Holding the line against artemisinin partial resistance in Africa: the case for preemptive treatment policy

1 Introduction

Every effective malaria treatment in use today rests on a single chemical foundation. Artemisinin-based combination therapies (ACTs) pair a fast-acting artemisinin derivative, which clears most of the parasite biomass within hours, with a slower partner drug that removes what remains. For 20 years, that pairing has held, sparing Africa the kind of collapse that ended the useful lives of chloroquine and sulfadoxine-pyrimethamine, both of which were lost across the continent within years of resistance taking hold. Each time, policy reacted only once failure was undeniable, and each time the reaction came too late to save the drug. The foundation is now cracking at the same moment as the wider response stalls. The World Health Organization (WHO) estimates 282 million malaria cases and 610,000 deaths in 2024, with the African Region carrying approximately 94% of cases and 95% of deaths, roughly three-quarters of them in children under five (World Health Organization, 2025a). The 2025 report names antimalarial drug resistance, together with deletions of the histidine-rich protein 2 and 3 genes that blind many rapid diagnostic tests, as a direct and growing threat to case management (World Health Organization, 2025a).

This Perspective makes a single argument and follows it to its consequences. The way African programs are expected to respond to resistance is mismatched to the way resistance behaves. Treatment policy is reactive by design: a national first-line regimen is changed only after measured failure crosses an agreed threshold. On a continent of intense transmission and very large populations at risk, the delay between a drug beginning to fail and a program being able to react is paid in lives. The alternative is to act on the trajectory of resistance rather than wait for its clinical endpoint. The contribution here is to reframe the familiar 10% treatment-failure threshold not as a safeguard but, during the establishment phase of resistance, as a structural liability, and to set out a preemptive agenda that is largely deliverable with tools that already exist.

2 A second continent, but not a milder one

For years, the working assumption was that Africa might escape artemisinin resistance, or that it would at least have to be imported and could be intercepted at a border. That assumption is no longer tenable. Resistance has emerged inside Africa, independently and at multiple points (Figure 1A). The kelch13 R561H mutation was confirmed in Rwanda in 2020 (Uwimana et al., 2020); clinical and molecular evidence of artemisinin-resistant parasites followed in Uganda (Balikagala et al., 2021; Conrad et al., 2023); and in Eritrea, resistant parasites were found already carrying the diagnosis-evading trait of HRP2 deletion (Mihreteab et al., 2023). Validated and candidate kelch13 polymorphisms, among them P441L, C469Y, R561H, R622I, and A675V, are now distributed across the Great Rift Valley, and the WHO records confirmed or suspected partial resistance in at least eight African countries (Rosenthal et al., 2024; World Health Organization, 2025a). Emergence has been de novo and multicentric, thus programs face local establishment at many points rather than a single advancing front that might be contained, and modeling of the selective pressures acting on kelch13 finds no reason to expect slower spread in Africa than in Asia (Rosenthal et al., 2024). The WHO distinguishes validated kelch13 markers, those experimentally and clinically confirmed to delay clearance, from candidate markers still under study, and the validated list relevant to Africa has only lengthened since 2020 (Rosenthal et al., 2024).

It would be comforting to assume that Africa is a milder version of the Greater Mekong, where resistance first arose. The opposite is closer to the truth. Transmission across much of sub-Saharan Africa is far more intense, which means more infections, more drug exposure, and faster spread of any resistant lineage once it gains a foothold. Partial immunity acquired through repeated infection helps many African patients clear parasites even when a drug is failing, which masks treatment failure and can delay its detection precisely where the threshold matters most. A larger, more diverse parasite population offers more genetic raw material for resistance to arise and recombine. Africa is also not one epidemiological place: transmission ranges from holoendemic to near-elimination, so resistance can establish in one zone while a national average stays below the trigger threshold; preemption must therefore be stratified sub-nationally. The WHO recognized the danger in its 2022 strategy to respond to antimalarial drug resistance in Africa (World Health Organization, 2022); the open question is not whether to act, but whether to act before or after first-line treatment begins to fail.

