Conceived and designed the experiments: MAP NM TS. Analyzed the data: MAP NM. Wrote the paper: MAP NM TS. Interpretation of the results: MP NM TS. Editing of the manuscript: ASchapira. Programming the simulations: AStuder NM. Running the simulations: AStuder NM. Programming: MP.
The authors have declared that no competing interests exist.
A number of different malaria vaccine candidates are currently in preclinical or clinical development. Even though they vary greatly in their characteristics, it is unlikely that any of them will provide longlasting sterilizing immunity against the malaria parasite. There is great uncertainty about what the minimal vaccine profile should be before registration is worthwhile; how to allocate resources between different candidates with different profiles; which candidates to consider combining; and what deployment strategies to consider.
We use previously published stochastic simulation models, calibrated against extensive epidemiological data, to make quantitative predictions of the population effects of malaria vaccines on malaria transmission, morbidity and mortality. The models are fitted and simulations obtained via volunteer computing. We consider a range of endemic malaria settings with deployment of vaccines via the Expanded program on immunization (EPI), with and without additional booster doses, and also via 5yearly mass campaigns for a range of coverages. The simulation scenarios account for the dynamic effects of natural and vaccine induced immunity, for treatment of clinical episodes, and for births, ageing and deaths in the cohort. Simulated preerythrocytic vaccines have greatest benefits in low endemic settings (<EIR of 10.5) where between 12% and 14% of all deaths are averted when initial efficacy is 50%. In some high transmission scenarios (>EIR of 84) PEV may lead to increased incidence of severe disease in the long term, if efficacy is moderate to low (<70%). Blood stage vaccines (BSV) are most useful in high transmission settings, and are comparable to PEV for low transmission settings. Combinations of PEV and BSV generally perform little better than the best of the contributing components. A minimum halflife of protection of 2–3 years appears to be a precondition for substantial epidemiological effects. Herd immunity effects can be achieved with even moderately effective (>20%) malaria vaccines (either PEV or BSV) when deployed through mass campaigns targeting all agegroups as well as EPI, and especially if combined with highly efficacious transmissionblocking components.
We present for the first time a stochastic simulation approach to compare likely effects on morbidity, mortality and transmission of a range of malaria vaccines and vaccine combinations in realistic epidemiological and health systems settings. The results raise several issues for vaccine clinical development, in particular appropriateness of vaccine types for different transmission settings; the need to assess transmission to the vector and duration of protection; and the importance of deployment additional to the EPI, which again may make the issue of number of doses required more critical. To test the validity and robustness of our conclusions there is a need for further modeling (and, of course, field research) using alternative formulations for both natural and vaccine induced immunity. Evaluation of alternative deployment strategies outside EPI needs to consider the operational implications of different approaches to mass vaccination.
The demand for an effective vaccine against
The impact of a vaccine will depend not only on average efficacy, but also on the extent of heterogeneity of the host response including its duration. Other determinants include the natural force of infection and its seasonal variation, the vaccination coverage which could be achieved, especially in the most exposed and the most vulnerable groups and the efficacy and coverage of other malaria control interventions, preventive or curative. As for all public health interventions, safety, cost, operational feasibility and acceptability also need to be considered when deciding which candidates to prioritize, which ones to consider for combination, and which ones to develop for specific target groups or deployment strategies.
Field trials of malaria vaccines are generally designed to evaluate the effect on morbidity or on infection rates in the vaccinated population
In line with our previous simulations, we find that moderately efficacious preerythrocytic vaccines applied via EPI do not have any substantial effect on malaria transmission, (results not shown), because only a small proportion of the population is protected. If the initial efficacy of PEV is high then effects on transmission are observed for EPI and EPI with boosters (
Results obtained assuming a vaccine efficacy of 80%, halflife of 10 years and homogeneity value of 10. Note that the blue and black lines almost overlap.
