Built for the village, breached in the city: redesigning malaria vector control for Anopheles stephensi and insecticide resistance

1 Introduction

Vector control has carried much of the progress against malaria this century, though not all of it: expanded diagnosis with rapid tests, subsidized artemisinin-based combination therapy, seasonal and perennial chemoprevention, and stronger case surveillance, including across borders, have each contributed. 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, most of them in children under five (World Health Organization, 2025). That toll, large as it remains, sits beneath a far larger one that was averted: across endemic Africa between 2000 and 2015, infection prevalence roughly halved, and an estimated 663 million clinical cases were prevented, with insecticide-treated nets the single largest contributor among the interventions that analysis partitioned, responsible for about two-thirds of the cases averted (Bhatt et al., 2015). The achievement rested on a coherent design. The dominant vectors bred in rural water, people slept indoors under nets, and a pyrethroid on the netting killed mosquitoes on contact, with indoor residual spraying as backing. For two decades, that design held.

Those gains have not held. Estimated deaths fell to 558,000 in 2019, then rose to 627,000 in 2020, with roughly two-thirds of that increase, approximately 47,000 deaths, attributed to COVID-19-related disruption of prevention, diagnosis, and treatment (World Health Organization, 2021). Cases have risen in every subsequent year, reaching 282 million in 2024, and mortality remains above the prepandemic level (World Health Organization, 2025). Two points of interpretation matter. First, the WHO revised the whole mortality series upward in 2021 after adopting a new cause-of-death method for children, so figures near 400,000 published before that revision are not comparable with those quoted here. Second, and more consequential for what follows, the architecture described above was already losing ground before either of the threats considered in this Perspective took hold: resurgence has been driven by the interruption and underfunding of programs as much as by biology, and a redesign that ignores the financing dimension will fail for the same reason.

It is now under attack from two directions at once. Progress has stalled since the mid-2010s, and the WHO names insecticide resistance and the invasive vector Anopheles stephensi among the biological forces shaping the present picture (World Health Organization, 2025). The two share a property that should change how programs respond. Each dismantles one of the founding assumptions of the model: A. stephensi introduces a mosquito the architecture was never built to stop, while pyrethroid resistance corrodes the chemistry the architecture runs on. Neither can be met by shipping more of the same commodity. The argument of this Perspective is that vector control in Africa has to be redesigned rather than resupplied, that the redesign is largely achievable with existing tools, and that its weakest point is not technology but the surveillance and financing needed to implement it.

2 Anopheles stephensi: an urban vector spreading through African cities

Anopheles stephensi is a competent vector of both Plasmodium falciparum and Plasmodium vivax, and unlike the rural species that drive African transmission, it is one of the few anophelines at home in the center of cities. It breeds readily in tanks, cisterns, rooftop reservoirs, and construction water of dense settlements; tolerates the hot and dry conditions that exclude many native vectors; and shows behavioral flexibility—resting and feeding both indoors and outdoors and taking blood from livestock as well as people—that allows it to slip past tools designed for a mosquito that bites a sleeping person indoors at night (Takken and Lindsay, 2019). A control system optimized for rural Anopheles gambiae and Anopheles funestus was never configured to find or kill it.

A little over a decade ago, the range of A. stephensi extended only across South Asia and parts of the Arabian Peninsula. Its first African detection came in 2012 at a seaport in Djibouti, and it has since been recorded in Ethiopia and Sudan (2016), Somalia (2019), Nigeria (2020), Eritrea, Kenya and Ghana (2022) and, most recently, Niger (2024), with detections beyond Africa in Sri Lanka (2017) and Yemen (2021), so that the WHO now reports its presence in nine African countries (Figure 1A) (World Health Organization, 2022a; World Health Organization, 2025). The WHO issued a vector alert and launched a dedicated initiative to slow the spread, classifying the mosquito as a major threat to malaria control and elimination (World Health Organization, 2022a; World Health Organization, 2022b). Its African detections have clustered around seaports and trade corridors, consistent with introduction through maritime traffic, and because it exploits the same urban container habitats as Aedes aegypti, the established machinery for Aedes surveillance can often be turned to finding it (Takken and Lindsay, 2019).