3 The partner drug is where treatment failure is decided

Southeast Asia has already run this experiment to its conclusion. In Pursat, Cambodia, the PCR-corrected efficacy of dihydroartemisinin-piperaquine fell from 98% in 2005 to 63% by 2013, while over the same period the prevalence of kelch13 mutants rose from 40% to 77% (Amaratunga et al., 2016) (Figure 1B). From 2015 to 2018, multi-country efficacy of the same regimen had fallen to 50%, with treatment failure exceeding 85% at the worst sites (van der Pluijm et al., 2019). The reactive response was a treadmill: as one ACT failed, programs switched to another, only to watch resistance follow, until the region’s strategic answer became an outright push to eliminate falciparum malaria before it became untreatable.

That collapse began with the partner drug rather than the artemisinin. Artemisinin partial resistance, on its own, rarely causes an ACT to fail: the kelch13 propeller mutations that define it slow the clearance of young ring-stage parasites, leaving more of them for the partner drug to remove, but a fully effective partner drug can still cure the infection. Failure arrives when resistance to the partner drug joins it (Figure 2A). Once artemisinin can no longer reliably suppress the parasite burden, the survivors meet the partner drug almost as a monotherapy, which accelerates selection for partner-drug resistance (Bjorkman et al., 2024; Rosenthal et al., 2024). The molecular basis of that second step is now well characterized: amplification of the plasmepsin-2 and plasmepsin-3 genes and mutations in the chloroquine resistance transporter underlie piperaquine failure, while changes in mdr1 and related loci shift susceptibility to lumefantrine and amodiaquine (van der Pluijm et al., 2019). Partial resistance is itself defined not by treatment failure but by delayed parasite clearance, defined as a clearance half-life of at least five hours or persistent parasitemia on day three, and confirmed by validated kelch13 markers (Rosenthal et al., 2024).

Africa is now showing the first signals along the same axis. In northern Uganda, parasites with reduced susceptibility to both dihydroartemisinin and the partner drug lumefantrine have been documented (Tumwebaze et al., 2022), and validated kelch13 mutations have climbed to substantial frequencies at several Ugandan sites, accompanied by reports of artemether-lumefantrine treatment failure (Conrad et al., 2023). The diagnostic threat travels in the same convoy: where pfhrp2 and pfhrp3 deletions spread, infections evade HRP2-based rapid tests, are treated late, and seed onward transmission, and the Eritrean and Ethiopian findings show drug-evading and diagnosis-evading parasites moving together (Emiru et al., 2023; Mihreteab et al., 2023). This convergence is especially dangerous in Africa, where case management largely now depends on HRP2-based rapid tests, so a parasite that evades both the test and the drug can spread while appearing, on paper, to be neither present nor resistant.

4 Why reactive policy is the wrong design for Africa

The conventional trigger for changing first-line therapy is a treatment failure rate above 10%. As a brake on premature and costly policy change, the threshold is sensible. As a strategy for a continent at the start of a resistance trajectory, it carries a structural defect. By the time failure is common enough to be measured reliably at 10%, resistance is already widespread, and changing a national regimen takes years to plan, finance, procure, and deliver (Figure 2B). The Southeast Asian sequence measures the ground lost during that delay, and Africa’s partial immunity is likely to lengthen it, since failures are masked and detected late (Rosenthal et al., 2024).

The deeper problem is an asymmetry of errors. Switching first-line therapy too early wastes money and procurement effort, but it preserves both lives and the long-term efficacy of the drugs held in reserve. Switching too late costs lives during the lag and can cost the drug itself, as resistance becomes established in the parasite population. Drug efficacy is best understood as a shared, exhaustible resource: every course that exposes resistant survivors draws the reserve down, and once it is spent, it cannot be quickly replaced. When the two kinds of error carry such unequal consequences, waiting for certainty is not caution but risk. The calculus here is not only epidemiological; it is also ethical. The cost of a lost first-line drug falls hardest on young children in high-transmission, low-income settings, who have neither the immunity nor the reliable access to second-line care that might cushion a failure; deferring action until failure is statistically unambiguous means, in practice, discovering the loss through their deaths.

