Cost-Effectiveness of Live Attenuated Versus Inactivated Influenza Vaccine Among Children

Annual vaccination has been accepted as the most effective method for preventing influenza illness and its complications, and has been recommended in the United States for all individuals 6 months and older without contraindications to vaccination.¹ Seasonal influenza epidemics lead to significant direct medical and societal costs in addition to heightened mortality risks.^2,3 Previously published studies have demonstrated that vaccination reduces costs associated with influenza.^4-9

Currently, 2 types of influenza vaccines are available in the United States: live attenuated influenza vaccine (LAIV) and trivalent inactivated influenza vaccine (IIV). LAIV is approved for eligible individuals aged 2 to 49 years, and some IIV formulations are approved for use in eligible individuals 6 months and older (IIV products differ with respect to eligible populations).

LAIV has demonstrated higher efficacy than IIV in 3 randomized controlled trials in children.¹⁰ Prior studies have examined the cost-effectiveness of vaccinating children with LAIV versus IIV. To the best of our knowledge, only 1 study has examined cost-effectiveness using efficacy data from a randomized trial that directly compared the 2 vaccines; this analysis was limited to children aged 24 to 59 months.⁴ New data on influenza mortality, resource utilization, and costs, along with the expanded recommendations for annual influenza vaccination, may impact previous results. We sought to evaluate the impact of this data on the relative cost-effectiveness of LAIV versus IIV.

The objective of this analysis was to assess the cost-effectiveness of trivalent LAIV (T/LAIV) versus trivalent IIV (T/IIV) in children aged 2 to 17 years across differing influenza season severities from US societal and payer perspectives, using efficacy data from the 3 randomized controlled clinical trials that directly compared the 2 vaccines.

METHODS

Model Structure

A decision-tree model with probabilistic sensitivity analysis (PSA) functionality was developed in Microsoft Office Excel 2007 with Visual Basic for Applications macros to analyze the cost-effectiveness of influenza vaccination with T/LAIV versus T/IIV in United States children aged 2 to 17 years. The adapted model structure, based on published influenza models,^4,11 is presented in

Figure 1

.

Influenza Infection and Complication Rates

The T/LAIV and T/IIV influenza infection rates for our base case analysis were taken from 2 clinical trials^12,13 performed in the 2002-2003 influenza season¹⁴ (

Table 1

^10,11,14-30). These trials provided a unique opportunity to evaluate T/LAIV and T/IIV in children aged 2 to 17 years in the same influenza season. There is no defined average influenza season in the United States. In a separate trial of T/LAIV versus T/IIV,¹⁸ attack rates in the T/IIV arm were 8.6% compared with the 5.8% seen in the T/IIV arm of the trials used in the current model. The influenza season of that study (2004-2005) has been referenced as mild to moderate by the US Vaccines and Related Biological Products Advisory Committee.³¹ Because the attack rate in the TIV arm used for our model was lower than that described by Belshe and colleagues,¹⁸ we considered the base case clinical trial—sourced scenario as mild. Because influenza severity can differ significantly by season and there is no defined attack rate for an average season, analyses were performed for hypothetical US moderate and severe influenza seasons to provide vaccine decision makers with estimates of the impact of T/LAIV and T/IIV across potential severity scenarios. T/LAIV versus T/IIV relative risk reductions from the clinical trials were applied to the defined moderate and severe influenza seasons. In the absence of clinical trial data, CDC surveillance data were used to estimate influenza attack rates. The CDC data provide the most comprehensive and quantitative information available on influenza seasons in the United States. The trial-based mild season attack rates were proportionally scaled using CDC influenza-like illness (ILI) count data for patients aged 0 to 24 years¹⁶ from more severe influenza seasons. Qualitative CDC surveillance influenza season descriptions were used to ensure that the ILI count data were reflective of season severity. The 2003-2004 season was designated as “moderately severe” and in the 2005-2006 season “activity remained elevated for a longer period of time.”¹⁴ The ILI counts for the 2003-2004 and 2005-2006 seasons were approximately double those of the 2002-2003 season,¹⁶ so the mild season attack rates were doubled to model the hypothetical moderate season. The 2007-2008 season was “associated with greater overall mortality, and higher rates of hospitalization.”¹⁴ The ILI counts for the 2007-2008 and 2008-2009 seasons were approximately 4 times those of the 2002-2003 season.¹⁶ To model the hypothetical severe season, the trial-based mild attack rates were multiplied by a factor of 3 to be conservative instead of the approximate ILI count-derived ratio of 4.

