Main

In 1962, Rachel Carson’s ‘Silent Spring’ revealed the ecological costs of DDT, an insecticide that had previously revolutionized agriculture and malaria control1,2,3. Carson’s revelations marked a pivotal moment in environmentalism, highlighting the necessary balance between immediate benefits and potential long-term ecological implications. This awakening paved the way for the holistic strategy called ‘One Health’—an interdisciplinary, multisectoral approach that ‘aims to sustainably balance and optimize the health of people, animals and ecosystems’4. However, some One Health engagements sidelined environmental aspects4,5,6. Our study intends to integrate environmental health considerations into malaria control efforts using ivermectin-treated cattle.

In 2022, 249 million malaria cases were reported, primarily in sub-Saharan Africa, where malaria impairs public health and development7. Combating malaria includes early diagnosis of the causative Plasmodium parasites, antimalarial drugs and control of Anopheles mosquitoes, the malaria vectors7,8. Control measures widely rely on insecticides targeting adult mosquitoes since modest reductions in survival may substantially lower parasite transmission9. Central control tools are long-lasting insecticidal nets (LLINs) and indoor residual spraying10.

Persistently high incidences are partly explained by Anopheles mosquitoes adapting to environmental changes and extensively used insecticides3,7,11. Some mosquitoes show increased survival against insecticide contact (physiological resistance12) or avoid contact by biting humans who are not protected by LLINs (behavioural resistance13). Mosquitoes also exhibit behavioural resistance by biting animals instead of humans, maintaining vector populations at levels enabling parasite transmission14. This largely neglected gap in current control measures motivates this study.

Using the endectocide ivermectin, we aimed to render cattle blood toxic to behaviourally resistant malaria mosquitoes of opportunistic or zoophagic feeding habits (Fig. 1a). With endectocide-treated livestock (ETL), vectors are exposed to mosquitocidal drugs in animal blood, and ivermectin is considered suitable for ETL15,16,17,18. It is safe and effective in treating parasites in animals (and humans) and improves animal health while boosting livestock productivity15,19,20. From a One Health perspective, ETL already links human and veterinary health. However, animals excrete ivermectin mainly unmetabolized in faeces21. These residues can enter agroecosystems and harm dung and soil organisms22,23,24,25. After injection, most commercialized veterinary ivermectin formulations reach peak plasma levels within days, with limited terminal half-life21. These formulations provide mosquitocidal protection for approximately 2 weeks, fostering the popularity of long-acting formulations to fully cover malaria transmission seasons16,26,27. However, long-acting endectocide formulations can also prolong release via cattle dung, with longer-lasting risks to non-target fauna28.

Fig. 1: Building a path towards environmental safety in endectocide-based approaches to control opportunistic and zoophagic malaria vectors.
figure 1

a, Conceptual framework illustrating the predicament and the three main study objectives. Once excreted, residual ivermectin that was originally repurposed to control zoophagic mosquitoes enters agroecosystems, followed by complex interactions in the dung–soil–biota–plant environment. b, Experimental treatment and sampling scheme. Not all cattle were sampled for plasma and dung at every time point. The IVM-BEPO depot is a formulation designed to solidify after injection. The formed polymeric matrix (the depot) entraps the drug, which is released in a controlled and sustained way. Figure 1 was created with BioRender.com.

Full size image

We investigated cattle and Anopheles coluzzii mosquitoes from Burkina Faso and compared a commercialized injectable ivermectin veterinary formulation and a long-acting depot formulation, repurposed to control vectors for at least 6 months. First, we characterized pharmacokinetic profiles in cattle plasma and dung. Second, we assessed mosquitocidal efficacy through direct skin-feeding assays on cattle (Fig. 1b). Modelling approaches were used to predict field effectiveness. Third, considering environmental risks, we monitored ivermectin dissipation in dung and performed soil sorption studies with soils from Burkina Faso. We related measured dung concentrations to published ecotoxicological observations on dung organisms and propose risk mitigation measures.

While ivermectin-based strategies promise notable advantages in malaria vector control and potential benefits against other arthropod-borne zoonotic diseases, these must be balanced with environmental implications. Considering these dual perspectives, we promote a holistic view of ivermectin administration to sustainably harness its potential.

Empowered by an extensive dataset, our vector population model underscores the transformative potential of ivermectin-treated cattle and long-acting ivermectin formulations. By bridging cutting-edge vector control with ecotoxicological insight, our research aligns with a comprehensive vision that encompasses animal health, community well-being and ecosystem preservation, reflecting the integrated, unifying approach4 necessary for sustainable development.

