Under-dosing of ACTs in children

Since pharmacokinetics are affected by body weight, we expect maximum plasma drug concentrations to vary according to both patient age and weight. Antimalarial drugs are administered to children under dosing regimens determined by age (Table 2). Age-dependence of drug concentrations will vary between drugs according to the nature of each drug’s pharmacokinetic dependence on body weight.

Using age-weight charts from the CDC and WHO, we show in Figure 1C that age-based dosing achieves approximately uniform maximum plasma drug concentrations for AM, DHA, and PQ. While lower maximum concentrations of LF are observed in younger children, patients of all ages are able to achieve maximum LF concentrations above LF’s estimated C50 of 280 μg/L. In contrast, under current dosing recommendations children below the age of 2 have maximum PPQ concentration much closer PPQ’s estimated C50 of 5 μg/L; prophylactic effects of PPQ will therefore have much shorter duration. The under-dosing of children with piperaquine under current dosing recommendations has been well-documented in literature [65], but as amended guidelines have yet to be released, we have chosen to use the under-dosed treatment recommendations in our simulations.

Pharmacodynamics of the combination therapies AL and DP as well as gametocyte killing of PQ were tuned to match parasite clearance times, recrudescence rates, and reinfection rates to clinical data (Figure 2). As predicted by the age- and weight-based models of maximum drug concentrations, children treated with DP had higher recrudescence and reinfection rates than adults did due to PPQ levels insufficient for cure and prophylaxis (Figure 2C). Young children were especially vulnerable to recrudescence, while all infections were cured in adults, indicating that piperaquine was failing to kill the remaining parasites untouched by DHA. Likelihood of reinfection decreased with age among individuals treated with DP. Field studies have also shown that DP is likely under-dosed in children, and a higher dosage has been suggested [65-67]. In contrast, young children receiving AL were not more likely to be reinfected than adults [68].

Interpretation of PQ pharmacodynamics

Experimental work suggests that PQ may render mature gametocytes incapable of infecting mosquitoes even while the gametocytes remain in the host bloodstream [13]. In our model, we are most interested in PQ’s transmission-blocking activity and thus do not explicitly model inactivated gametocytes, instead considering them to be killed and cleared. We find that to replicate the observation that even a low PQ dose of 0.065 mg/kg PQ results in very few infected mosquitoes 2 days post treatment when given with a full course of AL (Figure 2E) [44], the maximum kill rate of PQ on mature gametocytes must be very high in order to compensate for PQ’s short half-life of 8 hours. As currently parameterized, PQ completely clears gametocytes (Additional file 2: Figure S1D), making our predictions of its impact on prevalence an upper bound on its actual effect. However, incomplete knowledge of PQ’s active metabolites and their pharmacokinetic profiles means that our model of PQ’s killing action is only a gross approximation, and further refinement will be necessary as we learn more about PQ.

ACTs in mass drug campaigns: comparison of AL and DP, 1 month post campaign

We test the effects of distributing AL and DP in a mass administration context. Case management and vector control were excluded in order to focus explicitly on the effects of the drug campaigns. Using an isolated population experiencing seasonal transmission, a multi-round drug administration campaign was simulated during the dry season, when prevalence drops to 30%. Both AL and DP reduce prevalence while the campaign rounds are ongoing and continue suppressing prevalence for a few months into the high transmission season (Figure 3A). Due to PPQ’s long prophylactic tail, campaigns with DP result in lower prevalence that extends longer into the high transmission season than campaigns with AL as has been previously observed with generic short- and long-acting ACTs [25]. At this level of EIR, prevalence returns to baseline levels within 18 months of the MDA campaign even when DP is used. Using antimalarials to reduce prevalence in the long term requires repeated mass drug campaigns, permanent scale-up of case management with effective drugs, or simultaneous deployment of vector control with high coverage [12,23-25]. Here we focus on one aspect of a complex malaria control strategy: how to optimize elements of a single drug campaign to reduce parasite prevalence up to 4 months after the end of the campaign.

ACTs in mass drug campaigns: comparison of 3-round and 2-round campaigns, 1 month post campaign

When malaria prevalence is seasonal, mass administration campaigns are conducted during the low transmission season in order to most effectively deplete the infectious reservoir and possibly interrupt transmission [23,69]. We compared the efficacy of three-round and two-round mass drug administration (MDA) campaigns at reducing prevalence one month after the end of the campaign. While 3-round campaigns cost more than 2-round campaigns, the third round offers an additional chance to reach people who were not covered in previous rounds. At a coverage level of 70%, and independent coverage between rounds, only 9% of individuals will have never been treated. We investigate how critical it is to reach another 70% of the last 9%.