The model implemented acute otitis media (AOM) and lower respiratory infection (LRI) complication rates given influenza infection differentiated by vaccine and age (Table 1). The model assumed that the complication rates were constant across influenza attack rates. AOM complication rates were taken from a pooled analysis of 2 T/LAIV versus T/IIV controlled trials.¹⁷ The rates for children aged 2 to 7 years were used for children aged 2 to 8 years. A postmarketing observational study (R. Baxter and S. L. Toback, MedImmune, unpublished observations, 2011) documented an incidence of medically attended AOM in older children (9-17 years) that was approximately 40% of the incidence for children aged 5 to 8 years. This proportion was applied to the published rates and used as the AOM complication rate for children aged 9 to 17 years. LRI complication rates were taken from a clinical trial of T/LAIV versus T/IIV in children aged 6 to 59 months.18 LRI given influenza values for children aged 6 months to 5 years were applied to children aged 2 to 5 years in the model. Prosser and colleagues11 noted a lower LRI infection rate for children aged 5 to 11 and 12 to 17 years compared with younger children. In the absence of LRI infection rates for older children from clinical trials, age-based ratios of LRI rates calculated from Prosser and colleagues’ data were used to approximate LRI rates for the older children.

Adverse events associated with vaccination, such as reactogenicity, injection site events, and medically significant wheezing, were not included in the model. The most common adverse events associated with vaccination are low-cost, low-utility impact events such as runny nose, cough, or injection site pain or swelling.

Influenza Vaccine Efficacy

The influenza vaccine efficacies for children aged 2 to 5 and 6 to 17 years are from 2 randomized, open-label studies conducted in Europe and Israel during the 2002-2003 influenza season. The trial for the younger age group (2-5 years) was conducted in children aged 6 to 71 months with recurrent respiratory tract infections¹²; a meta-analysis compared T/LAIV and T/IIV influenza incidence rates in those trials for children aged 24 to 71 months, and the relative risk reduction reported for T/LAIV versus T/IIV was 48%.¹⁰ The trial for the older age group (6-17 years) was conducted in children and adolescents aged 6 to 17 years with medically stable asthma¹³; a meta-analysis was published comparing the T/LAIV versus T/IIV incidence rates for children aged 6 to 11 and 12 to 17 years, and the reported relative risk reductions for T/LAIV versus T/IIV were 31% and 30%, respectively. Although children with asthma had an increased risk of influenza complications, their risk of influenza illness was judged to be similar to that of healthy children of the same age.¹⁵ In the United States, T/LAIV currently is not recommended for use in children with recurrent wheezing or asthma.

Influenza Resource Utilization and Mortality

The model assumed that the utilization of medical resources given influenza infection is vaccine independent (Table 1). Uncomplicated influenza patients had an age-dependent risk of seeking outpatient medical care.¹¹ Complicated influenza patients were assumed to seek outpatient care, and complicated influenza patients with LRI had an age-dependent risk of inpatient hospitalization.¹¹ The model inputs estimated mortality rates for uncomplicated and complicated influenza.

Costs

Costs in the model included direct medical and indirect costs (Table 1). Costs were inflated to 2010 US dollars using the January 2010 values from the medical care component of the Bureau of Labor Statistics’ Consumer Price Index.³²

The model includes 3 categories of influenza-related direct medical costs: vaccination, outpatient, and hospitalization/emergency department (ED) visit (Table 1). Vaccination costs included private-sector vaccine acquisition and administration reimbursement costs.²¹ The default T/IIV costs were derived by weighting the private sector prices for preservative-free pediatric and adult doses based on the modeled age distributions for ages 2 to 17 years.³³ Administration was based on published average Medicare reimbursement for Current Procedural Terminology code 90460 (immunization administration through 18 years of age via any route of administration, with counseling by physician or other qualified healthcare professional; first vaccine/toxoid component).²² Costs were obtained from a claims analysis for ILI patients who sought outpatient care (MedImmune, unpublished observations, 2011.

