| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
 |
ABSTRACT |
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Cost-effectiveness analysis (CEA) usually is the method of choice in evaluating alternative health interventions because health decision makers are primarily interested to know what health improvements can be bought with a given budget, and not the overall economic impact per se.1,2 The present paper models the cost-effectiveness of a pre-erythrocytic malaria vaccine, using a dynamic stochastic simulation model of the epidemiology of Plasmodium falciparum in malaria-endemic areas and of case mangement in Tanzania.3,4 Our objective is to assess the potential cost-effectiveness of introducing this malaria vaccine into the Expanded Program on Immunization (EPI) under a range of scenarios, conditions, and assumptions.
We present the vaccine cost-effectiveness for one country, Tanzania. This first stage enables us to specify model inputs without having to consider simultaneously many heterogeneous settings, as would be the case for sub-Saharan Africa. Even one country does not present a single uniform context for ecologic, epidemiologic, socioeconomic, and health system inputs, but there is less heterogeneity than at the multi-country level.
|
MATERIALS AND METHODS |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
A societal perspective in CEA also requires that direct economic impacts of the intervention should be taken into account. In the case of a vaccine that reduces morbidity episodes as well as mortality, there is a clear impact on productive time either leading to higher income (in the case of market work) or higher unsold production (in the case of non-market work). This can either be through a gain in production of the averted malaria case, or where the patient is a child, the production gained of the care giver who would have cared for the averted malaria case. Therefore, the results include these hypothesized economic impacts.
Given the dynamic nature of the epidemiologic model, and the lower transmission rates to other non-vaccinated individuals associated with an effective vaccine, the health effects of the intervention can also include changes in morbidity and mortality of the non-vaccinated population as a result of reduced transmission. However, our epidemiologic analysis implies that these impacts will be minimal in the epidemiologic settings that we have analyzed.5
Model overview. To predict the cost-effectiveness of the malaria vaccine, we use a stochastic simulation model of the epidemiology of P. falciparum in malaria-endemic areas of Africa.6 This includes a sub-model for the case management of malaria in Tanzania.3 We link these elements with costing of vaccine delivery in the Tanzania setting.4
The epidemiologic model is a stochastic individual-based simulation of malaria infection in disease-endemic areas that uses a five-day time step. It takes as its input the pattern of the entomologic inoculation rate (EIR) in the absence of interventions, with separate values of the EIR specified for each of the 73 five-day periods during the year. We simulate the reference case management scenario in Tanzania3 to provide a baseline with which to compare simulations where a vaccine is introduced.
The simulated population is maintained as a steady state, and includes individuals of all ages, with immune status depending on their simulated exposure. The denominators for calculation of overall health impacts include individuals who were too old to be vaccinated, and 20-year simulation is thus influenced by cohort effects due to gradual increase in the proportion of the population vaccinated, and by dynamic effects of reduction in exposure on acquisition of natural immunity to asexual parasites.
