Perspective and boundary

The simulations refer to CEA under the perspective of the society as a whole, although only direct costs are included. They thus consider all relevant direct resource inputs and costs to the interventions, and resource consequences, costs, and health impacts resulting from the interventions. The indirect economic impact of malaria such as potential earning forgone of patients and unpaid carers is not included, as its inclusion is controversial in CEA[15, 16]. The CEA follows standard cost-effectiveness methodology[8, 17–21].

Interventions being compared

The simulations of vaccines are based on a previously described model for the natural history and epidemiology of P. falciparum malaria[13]. This model uses an underlying model based on descriptions of the course of parasite densities in malaria therapy patients[22]. The parameterization of the model for the present simulations is described in the companion article[14] and reviewed in more detail elsewhere[23]. Briefly, each simulated vaccine is characterized by an average initial efficacy (the efficacy reached after completion of a schedule of three doses), and by a half-life of this efficacy, which is assumed to decay exponentially with time. The vaccinated population is assumed to be heterogeneous in the response to vaccination, and to allow for this we assign initial values for efficacy drawn from a beta-distribution[11] (simulated vaccines are delivered at specified ages, and a range of coverage values is considered for vaccination to allow for individuals who do not complete the full schedule).

The effects of the three different vaccine types and four combinations were modeled as follows (see also [14]):

Vaccine delivery modalities

The delivery of the three vaccine types and their combinations through the following three strategies are simulated:

Vaccine coverage

For vaccines delivered via EPI, the assumed coverage of full vaccination (three doses) corresponds to that reported in Tanzania for three doses of diphtheria tetanus pertussis-hepatitis B virus vaccine in the year 2003, which stood at 89%. The assumed dropout rate from the first to the third dose is 6% since coverage for the first dose of DTP-HBV vaccine was 95%. When booster doses are included, it is assumed that 99% of those that receive the third EPI dose will be given a booster dose 1, 2, 3 and 4 years after the last EPI dose. For mass vaccination the coverage levels of 30, 50 and 70% are simulated.

Case management model

As detailed in the companion article[14], the simulations of the effects of vaccine interventions use a case management model including both formal and informal treatment, similar to a previous study of the authors[24]. An artemisinin-based combination therapy (ACT), artemether-lumefantrine, is used as first line treatment for uncomplicated malaria, as per recent policy changes, and the drug action model was modified accordingly, both in terms of the potential to reduce rates of severe disease, sequelae and death, and the impact on transmission intensity. The model assumes that 90% of patients comply with the ACT treatment schedule and have a cure rate of 85%, while in non-compliers there is no effect.

Measurement of health impact

Each simulation is repeated three times using different seeds to initialize the random number generator, and each of these simulations is compared with an independent simulation of a reference scenario: a control population with no vaccine, but with the same human demography, baseline transmission, and health system.

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. Weights for different malaria attributable disease conditions have been obtained from the Global Burden of Disease (GBD) study[25]. DALYs are computed with no age weighting to follow standard cost-effectiveness practices[26]. 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[27]. 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[28]. YLLs for each simulated death are computed under the assumption that this life table would apply in the absence of malaria.

Vaccine delivery costs

The vaccine delivery costs are estimated using the methodology of a previous study by the authors[29], which was based on an ingredient approach requiring information on the quantities of physical inputs needed and their unit 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 costs of distribution, cold chain, management, vaccine delivery by health facilities, training, and social mobilization[30] (Table 1). For booster doses, the per-dose delivery cost is assumed to be the same as that of routine EPI.

Potential cost savings of the interventions

The numbers of uncomplicated and severe malaria episodes averted due to vaccination are multiplied by the case management unit costs, as described above, to estimate the potential cost savings for both the health system and households. The cost savings are subtracted from the vaccine costs to compute the net costs of the interventions.

Calculating cost-effectiveness ratios and interpreting the results

The cost-effectiveness ratios presented are calculated for four health outcomes: uncomplicated and severe malaria episodes averted, DALYs averted, and deaths averted. Future costs and benefits are discounted at 3%. The cost-effectiveness ratios are presented for assumed vaccine prices of US$ 2 or US$ 10 per dose for all the vaccines and vaccine combinations. The results are presented as cost-effectiveness on a 10-year period.

The cost-effectiveness ratios are to be interpreted as incremental cost-effectiveness ratios of implementing the interventions in the simulated scenarios relative to a do-nothing scenario, which corresponds to maintaining only the case management model described above.

Accounting for uncertainty

The cost-effectiveness results are based on an advanced modeling methodology aimed at representing reality as accurately as possible. The large number of scenarios simulated includes some sensitivity analyses of results for key variables, for instance for vaccine efficacy levels. The likely impact on results of other key features of potential malaria vaccines is explored in the companion article[14].

However, many sources of uncertainty cannot be captured by these sensitivity analyses. Probabilistic sensitivity analysis and expected value of information analysis could serve to further assess the impact of uncertainty on the simulation results [37–40], but there are technical problems in implementing and presenting such analyses for a large set of interventions and scenarios. It has been, therefore, planned to run an expected value of information analysis on a sub-set of simulations that will be reported in another article.