We find that reducing campaign rounds from 3 to 2 increases parasite prevalence one month post-campaign for both AL and DP (Figure 3B). DP is more sensitive to number of rounds than AL. At 70% coverage, prevalence one month post-campaign is more than twice as high for a 2-round campaign as a 3-round campaign when DP is administered. Because DP has long-lasting prophylactic effects, additional campaign rounds are particularly beneficial as previously uncovered individuals become protected against reinfection and prevalence reduction is cumulative from round to round. In contrast, individuals treated with AL are not protected from reinfection for as long, and to some extent prevalence is able to reset to higher levels between rounds, leading to a smaller cost of reducing a campaign to fewer rounds.

When coverage is 100%, prevalence continues to be higher for the 2-round campaign, but the difference between 2- and 3-round prevalence is not as great (Additional file 2: Figure S2A). Switching from a 3-round campaign structure to a 2-round campaign is therefore recommended only for situations where very high coverage can be achieved. In all cases, treatment with DP showed lower prevalence post-campaign than treatment with AL.

ACTs in mass drug campaigns: comparison of MSAT and MDA, 1 month post campaign

Campaigns may distribute antimalarial drugs to all individuals (mass drug administration, MDA), or they may choose to give drugs only to individuals testing positive for parasites (mass screen-and-treat, MSAT). If an insensitive diagnostic is used in an MSAT to identify individuals carrying parasites, then the MSAT campaign will fail to eliminate a large portion of the parasite reservoir, as many individuals who harbour sub-patent infections will fail to receive treatment. However, MSATs may be preferred because they avoid unnecessary dosing of uninfected individuals and may be less likely to lead to drug resistance in parasites.

Current rapid diagnostic tests (RDTs) are sensitive only above 50–200 parasites/μL [70], which has been shown to be inadequate for reducing prevalence in a mass campaign context [71-73]. To identify a minimum diagnostic sensitivity necessary for an MSAT campaign to reduce parasite prevalence with efficacy comparable to an MDA, we tested MSAT screening sensitivities from 0.01 to 200 parasites/μL (Figure 3C). Only diagnostics capable of detecting parasites below 0.1 parasites/μL result in prevalence reduction on par with an MDA campaign; current RDTs are nowhere near sensitive enough and new technologies are necessary if MSATs are to become the campaign of choice in the future. For DP, even MSATs with sensitivity of 0.01 parasites/μL cannot achieve the degree of prevalence suppression seen in an MDA because treating uninfected individuals with DP confers protection against infection. In contrast, AL has little prophylactic effect, and an AL MSAT with a highly sensitive diagnostic can be a good alternative to an AL MDA.

ACTs in mass drug campaigns: effects of coverage and compliance, 1 month post campaign

Success of an MDA campaign may depend on both coverage, the fraction of the population reached by the campaign, and compliance, individuals’ rate of adherence to the correct drug dosing regimen. We tested the influence of coverage and compliance rates on outcomes of campaigns with AL and DP by comparing asexual parasite prevalence one month after the end of campaign rounds.

Increasing coverage up to 90% results in reduced prevalence for both AL and DP campaigns (Figure 3D). Under high transmission conditions with annual EIR of 50, each additional 20% increase in coverage reduces prevalence by half. Beyond 90%, further increase in coverage nets little additional gain. Campaigns deploying DP result in lower prevalence than campaigns using AL. To achieve the same prevalence reduction one month post-campaign, an AL campaign requires coverage of 15-20% higher than the DP campaign.

A sweep over compliance rates shows that most of the reduction in parasite prevalence is accomplished by taking the first dose of AL or DP (Figure 3E). In a semi-immune population, most infected individuals have low parasitaemia (Additional file 2: Figure S1), and a sub-curative dose of ACT is often capable of curing infections with low parasite density. Increasing compliance with unobserved doses from 10% to 100% reduces prevalence only by around 5%.

Coverage and compliance do not complement each other, and increasing compliance can compensate for deficiencies in coverage only to a very limited extent (Additional file 2: Figure S2C). For both AL and DP, increasing coverage results in stronger prevalence reduction than a similar increase in compliance. Compliance exerts a stronger effect on prevalence 4 months post-campaign for DP (Additional file 2: Figure S2D), when improving very low compliance for high coverage campaigns can reduce prevalence by 20% when compliance is increased from 0% to 20%. Improving compliance with DP campaigns lengthens the duration of protection against reinfection, which is a critical component of DP’s efficacy in suppressing prevalence.

Prophylaxis is most beneficial when transmission is high and coverage is moderately high: comparison of AL and DP, 4 months post campaign

In the simulated campaigns discussed above, administration of DP results in equal or lower prevalence than administration of AL under identical conditions. DP and AL differ primarily in the choice of partner drug to the artemisinin-based component. While lumefantrine is the more effective killer of asexual parasites, piperaquine possesses a much longer half-life and therefore confers a longer window of protection against reinfection.

The power of DP’s long prophylactic tail is most striking when considering asexual parasite prevalence four months post-campaign in a high transmission setting (Additional file 2: Figure S2B). A DP campaign with only 45% coverage achieves on average the same prevalence reduction as an AL campaign with 100% coverage at this timepoint. Difficulties achieving good coverage can be overcome by choosing an antimalarial with superior prophylactic qualities.