We observe elimination with PEV alone, at very high efficacy and mass vaccination coverage and at the lowest transmission levels (
For highly efficacious BSV, we observe effects on transmission, particularly at high transmission settings (
As expected, for vaccine combinations with MSTBV we observe greater reductions in transmission over PEV or BSV alone (
Results obtained assuming a vaccine efficacy of 52%, halflife of 10 years and homogeneity value of 10.
Elimination is generally simulated at the lower initial transmission intensities with vaccine combinations containing MSTBV and/or for highly efficacious vaccines delivered via EPI with mass vaccination. In the simulations that we examined in detail (results not shown) elimination is more likely when the homogeneity parameter is very low or if the vaccine halflife is very large.
The time to elimination, dependent on initial vaccine efficacy, is considered in
All results are for vaccines delivered via EPI with mass vaccination, no elimination is achieved under these conditions for vaccines delivered via EPI or EPI with boosters. Results obtained assuming vaccine halflife of 10 years and homogeneity value of 10.
Results obtained assuming vaccine efficacy of 52%, a vaccine halflife of 10 years and homogeneity value of 10, unless the values are varied along the xaxis. Vaccines are distributed via EPI (circles), EPI with boosters (*) and EPI with 70% mass vaccination (squares).
Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
As reported previously, the effectiveness of PEV depends strongly on the duration of protection for vaccines with halflife less than 2–3 years
Results obtained assuming a an initial vaccine efficacy of 52% and homogeneity value of 10.
Previous simulations showed that a PEV averts a higher proportion of clinical episodes and deaths if there is heterogeneity in the response to vaccination among individuals, namely if PEV concentrates its effects in some individuals (low value for the homogeneity parameter), who thus never become infected, than one that spreads protection more evenly across the population. This result is confirmed here (
Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
The curves relating the events averted (not shown) and effectiveness of BSVs to the primary efficacy (
As with PEV, the effectiveness of BSV strongly depends on the duration of protection for vaccines with halflife less than about 2–5 years (
As with PEV, the results predict a very small improvement if the vaccine effect is concentrated in some individuals. (
The biggest improvement to effectiveness by adding MSTBV to BSV, PEV or BSV with PEV, is observed at very low transmission settings, where almost all deaths, severe and uncomplicated events tend to a situation where they may all be averted over the 10 years for very high efficacy and mass vaccination coverages, suggesting local elimination could be achieved under such “ideal” conditions. (
The results indicating significant improvement to the number of cases averted for vaccine combinations with MSTBV is achieved only when delivered via EPI with mass vaccination (
With mass vaccination, the effectiveness of combination vaccines continues to increase with duration of protection even for halflives beyond 5 years, in contrast to other delivery modalities (compare
In most contexts, EPI with booster does not significantly improve cumulative effectiveness or cases averted over EPI alone. In contrast, mass campaigns increase effectiveness even at relatively low coverage, especially in low transmission settings. However, in high transmission settings the increase in effectiveness of delivery via EPI with mass vaccination is not as large. For very high transmission, a PEV with low efficacy is less effective under EPI with mass vaccination compared to EPI alone (compare
Under EPI with mass vaccination, increasing coverage increases effectiveness and cases averted in most transmission scenarios. The exceptions are in higher transmission settings for severe episodes averted by BSV with MSTBV, PEV alone and PEV with MSTBV, where very high coverage levels are associated with small reductions of effectiveness (
Results obtained assuming an initial vaccine efficacy of 52%, a vaccine halflife of 10 years and homogeneity value of 10.
We have presented the results of stochastic simulations of the likely population effects of different malaria vaccine profiles, distributed via different distribution modalities with the aim to assess their usefulness in multiple transmission settings. We examine the effects in controlling morbidity and mortality as well as impact on malaria transmission, thus bridging a gap between field trials of efficacy and malaria transmission models. The simulation results have implication for vaccine developers concerning which vaccines to develop and with what minimal profile, and in which transmission settings these vaccines are likely to provide most benefit. The results highlight the benefit of considering alternative methods of deployment outside of EPI. All these considerations need to be considered by vaccine developers along with issues of safety, immunogenicity, feasibility, and costs that fall outside the scope of this analysis.
