There is an interest in simplifying the standard artesunate dosing regimen (currently intravenous artesunate administered 5 times over 72 hours) for the treatment of severe malaria patients. Regimens requiring fewer doses would be preferable in resource-poor settings where ensuring intravenous access for patients in need can be very difficult. Within-host pharmacokinetic-pharmacodynamic (PK-PD) models characterise the relationship between the antimalarial drug concentrations in the blood and the change in the number of parasites (pharmacodynamic (PD) response) within an individual host over time. PK-PD models can be used to investigate the treatment of sensitive and resistant severe malaria infections with simplified intravenous artesunate dosing regimens.
Data were pooled from six studies conducted in Africa and Asia, on 265 severe malaria patients (70 adults and 195 children) treated with intravenous artesunate. A mean of 6 (range 1 to 14) parasite count measurements were available per patient. Bayesian hierarchical modelling was used to fit a within-host PK-PD model of the blood stage infection in the presence of treatment to the pooled parasite measurements. Clinical outcomes for hypothetical patients with either sensitive or resistant severe malaria infections were simulated from the posterior predictive distribution and used to explore whether the efficacy of two simplified regimens was comparable to the standard dosing regimen.
The proportion of hypothetical patients with 99% parasite clearance at 24 hours and the distribution of the simulated parasitaemia at day 3 demonstrated that the standard dosing regimen was optimal for sensitive infections, but for resistant infections neither the standard or simplified regimens had a substantial impact on parasite clearance.
Our within-host PK-PD modelling approach indicates that a change to simplified regimens for the treatment of sensitive severe malaria infections could compromise patient health, and that for severe resistant infections the standard 5 dose regimen may be sub-optimal.