To quantify the effect of a long-lasting prophylactic such as piperaquine, we assumed that all prevalence reduction observed after a campaign with AL is due to curing individuals, while any additional reduction in prevalence achieved by DP compared to AL is due to protection against reinfection (Figure 4A). Since both AL and DP cure nearly all individuals when correctly administered in the absence of reinfection (Figure 2), this assumption is reasonably accurate.

In a highly endemic setting, prophylaxis can suppress prevalence far beyond what can be achieved by cure alone (Figure 4B left). Even after an MDA campaign with good coverage, the infectious reservoir remains substantial and treated individuals remain vulnerable to reinfection, so prophylaxis is very powerful. Addition of a transmission-blocking drug such as PQ to an AL- or DP-based campaign further reduces prevalence, but prophylaxis has the larger effect.

Under low transmission conditions, administration with AL is able to deplete the infectious reservoir through cure. Because very few infected individuals remain after an MDA campaign with good coverage in a region with low transmission, protection against reinfection post-campaign offers little additional benefit (Figure 4B right). In contrast to high transmission settings, transmission-blocking drugs are approximately as effective at reducing transmission as prophylactics when coupled with a curative antimalarial in a low transmission setting.

The dependence of prevalence reduction due to prophylaxis on both coverage and EIR is shown in Figure 4C by measuring the difference in prevalence 4 months after MDA campaigns using AL and DP. As discussed above, prophylaxis reduces prevalence most for high EIR, where risk of reinfection is highest and protection against reinfection confers the most benefit. For coverage between 50 and 80%, prophylaxis reduces prevalence by an increasing amount as more individuals are protected against reinfection. For coverage above 80%, prophylaxis offers less additional gain in prevalence reduction because the infectious reservoir is already depleted through cure. Thus, each individual is challenged with few infectious bites, and prophylactic benefits are smaller.

Addition of primaquine to MDA with ACTs confers a small additional reduction in prevalence: comparison of AL and DP with and without PQ, 4 months post campaign under variable EIR

Current WHO guidelines recommend the addition of primaquine (PQ) to an ACT regimen as a gametocytocide against P. falciparum infections [61]. To systematically quantify the impact of PQ on ACT campaigns, we compared prevalence four months after MDA campaigns with and without PQ over a range of transmission intensities and coverage levels. While lowering EIR is not a perfect proxy for altered conditions under vector control, it is an acceptable approximation for our study of relatively short-term outcomes. Our results agree with previous modelling observations that single-dose PQ as part of an MDA reduces prevalence only a small amount beyond that achieved by ACTs alone [23,25,27].

We can measure the relative effect of PQ by calculating the slope of a linear regression between prevalence outcomes for campaigns with and without PQ at constant coverage over a range of EIRs (Figure 5A) (see Methods). For the example in Figure 5A, this relative reduction is 0.025, indicating that at any level of annual EIR in our sampled range, addition of PQ to a campaign with AL that has 70% coverage will, on average, reduce prevalence at the 4-month mark by 2.5% relative to the prevalence that would have been observed had only AL been deployed. See Additional file 2: Figure S3 and Additional file 1: Table S4 for prevalence correlations at all coverage levels for both AL and DP.

Higher coverage results in PQ reducing prevalence by a higher relative amount for both AL and DP campaigns (Figure 5B). For AL, complete coverage of 100% yields a relative prevalence reduction of 13% when PQ is present. Thus, we expect that for mass drug campaigns that do not employ a long-lasting prophylactic, dosing with PQ can further reduce prevalence by only 13% or so. Because we did not account for some individuals’ inability to be administered PQ or metabolize PQ to its active product, we expect 13% to approximate an upper bound on PQ impact. However, even at 100% coverage, MDA with AL results in prevalence of 30% at the 4-month mark, so a 13% additional relative reduction in prevalence can still reduce absolute prevalence by 4% overall.

When coverage is sufficiently high, a MDA with DP alone can result in very low prevalence 4 months after the end of campaign. Most individuals are protected against reinfection for many weeks after their last round of DP, and only those who were not reached by MDA and those experiencing drug failure are vulnerable. Addition of PQ to the MDA can be very powerful in this context: PQ reduces prevalence another 60% relative to what was achieved by DP alone. By clearing mature gametocytes, PQ reduces the infectious period of an ACT-treated gametocyte carrier from 7 days to only 1 day. Thus, most individuals in the DP + PQ campaign are both non-infectious and protected against reinfection, leaving only the small fraction of uncovered individuals to propagate the parasite.

Increasing coverage increases the likelihood of local elimination for AL- and DP-based campaigns at all sampled EIR levels (Figure 5C). At low EIR and high coverage, ACTs alone eliminate malaria in nearly all simulations, but only DP can eliminate when transmission is high. Addition of PQ increases the likelihood of elimination but exerts its strongest effect under high EIR when deployed in combination with DP with high coverage.