Alternatives being compared. We compare health outcomes, direct costs, and productivity gains of a combined strategy of a new malaria vaccine delivered through EPI in combination with the reference case management scenario for Tanzania with only the reference case management scenario.3 The EPI was chosen as the channel for vaccine delivery because in most African countries EPI is well established and achieves reasonably high levels of coverage amongst the target population group. Therefore, it is the only reliable mechanism to deliver a vaccine to a high proportion of infants less than one year of age.7–9
The vaccine modeled is a pre-erythrocytic stage vaccine requiring three doses to fully immunize a child. These doses are administered when infants are one, two, and three months of age, at the same time as the hepatitis B vaccine. Many of the inputs for the CEA are based on data from the case management model3 and epidemiologic scenarios.10
The cost-effectiveness model simulates the health system typical for a rural area of Tanzania.3 A set of different scenarios were constructed to reflect different malaria transmission intensities representing the stable, annually recurring pattern of malaria transmission. In all simulations, the seasonal pattern of transmission was assumed to be that recorded in the village of Namawala, Tanzania during 1989–1991 where exceptionally precise estimates of dry season transmission were made.11 The annual EIR for this site was 329 infectious bites per year. For the reference scenario, we use a seasonal pattern of transmission for a mesoendemic site, which was obtained by dividing the EIR from Namawala for each five-day period by 16. Direct measurement of dry-season transmission in mesoendemic areas is impracticable because of low mosquito densities. To simulate a high-transmission area, we use an EIR four times that of the reference scenario. This is probably more typical of high-transmission sites in Africa than the extremely high transmission in Namawala. This gives an overall annual EIR of 21 infectious bites per year, which is typical for a mesoendemic area in sub-Saharan Africa.12 The simulations were first run for a warm-up period of 90 years of exposure to define the baseline immune status of the simulated populations, which is highly age dependent. For the present analyses, the simulations are run in populations of 100,000 individuals, with an approximately stationary age distribution matching that of the demographic surveillance site in Ifakara.13
Measuring health gains. To estimate the number of disability-adjusted life years (DALYs), years of life lived with disability are calculated on the basis of the duration of disability and respective disability weights.3,14 Weights for different malaria attributable disease conditions have been obtained from the global burden of disease (GBD) study,15 and age-weighting is applied as in the GBD method. However, to assess how sensitive results are to the life table used, DALYs are also computed assuming a zero age weighting. The disability associated with anemia is assigned to the same time period as the malaria infections causing it.
Years of life lost (YLLs) and DALYs are calculated assuming age-specific life expectancies based on the life table from Butajira, Ethiopia, with an average life expectancy of 46.6 years at birth.16 This life table represents that of an east African setting with low malaria transmission and is very similar to that for Hai District, a high altitude and low malaria prevalence site in Tanzania.13 We thus compute YLLs for each simulated death under the assumption that this life table would apply in the absence of malaria.
Assumptions on vaccine efficacy. In the reference scenario, the efficacy of this hypothetical pre-erythrocytic malaria vaccine is assumed to be 52% reduction in infections in naive individuals,10 decaying exponentially with a half-life of 10 years. Since it is likely that the degree of protection provided varies between individuals, in the reference scenario, a value for the initial efficacy is drawn from a beta distribution with parameter b =10 and assigned to each vaccinated individual.5
Coverage. In the reference scenario, it is assumed that the coverage rate is the same as that reported in Tanzania for three doses of diphtheria tetanus pertussis–hepatitis B virus (DTP-HBV) vaccine in the year 2003, which stood at 89%. Given that the coverage for the first dose of DTP-HBV vaccine was 95%, the dropout rate from the first to the third dose is 6%.4
Case management. The case management model, including both formal and informal treatment, is described elsewhere.3 It has implications for health outcomes, both in terms of the potential to reduce rates of severe disease, sequelae and death, but also in the impact on transmission intensity and therefore the potential for new infections in the entire population. The rate of treatment seeking among uncomplicated malaria episodes was assumed to be 5%, which although apparently low, is justified due to the very sensitive definition of clinical episodes used. The clinical episodes simulated thus include very mild fevers that would be unlikely to elicit attendance at a health facility. The model assumes in the reference case a cure rate of 93% for the first-line drug sulfadoxine-pyrimethamine (SP) for uncomplicated malaria.5
Costs presented. We considered both marginal and average costs. The marginal cost reflects most closely the additional financial costs that would be incurred when introducing a new intervention. The average cost includes all those costs involved in delivering a health intervention, including the use of spare capacity or slack in the system, those health care resources diverted from other uses, and existing health sector resources that are shared with other health programs. All cost data are expressed in US$ 2004.
