^{1}

Population prepandemic vaccine and antiviral treatment strategies may be cost-effective.

We used a hybrid transmission and economic model to evaluate the relative merits of stockpiling antiviral drugs and vaccine for pandemic influenza mitigation. In the absence of any intervention, our base-case assumptions generated a population clinical attack rate of 31.1%. For at least some parameter values, population prepandemic vaccination strategies were effective at containing an outbreak of pandemic influenza until the arrival of a matched vaccine. Because of the uncertain nature of many parameters, we used a probabilistic approach to determine the most cost-effective strategies. At a willingness to pay of >A$24,000 per life-year saved, more than half the simulations showed that a prepandemic vaccination program combined with antiviral treatment was cost-effective in Australia.

Influenza pandemics of varying severity occurred 3 times in the last century (1918, 1957, and 1968); the first influenza pandemic of the 21st century occurred in 2009. Before this latest pandemic, awareness had been heightened by the emergence of the highly pathogenic (H5N1) strain (

The stockpiling of prepandemic vaccine is also an area of active consideration (

Mathematical models of disease transmission have been used to assess the feasibility of pandemic mitigation strategies (

An age-stratified transmission model (susceptible, exposed, infected, removed) was used to calculate clinical attack rates (CAR) and antiviral drug consumption, which became inputs in a decision analytic economic model as represented in

Schematic of hybrid transmission and decision analytic economic model. [+] indicates a cloned subtree with the same structure as the branch above. In sensitivity analysis, the probabilities of healthcare utilization and death were independent of each other but dependent on the probability of clinical infection. We assumed those with serious complications would seek primary healthcare. SEIR, susceptible, exposed, infected, removed.

We considered prepandemic influenza vaccination in isolation and in combination with antiviral treatment. Four strategies for pandemic mitigation were examined (

Strategy no. | Description |
---|---|

1 | Minimum pharmaceutical intervention |

2 | Antiviral treatment of those clinically infected |

3 | Population prepandemic vaccination |

4 | Strategies 2 + 3 |

*All strategies included an initial antiviral containment effort and distribution of a matched vaccine to the population 180 days after the first locally acquired case.

We divided the Australian population into 3 age groups: 0–19 years (26% [5,513,878], 20–64 years (61% [12,744,215], and

Immunogenicity data provide evidence of prepandemic vaccine efficacy (VE) in humans (

The first dose of prepandemic vaccine was assumed to be given coincident with the first local case-patient, followed 21 days later by the second dose. In all strategies, the first dose of matched vaccine was provided 180 days after the first local case-patient was identified in Australia; the second dose was administered 21 days later in strategies 1 and 2 only. Although 2 doses of a matched vaccine would be ordered under all strategies, in base-case only 1 dose of matched vaccine was administered in strategies involving 2 doses of a prepandemic vaccine (strategies 3 and 4). Population vaccine coverage (both prepandemic and matched) was assumed to be 80% (

In the base-case model, we estimated the efficacy of antiviral treatment for preventing hospitalization as 59% (

The CAR during a pandemic was determined by using the transmission model and was primarily a function of _{0} (1.7) (_{0}) represents the number of secondary case-patients that a representative person with influenza would infect in a fully susceptible population. Asymptomatic persons were assumed to be two thirds as infectious as symptomatic persons (

The rates of hospitalization and death were defined relative to the CAR by using a patient-hospitalization rate and patient-fatality rate. We used age-specific case-hospitalization rates (0–19 years = 1.875%, 20–64 years = 2.5%,

As recommended by Australian pharmaceutical funding guidelines, we focused on direct healthcare costs (

In the base-case model, we assumed a stockpile purchase price for pharmaceuticals of $12 per vaccine dose and $32 per antiviral course; a range of values was considered in sensitivity analysis. The limited shelf-life of the pharmaceuticals requires the renewal of stockpiles for prepandemic vaccine every 3 years and antiviral drugs every 5 years. The number of times stockpiles were replaced was based on the expected time to a pandemic. We assumed partial replacement of the stockpile annually on a continuous basis. An annual storage cost for vaccines ($1, refrigeration) and antiviral drugs ($0.5, no refrigeration) was included.