Results

Ivermectin pharmacokinetics in cattle plasma and dung

Mean ivermectin concentrations in cattle plasma and dung are shown in Fig. 2. After a single injection with the commercial IVOMEC-D formulation (not long-acting), ivermectin was detected in plasma and dung for up to 28 days (Fig. 2a,b). In plasma from the long-acting IVM-BEPO-treated cattle, ivermectin levels exceeded the limit of quantification for at least 211 days, our study duration (Fig. 2c). All control samples before and during treatment were free of detectable ivermectin. Peak mean ± s.d. concentrations for IVOMEC-D cattle were 122 ± 35 ng ml−1 in plasma (DAI (day after injection) 7 after the sixth injection, 160 days after study start) and 1,927 ± 912 ng g−1 dry weight (dw) in dung on DAI 2 after the first injection. After treatment with the long-acting formulation, maximum mean plasma concentration in the first month was documented on DAI 7 (22.9 ± 8.9 ng ml−1) and, overall, on DAI 204 (32.7 ± 6.2 ng ml−1). The highest mean concentrations in dung throughout the study were measured on DAI 7 (530 ± 327 ng g−1 dw) after depot injection. The Cmax/dose index for plasma and dung is considerably lower for IVM-BEPO since the injected dose was six times higher compared with IVOMEC-D. Individual data points, sample sizes and pharmacokinetic parameters for ivermectin in cattle plasma and dung are summarized in Supplementary Tables 1–3.

Fig. 2: Pharmacokinetic profiles of ivermectin in cattle plasma and dung.
figure 2

a,b, Arithmetic mean ± s.d. ivermectin concentrations over 2–28 days in cattle plasma (a) and cattle dung (b) after one subcutaneous injection of a commercial (0.4 mg kg−1, IVOMEC-D) or a long-acting (2.4 mg kg−1, IVM-BEPO) ivermectin formulation. On DAI 21 and 28, several IVOMEC-D concentrations were below the limit of quantification in plasma (Supplementary Table 1). c, Continuation of IVM-BEPO arithmetic mean ± s.d. ivermectin concentrations over 2–211 DAI in cattle plasma and dung. All points display quantifiable concentrations. a–c, Error bar areas show the smoothed (B-spline) s.d. of 4–8 measured samples from 2–4 treated cattle. The y axes in c use different scales.

Source data

Full size image

Mosquitocidal activity

Overall, 26,637 adult female mosquitoes were allowed to blood-feed on cattle in different experimental arms, 94% of which successfully fed, without differences between arms (Wald Χ2 = 1.70, P = 0.42; Supplementary Fig. 1). For survival measurements, we followed 5,384, 4,852 and 5,400 females blood-fed on control, IVOMEC-D and IVM-BEPO cattle, respectively. Mosquitoes fed on IVM-BEPO cattle always had higher mortality probabilities than control mosquitoes (all hazard ratios >1; P < 0.001; Supplementary Table 4). IVOMEC-D-fed mosquitoes repeatedly showed similar mortality probabilities to the control (hazard ratio not significantly different from 1 on DAI 120, 204, 211). Mosquitoes fed on IVM-BEPO-treated cattle rarely survived beyond 10 days (Fig. 3a), the median extrinsic developmental period for Plasmodium falciparum in field mosquitoes29. With this treatment, only mosquitoes that ingested Plasmodium before ivermectin exposure could become infectious, transmitting parasites during only 1–3 gonotrophic cycles. However, monthly IVOMEC-D treatments often resulted in mosquito survival over 10 days, failing to block transmission. Ivermectin concentrations killing 50% of mosquitoes after 10 days were equal between formulations (IVOMEC-D LC50 ± s.e.: 7.5 ± 0.7 ng ml−1; IVM-BEPO LC50: 8.3 ± 0.3 ng ml−1; t-test comparison between formulations: t = −1.17, P = 0.23). Unlike IVOMEC-D, the depot formulation reached this level over 211 days (Fig. 2c, Supplementary Table 5 and Supplementary Fig. 2). The 10-day LC90 differed significantly, with lower concentrations needed to cause 90% mortality with the depot (IVOMEC-D LC90: 31.8 ± 2.8 ng ml−1; IVM-BEPO LC90: 19.2 ± 0.6 ng ml−1; t = 4.20, P < 0.001).

Fig. 3: Efficacy endpoints of the ivermectin treatments measured experimentally through mosquito survival and simulated in silico using a population model.
figure 3

a, Median survival times ± 95% CI of mosquitoes after feeding on cattle (total n = 5,384, 4,852 and 5,400 female mosquitoes, blood-fed on control, IVOMEC-D and IVM-BEPO cattle, respectively; mean ± s.d. mosquitoes per respective sampling date: 269 ± 21.7, 269 ± 19.4, 270 ± 20.7). The shaded area represents median survival times mosquitoes should reach after feeding to become infectious. Plasmodium extrinsic incubation period is considered 10 days, gonotrophic cycle 3 days29. Survival time range encompasses two scenarios where mosquitoes feed on treated cattle 3 days before or after an infectious blood meal27. Because of high survival of control mosquitoes at 21 and 45 days after the experiment started, reliable upper confidence intervals could not be computed. b,c, Model prediction of IVOMEC-D (b; the model simulates up to 8 injections) and IVM-BEPO (c) formulation relative effects on mosquito field population densities after mass treatment of cattle. Treatments were simulated as in this study: 0.4 mg kg−1 bw monthly (b) or on a single dose of 2.4 mg kg−1 bw (c). Different parameters were given arbitrary values to simulate treatment efficacy in different scenarios: r = cattle versus human ratio; a = vector preference for humans as measured in choice test experiments; LLIN coverage = 0.5; HBI = calculated index representing the number of mosquitoes that have taken a blood meal on humans considering all other parameters. HBI values are for an LLIN coverage index of 0.5. Refined modelling scenarios are shown in the Supplementary Information Modelling chapter.

