Did the H1N1 Vaccine Reduce the Risk of Admission with Influenza and Pneumonia during the Pandemic?

Background The extent to which A(H1N1)pdm09 influenza vaccines prevented hospital admissions with pneumonia and influenza (P&I) during the 2009 pandemic remains poorly understood. We evaluated the effectiveness of the A(H1N1)pdm09 and seasonal influenza vaccines (TIV) used during the 2009 mass vaccination campaign in Manitoba (Canada) in preventing P&I hospitalization. Methods A population-based record-linkage nested case-control study. Cases (N = 1,812) were persons hospitalized with influenza (ICD-10:J09-J11) or pneumonia (ICD-10:J12-J18) during the study period. Age-, gender- and area of residence-matched controls (N = 7,915) were randomly sampled from Manitoba’s Population Registry. Information on receipt of A(H1N1)pdm09 vaccine and TIV was obtained from the Manitoba Immunization Monitoring System, a province-wide vaccine registry. Results Overall, the adjuvanted A(H1N1)pdm09 vaccine was 27% (95%CI 13–39%) effective against P&I hospitalization ≥ 14 days following administration. Effectiveness seemed lower among older (≥ 65 years) adults (10%; −16–30%), particularly when compared to under-5 children (58%; 30–75%). The number-needed-to-vaccinate to prevent 1 P&I admission was lowest among <4 year-olds (928) and ≥65 years (1,721). VE against hospitalization with laboratory-confirmed A(H1N1)pdm09 was 70% (39–85%) overall and (91%; 62–98%) ≥ 14 days following vaccination. Discussion Our data suggest that the adjuvanted A(H1N1)pdm09 vaccine was effective in preventing about 55–60% of P&I hospitalizations among children and younger adults who were at much higher risk of infection. Unfortunately, the vaccine was less effective among 65 or older adults. Despite that the vaccine still had a significant population-based impact especially among the very young (<5) and the older (≥ 65 years).


Introduction
Worldwide, more than 26 different monovalent A(H1N1)pdm09 vaccines were used during the 2009 influenza pandemic to control the spread of infection and reduce disease burden [1]. In Manitoba (Canada), we found that both adjuvanted and non-adjuvanted vaccines were effective in preventing laboratory-confirmed A(H1N1)pdm09 infections [2], and similar findings were reported for other jurisdictions. [3] However, the extent to which vaccination prevented more clinically meaningful outcomes, such as acute hospital admission due to influenza and pneumonia (P&I), is not known.
Rates of laboratory-confirmed influenza can vary significantly between jurisdictions and with epidemic phase in the same jurisdiction as a function of clinical guidelines, epidemic phase and laboratory workload. By comparison, decisions for hospitalization are less discretionary, and as a result hospitalization rates are a robust indicator of epidemic severity. [4] Arguably, avoiding hospitalization is a more relevant outcome for influenza control efforts than the mere reduction in numbers of laboratory-confirmed infections.
We evaluated the effectiveness of the A(H1N1)pdm09 and seasonal influenza vaccines (TIV) used during the 2009 mass vaccination campaign in Manitoba (Canada) in preventing P&I hospitalization using a case-control design and data from Manitoba's provincial administrative and laboratory databases. We also assessed vaccine effectiveness (VE) for different age groups and among high-risk populations, e.g., immunocompromised persons.

Design
We conducted a nested case-control study using de-identified records obtained by linking the electronic database of the Manitoba Immunization Monitoring System (MIMS) with the Hospital Separation Abstract database and other Manitoba Health (MH) administrative databases, housed at the Manitoba Centre for Health Policy. MH provides publicly-funded universal healthcare insurance to virtually all of Manitoba's 1.2 million residents regardless of age or

Key Points
In the recent pandemic, the A(H1N1)pdm09 vaccine was effective in preventing 55-60% of influenza and pneumonia hospitalizations among children and younger adults. Although less effective among 65 adults(~10%), the vaccine benefited this group the most as measured by the number-needed-to-vaccinate. income [5]. Insured services include hospital, physician and preventive services (e.g., immunizations). For administrative purposes, MH maintains several centralized electronic databases that can be linked using a unique health services number. The study was approved by the Research Ethics Board of the University of Manitoba and the governmental Health Information Privacy Committee.