Our previous simulations of PEV, aligned with the results of field trials of the RTS,S/AS02A vaccine
The simulations suggest that PEV in general will be much more effective in low transmission settings than in high transmission settings. If delivered via EPI with mass vaccination in high transmission settings and high coverage our models predict that low efficacy PEV may even lead to increases in severe morbidity over a 10 year period (or longer) by shifting the morbidity patterns to those observed in lower exposure settings, i.e. higher risk in higher agegroups. This predicted negative benefit of low efficacious PEV highlights a possible risk involved with introducing PEV in high transmission settings.
The model for the action of asexual BSV represents one of the more tentative components of our integrated model. Although one candidate BSV has shown a substantial effect on parasite densities
More generally, there is a need for further alternative simulations of both natural and vaccine induced immunity, including models of natural boosting. When field data become available they may require us to consider exposuredependence of vaccine efficacy or intrinsic age dependence of the primary effect of the vaccine.
In comparing different vaccine types it is important to remember that the measure of efficacy used to define the simulated vaccines differs for the different vaccine types; a 50% efficacy BSV is equivalent to a 50% efficacy PEV only in the sense that both represent imperfect vaccines. The present simulations of combination vaccines, with matched values of the efficacy parameters for the PEV and BSV components represent only one of an infinite number of possible combinations, and it will be most useful to simulate actual candidates, once their likely profiles become available. It seems unlikely that there would be much interest in combining very efficacious PEV or BSV with a rather poor efficacy partner. Combination vaccines with PEV and BSV components seem to have some potential in mass vaccination scenarios, even when efficacy is modest, but do not look much more promising for use in the context of EPI or EPI with boosters than the best of the individual components. In general, the effectiveness of such combinations seems similar to or lower than that of BSV in high transmission settings. This could be attributed to the PEV component lowering the exposure of vaccinated individuals so that the combination vaccine effectiveness is similar to that at a slightly lower transmission level for BSV alone (see
The most realistic scenarios for MSTBV however are clearly situations where high efficacy MSTBV might be deployed in mass vaccination to supplement the effects of moderate efficacy PEV or BSV. This approach is supported by the simulation results. In such situations a rather poor effectiveness of the PEV or BSV component on its own may become very substantial depending on the transmission intensity. So far we simulated combinations with matched durations of efficacy, however it may well be the case that MSTBVs have only very shortterm effects because of an absence of natural boosting
A number of the mass vaccination scenarios predict local elimination of the parasite. However, malaria is much easier to eliminate in computer simulations than in reality. This is particularly the case because transmission in nature is highly heterogeneous and this would be particularly important in low transmission settings
Interestingly, we found that PEV vaccines in which the effect is concentrated in some individuals are more likely to achieve elimination. This arises because of the convex shape of the effectiveness vs initial efficacy curve (e.g.
The simulations we present here should have implications for vaccine developers concerning the definition of minimal requirements for malaria vaccines to be used in public health. A PEV or BSV with a halflife of efficacy of less than 2–3 years will be of limited value and assessment of duration of protection is of great importance. Unfortunately Phase II trials are generally not designed to estimate duration. Since incidence declines steeply with age in young children, claims that efficacy is sustained in extended followups
In addition to possible effects on transmission, vaccine developers would also be interested in determining whether substantial herd immunity effects are likely, and thus clinical development plans need to evaluate effects on transmission to the vector. We have found that, as could be expected, substantial transmission impact is generally achieved only if EPI delivery is supplemented by mass campaigns. Developers thus need to consider how vaccines are to be deployed. Probably, an important criterion for whether vaccines can be relatively easily deployed widely outside EPI is whether protection can be achieved with only 1 or 2 doses. In this respect, further analysis needs to focus on scenarios that are aligned with realistic distribution systems, using field data to identify realistic correlations between vaccination at successive rounds, rather than assuming these to be independent, and to assess what are feasible intervals between rounds.