Source data

Full size image

Our vector population simulation model demonstrates that using ivermectin-treated cattle, combined with bed nets, reduces mosquito populations by 25 to over 90% (Fig. 3b,c; refined modelling scenarios in Supplementary Fig. M1). Model outputs predict increased effectiveness for both formulations when cattle outnumber humans as the host population and when vectors prefer animals. Treatments primarily affect infectious mosquitoes aged 10 days old or older since feeding likelihood on treated cattle and dying increases with age. The long-acting depot consistently reduces epidemiologically relevant vectors by over 50% for at least 6 months and peaks at 95% reduction. Multiple injections with IVOMEC-D are less effective, with vector populations cyclically increasing after peaks of 5–72% reduction.

Mosquito toxicity and ecotoxicological effects

Figure 4a shows ivermectin correlation in cattle plasma and dung after one injection of the depot formulation. Calculated 10-day LC50 and LC90 for A. coluzzii translate to predicted ivermectin concentrations in cattle dung ranging from 138–319 ng g−1 dw. The mean ± s.d. excreted concentration, however, was higher at 337 ± 185 ng g−1 dw (n = 86), corresponding to a mean plasma concentration of 19.1 ± 9.2 ng ml−1, regardless of time. These concentrations were descriptively compared to ecotoxicological half maximal effective concentration (EC50) data from standard tests (for example, refs. 30,31) with non-target dung beetles and flies (Fig. 4b and Supplementary Table 6). For a comparable formulation with similar release pattern under large-scale field conditions, predicted ivermectin concentrations in dung could exceed EC50 values for most tested dung flies. For flies and beetles, only EC50 data were compared, meaning that statistically, even lower concentrations could endanger a proportion of coprophagic arthropod communities32. Effectively, Fig. 4b serves as preliminary risk estimation for dung fauna based on laboratory data.

Fig. 4: Connection of ivermectin in cattle plasma and cattle dung and impact on non-target dung fauna.
figure 4

a, Simple linear regression (Y = 16.585 X) through the origin (R2 = 0.83), with 95% CI bands demonstrating the correlation between ivermectin concentrations in cattle dung (ng g−1 dw) versus cattle plasma (ng ml−1) for the long-acting IVM-BEPO formulation (data from 2‒211 days after injection where both plasma and dung data are available). Dashed anchor lines indicate predicted concentrations in dung for previously determined 10-day LC50 and LC90 for A. coluzzii. b, Half-violin plots of aggregated EC50 data from ecotoxicological single-species testing with dung beetles (mortality and reproduction) and dung flies (mortality); n = number of findings from reviewed literature. The EC50 dataset is detailed in Supplementary Table 6 and refers to the dry weight concentration in dung. Dotted lines and the shaded area are extended from the LC50 and LC90 values in a and illustrate a corridor of potential ivermectin release through dung of treated cattle. Whiskers in the kernel density plot show the standard deviation around the median (black line). All dung concentration measurements and EC50 are reported on a dry weight basis to facilitate accurate comparisons across studies.

Source data

Full size image

Environmental fate of ivermectin

Figure 5a presents the particle size distribution and organic carbon (Corg) content of 30 agricultural soils from three Burkina Faso villages. Since sorption studies33 were performed with fine soil (<2 mm), boxplots display the proportion of fine soil, considering high coarse-soil contents in many samples. Individual fine-soil analysis34 revealed mostly loams, sandy loams and silt loams in the study region. The KD sorption coefficients for six selected soils ranged from 55–123 ml g−1, averaging 84.5 ml g−1. Normalized to soil Corg, the KOC sorption coefficients varied from 6,630–10,870 ml g−1 (Supplementary Table 7). Sorption correlated significantly (P < 0.05) with Corg and pH value (r = 0.86 each) (Fig. 5b). Regarding dung storage for studying ivermectin dissipation (Fig. 5c), results were categorized into external and internal storage. Significant differences (P < 0.01) in ivermectin concentrations were observed in samples stored under internal laboratory conditions. Figure 5d displays mean ivermectin dissipation in internally stored samples. First-order kinetics describe the dissipation in stored dung. Including the day 0 concentration (normalized to unity), the fit through days 0, 30, 60 and 90 resulted in a degradation constant k1 = 0.00145 (adj. R2 = 0.059), 50% dissipation time (DT50) = 478 and DT90 = 1,588 days. With only days 30, 60 and 90, this fit yielded k1 = 0.00381 (adj. R2 = 0.968), DT50 = 182 and DT90 = 604 days in dung.