Definition of cases and controls
Anyone 6 months or older who was registered with MH during the study period was eligible for inclusion in the study. The study spanned the period from November 2, 2009 (1 week after the start of mass immunization in Manitoba) to February 7, 2010, two weeks following the last reported A(H1N1)pdm09 case in the province.
Any eligible person who was admitted for 24 hours to a hospital in Manitoba with a diagnosis of influenza (ICD-10: J09-J11) or pneumonia (ICD-10: J12-J18) during the study period was included in the cases group. Cases were identified using the Hospital Separation Abstract database, which, since 1971, has recorded all services provided by hospitals in the province, including admissions and day surgeries [6]. The data collected include clinical information such as admission and discharge dates and up to 25 diagnoses and 20 services or procedures, coded using the International Classification of Diseases (ICD-10-CA [7]) and the Canadian Classification of Health Interventions (CCI) [8].
Using risk-set sampling, we matched each case to five controls (persons who have not been admitted to hospital by the index date) who were of the same age, gender and area of residence. The date of admission was considered as the "index date" for cases and for their matched controls. Controls were randomly selected from MH's Population Registry, a continuously updated database that tracks dates and reasons for the initiation and termination of coverage (e.g., death/migration) for all insured persons.

Determination of vaccination status
For all cases and controls, information on the receipt of the pandemic, seasonal influenza and pneumococcal vaccines during and before the 2009/10 season was obtained from MIMS, a population-based province-wide registry of virtually all vaccines administered to Manitoba residents since 1988 [9]. Information, including vaccine type and date of vaccination, is captured either through direct data entry for vaccines administered by public health staff (who administered most influenza vaccines during the pandemic) or using physician claims data for vaccines administered by physicians [10].
In Manitoba, most pandemic vaccines were administered during a mass immunization campaign that began in October 26, 2009 (Week 43), just one week before the peak of the second pandemic wave [2]. Initially, the Canadian-manufactured Arepanrix 1 (GlaxoSmithKline), an AS03-adjuvanted split virion monovalent vaccine, was used to vaccinate adults and children over 6 months of age. Later on, two nonadjuvanted vaccines, from GlaxoSmithKline and CSL Limited, were offered to pregnant women and children 10 years or older [11]. All vaccines contained the A(H1N1)pdm09 hemagglutinin antigen derived from the influenza A/California/7/ 2009 strain recommended by the WHO. A single vaccine dose (15μg/0.5 ml) was recommended for those aged >9 years and 2 half doses given 21 days apart were recommended for children 6 months-9 years old.
All vaccines were offered free of charge, but due to limited supply at campaign start priority was given to certain groups including health care workers, Aboriginal persons, residents of remote communities, pregnant women, 6-60 months-old children, persons under 65 years with chronic medical conditions and all immunocompromised persons. On November 18, 2009, vaccines became available to the whole population [2].

Potential confounders
Individuals were assigned to a neighbourhood of residence (neighbourhood clusters within the capital city of Winnipeg and regional health authorities in the rest of the province) based on their postal code as recorded in MH's Population Registry. Household income quintiles, measured at the level of Census Dissemination areas, were determined using 2006 Canadian census data. Information on pregnancy, comorbidities, propensity to seek health care (measured as the number of hospital and family physician visits in the previous 5 years) was obtained from the Hospital Separation and Physician Claims databases. Previously validated algorithms were used to identify various chronic diseases and other indications for vaccination [12][13][14][15]. In addition, Charlson comorbidity scores were calculated using an algorithm validated for administrative databases [16]. Information on the use of antivirals and other medications was obtained from the Drug Program Information Network database, the comprehensive database of all outof-hospital prescriptions dispensed in Manitoba since 1995 [17].
Influenza testing results were obtained from the database of Cadham Provincial laboratory (CPL), the province's only public health laboratory. During the study period, influenza testing in Manitoba was completed at CPL using a real-time multiplex reverse-transcription polymerase chain reaction (RT-PCR) assay developed by the National Microbiology Laboratory [18].