Pre-erythrocytic vaccines

At a reference transmission setting with annual entomological inoculation rate (EIR) of 21, the simulations predict that a PEV with 52% initial efficacy could be very cost-effective when delivered via EPI alone. At a vaccine price of US$2 per dose, the cost per uncomplicated malaria episode averted would be around US$ 5, the cost per severe malaria episode averted US$ 269, the cost per DALY averted around US$ 35 and the cost per death averted US$1057 (see table S1 and S2, Additional file 1). The cost-effectiveness ratios are lower for higher effectiveness levels (Figure 1). They increase almost proportionally with vaccine price reaching US$ 160 per DALY averted and US$ 4869 per death averted for a vaccine price of US$ 10 per dose (see table S3 and S4, Additional file 1).

The proportion of events averted by PEV delivered via EPI with booster doses is slightly higher, but the cost per uncomplicated episode averted is 20% higher (see table S1, Additional file 1), and cost per DALY and death averted is around 31% higher (see table S2, Additional file 1).

With EPI and mass vaccination the proportion of events averted is 5% higher for mass vaccination coverage of 50% and 8% higher for coverage of 70%[14], and the cost per uncomplicated episode averted is slightly lower. However, the costs per DALY and death averted are around 60%–66% higher (see table S1 and S2, Additional file 1). For higher efficacy levels the pattern is similar, showing that the incremental benefits of these deployment modalities, in this transmission setting, are modest (Figure 1).

In low transmission settings, while the cost per uncomplicated episode averted under EPI alone is similar to that in the reference transmission setting (see table S1 and S2, Additional file 1), the cost per DALY and death averted are lower at US$ 31 per DALY averted and US$ 925 per death averted at a vaccine price of US$ 2 per dose (see table S2 and S4, Additional file 1). Adding booster doses leads to higher cost-effectiveness ratios for efficacy levels up to around 60%, but at near 100% efficacy the cost-effectiveness ratios become similar (Figure 1). In contrast, when mass vaccination is added to EPI, the cost-effectiveness ratios decrease substantially, by around 70% for the cost per uncomplicated case averted (see table S1 and S3, Additional file 1), and by 24% to 28% for the cost per DALY and death averted (see table S2 and S4, Additional file 1).

In high transmission settings, the effectiveness of PEV is low[14] and the cost-effectiveness ratios are therefore higher than in the other transmission settings irrespective of delivery modality. For some outcomes, vaccination even leads to an increase in the number of clinical events[14], and, therefore, to negative cost-effectiveness ratios and negative case management cost savings (see table S5, Additional file 1).

Across all transmission settings, the incremental benefits of booster doses are small and the cost-effectiveness ratios are higher. Adding mass campaigns has little impact on overall effect when the primary efficacy is low. However, for high vaccine efficacy and high coverage, this strategy is predicted to lead to local elimination of the parasite in low transmission settings and substantially reduce transmission in medium transmission settings[14] at low additional costs. Under these conditions, because of the effects of the vaccine on transmission, delivery via mass campaigns plus EPI becomes a cost-effective alternative to EPI alone.

Blood-stage vaccines

At the reference transmission intensity, BSV of moderate efficacy with a price of US$ 2 per dose applied through EPI achieves a cost per uncomplicated episode averted of about US$ 9 (see table S1, Additional file 1), which is higher than for the corresponding PEV, but the costs per DALY averted (US$ 21) and per death averted (US$ 630) are lower than for PEV (see table S2, Additional file 1). At higher efficacy levels, the cost-effectiveness ratios decrease, following the same patterns as for PEV (Figure 2). Adding booster doses increases the cost-effectiveness ratios somewhat. Mass campaigns also increase the cost-effectiveness ratios except for uncomplicated episodes, where they decrease.

At low transmission intensity BSV averts a lower proportion of uncomplicated and severe cases and deaths than PEV[14] and the cost effectiveness ratios are higher for all outcomes. Adding booster doses leads to slightly higher costs per uncomplicated episode averted (see table S1 and S3, Additional file 1), and much higher costs per DALY and death averted (see table S2 and S3, Additional file 1, and Figure 2). Adding mass campaigns to EPI leads to a dramatic reduction in the cost per uncomplicated episode averted, but the costs per DALY and death averted are only slightly lower (see table S1, S2, S3, Additional file 1, and Figure 3, 4).

In high transmission settings BSV is more effective than PEV especially in averting severe and mortality events[14] and it is also more efficient. Under EPI alone the cost per uncomplicated episode averted, in the highest transmission setting, is US$ 3.8, the cost per DALY averted is US$13.5 and the cost per death averted is US$401, at vaccine price US$ 2 per dose (see table S1 and S2, Additional file 1, and Figure 3). Adding boosters or mass campaigns, leads to higher incremental costs than incremental benefits (see table S1, S2, S3, Additional file 1, and Figure 2).