Vaccine delivery costs. The costs of introducing a malaria vaccine into the EPI in Tanzania include those related to an assumed range of vaccine purchase costs, and data collected from Tanzania on likely distribution and cold chain storage costs, management costs, vaccine delivery costs at health facility level, training costs, and social mobilization costs. A detailed description of the methodology used to estimate vaccine delivery costs can be found in an accompanying paper.4
The CEA is run under various vaccine price hypotheses ranging from US $1.0 to US $20 per dose. The vaccine delivery cost estimates according to the different price hypotheses are shown in Table 1
.
|
The direct costs of care seeking for an uncomplicated malaria episode at official facilities include the cost of an outpatient visit (US $1.02 dispensary; US $1.27 health center), a diagnostic test in a proportion of outpatient cases (US $0.30), the cost of a course of SP treatment (varying from US $0.012 to US $0.071 depending on age and weight), the cost of a course of amodiaquine treatment (varying from US $0.018 to US $0.114 depending on age and weight), and other costs incurred by patients when visiting an official health facility (US $0.30).
The direct costs of a severe malaria patient include inpatient hotel costs per day (US $7.80), drug treatment cost during hospitalization (varying from US $0.56 to US $3.74 depending on age and weight), average length of stay (4.5 days with full recovery), and the costs that patients incur when visiting an official inpatient facility (US $1.29 for the average length of stay). The case management cost inputs are presented in detail elsewhere.3
Measuring productivity gains. The productivity costs of malaria relate to the productive time lost due to illness, whether it is the patient or the patient care giver (especially if a child or elderly patient). In this analysis, productivity costs included are those related to time spent by adults seeking official care for their children, time spent by adults caring for children at home, and the time forgone by sick adults due to malaria episodes. Given that inclusion of productivity gains in CEA remains controversial, the reference case results do not include these hypothetical productivity gains.
To measure the value of productive time lost, we use the wage rate method that involves multiplying the time lost per episode (for adults only) by the average daily wage in Tanzania. These estimates are adjusted downwards by an estimate of the unemployment rate in Tanzania, thus taking into account that not all those sick or those caring for the sick would have been working.
The time lost per malaria episode is expected to be highly variable. For example, a recent review of the literature available found that for a sick adult the time off work ranges from one to five days, depending most importantly on severity of disease.17 For this study, uncomplicated adult malaria cases are assumed to lose two working days, while a care taker of a sick child loses one working day. Adults with a severe malaria episode are assumed to lose 4.5 days if not hospitalized, or if hospitalized, 1 day more than their length of stay in hospital. A care giver of a child with severe malaria is assumed not to be able to work during the hospitalization period.
For uncomplicated episodes, productivity costs are computed under two scenarios. In the first scenario, a productivity cost is attached to only those uncomplicated episodes that get treated, presumed to correspond in general to the more severe episodes. In the second scenario, a productivity cost is attached to all malaria episodes. These two scenarios represent the likely upper and lower bounds on the true productivity costs avertable through the introduction of an efficacious vaccine. The formulae for calculating productivity costs are presented below, and the data inputs are provided in Table 2.
|
 |
 |
 |
 |
where Icu and Ics are the productivity costs of the care taker for uncomplicated and severe malaria, respectively; Iau and Ias are the productivity costs of sick adults with uncomplicated and severe malaria, respectively; Tcu and Tcs are the time lost in days per episode by care taker of sick child for uncomplicated and severe malaria, respectively; Tau and Tas are the time lost in days for sick adults for uncomplicated and severe malaria, respectively; w is the minimum gross daily wage in Tanzania (US $3); and U is the assumed unemployment rate in Tanzania (40%).
Net cost calculations. The net costs associated with current case management and adding the vaccine to case management is computed over time as follows:
 |
where NC is net costs including only direct costs; DC ( cmv + v) is the direct costs in the case of the vaccine plus case management; DC (cmnv) is the direct costs of current case management under a no vaccine scenario; n is the time period of intervention (20 years); and r is the annual discount rate for future costs and health effects.
