Review of seasonal influenza in Canada: Burden of disease and the cost-effectiveness of quadrivalent inactivated influenza vaccines

ABSTRACT In the 2015/16 influenza season, the Canadian National Advisory Committee on Immunization (NACI) recommended vaccination with quadrivalent inactivated influenza vaccine (QIV) for infants aged 6–23 months and trivalent inactivated influenza vaccines (TIVs) or QIVs in adults. The objective of this review (GSK study identifier: HO-13-14054) is to examine the epidemiology and disease burden of influenza in Canada and the economic benefits of vaccination. To inform this review, we performed a systematic literature search of relevant Canadian literature and National surveillance data. Influenza B viruses from phylogenetically-distinct lineages (B/Yamagata and B/Victoria) co-circulate in Canada, and are an important cause of influenza complications. Modeling studies, including those postdating the search suggest that switching from TIV to QIV in Canada reduces the burden of influenza and would likely be cost-effective. However, more robust real-world outcomes data is required to inform health policy decision makers on appropriate influenza vaccination strategies for Canada.


Introduction
Many countries have implemented influenza vaccination as targeted programs for people at high risk for influenza-related complications, while some countries such as Australia, Canada, and the United States (US) recommend universal vaccination against seasonal influenza. [1][2][3] The aim of influenza vaccination in Canada is to prevent serious outcomes such as hospitalization and death. As such, the National Advisory Committee on Immunization (NACI) stresses the need for vaccination, particularly for those people at high risk for influenza-related complications, including infants 6 to 59 months of age, adults 65 y and older, and people with medical co-morbidities. 4 Two phylogenetically-distinct influenza B lineages (B/Yamagata and B/Victoria) emerged globally in the early 1980s, and have cocirculated in the US since 2000. 5 Between 2000 and 2010, the influenza B strain in the seasonal trivalent influenza vaccine (TIV) did not match the prevalent B lineage in around half of the European and North American influenza seasons. 5 As a result, the World Health Organization (WHO) recommended the inclusion of an influenza strain from each of the influenza B lineages for the 2012/ 13 influenza season onwards. 6 Traditional TIVs include either B/ Victoria or B/Yamagata whereas quadrivalent influenza vaccines (QIVs) include strains from both lineages. Thus the new QIVs are expected to reduce the burden of influenza B disease, 7,8 and this is supported by the available evidence. The first efficacy study of a QIV was a randomized, observer-blinded assessment of FluLaval TM TETRA in children from 3 to 8 y of age. 9 The study showed that the QIV versus control reduced real-time polymerase chain reaction (RT-PCR)-confirmed influenza A and B cases by 59.3% and reduced moderate-to-severe influenza cases by 74.2%. 9 A recent meta-analysis of randomized controlled trials of QIVs found that inactivated QIVs have equivalent immunogenicity and efficacy against the shared 3 strains included in TIVs, with superior immunogenicity against the B lineage not included in TIV. 10 In Canada, a number of influenza vaccines are currently licensed for use. These include 6 1 TIVs and QIVs are recommended for use in healthy adults and those with chronic health problems from 18 y of age. 1 NACI preferentially recommends QIV for use in infants from 6 to 23 months of age and LAIV for children 2 to 6 y of age. In the 2015/16 season, NACI offered no preference between TIV, QIV, or MF59-adjuvanted TIV for those 65 y of age and older. 1 The principal objectives of this literature review are to examine the epidemiology and disease burden of influenza in Canada and the economic benefits of vaccination. To facilitate this, we performed a comprehensive literature review using a detailed search strategy conducted to identify original research of influenza in the general population in a Canadian setting (GSK study identifier: HO-13-14054).

Search strategy
Search strings were developed to identify potentially relevant sources based on incidence, prevalence, morbidity and mortality of influenza in Canada, along with costs and cost-effectiveness or cost-utility analyses. With these, searches were performed in PubMed, EMBASE and the Cochrane databases for the period from January 2002 to December 2013. The detailed search strategies for PubMed are described in Tables S1, S2 and S3. Search criteria used to identify studies of interest were broad to include epidemiological, clinical and cost-effectiveness studies reporting on influenza in the general population (not pregnant women, diabetics, or HIVC [human immunodeficiency virus-positive] patients) in the Canadian setting. Identified literature was assessed independently by 2 reviewers for relevance, with any disagreements resolved by discussion and consensus. From these searches, we identified a total of 1,580 citations in the search of electronic databases. Following article appraisal, 64 articles were selected to inform this review. Of these, 27 studies contained relevant data on disease epidemiology, 42 studies on the burden of disease and costs of illness, and 7 studies on the economic evaluation of vaccines. An overview of the search results and selection is presented in Figure S1. We also searched 'gray' literature, including annual reports from Canada's influenza surveillance through the Canadian FluWatch surveillance system for the influenza seasons beginning 1999/00 up until 2009/10, including final weekly reports for 2010/11, 2011/12, and 2012/13, along with NACI statements describing vaccine match-mismatch data.