Indirect costs were accumulated for parent work days missed caring for sick children aged 2 to 12 years.²⁵ The model assumed 1 missed work day for each instance of dependent influenza illness.^2,26

Utilities and Quality-Adjusted Life-Years

Quality-adjusted life-year (QALY) reductions were calculated to reflect disutilities and illness times associated with influenza events (Table 1). The baseline healthy utility value was obtained from a study of patients without a chronic condition in a community-dwelling population.²⁷ The influenza relative utility value was from an influenza cost-effectiveness model²⁸ and reflects the loss of utility due to home confinement, limited physical and social activity, fatigue, and so forth. QALYs lost to influenza-related mortality were based on rates for influenza by complication,^11,20 average remaining life expectancy,²⁹ and healthy baseline utility at death; they were discounted 3% annually.³⁰

Analyses

The primary model outcome was the incremental costeffectiveness ratio (ICER) in US dollars per QALY for T/LAIV versus T/IIV. Additional outcomes included the numbers of influenza cases and outpatient resource utilization visits avoided per 10,000 vaccinated children. Costs per vaccinated child were calculated for both payer (direct medical costs) and societal (direct and indirect costs) perspectives.

One-way sensitivity analyses were performed to note the impact of key model inputs, including influenza season severity, T/LAIV vaccine efficacy, direct medical cost, societal cost, and demographic inputs. PSAs were also performed. Analyzed inputs included T/LAIV vaccine efficacy, direct medical cost, societal cost, and influenzarelated disutility inputs. T/LAIV vaccine efficacy was analyzedby varying T/LAIV influenza attack rates to simulate overall T/LAIV versus T/IIV relative risk reduction values greater than and less than the base of value of 35%.

This model did not incorporate dynamic transmission or the potential impact of herd immunity.

RESULTS

Base Case Results

T/LAIV resulted in fewer influenza cases, hospitalizations, outpatient physician visits, and ED visits versus T/IIV for all 3 modeled influenza seasons (

Table 2

). For the trial-based mild influenza season, use of T/LAIV prevented (per 10,000 vaccinated patients) 214 overall influenza cases (30 complicated cases), 100 physician visits, 9 ED visits, and 1 hospitalization. The relative benefit of T/LAIV versus T/IIV increased with the severity of the influenza season. In the hypothetical moderate and severe seasons, T/LAIV prevented 428 and 642 overall influenza cases, respectively. Use of T/LAIV resulted in QALY gains versus T/IIV for all modeled seasons.

For the trial-based mild influenza season using the societal perspective, the ICER of T/LAIV versus T/IIV was $27,875 per QALY. In the hypothetical moderate and severe seasons, T/LAIV was dominant (more QALYs gained, cost savings) over T/IIV. T/LAIV had a higher average per patient vaccination cost versus T/IIV of $8.96 due to higher vaccine acquisition costs; however, T/LAIV had lower relative average direct medical costs for the mild, moderate, and severe influenza seasons of —$1.46, –$2.93, and –$4.39, respectively, reflecting the higher modeled efficacy of T/LAIV (Table 2). The incremental direct per patient cost impact of T/LAIV versus T/IIV for the influenza seasons was $7.50, $6.04, and $4.57 for the mild, moderate and severe seasons, respectively. Indirect cost savings associated with T/LAIV were –$3.89, –$7.79, and –$11.68. The resulting total incremental costs per patient were lower for T/LAIV versus T/IIV for the moderate and severe seasons (–$1.75 and –$7.11), and slightly higher for the trial-based mild influenza season ($3.61.

From a payer perspective, T/LAIV remained more costeffective than T/IIV but had higher direct costs. The ICER of T/LAIV versus T/IIV was $57,986 per QALY for the trial-based mild influenza season and became more costeffective as the severity of the influenza season increased ($23,338 and $11,788 per QALY for the hypothetical moderate and severe seasons, respectively.