Reference scenario. In the reference case, results are presented to show cost-effectiveness at four different five-year time periods during the 20-year follow-up period (1–5 years, 6–10 years, 11–15 years, and 16–20 years) to reflect the possible fact that cost-effectiveness changes depending on time after vaccine introduction. The cost-effectiveness ratios (CERs) are presented under seven vaccine price assumptions (in US$): 1, 2, 4, 6, 8, 10, and 20. Incremental CERs are presented using two different definitions of cost: marginal cost to reflect the likely short-term financial impact of the intervention, and average cost to reflect the long-term and full opportunity cost associated with the intervention. In the reference case, only direct costs are included.
Incremental CERs are calculated under four health outcomes relevant for decision making: cost per episode averted, cost per DALY averted, cost per YLL, and cost per death averted. Future costs and benefits are presented both undiscounted and at a discount rate of 3% to reflect time preference.18
Sensitivity analysis. In addition to the reference case data assumptions and scenarios, the sensitivity analysis runs these same simulations under different assumptions. The rationales for these scenarios and the epidemiologic patterns associated with them are described in the accompanying paper.5
The different transmission intensity patterns used are low stable transmission (Namawala/64, equivalent to 5.2 infectious bites per year and high transmission (Namawala/4, equivalent to 83 infectious bites per year). The reference case is an EIR of 21 infectious bites per year, corresponding to Namawala/16. The different levels of vaccine efficacy are 30%, 80%, and 100%. The reference case is 52% entered in the model. Different decay rates for the efficacy are half-lives of 6 months, 1 year, 2 years, 5 years, and 10,000 years. The reference case is 10 years. Different distributions of vaccine effect in the population are b equals 0.01 and 100,000. The reference case is 10. Different vaccine coverage rates are a low coverage rate, with 70% of the infants receiving their first dose, and 50% receiving their third dose and complete coverage, with 100% of the infants receiving three doses. The reference cases were 89% for the third dose and 95% for the first dose. Inclusion of productivity cost savings were low productivity costs, where those with uncomplicated episodes who do not seek care are assumed not to lose productive time and high productivity costs, where all those predicted by the model to have a malaria episode are assumed to lose productive time.
|
RESULTS |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The number of deaths prevented over 20 years is 942. The number of undiscounted DALYs averted over 20 years is 58,579, which corresponds to a rate of 0.029 per capita per year. When DALYs are discounted at 3%, the number of DALY averted is 26,892, or 0.013 per capita per year. The number of undiscounted DALYs with no age weighting applied is 48,299 DALYs averted, or 0.024 per capita per year. Since most of these DALYs are due to the mortality effects, the number of YLL is very close to that of DALYs. Figure 1 presents the distribution of DALYs averted over the 20-year model period, indicating that the health effects of introducing the vaccine vary over time.
|
shows that the number of uncomplicated episodes averted is higher in the second and third five-year time periods and lower in the first and fourth five-year time periods since the start of vaccination. Most of the severe episodes averted occur in the first 10 years of the intervention, with a sharp decrease in the third five-year period, even registering a higher number of severe episodes under vaccination scenario in the fourth five-year period. Also, a higher proportion of deaths prevented are concentrated in the first 10 years after vaccine introduction. When health outcomes are discounted, this effect is stronger.
|
for discounted and undiscounted costs and at different vaccine price assumptions.
|
shows that the contribution of different cost components remains stable over the 20-year time period after introduction of the vaccine, comprising inpatient costs, out-patient costs, drug costs and patient costs. Before introduction of the vaccine, approximately 30% of direct costs are due to outpatient visits, approximately 10% to drugs, 40% to hospital care, and 20% to patient costs. After the introduction of the vaccine, more than 50% of total direct costs, at a vaccine price of US $ 1 per dose, would be due to the vaccine delivery costs. This proportion increases significantly as the vaccine price increases.
|
). The number of first-line drug treatments averted by the vaccine reaches a maximum in the second five-year interval, then decreases. The number of second- and third-line treatments averted is high in the first five-year period, after which it decreases to close to zero after 15 years. In the last five-year period, the number of first-, second-, and third-line drug treatment is higher when the vaccine is introduced. This is because in the last five-year period the vaccine does not prevent any severe episodes. There is also a shift in uncomplicated episodes to older ages, where higher drug costs are incurred due to the requirement for a greater dose.