General practitioner (GP) and emergency room (ER) visits
GP visit data for influenza infection were reported in 3 studies. 28,33,37 In a study of an elderly population in a nursing home with seasonal influenza, 66% of those with influenza received visits from a physician, 28 and a telephone survey of 1,009  Total cases of influenza B  43  254  152  128  40  214  472  119  673  570  7  586  965 Note: Data obtained from The Canadian FluWatch surveillance system. 11-22 y There was 1 case of B/Beijing-like influenza in 2000/01 (0.4% of cases). Data not shown due to rounding. Ã Data obtained from FluWatch final cumulative weekly report in absence of a full annual report. [20][21][22]  Canadian households reported a rate of 0.7 GP visits related to influenza-like illness (ILI) per household. 33 Information on ER visits was provided by one study which reported that the rate of influenza-attributable ER visits during a non-pandemic influenza season was 460 per 100,000 population. 62

Influenza and productivity loss
Few studies with data on loss of work, or absence from school, or quality-of-life associated with influenza illness were identified in the literature. In an ILI telephone survey, it was estimated that 2 d of work or school were lost per household. 33 In another study, it was estimated that during non-pandemic influenza seasons from 1997/98 to 2008/09, on average 14 work hours were lost per influenza patient, which translated to an aggregated 20 work days lost per 100 full-time employees. For pandemic strain, while absenteeism rates were similar to those

Modeling the cost of influenza vaccination
Between 2002 and end of 2013 a number of health economic influenza vaccine studies from Canada were reported, evaluating cost-effectiveness/cost-utility studies of seasonal TIVs, cost of vaccination strategies in targeted groups, and estimated costutility analyses and willingness-to-pay (Table 4). [69][70][71][72][73][74][75] A study of MF59-TIV compared with TIV used an age-structured model in which the population was divided into 5 different compartments or disease states: susceptible, vaccinated, exposed, infectious, and recovered. 69 The probability of moving between compartments was derived from epidemiological data and calibrated to Canadian influenza data (1997 to 2004 seasons). The simulation allowed new individuals to enter through birth or leave through non-influenza deaths over a 10 y time horizon. In the base case, the cost of using MF59-TIV over 10 y was higher (CAN$837.0 million) than TIV (CAN$730.5 million). These costs were offset by reducing the healthcare cost of influenza from CAN$501.76 million with TIV to CAN$473.50 million with MF59-TIV. The incremental cost per quality-adjusted life year (QALY) gained of MF59-TIV relative to TIV was CAN$2,111 in older adults (65 years). 69 In another modeling study, a decision tree was used to assess the cost-effectiveness of trivalent LAIV vs injectable inactivated TIV in children aged 2-17 y with a 1-year time horizon. The study showed that overall, trivalent LAIV dominated TIV and provided an estimated cost savings per vaccinated child of CAN$4.20 from a healthcare perspective and CAN$35.34 from a societal perspective. 70 Several studies discuss vaccinating different population subgroups or regions. A cost-utility analysis of the A(H1N1) pdm09 mass immunization program in Ontario predicted a cost of CAN$118 million (CAN$1,645/QALY gained) to immunize 30% of the provincial population. 71 In a cost-utility analysis of Ontario's universal vaccination program, universal versus targeted group vaccination predicted a cost of CAN$10,797/QALY gained, which is below the threshold usually considered cost-effective in Canada (CAN$40,000-50,000/ QALY gained). 72 In another study published in 2006, a decision analysis was used to assess the cost-effectiveness of universal vaccination of infants/toddlers aged 6-23 months compared with vaccinating high-risk children only. 73 Analyses were conducted for a cohort of 500,000 children assuming different vaccination coverage between influenza seasons (100% in year 1, 33% thereafter). In the first year, the incremental cost of vaccination from the healthcare system perspective was CAN$17 per day of illness averted, CAN$900,000/QALY gained, and CAN$6 million per death averted. In subsequent years, costs among those vaccinated would equal those unvaccinated (break-even) at a cost of CAN$6.81 per vaccine dose from a healthcare perspective, and CAN$11.90 per vaccine dose from a societal perspective. 73 Contingent valuation has been used to determine the willingness-to-pay for access to a pandemic influenza vaccine, and found that households were willing to pay CAN$417.35 for immediate A (H1N1)pdm09 vaccination. 74 In another analysis to determine key cost drivers in providing influenza vaccines within budget conducted in a health unit in Ontario, costs relating to labor (nursing and clerical), supplies, disposals, facilities, mileage, and program costs for influenza clinics (including labor costs) were estimated. This study reported that the most significant cost variables were labor costs and number of vaccines given per nurse per hour. 75