Djibouti shows where this can lead. The country was approaching elimination when A. stephensi arrived; an unusual urban outbreak in 2012 was followed by increasingly severe annual epidemics, and reported cases climbed from a few thousand to tens of thousands within a decade (Sinka et al., 2020). Spatial modeling of the mosquito’s preferred habitat suggests that, if it spreads unchecked across the continent’s cities, an additional 126 million urban Africans could be brought within reach of malaria (Sinka et al., 2020). The exposure is growing on its own: more than 4 in 10 Africans already live in towns and cities, a share projected to reach approximately 6 in 10 by mid-century, much of it in exactly the dense, water-storing settlements the vector favors (World Health Organization, 2022c) (Figure 1B). Because A. stephensi also persists through the dry season in man-made habitats, when rural species such as Anopheles arabiensis recede, it raises the prospect of year-round urban transmission in places that previously had a quiet season.

The resemblance to A. aegypti is the most operationally useful feature of this vector, and it deserves to be stated as a strategy rather than an analogy. The two species share container habitats: in Djibouti, A. stephensi was found breeding alongside A. aegypti and Culex quinquefasciatus in urban containers (Pommier de Santi et al., 2021); in Kebri Dehar, in eastern Ethiopia, both species were recovered from the same cohort of household water reservoirs (Yared et al., 2025); and in Aden, Yemen, A. stephensi was first detected in container surveys that were targeting A. aegypti (Allan et al., 2023). Three practical consequences follow. Container larval surveys, ovitraps, and household container indices that arbovirus programs already run can find A. stephensi without building new field infrastructure. Multiplex molecular assays, including environmental DNA recovered from breeding-site water, can identify both species and their resistance markers from a single sample (Kristan et al., 2023). Moreover, container-directed larval source management can likewise serve dengue, chikungunya and malaria control at once, which matters where budget lines are separate but field teams are not. The overlap is partial rather than complete, but the reason is local ecology rather than a fixed division of containers between the two species. In Kebri Dehar, the most productive A. stephensi habitats were large ground-level cisterns holding domestic and construction water, while A. aegypti was concentrated in tires and small containers (Yared et al., 2025). That split is not a general rule, and A. aegypti readily occupies large-storage vessels where piped supply is absent or unreliable. In a survey covering all nine inhabited islands of Cabo Verde, water drums were the most prevalent container type recorded, though puddles were the most productive habitat found (Leal et al., 2020); in Ouagadougou, drums and barrels ranked second to tires for larval production in two of three health districts and were the most productive pupal habitat in one of them, with container use differing significantly between districts of the same city (Ouédraogo et al., 2022). The operational consequence is not that either program can borrow the other's container list, but that neither list can be written in advance; each must be built from what local surveys actually find. Which container types carry the productivity has to be established by local larval and pupal surveys, city by city and species by species, before larval source management is targeted. Integration is an efficiency, not a license to assume one species proxies for the other.

3 Dire Dawa 2022: invasive vector, artemisinin partial resistance, and diagnostic escape in one outbreak

The sharpest warning so far has come from Dire Dawa, the second-largest city of Ethiopia. In 2022, a prospective case–control investigation tied a dry-season malaria outbreak in the city to A. stephensi, which was breeding in urban water containers and feeding in the city (Emiru et al., 2023). The episode mattered for more than the vector. The parasites recovered carried molecular signatures of both artemisinin partial resistance, in the kelch13 gene, and deletions of the histidine-rich protein 2 and 3 genes, allowing P. falciparum parasites to evade the rapid diagnostic tests on which most African programs rely (Emiru et al., 2023). In a single outbreak, three threats usually discussed separately arrived together: an invasive urban vector that the net-and-spray model does not stop, parasites against which the first-line medicine is beginning to fail, and parasites that the standard diagnostic can miss. A tool that addresses only one of the three leaves the others free to operate. It also locates the danger precisely where surveillance is thinnest—in fast-growing cities that have not historically been treated as malaria-transmission settings at all.