Statistical analysis
In the primary analysis, we used conditional logistic regression models to estimate the odds ratio (OR) for the association between the receipt of the adjuvanted A(H1N1)pdm09 vaccine and P&J hospitalization while adjusting for matching and confounding variables. Models were adjusted for income, comorbidity, receipt of seasonal and pneumococcal vaccines, use of neuraminidase inhibitors, frequency of contact with healthcare providers and belonging to a vaccine priority group. VE was estimated as (1-OR) x 100. Similar but separate models were used to estimate the VE of nonadjuvanted pandemic, seasonal and pneumococcal vaccines.
In primed healthy adults, the peak serum antibody levels are typically observed >2 weeks after vaccination [19]. To account for differences in effectiveness by time since vaccination, we classified vaccinated individuals into three groups depending on whether vaccination occurred 1-6, 7-13, or 14 days before the index date, and contrasted the odds of A(H1N1)pdm09 infection in each group with the odds of infection among the unvaccinated.
In addition, we repeated the adjuvanted A(H1N1)pdm09 vaccine analyses after stratifying by age group, place of residence, epidemic phase (admission before and after the peak), presence of high-risk conditions and belonging to a vaccine priority group. We also assessed for possible effect modification by use of seasonal and pneumococcal vaccines. The statistical significance of adding the interaction terms was assessed using the likelihood ratio test [20].

Results
A total of 1,812 persons met the case definition and were matched to 9,060 controls. About 54% of cases were 65 years or older at diagnosis, and 39% resided in the poorest parts of the province ( Table 1). As expected, cases tended to be generally sicker than controls with higher Charlson comorbidity scores (average score of 3 compared to 0 for controls), higher overall chance of having 1 diagnosed chronic medical conditions (73% compared to 39% among controls) and more frequent hospitalizations and physician consultations. As a result, cases were also more likely to belong to A(H1N1)pdm09 or TIV priority groups (Table 1).
About 8% of the cases and none of the controls tested positive for A(H1N1)pdm09 (Table 4), which is not surprising given provincial guidelines discouraging viral testing unless the patient is very ill. Only 6 controls were tested within 2 weeks of the index date compared to 921 cases. In analyses limited to cases hospitalized with laboratory-confirmed A(H1N1) pdm09, 13% of the cases received an adjuvanted A(H1N1)pdm09 vaccine compared to 25% of the controls (Table 4), corresponding to a VE estimate (adjusted for matching covariates) of 59% (30-76%). The corresponding fully adjusted estimate was 70% (39-85%). VE among those who were vaccinated 14 days before the index date was higher (91%; 62-98%) than among those who were vaccinated <7 days (46%; −48-81%) or 7-13 days before the index date (61%; −11-86%). Despite small numbers, estimates of VE of the nonadjuvanted A (H1N1)pdm09 vaccine were comparable. On the other hand, there was no evidence that the 2008/09 or the 2009/10 TIVs reduced the risk of hospitalization with laboratory-confirmed A (H1N1)pdm09.

