Stochastic models for the consequences of "Plasmodium falciparum" infections in the human host : malaria morbidity, mortality and infectivity

The consequences of Plasmodium falciparum infections for humans range from selflimiting asymptomatic parasitaemia to rapid death. Annually, P. falciparum malaria is estimated to cause 0.5 billion acute febrile episodes, 2-3 million severe episodes warranting hospital admission and one million deaths. This enormous burden demands effective control strategies. A number of different interventions are available, but policy-makers need a rational basis for discriminating between them. The likely consequences of each intervention, or combination of interventions, must be considered. Trials can provide estimates of the impact of interventions on acute episodes over a short timespan. Predictions are required where field data are not available: over longer time periods, for severe outcomes, for many combinations of interventions, or for interventions which do not yet exist. A model intended for making quantitative predictions of the impact of interventions must relate transmission and infection to the key outcomes used by health-planners such as morbidity, mortality and cost-effectiveness. It must also allow for dynamic effects on transmission and acquired immunity, and incorporate the effect of the health system. Until recently, there was no such model. In the past, emphasis had lain with the transmission cycle. In addition, the most practical models would be individual-based with high computational demands. In response to this need, a stochastic individual-based integrated model has been developed at the Swiss Tropical Institute. The core of the integrated model is a description of asexual parasite densities, providing a basis for the effects of acquired immunity on reducing densities and for the density-based consequences of infection. This thesis contributes those elements that consider the immediate consequences of human infection: morbidity, mortality and transmission to the vector. The integrated model is then applied to questions concerning a new intervention, intermittent preventive treatment in infants (IPTi). A framework for morbidity and mortality is proposed. The probability of an acute febrile malaria episode is related to parasite densities via individual- and time-specific pyrogenic thresholds that respond dynamically to recent parasite load. Severe episodes result either from overwhelming parasitaemia, or from acute episodes in conjunction with a co-morbidity which acts to weaken the host. Both direct and indirect mortality were considered. Age-dependent case fatality rates estimated from field data were used to quantify the probability of direct mortality. Indirect deaths occur following an acute episode with subsequent co-morbidity after the parasites have cleared, or within the neonatal period as a consequence of maternal infection. Co-morbidity is assumed to be age-dependent. The model components are fitted to field data or to summaries of field data, and can simultaneously account for the observed age- and exposure-specific patterns of paediatric malaria and malaria-associated mortality. The model component for infectivity relates asexual parasite densities to the probability of infecting a feeding mosquito, taking into account the delay resulting from the timecourse of gametocytaemia and the need for both male and female gametocytes in
the blood meal. This component is fitted to data from malariatherapy patients and can
account for observed patterns of human infectivity. The integrated model is validated
against published estimates of the contribution of different age groups to the infectious
reservoir.
The integrated model, in conjunction with an added component for the action of
sulphadoxine-pyrimethamine (SP) and site-specific inputs, reproduced the pattern of
results of the IPTi trials reasonably well. The model was modified to represent different
hypotheses for the mechanism of IPTi. These hypotheses concerned the duration of
action of SP, the empirical timing of episodes caused by individual infections, potential
benefits of avoiding episodes on immunity and the effect of sub-therapeutic levels of SP
on parasite dynamics. None of the modified versions improved the fit between the
model predictions and observed data, suggesting that known features of malaria
epidemiology together with site-specific inputs can account for the pattern of trial
results. Predictions using the integrated model suggest that IPTi using SP is effective
over a wide range of transmission intensities at reducing morbidity and mortality in
infants. The predicted cumulative benefits were proportionately greater for mortality
and severe episodes than for acute episodes, due to the age-dependent co-morbidity
functions in the model. IPTi was predicted to avert a greater number of episodes where
IPTi coverage was higher, the health system treatment coverage lower, and for drugs
which were more efficacious and had longer prophylactic periods. Additionally, IPTi
was predicted to have little impact on transmisison intensity.
This is the first major attempt to model the dynamic effects of malaria transmsission,
parasitological status, morbidity, mortality and cost-effectiveness using model
components which were fitted to field data. The model can be extended to predict the
dynamic effects of different interventions, and combinations of interventions. The
ability to compare the likely impact of different interventions on the same platform will
be a valuable resource for rational decisions about strategies to control the intolerable
burden of malaria.