Sensitivity Analyses

T/LAIV was cost saving and dominant versus T/IIV from a societal perspective for a number of the 1-way sensitivity analyses and was cost-effective within the range of other Advisory Committee on Immunization Practices (ACIP)- recommended vaccines for the rest of the analyses (

eAppendix Table

available at ajmc.com).³⁴ The severity of the influenza season had the largest impact on QALYs gained, because use of T/LAIV prevented more influenza illnesses as the season severity increased. Medical costs had minimal impact on total cost per vaccinated child, whereas changes in vaccine price, relative vaccine efficacy for T/LAIV versus T/IIV, and proportion of vaccine-naïve patients missing their second dose had moderate impact; however, from a societal perspective, use of T/LAIV remained cost saving or cost-effective. Changes in season severity, vaccine price, and relative vaccine efficacy were the most influential inputs to the ICER from the payer perspective.

From a societal perspective, the mild season PSA results (

Figure 2

) show that use of T/LAIV versus T/IIV was consistently cost-effective and sometimes dominant within the ranges of the modeled uncertainty. T/LAIV wasdominant (more QALYs gained, total cost savings) in 3% of the simulations. For the moderate season PSA results (

Figure 3

), T/LAIV was dominant versus T/IIV in 70% of simulations from the societal perspective. Consistent PSA results were seen for the severe season (

eAppendix Figure

available at ajmc.com).

DISCUSSION

The current healthcare environment is focused on reducing costs while improving overall patient care. Influenza continues to present a significant burden to the healthcare system, and influenza vaccination rates should continue to be supported by all healthcare providers. Despite the increased cost of T/LAIV relative to T/IIV, immunization with T/LAIV in our model was cost-effective or cost saving for the pediatric population depending on season severity and should be made available for all eligible children by healthcare providers and health plans. T/LAIV reduced the number of influenza illnesses and healthcare resource use versus T/IIV in children aged 2 to 17 years in all modeled influenza seasons. From a societal perspective, T/LAIV became dominant over T/IIV as the modeled influenza season became more severe. When only direct costs were included in the model, the ICER associated with the use of T/LAIV versus T/IIV ranged from $11,788 to $57,986 per QALY, with T/LAIV more cost-effective as the season severity increased. The ICERs from our model are in line with those for other ACIP-recommended vaccines, for which societal costs have been shown to range from cost savings to $180,000 per QALY.³⁴ Additionally, our results do not include the impact of indirect protection afforded through reduction in transmission, which would only improve cost-effectiveness. Reduced transmission of influenza because of higher vaccination rates would lead to cost offsets because influenza would be prevented in unvaccinated individuals. Therefore, our model presents a conservative value for influenza immunization.

With the 2008 US recommendation to vaccinate all children 6 months to 18 years, followed by the 2010 universal recommendation for annual influenza vaccination for all eligible individuals aged 6 months and older, increased attention has been focused on vaccinating all children against influenza. Previous economic analyses have reviewed the comparative efficacy of T/LAIV versus T/IIV using indirect comparisons or have used primary efficacy data focusing on younger children (2-5 years). Our model built on those previous efforts to describe the value of T/LAIV use in the pediatric population, but is the first to use direct clinical efficacy data for children aged 2 to 17 years. The clinical trial data used in our models compared T/LAIV with T/IIV with no placebo arm; therefore, our model utilizes relative efficacy compared with T/IIV.

The key driver of model results was influenza season severity, followed by parental missed work and vaccine price. The sensitivity of the results to season severity highlights that the cost-effectiveness of vaccination can vary considerably between seasons. Seasons with lower influenza attack rates show higher ICERs for T/LAIV versus T/IIV than seasons with higher attack rates. Lower attack rates result in fewer potential influenza cases to prevent, thereby decreasing the value of a more efficacious vaccine. This is consistent with results demonstrated by Prosser and colleagues,11 who found higher ICERs for T/IIV versus no vaccine in seasons with lower influenza attack rates. In seasons with higher influenza attack rates, the cost-effectiveness of T/LAIV would be more pronounced because the difference in efficacy would be more evident. Given the inability to predict the severity of an influenza season, vaccine decision makers, payers, and healthcare providers should consider all potential severity scenarios along with patient eligibility and preference when deciding on the type of influenza vaccine to provide to patients.

