Spread of anti-malarial drug resistance: mathematical model with implications for ACT drug policies

Abstract

Background: Most malaria-endemic countries are implementing a change in anti-malarial drug policy to artemisinin-based combination therapy (ACT). The impact of different drug choices and implementation strategies is uncertain. Data from many epidemiological studies in different levels of malaria endemicity and in areas with the highest prevalence of drug resistance like borders of Thailand are certainly valuable. Formulating an appropriate dynamic data-driven model is a powerful predictive tool for exploring the impact of these strategies quantitatively.

Methods: A comprehensive model was constructed incorporating important epidemiological and biological factors of human, mosquito, parasite and treatment. The iterative process of developing the model, identifying data needed, and parameterization has been taken to strongly link the model to the empirical evidence. The model provides quantitative measures of outcomes, such as malaria prevalence/incidence and treatment failure, and illustrates the spread of resistance in low and high transmission settings. The model was used to evaluate different anti-malarial policy options focusing on ACT deployment.

Results: The model predicts robustly that in low transmission settings drug resistance spreads faster than in high transmission settings, and treatment failure is the main force driving the spread of drug resistance. In low transmission settings, ACT slows the spread of drug resistance to a partner drug, especially at high coverage rates. This effect decreases exponentially with increasing delay in deploying the ACT and decreasing rates of coverage. In the high transmission settings, however, drug resistance is driven by the proportion of the human population with a residual drug level, which gives resistant parasites some survival advantage. The spread of drug resistance could be slowed down by controlling presumptive drug use and avoiding the use of combination therapies containing drugs with mismatched half-lives, together with reducing malaria transmission through vector control measures.

Conclusion: This paper has demonstrated the use of a comprehensive mathematical model to describe malaria transmission and the spread of drug resistance. The model is strongly linked to the empirical evidence obtained from extensive data available from various sources. This model can be a useful tool to inform the design of treatment policies, particularly at a time when ACT has been endorsed by WHO as first-line treatment for falciparum malaria worldwide.

Figure 1

Schematic diagram of the biological model through time. The model progresses from steady state through to the introduction of resistance and changes in drug policy.

Figure 2

The model-predicted spread of SP resistance. The predicted spread of resistance to mefloquine when combined with artesunate at different level of resistance. Figure 2A shows the result in low transmission setting and Figure 2B shows the result in high transmission setting, assuming 85% coverage of the ACT (SP and artesunate). The dotted line shows the 50% resistance. Each curve represents the mean of ten simulations.

Figure 3

Delay in the spread of resistance to SP. The delay in the development of resistance is measured as the proportional increase in time to 50% resistance compared to the continued use of SP as a monotherapy.

Figure 4

The spread of artemisinin resistance at different levels of ACT coverage. Figure 4A shows the spread of artemisinin resistance at varying levels of ACT coverage from 0% (i.e. use of artemisinin monotherapy) to 100% use of ACTs. Each line represents the mean of ten simulations. The dotted line shows the 50% resistance level. Figure 4B shows the delay in resistance measured as the increase in the time to 50% resistance (t₅₀) compared with the t₅₀ when using artemisinin monotherapy. The dotted line represents an extrapolation of the curve when the resistance does not reach 50% within the 12-year timeframe.

Figure 5

Change in malaria prevalence and incidence at different rates of coverage with ACT.

Figure 6

Schematic diagram of how immunity influences age-stratified likelihoods in the biological model.