Discussion
Based on previous modeling of US surveillance data across 10 influenza seasons, the public health benefit of switching from TIV to QIV is likely to vary depending upon the epidemic intensity of influenza B during each season. 8 As such, many countries, including Canada, have established sentinel surveillance networks to help monitor influenza disease and to guide public health decision-making. Using data from the national surveillance system and published studies identified by a comprehensive literature search across 2002-2013, we have described the burden of seasonal influenza disease and the cost-effectiveness of vaccination in Canada, with the aim of developing a dataset that may be used to develop future analytical frameworks to model the effect of QIV in Canada.
According to Canadian national surveillance, influenza A was the dominant subtype circulating in all but 2 seasons (2000/01 and 2011/12), whereas influenza B was less prevalent, representing an average of 17% of all laboratory-confirmed cases during the 2001/ 02 to 2012/13 influenza seasons. 4 However, the surveillance data showed that influenza B strains from the Yamagata and Victoria lineages co-circulated in 7 of the 12 influenza seasons, resulting in mismatch with the TIV. Viral circulation and the occurrence of B lineage vaccine mismatch reported in Canada was generally consistent with that observed in Europe and the US. 5,[76][77][78] Studies in the US and Europe suggest that A/H3N2 is the most common subtype associated with complications and death in older adults, followed by influenza B, and then A/H1N1. [79][80][81] In children and young adults, influenza B is suggested to pose the highest risk of severe outcomes. 5,7,82 As we describe above, laboratory confirmed influenza infection rates in Canada were generally highest among elderly populations, followed by children aged <5 years. These two groups, along with people with medical co-morbidities, are at increased risk for influenza complications and are recommended for annual seasonal vaccination in most industrialized countries. 4 The WHO reports that the global attack rate of influenza is about 10-20%, 83 while an analysis based on US hospital discharge records collected for 22 seasons (1979-2001), estimated that 230,000 influenza-related hospital admissions occurred annually. 79 In a more recent surveillance study of laboratoryconfirmed influenza-related hospitalizations between 2005 and 2008 in the US, the annual age-adjusted rate of influenzarelated hospitalizations was 12.2 per 100,000 persons. 84 The rate of hospitalization reported in Canada is consistent with that reported in the US. For example, in an evaluation of influenza-related hospitalization in Toronto in which the median age of patients was about 76-77 years, the rate of hospitalization ranged from 2.6 per 100,000 population in the 2005/06 season to 27.1 per 100,000 population in the 2010/11 season. 27 We are aware that a number of Canadian studies have been published since the 2013 search cut-off data used to inform this review, some of which we describe below. From an epidemiological perspective, the Serious Outcomes Surveillance Network of the Canadian Immunization Research Network reported an interim analysis for the 2014/15 influenza season. They found that the great majority (99%) of laboratory-confirmed influenza hospitalizations with a known influenza virus type reported in their sentinel hospitals was due to influenza A. Almost 70% of hospitalized laboratory-confirmed influenza cases were >75 years, and 7.9% of all hospitalized cases died in hospital. 85 From a health economic perspective, a recent systematic review provided a qualitative appraisal of all vaccine economic evaluations in Canada published between 1993 and 2013, including influenza cost-effectiveness studies. 86 While this review's aim was primarily to evaluate and report on study characteristics rather than outcomes per se, those identified for influenza were described either as cost-effective or dominant. Subsequent to the systematic search, 2 further economic evaluations of the benefits of switching from TIV to QIV have been published, one of which used a static model, and the second a dynamic model. 87,88 The first study used a probabilistic and static cost-utility model adapted for Ontario, and found that, in comparison to TIV, universal vaccination with QIV is estimated to realize reductions of an additional 2,516 influenza cases, 1,683 influenza-associated medical visits, 27 influenzaassociated hospitalizations, and 5 influenza-associated deaths. Furthermore, QIV would generate 76 additional QALYs with a net budget impact of CAN$4,784,112 from a societal perspective, and with an incremental cost-effectiveness ratio (ICER) of CAN$63,773/QALY relative to TIV. 87 The second study used an age-structured compartmental dynamic transmission model applied to the overall Canadian population. 88 This study found that switching from TIV to QIV across Canada would prevent an estimated 328 (6.8%) deaths; 1,876 (5.7%) hospitalizations; 3,395 (5.7%) ER visits; 52,200 (4.9%) doctor's office visits; and 135,538 (4.6%) symptomatic cases of influenza in an average season. The incremental cost-utility ratio (ICUR) for QIV compared with TIV was CAN$7,961 per QALY gained. 88 In 2004, the NACI recommended annual influenza vaccination in healthy infants/toddlers aged from 6 to 23 months in addition to the groups already recommended for vaccination and most Canadian provinces adopted the recommendation within their publically-funded immunization programs. 89 In the 2014/15 influenza season, universal influenza vaccination of healthy individuals from 6 months of age was recommended by NACI in Canada. 4 The incremental reduction in the burden of influenza B disease associated with switching from TIV to QIV was initially modeled by the US Centers for Disease Control and Prevention (CDC). 8 It was estimated that this switch would result in fewer influenza cases, hospitalizations, and deaths, dependant on the proportion of cases due to influenza B included in the analysis. As we describe above, modeling studies performed for the Canadian population show similar benefits. 87, 88 We should acknowledge that the present review has some limitations. While we performed a comprehensive systematic search to identify and select studies to inform this review, no formal systematic appraisal of these studies was performed. As such, bias, both in the study selection and in the original research (including reporting and selection bias) may exist, and this may impact upon our summary of the data. It may be noted however, that while reporting bias may exist, some consider these to have a particular impact when discussing vaccine effectiveness (VE) data. In our review, we specifically do not discuss VE data, in part as data is limited for the Canadian population, although we accept that this omission may also be a limitation of this review.
Another limitation was in the search cut-off date. While we have sought to mitigate the latter by making some mention of relevant more recent literature, it remains that other relevant recent literature has not been included in this review. In addition, reporting differences exist between epidemiological data reported in journals and those reported in surveillance reports ('gray' literature). The latter, inclusion of which we consider a strength of the review, often report preliminary, aggregated data, with less detail than formal reports and should be considered as complementary to those data reported in formal publications.
In summary, Canadian national surveillance data show that over 11 seasons, influenza A was the predominant influenza virus, and excluding the pandemic seasons, about 26-34% of cases were attributed to influenza B viruses. Published studies suggest that influenza-related complications are an important cause of hospitalization, antibiotic use, and mortality. Economic models indicate that universal vaccination with TIVs is cost-saving in Canada, with more recent models suggesting that switching from TIV to QIV may further reduce the burden of influenza B disease and be cost-effective. Nevertheless, it remains that outcomes data from the Canadian setting is limited and further real-world data is required to help inform health policy decision makers on appropriate influenza vaccination strategies for Canada.