Discussion
Our data suggest that the adjuvanted A(H1N1)pdm09 vaccine prevented about 55 to 60% of P&I hospitalizations among children and younger adults and a much lower percentage (10-15%) of hospitalizations among 65 or older adults and among those with pre-existing chronic diseases (14%). The vaccine was also effective (70% on average) in preventing hospitalizations with laboratory-confirmed A(H1N1)pdm09 with higher levels of protection achieved >14 days after vaccination. Excellent immune responses following even one dose of the monovalent split/subunit inactivated pandemic vaccines were documented in several immunogenicity trials [22,23]. In postmarketing studies, the vaccines were also effective in preventing laboratory-confirmed A (H1N1)pdm09 infections during the pandemic. In a systematic review that was limited to 5 observational studies which met stringent quality criteria, the median VE of monovalent A (H1N1)pdm09 vaccines was about 69% [3]. Similar estimates were observed in studies not included in the review [2,24]. To our knowledge, this is the first published study to evaluate the effectiveness of the A (H1N1)pdm09 vaccine against admission with P&I during the pandemic. On the other hand, VE against hospitalization with A(H1N1)pdm09 infection was examined in few studies. Using data from a Scottish general practice sentinel surveillance network, Simpson et al reported that the adjuvanted vaccine was 95% (76-100%) effective against laboratory-confirmed A(H1N1) pdm09 [25]. Emborg, et al. estimated, using Danish health databases, that the adjuvanted monovalent vaccine was 44% (−19%-73%) effective in preventing hospitalization with laboratoryconfirmed A(H1N1)pdm09 infection among younger (<65) chronically ill people [26]. Steens, et al. found that among persons with underlying medical conditions or 60 years of age, a single dose of the MF-59 adjuvanted A(H1N1)pdm09 vaccine had VE of 19% (−28-49%) [27].
Higher estimates of VE against P&I hospitalization among children and young adults in our study may reflect the greater contribution of A(H1N1)pdm09 to P&I hospitalization during the pandemic among this age group compared to older adults who generally were less likely to become infected [28,29]. Also, in a previous analysis from Manitoba we found that the vaccine was more effective in preventing A(H1N1)pdm09 infection in children compared to older adults. [2] This is also consistent with studies that examined VE against hospitalization with A (H1N1)pdm09 among children. In a Quebec study, VE of a single pediatric dose of the same AS03-adjuvanted vaccine that was used in Manitoba, was 85% (61-94%) among 6 month-9 year olds and slightly lower in [5][6][7][8][9] year-olds at 79% (−31-96%) [30]. In a smaller study conducted in New York, a single dose of the nonadjuvanted vaccine was 82% (0-100%) effective in in preventing hospitalization 14 days after vaccination in children aged 3-9 years [31]. Based on this evidence, it is reasonable to conclude that that the A(H1N1)pdm09 was effective in protecting younger children against hospitalization with P&I during the pandemic.
For older adults, our A(H1N1)pdm09 VE estimates were lower than estimates obtained from observational studies of the effectiveness of TIVs against P&I hospitalization during nonpandemic seasons. In a comprehensive Cochrane review, [32] 8 such studies had a pooled estimate of 26% (12%-38%) during seasons when the vaccine was well matched to the circulating strain. However, our estimates are more in line with the studies that attempted to control for confounding by the "healthy vaccinee effect" (seniors who get vaccinated are on average healthier than those who do not) which produced estimates between 8 and 14% [33][34][35].
Despite lower VE and lower incidence of A(H1N1)pdm09 among the elderly, our data suggest that only very young children (0-4 years) had a lower number needed to vaccinate (NNV = 928) to prevent one hospital admission for P&I. The NNV among 65 or older was 1,721 compared to 2,273 among 45-64 olds and 7,598 among 25-44 olds. This reflects the much lower overall rate of P&I admissions among the latter two groups (4-5/10,000) compared to 59/10,000 among the 65 age group and 19/10,000 among 0-4 year olds. Ignoring potential indirect benefits due to herd immunity, these figures suggest that vaccinating the <5 and 65 age groups was more cost-effective than vaccinating other groups.
In our study, the TIVs did not appear to protect against P&I admissions overall or due to A (H1N1)pdm09. This makes sense because in Manitoba the pandemic strain almost entirely replaced previously circulating influenza strains during the 2009/10 season and there was no evidence of significant cross-protective response in previous studies. [36] We observed that the A(H1N1)pdm09 vaccine was less effective among those who also received the 2009/10 TIV. It is unclear whether this reflects a biological effect or confounding by indication because the TIV is indicated to persons at higher risk of developing severe influenza illness and its complications [37]. The issue of whether TIV use before or during the pandemic increased the incidence or severity of A(H1N1)pdm09 remains controversial [38].

Strengths and Limitations
Because of its population-based design and the availability of accurate automated hospital admission and vaccination records, this study is less susceptible to selection and recall biases that commonly afflict conventional case-control studies [10]. Because cases and controls were identified on the basis of comprehensive hospital records with high standards of coding practices, [13] misclassification of hospitalization status is also not a major concern in this study.
We used proxies for access to health care (e.g., frequency of physician encounters) to adjust for factors associated with the likelihood of admission and influenza testing. We also adjusted for confounding by several vaccine indications such as immune status and pre-existing health conditions using information obtained from administrative databases. The completeness and accuracy of the MH database are well established [6,39], and these databases have been used extensively in studies of post-marketing surveillance of drugs and vaccines [2]. However, it is still possible that some variables were measured with error, which could result in residual confounding. We did not have information on functional status. However, the observed protective effects of A(H1N1)pdm09 vaccination are unlikely to be due to the healthy vaccinee effect, because vaccination was targeted at the higher-risk, and generally less healthy, groups, and we adjusted for vaccine indications in our models. Also, we did not observe any protective effects for the TIVs which are presumably subject to the same bias.
Confounding by herd immunity following the summer wave of the pandemic is also unlikely explanation for our findings, because we adjusted for area of residence in all models and because VE estimates were similar for northern and southern communities despite significant differences in A(H1N1)pdm09 seroprevalence between these communities at the end of the summer wave of the pandemic [28].

Conclusions
Our data suggest that the adjuvanted A(H1N1)pdm09 vaccine was effective in preventing about 55-60% of P&I hospitalizations among children and younger adults who were at much higher risk of infection. Unfortunately, the vaccine was less effective among 65 or older adults. Despite that the vaccine still had a significant population-based impact especially among the very young (<5) and the older ( 65 years).