We assumed that vaccination (and initial antiviral drug distribution) would be administered in mass clinics at a cost of $11.60 per course/dose (

Hospitalization costs were based on analysis previously conducted by our group, which reviewed records of patients hospitalized for influenza and pneumonia in Australia (

The costs of lost production were valued by using the human capital approach. Lost production was only valued for those employed in paid work (

The base-case analysis used an _{0} value of 1.7, which led to a CAR of 31.1% in the overall population in the absence of any intervention. The assumption of greater mixing in children meant that this group experienced the highest CARs, with 38.1% in persons 0–19 years of age, 30.4% in persons 20–64 years of age, and 20.4% in persons

Several parameters were influential in determining the CARs (_{0} (_{0} increased and declined as VE improved and coverage increased, with sharper transitions occurring as the number of secondary case-patients that a single case-patient infects approached 1. The CAR for prepandemic vaccination strategies also increased markedly when vaccination was delayed until after a local outbreak had commenced (_{0} increases (_{0}. The effect on prepandemic vaccination strategies is similar, provided the strategy is largely successful in containing the pandemic. However, for _{0} values >1.7, when this is no longer the case, prepandemic vaccinations strategies prevented fewer deaths (

Sensitivity analyses of clinical outcomes as key parameters are varied. In A–C, the clinical attack rate (CAR) is displayed as a function of _{0} and vaccine efficacy (VE) (A), vaccine coverage and VE (B), and the delay to vaccination (C). In D, deaths prevented per 100,000 population compared with no intervention is displayed as a function of _{0}.

The total discounted healthcare costs for a pandemic in the absence of any intervention was $31.1/person in the population. The gross discounted cost over 5 years (including purchase, replacement, storage, and administration) of a prepandemic vaccination program was $68.4/person in the population and the cost of an antiviral treatment program (purchase, replacement, and storage only) over the same period was $24.8/person.

The base-case results for the healthcare system perspective are shown in _{0} = 1.7, VE = 40%), strategies 2–4 each offered increased effectiveness at an increased cost when compared with the next best strategy. Under these conditions (theoretically), decision makers should first decide if strategy 2 offers value for money (incremental cost-effectiveness ratio [ICER] = $909/LYS) and then consider the value offered by each additional increase in spending, moving from strategy 2 to 3 (ICER = $1,084/LYS) and then from strategy 3 to 4 (ICER = $7,458/LYS).

Strategy no. | Net cost | Incremental cost | LYS | Incremental LYS | Incremental cost per LYS† |
---|---|---|---|---|---|

1 | 65.88 | – | – | – | – |

2 | 82.24 | 16.36 | 0.01803 | 0.01803 | 908 |

3 | 100.65 | 18.40 | 0.03501 | 0.01698 | 1,084 |

4 | 124.00 | 23.36 | 0.03814 | 0.00313 | 7,458 |

*LYS, life-year saved. Costs and life-years discounted at 5% annually; all costs calculated in 2005 Australian dollars. †Rounded to the nearest whole dollar.

From a societal perspective the least costly strategy was prepandemic vaccination (strategy 3), which was cost saving when compared with the minimum pharmaceutical intervention. Strategy 3 also dominated antiviral drug treatment alone (strategy 2), being more effective and less costly. The addition of antiviral drug treatment to prepandemic vaccination cost $7,404/LYS.

Key parameters affecting the cost-effectiveness of strategies included the _{0} value and factors impacting vaccine or antiviral effectiveness. Because strategies differed in their sensitivity to these parameters, the cost-effectiveness of strategies relative to each other varied. Dominance occurs when a strategy is considered superior to the alternative by being either more effective and less costly (simple dominance) or more effective and more costly but with a lower ICER (extended dominance) (_{0} <1.6, prepandemic vaccination alone was largely sufficient to contain the pandemic, and the addition of antiviral treatment offered only a minimal incremental effect at a high incremental cost (ICER >$1million per LYS). At lower values of VE (37%) or at higher values of _{0} (1.8), prepandemic vaccination alone was dominated by prepandemic vaccination combined with antiviral drug treatment, which offered reasonable value for money (ICER <$3,500/LYS) when compared with antiviral drug treatment alone. When we considered a VE of 20%, the addition of prepandemic vaccination to antiviral treatment alone cost ≈$9,000/LYS.

To be cost saving, prepandemic vaccine (strategy 3) and antiviral drug treatment (strategy 2) would have to be priced at <$3.1/dose and <$10.0/course when compared with the minimum pharmaceutical intervention. Variation in most other parameters did not affect the cost-effectiveness of strategies relative to each other. When the CAR was reduced (20% in the absence of any intervention) as a result of the percentage of asymptomatic infections, the ICER of all strategies increased. However, all strategies still had an ICER <$10,000/LYS. When we assumed the pandemic was relatively mild (patient-fatality and patient-hospitalization rates 5× less than base-case) and occurred 30 years later, all strategies had ICERs >$50,000/LYS. Varying the age distribution of severe clinical case-patients (patient-fatality and patient-hospitalization cases) had only a minor impact on the cost-effectiveness. When we varied the discounting rate (to be either 0% or 3% for costs and effects), ICER for all strategies decreased with no change to strategy order. Variation in other parameters was explored in probabilistic sensitivity analysis.