Acknowledgments
The authors thank Eva Razumienko (GSK) for her contribution to the literature review (writing of a chapter of the report). The authors would like to acknowledge Graeme Ball (OptumInsight) for screening the articles, Jatin Gupta and Robin Jha (OptumInsight) for extracting data from the selected articles into structured tables and Qing Gu (OptumInsight) for both articles screening and data extraction. The authors thank Gr egory Leroux (Business & Decision Life Sciences on behalf of GSK) for editorial support and publication development management, V eronique Gochet (Business & Decision Life Sciences on behalf of GSK) for editorial support, and Amrita Ostawal and Annick Moon (freelance medical writers, on behalf of GSK) for medical writing.

Funding
GlaxoSmithKline Biologicals SA funded all costs associated with the literature review (GSK study identifier: HO-13-14054) and with the development and publication of this manuscript.

Authors' contributions
EWT, MKr, Mko & RS contributed to the design of the literature review, MKr supervised article selection, acted as the arbitrator of any dispute, and supervised extraction of the data from the selected articles, EWT, MKr, & Mko drafted the report of the literature review, EWT, MKr, Mko, RS & SGN interpreted the data. All authors participated in the development of this manuscript. All authors had full access to the data and gave final approval before submission. All authors agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The work described was carried out in accordance with ICMJE recommendations for conduct, reporting, editing and publications of scholarly work in medical journals. The corresponding author had final responsibility to submit for publication.