Limited data exist on work time lost by caregivers associated with influenza in children aged 2 to 17 years. Our analysis assumed 1 missed day of work for parents of children aged 2 to 12 years with influenza irrespective of influenza complication, based on previous studies.2,26 Using the CDC-funded New Vaccine Surveillance Network, Ortega-Sanchez and colleagues2 examined work time missed in the United States for parents of children younger than 5 years with laboratory-confirmed influenza in the inpatient, ED, and outpatient settings. Parents missed an average of 73 hours (9.1 days), 19 hours (2.4 days), and 11 hours (1.4 days) in each setting, respectively. A prospective, multi-center European study reported that parents of children younger than 14 years with laboratory-confirmed influenza missed an average of 1.25 work days.²⁶ In a prospective observational cohort study examining the impact of household member respiratory illness on employee absenteeism in the United States, Palmer and colleagues³⁵ reported that employees with any influenza-like illness in a child household member (<18 years) missed significantly more work than employees with no acute respiratory illness in a child household member (0.9 vs 0.3 days; P <.001). However, these numbers may underestimate work days lost because they reflect only individual employee-reported work loss and not total work loss for the household. Our model conservatively assumed 1 work day loss for children aged 2 to 12 years with influenza based on the findings of Principi and colleagues,²⁶ which are supported by the findings of Ortega-Sanchez and colleagues. ² We assumed that older children were more likely to be allowed to stay home alone with influenza illness and did not apply parental missed work to their associated illness. Additionally, we did not differentiate work loss by influenza severity because we had limited or no data on impact to work loss by setting of care or severity for children older than 5 years. These assumptions likely underestimated total lost productivity by excluding work loss associated with influenza in older children (>13 years) and increased number of work days missed because of more severe illness.

Other published models used higher outpatient medical and hospitalization costs than those used in our model. Outpatient costs in our model ($77.80 for uncomplicated influenza and $82.40 for complicated influenza) were derived from a prospective, observational, cohort study among employees of 3 large US companies (MedImmune, unpublished observations, 2011). The study linked subjectreported ILI cases for themselves or household members with subjects’ healthcare claims. The use of claims data may underestimate the cost of influenza-related outpatient visits because of the inclusion of ILI instead of only confirmed influenza. Luce and colleagues⁴ modeled a single physician visit for all cases of medically attended influenza costed at $77, but included additional physician visits for AOM- and LRI-complicated influenzas costed at $77 and $116, respectively. Additionally, Luce and colleagues analyzed only children aged 24 to 59 months, as opposed to our model, which analyzed children aged 2 to 17 years. Costs derived solely from children younger than 5 years may not be applicable to older children. Molinari and colleagues³ stratified outpatient visit costs by age and high-risk conditions, from a mean of $95 for non—high-risk patients aged 5 to 17 years to $649 for high-risk patients of the same age range. Our model did not distinguish between risk groups. Molinari and colleagues also had substantially higher hospitalization costs of $15,014 and $41,918 for the 2 risk groups compared with our model value of $7116. Utilization of higher outpatient and inpatient medical costs would increase total costs of influenza and make use of T/LAIV more cost-effective than T/IIV in our model.

Previous models examined the cost-effectiveness of T/LAIV versus T/IIV in children. Analyses by Prosser and colleagues 11addressed comparative efficacy using an indirect evidence approach for children aged 6 months to 17 years. Prosser and colleagues found ICERs for T/LAIV versus no vaccination ranging from $15,000 per QALY in non—highrisk 2 year olds to $109,000 per QALY for non–high-risk children aged 12 to 17 years. Although our model included a similar population and used the same probabilities for seeking healthcare, the Prosser model used effectiveness data taken from meta-analyses examining the efficacy of T/IIV and T/LAIV versus placebo rather than the direct efficacy of T/LAIV versus T/IIV. Our model is the first touse direct efficacy data to compare T/LAIV with T/IIV for the broad pediatric population (2-17 years of age) taken from the same influenza season. Use of data from the sameinfluenza season limits confounding in vaccine effectiveness due to factors that may vary by influenza season such as season severity and vaccine match to circulating strains. Additionally, Prosser and colleagues did not consider indirect costs. Our model incorporated both direct medical and parental missed work costs to represent a more holistic view of influenza burden.