4 Pyrethroid resistance and the widening gap between net coverage and protection

The second front is chemical. For most of the period in which the insecticide-treated net carried the response, pyrethroids were the only insecticide class approved for treating nets, and the resulting selection pressure has produced the predictable result. Resistance, driven by both target-site mutations that reduce the binding of the insecticide and metabolic mechanisms in which overexpressed detoxifying enzymes break it down, is now widespread across the principal African vectors, including A. gambiae, A. coluzzii, and A. funestus (World Health Organization, 2025). The consequence is not theoretical. Mathematical modeling that combines entomological and trial data estimates that pyrethroid resistance has measurably eroded the protection nets provide, reducing the number of cases averted below what fully susceptible vectors would allow (Churcher et al., 2016). Metabolic resistance in particular, driven by overexpressed cytochrome P450 enzymes such as the CYP6P9 variants in A. funestus, is the mechanism the piperonyl butoxide synergist is designed to reverse. Resistance is also patchy, differing between species and between districts, so a national average can hide places where the standard net has already failed (Churcher et al., 2016).

Hut-trial data make the loss concrete. Against wild pyrethroid-resistant mosquitoes, a standard pyrethroid-only net kills only about one in five of the host-seeking vectors that contact it (Tchouakui et al., 2023). A net that kills a fifth of the mosquitoes it meets is still distributed, counted and reported as coverage, but it delivers a fraction of the protection the coverage figure implies. This is the central problem of the present moment: across large areas, coverage and protection have come apart, and the gap between them is invisible in the routine statistics that programs and donors use to judge progress. Continuing to push pyrethroid-only nets into highly resistant settings delivers steadily less per net, while appearing, on paper, to hold the line.

5 Next-generation nets: what the trials show and where the advantage fades

Two redesigned net classes have reached scale in response. The first adds the synergist piperonyl butoxide (PBO), which inhibits the detoxifying enzymes behind metabolic resistance and partly restores the killing power of the pyrethroid; in a large pragmatic trial embedded in Uganda’s national net campaign, PBO nets reduced infection prevalence relative to conventional pyrethroid nets (Staedke et al., 2020). The second goes further, pairing the pyrethroid with a second insecticide that acts by a different mechanism: Interceptor G2 combines the pyrrole chlorfenapyr with a pyrethroid, and Royal Guard combines the insect growth regulator pyriproxyfen with one. A cluster-randomized superiority trial in Benin found that the chlorfenapyr–pyrethroid net reduced malaria infection prevalence by 46% over 2 years relative to a pyrethroid-only net (Accrombessi et al., 2023), and a four-arm trial in Tanzania that tested all three next-generation net types head to head recorded a 55% reduction in the odds of infection for the chlorfenapyr net and ranked it the most effective and most cost-effective of the new designs (Figure 2A) (Mosha et al., 2022). In that same trial, neither the piperonyl butoxide net nor the pyriproxyfen net significantly outperformed the standard pyrethroid net, a reminder that the next-generation label spans products of very different value (Mosha et al., 2022). Experimental hut data against resistant vectors point the same way (Tchouakui et al., 2023). The market has moved with the evidence: next-generation nets rose from 59% of those delivered to sub-Saharan Africa in 2022 to 78% of 195 million nets delivered in 2023 (Figure 2B) (World Health Organization, 2024).

The limits matter as much as the gains, and they point back toward the same conclusion. Next-generation nets cost more, a real constraint as financing contracts and the pooled procurement that disciplines prices comes under pressure. The Tanzania trial that ranked the chlorfenapyr net most effective also judged it cost-effective despite its higher unit price, because the protection it bought outweighed the premium (Mosha et al., 2022); that calculus, however, rests on durability. Durability and wash resistance vary between products: the advantage of pyriproxyfen-based nets is inconsistent where target-site resistance dominates, and even the chlorfenapyr net is not permanent. The third-year analysis of the Benin trial found that the entomological and epidemiological advantage of the dual-ingredient nets had largely faded by their final year of intended use, partly through declining insecticide content and lower net use (Accrombessi et al., 2024). The operational implication is uncomfortable for programs accustomed to a single commodity. Net type now has to be matched to the local resistance mechanism, and that matching is only possible where resistance is actually measured. The net has stopped being generic; it has become a choice that depends on surveillance.

