Stochastic simulation models of "Plasmodium falciparum" malaria epidemiology and control

Every year malaria causes an estimated 1.3–3 million deaths and around half a billion clinical episodes. The majority of deaths occur in children under the age ofyears. Malaria today occurs mostly in tropical and subtropical countries, particularly in subSaharan Africa and Southeast Asia. In developing countries malaria may account for as much as 40% of public health expenditure, 30-50% of hospital admissions, and up to 50% of outpatient visits to health facilities. Malaria is a vector borne disease caused by the protozoan parasites of the genus Plasmodium. Plasmodium falciparum causes the most severe form of the disease, and is responsible for half of the clinical cases and 90% of the deaths from malaria. Malaria control interventions in countries where the disease is endemic currently include personal protection against mosquito bites, vector control, and prophylactic drugs. There is currently no registered malaria vaccine, but this is an active field of research. The vaccine that is furthest advanced in clinical development is called RTS,S/AS02A. This is a preerythrocytic vaccine, which aims to kill the parasites before they enter the red blood cells. Predictive models can provide a rational basis for decisions on how to allocate resources for malaria control. Mathematical modeling of malaria has a long history, starting with the first models of malaria transmission dynamics by Ross a century ago. At the Swiss Tropical Institute, a malaria modeling project has generated algorithms for rational planning of malaria control. This model is implemented as an individual-based discrete-time simulation model. The behaviors and state changes of simulated human individuals are governed by a minimal set of sub-models that are considered crucial for making quantitative predictions of the impact of malaria control interventions.
The integrated model includes components that capture relevant aspects of malaria transmission
and epidemiology in the absence of control: the relationship between the entomologic
inoculation rate and the force of infection; epidemiologic models for acute illness, severe
morbidity, and mortality; infectiousness of human population. Another central model
component, for natural immunity to asexual blood stages of P. falciparum, is described in
this thesis. The use of the model for making quantitative predictions requires reliable estimates
of the values of the parameters of the mathematical functions. The different model
components were therefore fitted to a number of datasets from studies in various ecological
settings and for various epidemiologic outcomes using a simulated annealing algorithm.
Comparison of the model predictions with field data show that the model appears to reproduce
reasonably well the parasitologic patterns seen in malariologic surveys in endemic
areas.
Epidemiologic patterns can be modified by control interventions. Because of the individualbased
approach chosen, a number of different simulated interventions can be introduced
by making assumptions on how they modify the processes described above. This thesis
describes a model for case management to predict the impact of improved case management
on incidence of clinical episodes and mortality while incorporating effects on persistence of
parasites and transmission. It allows the simulation of different rates of treatment coverage
and parasitologic cure rates, and makes it possible to look at how variations in transmission
intensity might affect the impact of changes in the health system. It also defines a baseline
environment that can be used the predict the impact of other control interventions.
The second part of the thesis focuses on the prediction of the impact of a pre-erythrocytic
stage vaccine. Different assumptions about how such a vaccine may lead to a measured
reduction in the incidence of new infections in vaccinated individuals are discussed. The
vaccine profile was chosen to match data from clinical trials of RTS,S/AS02A. The results
demonstrate that an adequate simulation of the first two RTS,S/AS02A trials published can
be achieved by assuming that vaccination completely blocks a certain fraction of infections
that would otherwise reach the erythrocytic stages.
The impact that such a vaccine would have on the epidemiology if introduced via the
Expanded Program on Immunization (EPI) is then predicted. This is the first major
attempt to combine dynamic modeling of malaria transmission and control with predictions
of parasitologic and clinical outcome. The results suggest a significant impact on morbidity
and mortality for a range of assumptions about the vaccine characteristics, but only small
effects on transmission intensities.
To make predictions of the cost-effectiveness of such a vaccination program, costing data
are incorporated into a model of a health system that is currently in place in a low-income
country context, based largely on data from Tanzania. Depending on the assumed vaccine
characteristics and cost, the predicted cost-effectiveness ratios would make vaccination
campaigns an attractive choice for health planners compared with other malaria control
interventions.
In addition to making quantitative predictions, the model points to data that may be
important to make accurate predictions. In order to make mid- to longterm predictions,
more data on the clinical epidemiology of malaria in adolescents and adults would be
desirable.
The work reported here creates a sound foundation for measuring the effects of introducing
new antimalarial interventions, or scaling-up those that are already known to be efficacious
and cost-effective. A challenge that remains is to make a comprehensive set of model
predictions available to a non-modeler audience so it can be valuable both for informing
malaria control strategies and research funding policy.