Fig. 5: Soil properties, sorption correlation and degradation of ivermectin in stored dung samples.
figure 5

a, Particle size distribution and organic carbon (Corg) in topsoils (n = 30). Sand, silt, clay and Corg percentages are normalized to 100% to complement the coarse-soil fraction. Shown fine-soil distributions do not describe soil texture. The black line in boxes is the median, with boxes stretching from the 25th to the 75th percentile. Whiskers display the upper and lower inner fences. See Supplementary Table 12 for village-specific details. b, Correlation (Pearson’s r) between linear KD sorption coefficients and soil properties (n = 6 selected soils). Circle sizes represent absolute values of the correlation coefficient; *P ≤ 0.05. EC, electrical conductivity. c, Relative ivermectin IVM-BEPO concentrations in stored dung (n = 13–23) related to storage time under external (semi-field) and internal (insectary) conditions. Concentrations in stored dung are from timepoints DAI 2, 7, 75, 174 and 204 to cover the study duration. Respective day 0 concentrations are normalized to unity for comparison (two-sided Tukey pairwise mean comparison, with specific comparisons revealing: 30 days versus 60 days (P = 0.02), and 30 days versus 90 days (P = 0.0002)); storage concentrations followed a normal distribution (P < 0.05). d, Modelled dissipation in stored dung (internal conditions) following simple first-order kinetics, with x axis from −3 to 93 and y axis abridged. DT50 and DT90 show the predicted time (days) for 50% and 90%, respectively, of the initial ivermectin concentration to dissipate. Day 0 concentration normalized to unity. Days 30, 60 and 90 show mean ± s.d. of n = 13, 23 and 17 stored dung samples, respectively.

Source data

Full size image

Discussion

Initially, we posed three questions (Fig. 1a) to also guide future ETL endeavours: (1) How can ivermectin-treated cattle complement malaria vector control? (2) What improvement on malaria vector control can be achieved with ivermectin-treated cattle? (3) How can we avoid ecological drawbacks in vulnerable agroecosystems?

Efficacy of a single-depot injection versus repeated injections

Considering our proof-of-concept26, we adjusted the ivermectin formulation and statistical design. The long-acting formulation sustains ivermectin release over at least 6 months. Although the release pattern and dose need adjustment to refine the release period, the potential long-term efficacy meets WHO criteria for malaria transmission control with endectocides35. The slow release of ivermectin from the subcutaneous depot leads to a rate of absorption in blood lower than the rate of elimination from the body. The lipophilicity of ivermectin causes high volumes of distribution in animals and accumulation in target tissues, acting as secondary drug reservoir21. Hence, concentrations of ivermectin and metabolites in subcutaneous fat tissue may remain high in IVM-BEPO-treated cattle, enhancing effectiveness.

The zoophagic proportion in mosquito populations, relevant for residual Plasmodium transmission11, can increase with widespread LLIN use36. In southwestern Burkina Faso, populations of the major vector A. coluzzii show opportunistic blood-feeding behaviour, with over 50% feeding on alternative animal hosts, while their innate feeding preference remains toward humans36. Our approach with ivermectin-treated cattle addresses these zoophagic vector compartments of malaria transmission, which are not yet considered in conventional anthropocentric malaria control programmes35. The long-acting ivermectin formulation mitigates logistical and cost challenges of repeated injections, offering a stronger barrier against malaria transmission by preventing vector population resurgence between treatments. While zooprophylaxis against malaria is a complex concept, leaning towards zoopotentiation depending on transmission context37, our approach capitalizes on existing cattle, rendering their blood toxic to opportunistic and zoophagic mosquitoes while improving cattle health and milk production20.

ETL impact on vector populations could be greater than modelled since ivermectin is also toxic to traits influencing fitness other than survival38. Moreover, opportunistic vectors failing to feed on net-protected humans and seeking alternative hosts are not addressed in our model. Overall, effectiveness depends on the human blood index (HBI), indicating the proportion of mosquito populations feeding on humans. Blood feeding tendencies should therefore be characterized before ETL field deployment, especially since livestock other than cattle could represent alternative hosts. Effectiveness also relies on local livestock density and the treated herd fraction.

Moreover, ETL evaluation should consider veterinary benefits through parasite control and incorporate realistic animal populations into models, including gestating and lactating females, and animals for consumption (meat and milk39). Following Joint FAO/WHO Expert Committee on Food Additives40 guidelines and involving animal owners is essential.

Ecotoxicological considerations

Ivermectin’s capacity to control vectors holds promise for mitigating malaria and improving animal health and productivity15,18. However, vigilance is required concerning potential environmental consequences. Comprehensive research spanning over 30 years has consistently flagged non-target effects of ivermectin and other avermectins on dung and soil biota22,23,24,25,41.

In a preliminary assessment via literature review, plasma concentrations lethal to 50–90% of mosquitoes (Fig. 4) could produce excreted ivermectin levels harmful to most dung insects. However, it is important to recognize limitations of laboratory data and seek validation in field scenarios. Notably, a landscape-field study25 confirmed ivermectin effects on non-target fauna and emphasized the importance of spatio-temporal realism in environmental assessments. For the related endectocide eprinomectin, cattle treated with a long-acting injectable formulation excreted residues for over 25 weeks, suppressing dung-breeding insects and reducing insect diversity28. While much current literature centres on temperate and Mediterranean species, a clear knowledge gap persists concerning West Africa, and few studies assess impacts on (sub-)tropical entomofauna, indicating an urgent research objective.

The use of veterinary ivermectin reportedly carries adverse implications for dung fauna biodiversity and ecosystem functioning42,43,44. Given these potential ecotoxicological impacts, the precautionary principle becomes a guiding factor to proactively address and mitigate risks to biodiversity and ecosystem health in West African agroecosystems. This is especially imperative considering the scarcity of research on the ecological impacts of long-acting injectable ivermectin formulations.