6 Beyond the net: larval source management, house modification, and non-chemical tools

Redesigning the net is necessary but not sufficient, because an urban vector that breeds in containers and an indoor net designed for rural sleeping spaces are only a partial match. Urban transmission calls for larval source management adapted to city ecology, including the management and treatment of the water-storage and construction-site habitats A. stephensi exploits, supported by entomological surveillance at points of entry to detect the vector before it establishes. Mauritius offers an instructive model: leveraging an existing program against Aedes, it strengthened surveillance at its seaport and airport and coordinated across health, veterinary, and port authorities, and although A. stephensi was not detected, the exercise built the multisectoral capacity needed to respond (Iyaloo et al., 2024), the kind of approach the WHO framework for the malaria response in urban areas is designed to guide (World Health Organization, 2022c). Indoor residual spraying remains a core intervention, but here the redesign principle applies as well: where pyrethroid resistance is intense, programs have moved to non-pyrethroid formulations, such as organophosphate and neonicotinoid products, to preserve its effect, and larviciding has a defined, if supplementary, role where breeding sites are few, fixed, and findable—as they often are in cities (World Health Organization, 2025).

Chemistry is not the only option, and in cities, it may not be the most durable one. Larval source management has the strongest evidence among the non-chemical approaches: a Cochrane review found that habitat modification and larviciding reduced malaria incidence by 49% to 74% and prevalence by 68% to 89% where habitats were few, fixed, and findable (Tusting et al., 2013), which is close to a description of A. stephensi in an African city. Targeting matters more than coverage. In Kebri Dehar, 77% of larvae came from 23% of sites, overwhelmingly large cisterns holding domestic and construction water, and a metapopulation model of that system predicted that treating those superproductive reservoirs at coverage above 60% could locally eliminate the population (Yared et al., 2025). However, which containers hold that productivity is itself a local finding rather than a species constant (Leal et al., 2020; Ouédraogo et al., 2022). Construction practice itself has been shown to create and sustain the habitat in Ethiopian cities (Yared et al., 2023). Biological larvicides based on Bacillus thuringiensis israelensis suit container habitats where chemical residues in stored drinking water are unacceptable. Structural measures have independent support: a Cochrane review of house modification found residents of screened or otherwise modified houses about a third less likely to carry P. falciparum, and the WHO recommends house screening as a supplementary measure (Fox et al., 2022). Not every alternative is equally supported, and the difference is instructive. A Cochrane review of larvivorous fish found the evidence insufficient to support their use for malaria control (Walshe et al., 2017), and attractive targeted sugar baits, despite strong entomological results in earlier work, produced no significant reduction in malaria incidence or prevalence in phase III cluster-randomized trials in Zambia and Kenya (Ashton et al., 2024; Ogwang et al., 2025). The discipline that rejects the generic pyrethroid net rejects the untested alternative on the same grounds: adoption follows trial evidence, not novelty.

Volatile pyrethroid spatial emanators, which release insecticide into the air around a person rather than relying on contact with a treated surface, address the outdoor and residual biting that nets do not capture, and a recent systematic review and meta-analysis supports their potential as a complementary intervention where nets alone are insufficient (Chen et al., 2025). Further out, endectocides that make treated people or livestock lethal to the mosquitoes that bite them, attractive targeted sugar baits, and genetic approaches such as gene drive are under development, but none is yet deployable at scale, and none removes the need for the core tools. The realistic near-term picture is not a single replacement for the pyrethroid net but a matched set of interventions, chosen for the local vector and the local resistance profile, which is a more demanding thing to run.

7 What surveillance requires, and why the financing structure obstructs it

Every element of this redesign rests on one capacity, so it is worth being specific about what that capacity consists of rather than invoking it.

For resistance, it means sentinel sites sampled at defined intervals rather than opportunistic collections. The WHO discriminating-concentration bioassays establish whether resistance is present; intensity assays at 5 and 10 times that concentration establish whether it is strong enough to matter operationally; and synergist assays with piperonyl butoxide establish whether the mechanism is metabolic, which is the question that determines whether a PBO net is the correct purchase or a wasted premium (World Health Organization, 2022d). Compounds that cannot be applied stably to filter paper, chlorfenapyr and clothianidin among them, require the WHO bottle bioassay, and molecular genotyping of knockdown-resistance mutations and CYP6P9 variants separates mechanisms that phenotype alone cannot. Results reported to the WHO Malaria Threats Map make the picture comparable across districts and countries rather than leaving it trapped in national reports (World Health Organization, n.d).