Luce and colleagues⁴ utilized direct efficacy data to compare T/LAIV with T/IIV, but only for children aged 24 to 59 months. Their analysis reported a societal perspective incremental cost savings of $45.80 per child vaccinated with T/LAIV versus T/IIV. Our model suggests a small cost increase of $2.41 per child vaccinated with T/LAIV versus T/IIV; however, our analysis differs in several key ways. Our model included a larger pediatric population of children aged 2 to 17 years. Older children are less likely to have influenza complications such as AOM and LRI, and therefore are lower utilizers of healthcare resources compared with younger children. Additionally, the modeled relative efficacy of T/LAIV versus T/IIV for children aged 6 to 17 years was slightly lower than that for children aged 2 to 5years. Luce and colleagues used a micro-costing algorithm for influenza costs that resulted in significantly higher costs per influenza infection compared with the claims analysis data used in our model. Although our data were based on ILI cases and not confirmed influenza, we think they reflect the real-world costs of managing influenza infection.

Previous cost-effectiveness analyses have focused on the use of influenza vaccination in general rather than the comparative cost-effectiveness of one vaccine versus another, or they have focused on the use of T/LAIV in children versus no vaccination in specific settings (such as schools), or they have focused on adult or elderly populations. ^36-39 This analysis is the first to examine T/LAIV versus T/IIV for US children aged 2 to 17 years using direct efficacy data.

Limitations

All computer simulation models are simplifications and cannot exactly replicate or predict real-world outcomes. Currently, there is no systematic collection of influenza attack rate data in the United States. The availability of headto-head clinical data for all pediatric ages from a single influenza season dictated our decision to use the 2002- 2003 influenza season as the base case analysis. The base case season should not be interpreted as an average or typical influenza season. It is extremely difficult to predict the severity of any influenza season given 2 cocirculating B lineages, the existence of animal hosts for influenza A providing an opportunity for pandemics through mutation or reassortment, the potentially limited cross-protection across strains and lineages, and changing behavioral trends in people, such as increasing urbanization and travel that facilitate the spread of influenza viruses. Therefore, the model used CDC historical descriptions and data to define and simulate more severe influenza seasons relative to the base case mild season to better understand and estimate the cost-effectiveness of LAIV in various severity scenarios. The definitions of the mild, moderate, and severe influenza seasons are for analysis purposes only, and do not imply either the frequency or likelihood of the occurrence of these season severities. Rather, they are intended to provide a broader picture of the value of LAIV use in a pediatric population across a variety of scenarios.

The model incorporated trial data from children with a history of respiratory conditions and applied the findings to the entire pediatric population. It is unknown how applicable the efficacy results are for the general population. The clinical trials used for the input estimates did not include placebo arms; therefore, the baseline unvaccinated influenza attack rate for the study populations is unknown. Additionally, influenza severity varies seasonally. We examined attack rates for 3 severity seasons to understand the impact of influenza attack rates.

The influenza-related outpatient costs in the model were lower than costs from published literature,^4,11 particularly for complicated influenza. The cost of influenza with AOM was used for all complicated influenza cases in the model. However, we do not believe this significantly impacted our results, because outpatient costs were not a major driver of model results. Finally, our analysis was limited to a US pediatric population and did not model the broader impact of reduced transmission, which has been significant in other analyses; cost-effectiveness results would likely be different for other age groups or countries.

CONCLUSIONS

In our model, use of T/LAIV versus T/IIV resulted in cost savings from a societal perspective in the moderate and severe influenza seasons, and acceptable costeffectiveness in the mild season. Vaccination with T/LAIV versus T/IIV reduced the number of influenza illnesses and healthcare resource utilization in children aged 2 to 17 years. Indirect costs associated with parental missed work were a significant driver of model results. Our model did not examine the impact of vaccination on influenza transmission. When indirect costs were removed from the model, T/LAIV was no longer cost saving; however, regardless of severity of season and without inclusion of indirect protection, use of T/LAIV remained cost-effective.