Ultimately, malaria vaccines will be deployed as part of integrated control strategies. We thus plan further analyses to explore the interactions of vaccination with other malaria control interventions and the implications for resource allocation and management within the health system.
We base our simulations of vaccines on our previously described model for the natural history and epidemiology of
For the present simulations we have recalibrated the model, using a genetic algorithm to parameterise it to 61 field scenarios from subSaharan Africa, comprising data on seasonality, agepatterns of infection, parasite density, clinical episodes, severe malaria and mortality
The simulations of the effects of vaccine interventions use a case management model, including both formal and informal treatment, based on that of a previous study to simulate existing case management in Tanzania
Each simulated vaccine is characterised by an average initial efficacy, which is reached after completion of a vaccination schedule of 3 doses and thereafter decays exponentially. For the reference vaccine initial efficacy 52% after the third dose we assume efficacies for dose 1 and 2 used previously, namely 40% and 46% respectively
Vaccine combinations:  PEV (Preerythrocytic vaccine) 
BSV (Bloodstage vaccine)  
PEV+BSV  
PEV+TBV (Mosquitostage transmissionblocking vaccine)  
BSV+TBV  
PEV+BSV+TBV  
Delivery modality  EPI (1,2,3 Months) 
EPI with booster 1,2,3,4 years after the last EPI dose,  
EPI+Campaign: Mass vaccination 3 doses at start of intervention period, then 1 dose at 5,10,15 years.  
Vaccine coverage  EPI: 89% 3rd dose; 95% 1st dose 
EPI with booster: 99% of previously vaccinated  
EPI+Campaigns: varying levels of coverage from 0% to 95% are considered.  
Initial protected efficacies after dose 3  0.1 to 1 (stepsize 0.1) (reference 
Half live of protective efficacy  0.5, 1, 1.5, 2, 2.5, 5, 
Between host variation in initial protective efficacy 

Transmission intensity  EIR (infectious bites per annum) = 5.25, 10.5, 
parameter
Figures in bold represent the value used for the reference scenario
We assume preerythrocytic vaccines lead to a reduction in the proportion of sporozoite inocula that lead to blood stage infection, where the efficacy is equal to the proportional reduction in incidence of infection. This model is justified by analysis of the effects recorded in trials of the RTS,S vaccine
The immediate effect of a blood stage vaccine is assumed to be reduction in parasite density levels at each time step, where efficacy is equal to the proportional reduction.
We model mosquitostage transmission blocking vaccines by defining the efficacy to be the proportion by which the probability that a mosquito becomes infected from any one feed is reduced. We assume the efficacy of MSTBV to be proportional to the number of doses. This if the initial efficacy after the third dose is 95%, for first and second doses we assume initial efficacies of 32% and 63%, respectively.
We consider combination vaccines of PEV with MSTBV, BSV with MSTBV, BSV with PEV and also a combination of all three. In each case we assume PEV and BSV to be matched in the initial efficacy and in rate of decay. Since it is unlikely that MSTBV with only moderate efficacy will be developed, we consider combinations of PEV, BSV and of PEVBSV with high efficacy MSTBV, and thus assume an MSTBV initial efficacy after the third dose of 95%. The rate of decay of MSTBV is matched to that of the other vaccine components.
We model three delivery modalities:
The first is the delivery through the EPI according to the usual DTP3 schedule (children 1, 2, 3 months of age).
In this delivery modality, booster doses are added to the normal EPI schedule with booster doses at 1,2,3,4 years after the last EPI schedule. We assume that the effect of a booster dose of vaccine is to restore the protective efficacy to that achieved after the 3^{rd} dose in the same individual.
The third delivery modality combines the delivery of the vaccine to infants through EPI and a populationwide mass vaccination campaign with three doses at the beginning of the intervention period followed by additional mass vaccinations with a single dose after 5, 10, and 15 years. The protective efficacy of the vaccine is assumed to increase linearly up to dose 3. Additional doses restore the efficacy to that achieved at dose 3.