For A. stephensi, it means larval container surveys and adult trapping at seaports, airports, and trade corridors; morphological identification confirmed by ITS2 or COI sequencing because misidentification against native anophelines is common; and case investigation capable of separating locally acquired from imported infection. For both, it means entomological data feeding the subnational stratification exercises that now determine intervention mixes, since a national average conceals the districts where the standard net has already failed.

The obstacle is not only the volume of financing but also its structure. The Global Fund, the largest external financier of malaria control, requires that a change to a more expensive net class be justified by the vector profile and robust entomological data (The Global Fund to Fight AIDS, Tuberculosis and Malaria, 2022). That requirement is correct in principle and perverse in practice, because the surveillance that generates the evidence competes for the same country allocation as the commodities themselves, and commodities win when the allocation contracts. Allocations are disease-specific, set on a 3-year cycle and calculated by a burden-weighted formula, so a program that wants to fund shared entomological capacity serving both Anopheles and Aedes, or to hold a surveillance system together between grant cycles, has little room to do so. The Fund’s own independent evaluation identifies the share of malaria funding reaching vector control as a constraint on impact (The Global Fund to Fight AIDS, Tuberculosis and Malaria, Evaluation and Learning Office, 2025), and analysts have proposed consolidated envelopes with eased disease earmarks precisely so that countries can allocate across these boundaries (Center for Global Development, 2026). The 2025 contraction of United States assistance falls on the same function, and modeling of two decades of that assistance links its sustained withdrawal to a substantial rise in avoidable deaths (Cavalcanti et al., 2025; World Health Organization, 2025).

The failure mode is specific and slow. A program can distribute nets from existing pipelines for a year or more, while its ability to detect rising resistance, find the new vector early, and judge whether the tools still work decays without appearing in any coverage statistic. By the time the loss is visible in cases, the decisions that would have prevented it are years past.

8 What redesign asks of programs

If vector control is to be redesigned rather than resupplied, four shifts follow. First, net choice has to become a decision rather than a default. Distributing the cheapest pyrethroid-only net by habit is no longer defensible where resistance is intense; programs need routine resistance data and the procurement flexibility to send PBO or dual-active-ingredient nets where the local mechanism warrants them (Churcher et al., 2016; Staedke et al., 2020; Accrombessi et al., 2023). Second, the urban dimension has to be built in rather than bolted on. Cities that never had an indigenous malaria vector now need entomological surveillance at ports and within neighborhoods; larval source management fitted to container habitats and to the container types that local productivity surveys identify, run jointly with Aedes programs where both vectors share the same containers; and case investigation able to separate local transmission from imported infection (Sinka et al., 2020; Ouédraogo et al., 2022; Iyaloo et al., 2024; Yared et al., 2025). Third, the toolbox beyond the net has to be diversified deliberately, with non-pyrethroid indoor spraying where resistance demands it, larval source management and house modification where the evidence supports them, and spatial emanators where outdoor and residual biting escape the net, each chosen on evidence, not on whatever happens to be available (Tusting et al., 2013; Fox et al., 2022; Tchouakui et al., 2023; Chen et al., 2025). Fourth, and underpinning the rest, the surveillance and procurement systems that make these choices possible have to be financed as core infrastructure, not as project add-ons that lapse when a grant ends, which requires financing instruments flexible enough to fund capacity rather than commodities alone (The Global Fund to Fight AIDS, Tuberculosis and Malaria, 2022; World Health Organization, 2025; Center for Global Development, 2026). None of these is a research aspiration; each is deliverable now. What they require is that programs trade the simplicity of a single commodity for the harder discipline of matching tools to a measured and moving threat.

9 Conclusion

None of this requires waiting for a breakthrough; the difficult parts are organizational and financial, not technological. The through-line is that vector control can no longer be run as the distribution of a single, generic commodity. It has become a problem of matching a widening set of tools to a moving target, which is harder to manage, more expensive, and far more dependent on data. The cheapest moment to contain A. stephensi and to stay ahead of insecticide resistance has already passed; it was before either took hold. The next cheapest moment is now, while the tools exist, the vector’s range is still incomplete, and the resistance is still partial. The window will not stay open on its own, and the surveillance that would tell us how fast it is closing is the very thing now most at risk.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI was used: Claude (Anthropic) was used for language editing during 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: Diego Alonso, São Paulo State University, Brazil

Reviewed by: Lara Ferrero Gómez, Universidade Jean Piaget de Cabo, Cabo Verde