Environmental fate of ivermectin and exposure assessment

Our KD and KOC soil sorption coefficients, the first for Burkina Faso, align with previous reports22,45. Strong, mostly irreversible ivermectin sorption45, emphasizes reduced mobility and availability. However, sorption tests are conducted with fine soil33, potentially overestimating sorption in tropical soils with abundant coarse soil and limited surface area.

Ivermectin is moderately persistent, and dissipation in soil and manure increases under warm, aerobic and humid conditions46,47. In soils, comparable models projected DT50 of 16–67 days at 20 °C and 89–105 days at 6 °C (ref. 47) and from 10–16 days48. In soil–manure mixtures, DT50 ranges from 7–240 days47. These authors concluded that sorption influences dissipation and even under aerobic conditions, ivermectin dissipation in soils is relatively slow47. Our dissipation results in cattle dung indicate persistence under these conditions. Storage during dry seasons may lead to reduced dissipation and could preserve ivermectin until the next cultivation period when dung is utilized as fertilizer. Increasing ivermectin concentrations in dung after 30 days (Fig. 5c) presumably show relative enrichments because readily degradable organic matter decomposes while ivermectin remains. In soils, ivermectin dissipation fluctuates with climatic conditions46,47,48, and its behaviour under sub-Saharan climates remains ambiguous. In West African soils, weaker sorption and higher temperatures might accelerate dissipation. However, data on ivermectin’s environmental fate under ETL scenarios, especially within soil–dung mixtures during seasonal rainfall, are lacking. Consequently, managing ivermectin residues should also account for climatic conditions, encompassing Sahelian, Sudanian and Guinean zones.

In the European Union, environmental risk assessment during veterinary pharmaceutical registration involves estimating a predicted environmental concentration (PEC)32,49. This process utilizes a toolbox spanning Phase I and Phase II assessments. An initial step in ecotoxicological oversight of ETL should establish PECs in dung, soils, dung–soil mixtures, surface waters and sediments (Fig. 6). Here, dung fauna exposure was estimated using measured ivermectin concentrations in dung (Fig. 4b).

Fig. 6: Exposure assessment as the initial step of environmental risk assessment for endectocide-based approaches to control opportunistic and zoophagic malaria vectors.
figure 6

a, Equation (1) for PECsoil initial is based on ref. 49. Exemplary parameters for a 211-day pasture scenario: D = 0.0019, Ad = 211, BW = 160, SD = 9.5, Fh = 1. Phase I screens products and includes initial PEC estimations, while Phase II conducts detailed assessments using refined data, field studies and modelling. Fh, the treated herd fraction, is notably relevant for refugia51 incorporation to manage resistance in vectors and parasites. b, Exposure assessment of endectocides used for vector control, such as ivermectin, must consider various pathways, including direct excretion or dung use as fertilizer for potential drug residue transfer into agricultural soils. Soil contamination is also conceivable on adjacent unmanaged soils. Contamination of surface water and sediment can occur through run-off, erosion or direct excretion, potentially exposing aquatic fauna and larval vectors to ivermectin. Figure 6b was created with BioRender.com.

Full size image

To transfer exposure from dung to soil, an exemplary PECsoil initial was calculated at 4.86 µg kg−1 for the long-acting formulation dosed at 2.4 mg kg−1. This compares to reported PECs for pasture grazing22. However, estimating and refining soil PECs varies with livestock treatment and data availability22,49, making assessments challenging without defined treatment scenarios. The longer release of long-acting formulations is especially critical. Since dung is a valuable agricultural resource in many malaria endemic zones where breeding and agricultural practices are integrated, responsible dung management remains imperative.

Environmental risk mitigation

Ecotoxicological guidance for ivermectin-based vector control should encompass exposure assessment, effect assessment, risk assessment and risk mitigation measures (RMMs). Integrating these tools with vector control measures can yield synergies and improve cost effectiveness. Should ivermectin contaminate puddles or surface waters, minor degradation is anticipated, but more likely rapid dissipation and sorption into the sediment22,45. Ivermectin degradation through photolysis22 in dung patties or soils is limited to upper layers. In our external storage experiment, samples were shielded from sunlight (Extended Data Fig. 1c). However, given common agricultural practices of incorporating dung into soils, substantial sunlight exposure for excreted drugs is unlikely.

For veterinary pharmaceuticals, RMMs to reduce environmental risks for non-target fauna are already conceptualized32,50. Importantly, RMMs synergize with measures to reduce resistance in livestock parasites and facilitate sustainable parasite control. Reducing ivermectin distribution with RMMs also limits contact for adult and larval vectors, mitigating resistance selection pressure. Modifying previous recommendations50, RMMs for ETL should respect local conditions and could be guided by these principles:

  1. (1)

    Prevent drug entry into surface waters.

  2. (2)

    Corral treated animals whenever possible.

  3. (3)

    Leave fractions of the herd periodically untreated to create refugia51. This allows safe consumption (meat and milk) and reduces resistance pressure and non-target fauna risks.

  4. (4)

    Implement measures of animal performance and routine diagnostics for early detection of ivermectin-resistant parasites, aligning livestock health and vector control goals while minimizing drug use and environmental exposure.