Across all transmission settings, the incremental costs of adding booster doses to EPI are higher than the incremental benefits and this is particularly true for severe episodes, DALYs, and mortality (see table S1, S2, S3, Additional file 1, and Figure 2). In low transmission settings, campaigns improve cost-effectiveness for uncomplicated episodes averted, but do not change cost-effectiveness estimates for DALYs and deaths averted. However, in moderate to high transmission settings, the incremental costs of campaigns are higher than the incremental benefits (see table S1, S2, S3, Additional file 1, and Figure 2).

Combination vaccines and MSTBV

Combining BSV with PEV (with matched efficacies) in general, improves or matches the cases averted over PEV alone for all transmission settings and vaccine delivery modalities[14]. The cost-effectiveness ratios for this combination are lower than those of PEV in all transmission settings particularly for the cost per DALY and per death averted and in moderate to high transmission settings (see table S1, S2, S3 in Additional file 1, and Figure 3, 4). Compared to BSV alone, the cost-effectiveness ratios of combining BSV with PEV are lower, though the difference is smaller than for PEV and in this case it is higher in moderate to lower transmission settings than in high transmission settings. Adding booster doses to EPI leads to higher cost-effectiveness ratios across all transmission settings for this combination – the costs per uncomplicated episode averted increases by around 19%–23% while those per DALY and death averted show even larger increases (around 30%–40%).

Adding mass campaigns in low to moderate settings lead to incremental uncomplicated episodes averted that are higher than the incremental costs. However, in terms of DALYs and deaths averted the benefits exceed the costs only in the lowest transmission setting, while they are significantly lower in the reference and in high transmission settings. In high transmission settings even the additional uncomplicated episodes averted are lower than the additional costs.

Combinations of MS TBV with PEV or BSV and the triple combination do not improve the effectiveness of the vaccines alone when delivered via EPI or EPI with boosters[14]. However, adding mass campaigns leads to greater effectiveness in all transmission settings (Figure 4). The additional benefits of these combination vaccines are then much higher than the additional costs compared to delivering the vaccines under EPI alone and to all delivery modalities of PEV and BSV alone. In the reference transmission setting, for instance, the cost per uncomplicated episode averted of combining BSV with MSTBV, delivered via EPI and mass campaigns, is (at a vaccine price of US$2) US$1.8 and US$2.3 for 70% and 50% coverage (see table S1, Additional file 1), while the cost per DALY averted is US$20 and US$ 22 for 70% and 50% coverage (see table S2, Additional file 1). The costs per DALY averted vary between US$ 12 and US$40 across transmission settings with the lowest value in the lowest transmission setting where the greatest improvement to effectiveness is observed. The very favourable cost-effectiveness ratios in low transmission settings are related to the case-management cost savings, which may compensate up to more than 50% of the costs of the vaccine intervention (see table S4, Additional file 1).

Effect of delivery modalities

Adding boosters to EPI does not improve effectiveness or cases averted over EPI alone by very much even at the very high coverage level modeled, but it does incur additional costs. This delivery modality does therefore not represent a cost-effective alternative to EPI alone in any scenario (see table S1, S2, S3, Additional file 1).

Delivering all vaccines and combinations via population based campaigns improves the effectiveness at mass vaccination coverage of 50%, especially in low transmission settings[14]. Depending on the transmission setting and the vaccine type considered, the incremental costs of delivering vaccines via population based campaigns can be lower than the incremental benefits, leading to a significant reduction in the cost-effectiveness ratios (see table S1, S2, S3, Additional file 1, and Figure 4). Disseminating vaccines via population-based campaign in these cases is predicted to be a more cost-effective way of delivering malaria vaccines than EPI alone. Increasing the coverage of the mass vaccination campaigns increases the effectiveness and cases averted for all vaccine and vaccine combinations under most transmission settings[14]. However, the incremental benefits of increasing coverage are often lower than the incremental costs of achieving it (Figure 5). In some cases, the predictions suggest an optimal cost-effectiveness ratio at intermediate values for the campaign coverage. This is not a consequence of non-proportionality of vaccine delivery costs as a function of coverage (which could be realistic, but not modeled in this study), but of the indirect effects of the vaccines.

Effect of vaccine price

Although the simulations focus on comparative cost-effectiveness of different candidate malaria vaccines and delivery modalities, and not on the sensitivity of cost-effectiveness ratios to vaccine prices, which are hypothetical, it is evident that the cost-effectiveness results are almost directly proportional to the vaccine prices. In fact, at an assumed vaccine price of US$ 10 per dose, most cost-effectiveness ratios are between 4 and 7 times higher than those obtained at US$ 2 per dose (see table S1, S2, S3, Additional file 1). At a vaccine price of US$ 2 per dose, most vaccines and delivery modalities simulated present cost-effectiveness ratios comparable to those of other malaria interventions[9, 10, 41–43], while at a vaccine price of US$ 10 per dose in many of the simulated scenarios the cost-effectiveness ratios are higher.