|
over the entire 20-year intervention period and by vaccine price. The cost per death averted by introducing the vaccine is US $475, under a vaccine price assumption of US $1 per dose, increasing to US $7,158 per death averted at a vaccine dose price of US $20. The CERs using the marginal cost are generally between 97% and 99% of the CERs at average cost. Furthermore, discounting costs and health effects makes only a marginal difference to the CER, as shown in Table 5.
|
shows the relationship between CERs (for deaths averted and DALYs averted) and vaccine price.
|
shows CERs for selected health outcomes over the four time intervals, and at different vaccine price assumptions.
|
). The cost per death increases almost linearly with the vaccine price and at US $20 per dose it ranges between US $6,028 and US $9,332 per death averted at different time periods.
|
). At vaccine price of US $1, the cost per DALY averted varied between US $6 and US $11 over time, but it increases with the vaccine price up to a range of US $92 to US $149 at US $ 20 per dose.
The discounted cost per DALY averted is higher, ranging between US $11 and US $16 at US $1 per dose (Figure 5). When DALYs are computed undiscounted and assuming zero age weighting, the average direct cost per DALY averted over the four time intervals ranges between US $7 and US $12 at this vaccine price. The cost per DALY averted by the vaccination program is thus lower in the first two five-year time periods than in the latter. The CER is much higher if both costs and DALYs are discounted at 3%, and excluding the age weighting from the DALY calculation leads to a cost effectiveness ratio that is somewhere in between.
Cost-effectiveness ratios for cost per episode averted demonstrate another pattern. Since most uncomplicated episodes are prevented a few years after vaccine introduction and before the end of the third five-year interval, the cost per uncomplicated episode averted is higher in the first and last five years (US $3 at a vaccine price of US $1 per dose) and lower in the second and third five-year periods (US $2 at a vaccine price of US $1 per dose). This finding is even stronger for the severe episodes since most are averted in the first 10 years, and in the last 5 years the number of severe episodes is higher in the vaccination scenario than under no vaccination. The cost per severe episode averted is US $106 in the first five-year period, US $123 in the second five-year period, and US $1209 in the third five-year period. In the fourth five-year period, the health effect is negative, thus giving a negative CER.
Effect of transmission intensity.
The number of deaths and DALYs averted in the first 10 years of the simulation is lower in a low-transmission setting (5.2 infectious bites per year) than in the reference scenario, while in a high-transmission setting (83 infectious bites per year) is close to the number reported in the reference scenario (Figure 6). However, in a high-transmission setting, almost all deaths prevented (approximately 90%) and DALYs averted (approximately 93%) occur in the first 10 years.
|
). In a low-transmission setting, the cost per death prevented and per DALY (both undiscounted and discounted) averted are twice as high as those in the reference scenario in the first five years, and lower in the subsequent years (Table 7).
|
shows the number of deaths averted over the 20-year period at different levels of vaccine efficacy. The effect on DALYs averted shows a similar pattern.
|
shows the cost-effectiveness results under different efficacy assumptions. If the efficacy of the vaccine is 30% instead of 52%, the direct costs per death prevented and per DALY averted would be considerably higher, with the highest difference being in the second and third five-year periods where it is more than 200%. Increasing the efficacy to 80% would reduce the CERs by approximately 50% to between US $3 and US $6 per DALY averted and to between US $200 and US $400 per death averted. The cost-effectiveness of a completely efficacious vaccine would result in a considerable further improvement in the CER to US $1.4–$3.2 per DALY averted.
Effects of decay of efficacy.
The reference scenario assumes a half-life of protection against infection of 10 years. The cost-effectiveness simulations are run assuming different duration of vaccine protection from 6 months up to 10,000 years, approximating a non-decaying efficacy. The impact on DALYs averted is shown in Figure 8.
|
). However, the improvements in cost effectiveness ratios are not linear. Improving the half-life from six months to five years leads to substantial improvements in the CER, but the differences in cost-effectiveness between 5 and 10 years efficacy duration are slightly smaller.