  5. (5)

    Capitalize on existing agro-pastoral systems of dung collection and store dung under conditions that promote drug degradation.

  6. (6)

    Avoid repeated application of dung with drug residues as fertilizer if long-term effects on non-target fauna are unknown.

In our study region of Burkina Faso, manure storage pits called ‘fosses fumières’ are common (Extended Data Fig. 1a). Pits are excavated near settlements and filled with dung, crop residues and kitchen waste during dry seasons. They are later emptied and the mixture is spread as fertilizer in the following cultivation period. We encourage exploring composting conditions in these pits for potential ivermectin degradation. However, ivermectin residues and their associated toxicity may not completely dissipate during storage52. Adapting RMMs designed for temperate climates and pasture scenarios while respecting agricultural practices in West Africa remains fundamental. This also requires training opportunities for herders and veterinarians.

Side effects on plants

Regarding potential effects of ivermectin on crops, limited research exploring phytotoxicity is available for real-world environments. Some reports24,53 considered that via livestock, ivermectin may affect plants. Since this concern is mostly undiscussed for West Africa, caution is appropriate. For Burkina Faso, potential effects on local varieties of corn, sorghum, cotton or millet (Extended Data Fig. 2b) should be investigated under field scenarios with tropical soils.

Potential ivermectin-induced mosquito resistance

Human and animal intakes can alter volatile plume signatures, influencing Anopheles vector host recognition54. With ETL, changed animal odour may affect host attractiveness or repellency, raising ethical concerns about redirecting opportunistic mosquitoes toward humans. Although ETL targets adult female mosquitoes, resistance to lethal ivermectin concentrations in populations may develop through selection pressures. These arise from adult females ingesting ivermectin during blood feeding and larvae exposure to ivermectin-contaminated dung in surface waters. Large-scale ETL deployments should also incorporate phenotypic diagnostic tools for on-field mosquito resistance monitoring55.

Notably, our approach inherently incorporates refugia: not treating entire herds creates a refuge for susceptible parasites, aligning with known strategies for mitigating anthelmintic resistance51. This fortifies the efficacy and sustainability of ETL solutions.

Towards sustainability in vector control

The One Health Joint Plan of Action (2022–2026) by FAO, UNEP, WHO and WOAH consolidates that ‘health of the environment is a critical foundation for the health and well-being of humans, animals and plants’4. Aligned with a modern, non-anthropocentric view of human health, this emphasizes the need for ecological equity56. Therefore, vector control approaches based on endectocides must equally consider socio-economic realities, as well as human, animal and environmental health dimensions.

Ecotoxicological guidance addresses environmental impacts of ETL and underscores the importance of biodiversity, soil health and sustainable resource management. Integrating traditional organic waste management (for example, fosses fumières) with scientific advancements could optimize composting and drug degradation. This engages communities and strengthens nutrient cycles and soil resilience against erosion and climate change. Protecting non-target fauna and enhancing soil carbon balances can mitigate risks to biodiversity and ecosystem services. Echoing Rachel Carson’s sentiments in ‘Silent Spring’ on health risks of insecticides, ‘…prevention is the imperative need’1 is a notion equally applicable to environmental health.

We demonstrate that long-acting ivermectin formulations for cattle can transform malaria vector control. This strategy meets WHO efficacy criteria, remarkably improves animal health and potentially extends control over other arthropod-borne zoonotic diseases.

Our unique multidisciplinary approach, embedding comprehensive modelling, an innovative formulation and ecotoxicological insights promises a new aspect of sustainable vector control. Before widespread implementation, potential ecotoxicological risks associated with ivermectin residues in dung demand careful consideration. We recommend: (1) studying ivermectin degradation in manure pits; (2) characterizing the entomofauna of West African agroecosystems; and (3) initiating ecotoxicological routines in West Africa to identify potential ecological consequences.

In addition, we propose actionable, bottom-up and region-specific risk mitigation strategies. If endectocide-treated cattle become routine in vector control, interdisciplinary research, generational knowledge and ecotoxicological insights must converge to create locally adapted, environmentally mindful solutions. Bridging the research–policy gap is essential to establishing necessary regulatory frameworks to translate research into sustainable action.

Methods

Layout of the study design

We randomly assigned 24 male cattle into three study arms: 8 were treated with monthly injections of a commercial IVOMEC-D ivermectin formulation, 8 received the long-acting IVM-BEPO ivermectin formulation and 8 had no ivermectin treatment (control). An animal from each arm was put aside as spare animal (data from 7 cattle per arm were then further collected). Doses were based on cattle body weight (bw) on injection day: 0.4 mg kg−1 bw for monthly IVOMEC-D injections and 2.4 mg kg−1 bw for a single IVM-BEPO injection at study start. These doses were doubled on the basis of our previous data26,27 to increase efficacy and remanence. Injections were administered by veterinary technicians. Study duration was 34 weeks (Fig. 1). We sampled cattle blood plasma and dung to determine ivermectin pharmacokinetics and performed skin-feeding assays on cattle to assess mosquito mortality as a proxy for formulation efficacy. Selected dung samples were stored to determine ivermectin degradation. Soils from southwestern Burkina Faso were sampled to assess ivermectin sorption.