Under delivery modality (i), the vaccination coverage is as detailed in
The simulated scenarios cover all three vaccine types, and the three combinations, delivered through the three modalities at a range of coverage levels (
The main analyses consider the aggregated effects over the first 10 years of the vaccination program. We consider the effect of each vaccine on simulated values of a standard set of epidemiological outcomes in the whole population (not just those vaccinated): the number of uncomplicated malaria episodes, the number of severe malaria episodes and the number of deaths caused by malaria. For each of these outcomes we compute the number of events averted per 1000 personyears by comparing the vaccine simulation with the corresponding control simulation. We define the effectiveness as the proportion of events of each type that are averted. In addition, at each time point of the 20 years of the simulation we consider the proportion of mosquitoes that become infected at each feed as a measure of the level of transmission. We present predictions via plots of outcomes or estimates of average effectiveness for particular scenarios (
Effect of initial efficacy (a–c), vaccine halflife (d–f) and degree of heterogeneity (g–i) on the effectiveness of PEV for the reference transmission setting of EIR 21. Results obtained assuming vaccine efficacy of 52%, a vaccine halflife of 10 years and homogeneity value of 10, unless the values are varied along the xaxis. Vaccines are distributed via EPI (circles), EPI with boosters (*) and EPI with 70% mass vaccination (squares).
(1.04 MB TIF)
Effect of initial efficacy on effectiveness of PEV for different transmission settings delivered via EPI with mass vaccination for 0% (a–c), 10% (d–f), 30% (g–i), 50% (j–l),7 0% (m–o) and 90% (p–r) coverage. Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
(1.46 MB TIF)
Effect of initial efficacy on effectiveness of BSV for different transmission settings delivered via EPI (a–c), EPI with boosters (d–f) and EPI with 70% mass vaccination (g–i). Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
(1.16 MB TIF)
Effect of initial efficacy on effectiveness of all vaccines for different transmission settings delivered via EPI and boosters (BSV (a–c), BSV/TBV (d–f), PEV (g–i), PEV/TBV (j–l), BSV/PEV (m–o) and BSV/TBV (p–r)). Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
(1.34 MB TIF)
Effect of vaccine halflife on effectiveness of all vaccines for different transmission settings delivered via EPI (BSV (a–c), BSV/TBV (d–f), PEV (g–i), PEV/TBV (j–l), BSV/PEV (m–o) and BSV/TBV (p–r)). Results obtained assuming an initial vaccine efficacy of 52% and homogeneity value of 10.
(1.14 MB TIF)
Effect of vaccine halflife on effectiveness of all vaccines for different transmission settings delivered via EPI with 70% mass vaccination (BSV (a–c), BSV/TBV (d–f), PEV (g–i), PEV/TBV (j–l), BSV/PEV (m–o) and BSV/TBV (p–r)). Results obtained assuming an initial vaccine efficacy of 52% and homogeneity value of 10.
(1.36 MB TIF)
Effect of the degree of heterogeneity on effectiveness of all vaccines for different transmission settings delivered via EPI (BSV (a–c), BSV/TBV (d–f), PEV (g–i), PEV/TBV (j–l), BSV/PEV (m–o) and BSV/TBV (p–r)). Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
(1.14 MB TIF)
Effect of initial efficacy on effectiveness of vaccine combinations with MSTBV for different transmission settings delivered via EPI with mass vaccination for 0% (a–c), 10% (d–f), 30% (g–i), 50% (j–l),7 0% (m–o) and 90% (p–r) coverage. Results obtained assuming a vaccine halflife of 10 years and homogeneity value of 10.
(1.62 MB TIF)
Effectiveness (%) of each vaccine or combination over 10 years
(0.13 MB DOC)
This work depended on the goodwill of the many thousands of volunteers who make their computers available to malariacontrol.net, and input to software development from the Africa@home team. We would also like to acknowledge the organizational support of Marcel Tanner, and helpful scientific discussions with Allan Saul and the Technical Advisory Group of the project.