The financing environment sharpens the point. The abrupt contraction of United States assistance in 2025, including the dismantling of the President’s Malaria Initiative and its implementing agency, has disrupted commodity supply, surveillance, and the technical partnerships on which resistance monitoring depends; modeling attributes roughly half of the malaria mortality reduction of the past two decades to United States development assistance and warns of large avoidable mortality if the cuts stand (Cavalcanti et al., 2025). With total malaria investment flat near US$4 billion against a WHO target of above US$9 billion (World Health Organization, 2024; World Health Organization, 2025a), the surveillance and procurement capacity that a reactive policy depends on is precisely what is most exposed. Sustained domestic resource mobilization, with surveillance budgets ring-fenced rather than treated as the first line of cuts, is now part of any serious resistance strategy. The alternative is to let the same contraction that threatens current case numbers also blind the system to the slower, larger threat building underneath, and push the global elimination targets further out of reach.

5 A preemptive treatment agenda

Acting ahead of clinical failure does not require tools that do not yet exist. It requires using available tools differently, reducing the pressure on them, and preparing the next generation before it is needed.

First, protect the partner drugs now through multiple first-line therapies. Deploying several different ACTs across a population at once, whether split by geography or health facility, rotated over time, or assigned at random, spreads drug pressure and slows selection for resistance to any one partner drug. A Rwanda-specific modeling study that evaluated 26 policy options found multiple first-line therapies to be the single most effective change, lowering projected treatment failures by approximately half relative to the status quo, and the gain was largest when the policy was adopted before resistance became common (Zupko et al., 2023). The building blocks are already in hand, since the WHO recommends six ACTs for uncomplicated malaria (World Health Organization, 2025a); this is a deployable change that does not wait on new products.

Second, prepare triple artemisinin-based combination therapies (TACTs) as a deployable contingency. Adding a second partner drug to an artemisinin forces a parasite to overcome both compounds at once. Artemether-lumefantrine-amodiaquine and artesunate-mefloquine-piperaquine have proved efficacious and well tolerated, including in multidrug-resistant settings (Peto et al., 2022), and modeling indicates that adopting them early, before resistance establishes, delays its emergence rather than merely reacting to it (Nguyen et al., 2023). The regulatory pathways and fixed-dose formulations that would make TACTs deployable should be advanced now, while they remain a contingency rather than an emergency. Feasibility studies in Asia have begun to map the regulatory, supply, and acceptability questions that such a switch would raise; doing that work in Africa now, rather than amid a crisis, is what would make it real (Peto et al., 2022; Nguyen et al., 2023).

Third, build the non-artemisinin bench. The most advanced candidate, the non-artemisinin combination ganaplacide-lumefantrine, met its primary endpoint in the Phase 3 KALUMA trial in 2025, the first such advance in more than two decades and one that held up against partial-resistance mutants, although regulatory approval and wide deployment still lie several years ahead (Medicines for Malaria Venture, Novartis, 2025). Planning introduction pathways and financing in advance avoids repeating the reactive scramble of every previous resistance transition.

Fourth, lower the pressure on the drugs themselves. Resistance is selected by drug exposure, so every infection prevented is also one less chance for it to spread. Two malaria vaccines are now WHO-recommended and in use: the R21/Matrix-M trial reported approximately 75% efficacy against clinical malaria with seasonal vaccination (Datoo et al., 2024), and an independent evaluation of the earlier RTS,S pilot found a 13% decrease in all-cause child mortality (World Health Organization, 2025b). By 2024, the two had reached children in more than 17 African countries, and the WHO is clear that they are most effective when layered onto existing tools rather used in place of them (World Health Organization, 2025b). None of this cures a resistant parasite, but seasonal chemoprevention, sustained vector control, and vaccination all cut the number of clinical cases, which lowers the volume of ACT use and buys the existing drugs time. Treatment policy and prevention policy are, in this sense, parts of the same resistance strategy.

