Continuing availability of artesunate monotherapy
The model predicts that if there is no intervention, and use of artemisinin monotherapies continues, there will be an exponential rise in the proportion of resistant infections and a slowly increasing prevalence of infection. By 2030, the model predicts that the prevalence of malaria will have doubled compared to 2008 and resistance to the artemisinins will be approaching 100% (Figure 2).
Eliminating artemisinin resistance
The model predicts that it is possible to achieve elimination of artemisinin resistance using medications alone, but in all scenarios elimination of malaria is required in order to eradicate artemisinin resistance.
The most effective single intervention to achieve elimination of artesunate resistance is the elimination of inadequate courses of artesunate monotherapy and replacement with ACT with high coverage (a). This would be achieved if all monotherapies were actively replaced by ACT in the private sector and there was adequate continued supply of good quality, affordable ACT. Using this strategy and with an introduction of the ACT over three months it is possible in the model to achieve elimination of artemisinin-resistant malaria within four years in 70% of cases (mean time to elimination 3.42 years (95% CI 3.32–3.60 years), (Figure 3). The stochastic model was run 200 times and elimination was achieved with this strategy in 100% of cases, suggesting this strategy is highly likely to be successful. The downside is that, because a lot of artemisinin is being used, the prevalence of artemisinin resistance just before elimination drastically increases. It reaches 82% in the deterministic model and 100% in 58% of cases in the stochastic model (Figure 4 illustrates one of these cases).
The proportion of infections resistant to artemisinins then remains at this high level provided that there is no importation of new infections from elsewhere. This suggests that if an intervention were to fail, or was discontinued prematurely, the prevalence of resistance would be much higher subsequently than if no intervention had taken place.
If artemisinin monotherapies remain available, one three-month pulse of MSAT or MDA with either ACT or AP (b or c) reaching 80% of the infected population would have a significant short-term impact, but a negligible long-term impact on the numbers of resistant infections. This is because although such interventions reduce infections they are not sufficient to completely interrupt transmission at the population level. For example, AP MSAT is the most effective of these strategies but one pulse of this only produces a 35% reduction in total malaria infections and 31% reduction in the number of artemisinin-resistant infections. If a three-month pulse of MSAT or MDA is used in addition to a switch to ACT for treatment then there is negligible additional effect. Repeated annual three-month pulses of MSAT or MDA are also insufficient to achieve elimination when used alone, regardless of the drugs used. This is because the number of infections increases again in the other nine months of the year with an increasing proportion of artemisinin-resistant infections each time due to ongoing artemisinin use. When used in this way, MSAT is more effective than MDA and AP has greater impact than ACT. For example, the maximum effect of long-term annual MSAT with AP on the numbers of artesunate resistant infections is at two years. At this time, the decrease in malaria infections is 55% and artemisinin-resistant infections have fallen to 39% compared to 2009. Following this trough, the number of artemisinin-resistant infections rises again although the total number of malaria infections does not reach a minimum until after 7 years (80% below that for 2009). For comparison, if ACT MSAT is used, the lowest number of artesunate resistant infections is in the first year with a 32% decrease in malaria prevalence and 31% fall in artesunate resistant infections compared to 2009. The number of artesunate resistant infections then increases again but the total number of malaria infections does not reach a minimum until after five years (61% lower than in 2009).
If the MDA or MSAT is carried out in the peak transmission season for malaria, the maximum decrease in the number of artemisinin-resistant infections is half that which can be achieved in the low transmission season. The frequency of MSAT and MDA was varied and it was found that if carried out twice a year then elimination became possible. If MSAT with AP was undertaken at a maximum of four times a year (it takes three months to complete one round) then elimination can be achieved in eight years. All this is predicated upon resistance not emerging to AP.
The addition of primaquine to annual MDA or MSAT (d) reduces the trough in the number of artemisinin-resistant infections by 20% and the total number of malaria infections by 40%. This extra effect is still insufficient to achieve elimination.
Assuming that malaria vectors bite after people are in or near their beds, then ITNs (e) have a relatively large effect and accelerate the eradication of resistance. The longer that nets are used for, the larger the effect; for example, if the effect of bed nets lasts around four years (the equivalent of long-lasting treated nets) and they reduce transmission by 30%, then time to eradication is reduced by about 50%. Thus a modest, but sustained, protective effect from bed nets or other transmission blocking methods can have a significant impact on the time to elimination.
When a fitness cost of artemisinin resistance is included in the model, the rate of increase of resistant infections is slower (around 33% slower for fitness cost of 5%) and the rate of elimination by an intervention is marginally faster (19% for a fitness cost of 5%). The fitness cost must be less than 7.5% for the number of artemisinin-resistant infections to increase over time.
Sensitivity analysis
For all interventions, the following were varied in turn: the coverage with ACT from 0 to 100%, the effectiveness of dihydroartemisinin and piperaquine against resistant infections and the cost of piperaquine resistance between 0 and 100%, the cost of artemisinin resistance in terms of fitness between 0 and 5%, the average time to receive anti-malarials during an intervention from 7 to 90 days, and time to natural recovery from infection (i.e. without treatment) from 60 to 200 days. Duration of drug effect was also varied for atovaquone from 10–15 days and for piperaquine from 20–30 days.
The model was most sensitive to ACT coverage, effectiveness of artemisinins against resistant infections, fitness cost of ACT resistance, and time to receive anti-malarials. There was a threshold coverage with ACT of 47% below which the time to eradication was more than a decade. With ACT coverage of <28%, elimination was impossible. The effectiveness of artemisinins against resistant infections was determined from field data to reflect current phenotypes [7] but it is likely that this will decrease over time with continued use of artemisinins. The model predicts that if this effectiveness halves compared to 2009, time to eradication will take 50% longer. If time to receive anti-malarials (ACT or AP) doubles (from 16 to 32 days after developing blood stage infection) time to elimination increases more than threefold, whereas if people receive anti-malarials in half the time (eight days after blood stage infection), time to elimination is three times less. The time to eradication was unaffected by natural recovery rate from infection, and the effectiveness of and fitness cost of resistance to piperaquine. Changing the duration of atovaquone and piperaquine effect had minimal impact on the results. To explore the effect of synergy between the components of ACT, the rate of clearance of infection was varied from 1/3 days-1 to 1/7 days-1. This had negligible effect on the time to elimination but decreasing the clearance rate of gametocytes by ACT to 1/7 days-1 decreased the percentage of artemisinin-resistant infections at the time of elimination from 82% to 54%.