Effects of variation in vaccine efficacy between individuals.
The distribution of vaccine efficacy among the vaccinated infants has a moderate effect on the number of deaths and DALYs that can be averted introducing the vaccine. The two alternative scenarios modeled, assuming either an all-or-nothing response or complete heterogeneity, show that the more the efficacy is concentrated in a few vaccinated subjects, the more deaths and DALYs can be prevented. This finding is also reflected in the CERs that are more favorable than the reference case if b = 0.01 (i.e., efficacy concentrated among fewer individuals), and less favorable if the effect was completely dispersed (Figure 9).
|
).
|
). A complete coverage would increase slightly the number of deaths prevented and more significantly the number of DALY averted compared to the reference scenario, while the cost per deaths and DALY averted would be slightly lower. Inclusion of productivity costs. The economic implications of reducing the burden of malaria go beyond the direct costs due to health care treatment. In the sensitivity analysis, we model the cost-effectiveness results including the productivity costs of productive time lost due to the disease. Results are presented for two assumptions of the proportion of malaria episodes where there is a productivity cost associated with the disease: the high productivity cost case where productivity costs are incurred by all episodes predicted by the epidemiologic model, and the low productivity cost case where there are no productivity costs associated with uncomplicated episodes unless the patient seeks treatment.
Over the entire 20-year follow-up period, introducing the vaccine would lead to savings in productivity costs of approximately US $263,634 in the high productivity cost scenario and US $28,443 in the low productivity cost scenario. However, since effects of the vaccine vary over time, the savings in productivity costs also vary over time (Table 8). Under the high productivity cost scenario, the savings in productivity costs reduce the total net cost of introducing the vaccine by between 49% and 90% in different time periods at a vaccine price of US $1 per dose. The savings are significantly reduced to an impact on total net cost of 3–4% when the vaccine price increases to US $20 per dose.
|
As a consequence, the cost per DALY averted (discounted) is lower when productivity costs are included, as shown in Table 9. Under the high productivity cost scenario, the total cost per DALY averted would be between US $1.7 and US $8.1 at a vaccine price of US $1 per dose. These figures represent a reduction in cost per DALY averted of between 63% and 89% when compared to the CER containing only direct costs. However, since the vaccine price increases, the cost per DALY becomes closer to the reference case analysis CER. For example, at US $20 per dose, the cost per DALY averted would be between US $148 and US $227 in different time periods, which is similar to that including only direct costs.
|
|
DISCUSSION |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Over the vaccine price range of US $1.0 to US $20 per dose, the CER is almost proportional to the price per dose, ranging (in the reference analyses) between US $12 and US $190 per (discounted) DALY averted. In the sub-Saharan African context, CERs towards the lower end of this range would be very attractive for health ministries.18,19 Up to a vaccine price per dose of almost US $10, the cost per discounted DALY averted remains less than US $100. When productivity costs due to morbidity are included, our CERs are even lower than those estimated including only direct costs. However, this difference decreases with the increase of vaccine price. There is little difference between marginal and average costs, which means that substantial savings cannot be achieved by taking up spare capacity in the health system; thus, the cost of the vaccine is the major determinant of costs.
These results should be interpreted in the context of CEA of other malaria control strategies. At vaccine price towards the lower end of the range used, our cost-effectiveness estimates of vaccination compare favorably with those of several other malaria control interventions estimated for the Global Forum for Health Research (GFHR),20 but these comparisons are problematical because of differences in the methodology. Although the GFHR study used DALY calculations based on an African life table with similar life expectancies to those that we use, our models differ by including dynamic effects that result in age and time shifts in the burden of disease.
