age of infection and effectiveness of seasonal vaccination

Seasonal influenza vaccine and protection against pandemic (H1N1) 2009-associated illness among US military personnel

Infection and death from influenza A H1N1 virus in Mexico: a retrospective analysis

Partial protection of seasonal trivalent inactivated vaccine against novel pandemic influenza A/H1N1 2009: case-control study in Mexico City

Outbreak of 2009 pandemic influenza A (H1N1) at a New York City school

Clinical and epidemiologic characteristics of an outbreak of novel H1N1 (swine origin) influenza A among United States military beneficiaries