Cost-effectiveness acceptability curves (

Cost-effectiveness acceptability curves. Panels A and B show the healthcare system perspective; C and D show the societal perspective. In B and D, we assumed that half of the time (Q = 50%) the emergent pandemic strain would be would be of a subtype to which the stockpiled vaccine offered no protection. We did not explore the use of such a vaccine in subsequent pandemics. Costs and life-years discounted at 5% annually. A$, Australian dollars; LYS, life-year saved.

Under the assumption of a severe pandemic occurring in the near future, the pharmaceutical-based mitigation strategies examined were generally estimated to be cost-effective. For at least some of the plausible range of transmission parameters, strategies involving population prepandemic vaccination were effective in containing an outbreak until the arrival of a matched vaccine. A combination of antiviral drug treatment and prepandemic vaccination offered the best protection for the population. From a societal perspective, prepandemic vaccination was estimated to be cost saving when compared with the minimum pharmaceutical intervention.

The cost-effectiveness of pandemic influenza mitigation strategies was quite resilient to major changes in influential parameters such as the value of _{0} and the effectiveness of vaccination and antiviral drugs. This resilience stems from 2 important assumptions: 1) we assumed that the pandemic would be severe (our base-case has similar characteristics to the 1918 pandemic); and 2) we assumed a pandemic would occur soon (5-year delay in base-case). The first assumption implies that the consequences of a pandemic would be large in terms of the number of deaths and the healthcare resources required, whereas the second assumption implies that the costs associated with maintaining a stockpile were limited and that the future benefits would not be dramatically reduced by discounting. Under these assumptions, even moderately effective interventions from a clinical perspective (e.g., a vaccine with 20% efficacy) may be cost-effective. When we assumed instead that the pandemic was relatively mild (patient-fatality and patient-hospitalization rates 5× less than base-case) and occurred 30 years later, pandemic mitigation strategies were borderline cost-effective at best. This mild scenario still assumes a disease incidence several times that of seasonal influenza.

We found that vaccination and antiviral strategies differed in their sensitivity to certain key parameters (

This analysis was restricted by a lack of accurate information on prepandemic VE. However, because any emergent pandemic strain is unknown, some level of uncertainty around VE is unavoidable. We assumed that a prepandemic vaccine would offer moderate protection (below that of a matched seasonal vaccine), and using probabilistic sensitivity analysis (

Our model approach was deterministic so that although stochastic variation in parameters was considered, identical parameter choices led to identical model outputs. Because our analysis was limited to assessing the effect on overall attack rates and the costs and benefits associated with this, rather than outputs such as daily case counts, the influence of stochasticity at the simulation level should be relatively minor. Furthermore, the importation of cases from outside the country is likely to rapidly increase counts to a level at which deterministic behavior dominates. A major advantage of a simple deterministic approach is that sensitivity analyses are not constrained by computational resources, enabling detailed uncertainty analysis.

We have largely ignored issues of capacity constraint. For instance, hospital bed day capacity is likely to be severely strained during the peak of an influenza pandemic (

Population prepandemic vaccine and antiviral drug treatment strategies offer substantial scope to be cost-effective strategies for pandemic influenza mitigation. Unlike antiviral treatment strategies, population prepandemic vaccination offers the possibility of containment until the arrival of a matched vaccine. The stockpiling of prepandemic vaccines should be carefully considered and take into account the current level of uncertainty and budgetary limitations.

Description of the transmission model

Parameters: base-case and sensitivity range

This material was compiled before the declaration of pandemic (H1N1) 2009 and concerns stockpiling for a future influenza pandemic.

This research was investigator driven and funded by a grant from GlaxoSmithKline (GSK) Australia. C.R.M. has received funding for investigator-driven research from GSK and CSL Biotherapie and has been on advisory boards for GSK, Merck, and Wyeth. A.T.N. and J.G.W. have received research funding from GSK.

Dr Newall is a lecturer at the University of New South Wales, Sydney, Australia. His research interests include the economic evaluation of prevention programs and the epidemiology of infectious diseases.