The indirect economic impact of malaria is clearly important and we aimed to capture these effects by including productivity costs in some analyses. However, there are many pitfalls in measuring potential or actual economic impact in the context of rural Africa where most of the population is subsistence farmers, child care is often performed by older siblings, work is seasonal, and work inputs may be shifted over time and between household members. We had no empirical studies available for our estimates of time use or on the impact of malaria episodes on productive capacity. Concerns about equity effects, inadequate data, or methods for estimating economic benefits mean that indirect costs are often excluded from CEA.2,17 Indirect costs were not included in the GFHR study,20 or in any other cost-effectiveness studies to date of malaria interventions, and were not included in the analyses underpinning World Health Organization guidelines for CER thresholds for considering health interventions as attractive or very attractive.18 Our analyses that include productivity costs are thus even less comparable with those of other studies.
A major impact of malaria on productivity is likely to be by the effects on premature mortality, but it is inappropriate to include in a CEA the costs of mortality, as available from estimates of life-time earnings forgone or willingness to pay studies, since this would result in double counting of the benefits of averting deaths.1,2,21 Among the microeconomic studies on the economic consequences of malaria published, only one22 has included productivity costs due to premature mortality. That study estimated the economic burden of malaria and not the cost-effectiveness of interventions.
A malaria vaccine may also have positive impacts on social and economic development that are not captured by the productivity cost savings. Endemic malaria is associated with substantially lower indices of economic development at the national level,23,24 and reducing the burden of malaria might have macroeconomic benefits that are not captured in microeconomic analyses. However the epidemiologic analyses5 clearly indicate that on their own vaccine programs with profiles like those we investigated will avert a proportion of illness events that is much lower than the primary vaccine efficacy, and will have little or no effect on malaria endemicity. In this context, it would be surprising if they had substantial effects on economic development.
We obtain only modest estimates of the wider economic benefits of a vaccination program if we apply the recently suggested approach25 of estimating these benefits by multiplying the number of DALYs averted by the average GDP per capita. Using our prediction that a pre-erythrocytic malaria vaccine would avert between 0.013 and 0.029 DALYs per capita per year and the GDP per capita of Tanzania (US $322 in 2005), the annual per capita economic benefits would be between US $4.2 and US $9.3 (according to whether DALYs are discounted and aged weighted).
These conclusions reflect the reference case, but the CER is highly sensitive to assumptions about the epidemiologic setting and vaccine characteristics including the transmission intensity, the efficacy, and duration of protection (Table 7). The CER varied with the time since the start of the vaccination program because the epidemiologic model does not reach equilibrium within the time scale of the simulation.5 In general, the cost per DALY averted is lower in the first phase of the vaccination program than later, with the highest cost per DALY in the third five-year time period after the start of the program. Extending the duration of protection increases the CERs in the third and fourth five-year time periods. A vaccine boost at some specified time point may have a similar effect, although this would involve additional costs that would be included in calculation of the CER. We have not addressed the emerging problem of drug resistance, which could be included in the case management model and would presumably increase the cost per DALY averted.
Our simulations considered only a limited set of sources of heterogeneity. In particular, we assumed that each person in the simulation was exposed to the same entomologic challenge, and that the chances of being vaccinated were independent of individual susceptibility to disease. We also assumed homogeneous probabilities of accessing health care. Over a period of 20 years, the introduction of a new malaria vaccine would have an impact on the health system and on the case management of malaria. It would be possible to simulate more realistic patterns of heterogeneity but the field data on which to base such models are very limited.
Some counter-intuitive behavior in CERs corresponds to health effects in the model. When episodes are delayed rather than averted, they occur in older individuals who may require larger drug dosages. Thus, the health benefit of delaying illness may be partially offset by increased costs. Since the epidemiologic model also corresponds with field data that suggests a maximum incidence of clinical episodes (though not mortality) at intermediate transmission intensities,26–29 it is possible for reductions in malaria transmission to lead to increased case loads.