Ivermectin formulations, dosage and injections

The long-acting injectable formulation was built on BEPO technology57, an injectable in situ-forming depot technology based on biodegradable copolymers. Following subcutaneous injection, the block copolymers precipitate upon contact with body fluids, thereby entrapping the therapeutic molecule within the formed polymeric solid matrix (the depot). The depot progressively bioresorbs while delivering the active pharmaceutical ingredient with the desired pharmacokinetics. The same technology has been used to assess the macro- and microfilaricidal efficacy of a 12-month ivermectin formulation prototype against Onchocerca ochengi58 and as a proof-of-concept, tested against A. coluzzii27.

The IVM-BEPO formulation was selected from a preliminary study27 and composed of 45% (w/w) copolymer comprising a tri-block PLA97-PEG45-PLA97 and a di-block mPEG45-PLA130, 5% (w/w) ivermectin and 50% (w/w) dimethyl sulfoxide (DMSO). The formulation was designed to release ivermectin for at least 6 months. Before preparation, the tri- and di-block copolymers of the formulation were preliminarily dissolved overnight in DMSO (Procipient, Gaylord Chemical) at room temperature and under continuous stirring. Then, ivermectin (Fagron) was added to the polymer solution until complete dissolution. The formulation was sterile filtered using 0.2 µm filters (Minisart SRP 15, Sartorius) and then administered to cattle according to their bw (2.4 mg of ivermectin kg−1 bw, that is, 48 mg of formulation kg−1 bw) with a hypodermic syringe capped with a 16-gauge needle. Volumes were adjusted to the weight each cattle was expected to reach 3 months after the injection (mid experimental time) to consider potential dilution of the product due to animal growth. Expected weight gains were calculated from weighing performed before treatments in January and April 2019 (Supplementary Table 8). A 25% increase in 3 months was estimated for cattle below 130 kg, 15% for cattle weighing 130–170 kg and 10% for the heaviest cattle (>170 kg). Weights and corresponding volumes of the long-acting formulation injected per cattle and per treatment were calculated accordingly (Supplementary Table 9). The long-acting formulation was imported into Burkina Faso under clearance provided by the ‘Direction Générale des Services Vétérinaires’ of Burkina Faso (Visa from the ‘Direction de la Santé Publique Vétérinaire et de la Législation’ issued 11 January 2019).

The IVOMEC-D injectable veterinary formulation was used at a dose of 0.4 mg kg−1 bw and injected each month subcutaneously (Fig. 1b) using a hypodermic syringe capped with an 18-gauge needle. Cattle were weighed before each treatment to adjust injection volumes (Supplementary Table 10). Remaining cattle from the experiment were weighed each month as well. Weight gain of all cattle was examined to ensure their well-being and to identify potential treatment effects (generalized linear modelling; Supplementary Fig. 3). For both formulations, the ivermectin dose was selected on the basis of experiences from our previous studies26,27.

Ethics oversight

The study received approval from the CIRDES ethics committee under letter no. 15/CE-CIRDES/16-10-2018.

Cattle care, blood and dung sampling

All cattle were crossbred Soudaneese Fulani Zebu × Baoulé Taurine and were kept at the Centre International de Recherche-Développement sur l’Elevage en zone Subhumide (CIRDES) in Bobo-Dioulasso, Burkina Faso. One month before the experiment started, animals were moved into a net-protected stable and received a veterinary examination and prophylactic treatment with diminazene aceturate (Berenil2000, 3.5 mg kg−1 bw) against trypanosomiasis and albendazole (one bolus Benzal 2,500 mg per animal) against gastrointestinal parasites. The latter was chosen over ivermectin to avoid cross-contamination in later sampling. At the beginning of the study, mean cattle age was 3 years and mean weight was 122 kg (cattle were weighed monthly). Stable openings were closed by fine-meshed filtering tissue to protect cattle from insects and outside animals. Cattle were checked daily by cowherds for signs of parasite infestation or disease symptoms and before each mosquito blood-feeding experiment by veterinary engineers (engineers checked for clinical signs of aphthae, cough, nasal discharge, slime, anaemia, recumbent position, anorexia, diarrhoea, fever, scabies). There was no incidence of any symptoms during the whole experiment. Unfortunately, an animal from the IVM-BEPO group died following an accident with a stable box barrier that fell on its head at night. This happened 3 months after treatment. The animal was replaced by the spared animal from the same treatment.

According to local recommendations59, cattle were fed twice daily with rice straw and cotton cake (5 kg and 1 kg in total per animal, respectively). Water and a mineral complement block were offered ad libitum. Blood samples were collected from the jugular vein with 5 ml heparin tubes at the timepoints given in Fig. 1b. Plasma was immediately collected after sample centrifugation at 1,500 g for 15 min, and stored at −20 °C until further processing. Dung samples were collected directly from the cattle rectum or on freshly deposited dung synchronously with plasma. We constituted samples (30 g each) for ivermectin dosage on fresh dung and on dung after storage. For IVM-BEPO dung samples, fresh dung was distributed in plastic vials in two groups and according to the intended storage duration (30, 60 or 90 days). One group was inside the insectary (to mimic steady, fresh and humid conditions) and the other outdoors (exposed to dry season conditions). Plastic vials were covered with a piece of bed net (Extended Data Fig. 1c). Environmental conditions in the insectary were measured daily and were 26 ± 2 °C and 75 ± 5% relative humidity. For external storage, daily and nightly temperatures during the respective storage periods are given in Supplementary Table 11 (from ref. 60). Mean outdoor temperatures ranged from 32.1–39.6 °C during the day and from 17.9–22.8 °C at night. After fresh sampling or storage, dung samples were kept at −20 °C for a minimum of 24 h, freeze dried for 24 h, ground for 15 min using a mortar and pestle, disposed into labelled plastic bags and stored at −20 °C before chemical analysis.