Fifth, fund molecular surveillance as a routine program infrastructure rather than grant-dependent research. Tracking validated kelch13 and partner-drug markers, and monitoring pfhrp2 and pfhrp3 deletions to protect diagnosis, is what lets policy read the trajectory early enough to act. It is the function most exposed to the current funding contraction and the least substitutable, and integrating it into national budgets and into the regional genomic-surveillance networks that already exist is the precondition for every other element here. Surveillance is not a research luxury to be restored once budgets recover; it is the sensor without which a preemptive policy is blind and a reactive one is merely slow. Because parasites and their resistance genes cross borders, these measures also work best when coordinated regionally and stratified by transmission setting within countries, since a mutation tracked within one national program offers little protection if it is spreading unmonitored next door.

6 Anticipating the objections

Several objections deserve a direct answer. The first is that preemptive change is premature and wasteful: why spend on switching while current regimens still cure most patients? The reply is the asymmetry set out above, made concrete by Southeast Asia, where the cost of waiting was measured in a first-line drug lost across a region. The WHO’s own 2022 strategy already calls for proactive resistance management in Africa (World Health Organization, 2022); preemption is an interpretation of that strategy, not a departure from it.

The second objection is cost and logistical complexity, particularly for multiple first-line therapies and triple combinations. Multiple first-line therapies, however, redeploy products that programs already buy and distribute, therefore the change is largely organizational, though its cost and feasibility remain setting-dependent (Zupko et al., 2023). Triple combinations do carry a price premium and a formulation challenge, which is precisely the argument for doing the regulatory and manufacturing groundwork now, when it is cheap relative to the cost of losing a first-line drug later.

The third objection is that triple combinations are not yet recommended by the WHO for routine use, and that the 10% threshold exists to prevent destabilizing, evidence-free policy churn. Both points are fair, and neither is a reason for inaction. Preparing TACTs for deployment is not the same as deploying them prematurely; it is ensuring they are ready when surveillance shows they are needed. Furthermore, preemption does not discard the threshold so much as add to it: routine, well-funded surveillance still triggers reactive change, while multiple first-line therapies, prepared contingencies, and a stocked pipeline shorten the lag once that trigger is pulled.

The fourth objection is that surveillance will catch resistance in time, or that African immunity will buffer its impact. Both are double-edged. The surveillance that would provide early warning is exactly the capacity now being defunded, and immunity that masks treatment failure delays detection rather than preventing spread, thus it makes the reactive threshold less reliable, not more. Neither offers grounds for waiting. A final concern is that acting early might itself hasten resistance to the reserve drugs by using them sooner. The modeling points the other way: spreading exposure across several regimens, and adding partner drugs rather than replacing them, lowers the pressure on any single compound compared with running one first-line drug until it fails (Nguyen et al., 2023; Zupko et al., 2023).

7 Discussion

The biology in Africa has not yet reached the Southeast Asian endpoint of routine ACT failure. That is the entire point, and the entire opportunity. The continent still holds a window in which to act before first-line efficacy is widely lost, and the measures that fit that window—multiple first-line therapies, prepared triple combinations, an advanced non-artemisinin pipeline, interventions that lower drug pressure, and surveillance funded as core infrastructure—are largely available today. Each costs money and political attention at a moment when both are scarce, and the populations who would bear the cost of inaction, young children in the poorest settings, are the least able to absorb it. The counterfactual is a continent-wide collapse of first-line treatment across the settings that carry most of the world’s malaria deaths, which is far more expensive in both money and lives. Whether resistance follows the Southeast Asian path remains, to a degree, within reach of policy. Holding the line is a decision about timing, and the cheaper time to decide is now, before the curve that closes this argument acquires an African counterpart.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Claude (Anthropic) was used for language editing during the final revisions.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Carlo Giannangelo, Monash University, Australia

Reviewed by: Karine Gaelle Le Roch, University of California, Riverside, United States