The proportion of clinical episodes averted varies by transmission intensity,5 as do the numbers of DALYs averted. The numbers of clinical episodes continues to decrease after 10 years of the vaccine introduction only in low-transmission scenarios. This is explained by the fact that in high-transmission settings there is an increase in severe malaria incidence in children more than five years of age due to reduced accrual of immunity to asexual blood stage parasites during early childhood. In addition, the pyrogenic threshold, which determines the parasite density that leads to acute illness, depends on the recent exposure to parasite and can be lower in vaccinated individuals.10 In the model, the lower level of acquired immunity in vaccinated individuals and the resulting inability to effectively control parasite densities also leads to higher proportion of the acute episodes being severe.
An extension to the current work will be to carry out a full probabilistic sensitivity analysis. This will enable us to present acceptability curves in addition to the presentation of CERs in this report. However, the present analyses already indicate that a pre-erythrocytic malaria vaccine, even one with moderate efficacy and minimal effectiveness in reducing transmission to the vector, could be a cost-effective intervention in reducing the intolerable burden of malaria in sub-Saharan Africa.
Received September 18, 2005. Accepted for publication April 9, 2006.
Acknowledgements: We thank the members of the Technical Advisory Group (Michael Alpers, Paul Coleman, David Evans, Brian Greenwood, Carol Levin, Kevin Marsh, F. Ellis McKenzie, Mark Miller, and Brian Sharp), the Project Management Team at the Program for Appropriate Technology in Health (PATH) Malaria Vaccine Initiative, and GlaxoSmithKline Biologicals S.A. for supporting this study.
Financial support: The mathematical modeling study was supported by the PATH Malaria Vaccine Initiative and GlaxoSmithKline Biologicals S.A.
Disclaimer: This publication and the contents hereof do not necessarily reflect the endorsement, opinion, or viewpoints of the PATH Malaria Vaccine Initiative or GlaxoSmithKline Biologicals S.A..
* Address correspondence to Thomas A. Smith, Swiss Tropical Institute, Socinstrasse 57, Postfach, CH-4002 Basel, Switzerland. E-mail: Thomas-A.Smith{at}unibas.ch 
Authors’ address: Fabrizio Tediosi, Guy Hutton, Nicolas Maire, Thomas A. Smith, Amanda Ross, and Marcel Tanner, Swiss Tropical Institute, Socinstrasse 57, Postfach, CH-4002 Basel, Switzerland, Telephone: 41-284-8273, Fax: 41-284-8105, E-mails: fabrizio.tediosi{at}unibas.ch, nicolas.maire{at}unibas.ch, Thomas-A.Smith{at}unibas.ch, amanda.ross{at}unibas.ch, guy.hutton{at}unibas.ch, and marcel.tanner{at}unibas.ch.
Reprint requests: Thomas A. Smith, Swiss Tropical Institute, Socinstrasse 57, Postfach, CH 4002 Basel, Switzerland.
|
REFERENCES |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  T. SMITH, G. F. KILLEEN, N. MAIRE, A. ROSS, L. MOLINEAUX, F. TEDIOSI, G. HUTTON, J. UTZINGER, K. DIETZ, and M. TANNER MATHEMATICAL MODELING OF THE IMPACT OF MALARIA VACCINES ON THE CLINICAL EPIDEMIOLOGY AND NATURAL HISTORY OF PLASMODIUM FALCIPARUM MALARIA: OVERVIEW. Am J Trop Med Hyg, August 1, 2006; 75(2_suppl): 1 - 10. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. ROSS and T. SMITH THE EFFECT OF MALARIA TRANSMISSION INTENSITY ON NEONATAL MORTALITY IN ENDEMIC AREAS. Am J Trop Med Hyg, August 1, 2006; 75(2_suppl): 74 - 81. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. HUTTON and F. TEDIOSI THE COSTS OF INTRODUCING A MALARIA VACCINE THROUGH THE EXPANDED PROGRAM ON IMMUNIZATION IN TANZANIA. Am J Trop Med Hyg, August 1, 2006; 75(2_suppl): 119 - 130. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||