Mosquito protocol

An A. coluzzii mosquito colony was established between February and March 2019 in the insectary of CIRDES in collaboration with the Institut de Recherche en Sciences de la Santé. Wild blood-fed females were collected in dwellings of the Kou Valley, 50 km northwest of Bobo-Dioulasso in southwestern Burkina Faso. Mosquito breeding and direct skin-feeding assays on cattle (Extended Data Fig. 2a,b) followed a previously reported procedure26, except that mosquitoes received a 10% glucose feeding solution instead of 5%. Exposure took place in 20 instances, including one pre-treatment to test for potential cattle confounding effects on mosquito survival (Fig. 1b). All 24 cattle were included in this pre-treatment study. Overall survival was assessed in 15,637 fully engorged females that were followed from days 1‒30 post feeding. The targeted number of females per cattle and time point was 40, distributed in lots of 10 in 4 white cardboard cages (V = 200 cm3) covered with an untreated mosquito net fixed using a rubber band. Each day, the number of dead females was counted and surviving females at the end of the experiment were also counted. Mosquito survival analysis data, sample sizes and the R code are available on GitHub at https://github.com/angeliqueporciani/ANIVERMATE.

Soil sampling and characterization

Soil samples were taken in Tuy Province, southwestern Burkina Faso, in the villages of Sébédougou, Waly and Kari (Extended Data Fig. 2a). We aimed to sample a representative variety of Burkinabé arable soils under comparable climatic conditions. At each village, we sampled ten fields currently used for regional crops. On each field, five disturbed soil samples were taken with an auger at 0–15 cm depth and mixed for a composite sample. Soil properties were derived from field notes and harmonized into WRB soil classification61. Laboratory analyses were performed for soil pH value (0.01 M CaCl2), electrical conductivity, coarse-soil fraction, particle size distribution and C:N ratio. Carbonate content was estimated following the semi-quantitative preparation step of EN ISO 10693:2014, which indicated no inorganic carbon present. Thus, Ctotal from C:N measurements with a UNICUBE trace CN elemental analyser was used as a substitute for organic carbon (Corg). A complete characterization of sampled locations and soils is given in Supplementary Tables 12 and 13. Furthermore, six soil profiles were described in the sampled locations and brief descriptions are provided in Supplementary Figs. 4–9. Soil samples were transported under clearance provided by the RP Gießen, Pflanzenschutzdienst, Germany (issued 26 September 2019) and the Direction de la protection des végétaux et du contrôle des végétaux alimentaires et des Pesticides, Burkina Faso (issued 5 November 2019).

Environmental studies

To address the environmental fate of ivermectin after excretion, we conducted soil sorption studies and monitored ivermectin degradation in stored dung samples over 90 days. Sorption experiments and derivation of linear distribution coefficients KD and KOC were based on ref. 33 and published methods45,46. Correlation (Pearson’s r) between linear KD sorption coefficients and soil properties was described with the OriginPro 2022 application Correlation Plot (v.1.31). Storage of dung samples was conducted in Burkina Faso under external, semi-field conditions and internal, laboratory storage (see above). These samples were eventually analysed for ivermectin after 30, 60 and 90 days of storage, and concentrations were compared (two-sided Tukey pairwise mean comparison, in OriginPro, v.2022 (64 bit)) to each other and their respective day-zero concentration. Concentrations in stored dung were normally distributed (Shapiro–Wilk, α = 0.05, all P > 0.05). Next, we applied simple first-order kinetics to approximate future ivermectin concentration changes in stored dung, using the results from the internal storage experiment. Underlying methods are elaborated upon in ref. 47. The resulting degradation constant, or rate constant, k1, was used to estimate ivermectin dissipation over time. The time to reach 50% of the initial concentration is expressed as DT50 = ln 2/k1, while the corresponding DT90 equals ln 10/k1 (ref. 47). OriginPro v.2022 (64 bit) was used to create Figs. 2a–c, 3a, 4a–b and 5a–d.

Chemical ivermectin analysis

Extraction and quantification of ivermectin were based on existing techniques46,62 and harmonized for plasma and dung samples in this study. For quantitative determination, we used high-performance liquid chromatography (HPLC) with fluorescence detection after derivatization. The complete protocols are listed in Supplementary Table 15 and the accompanying supplemental text. In essence, samples (thawed plasma, freeze-dried dung) were fortified with an internal surrogate standard (doramectin) and extracted with acetonitrile. After ultrasound-assisted extraction and then centrifugation at 3,865 g, an aliquot of each supernatant was evaporated, reconstituted in acetonitrile and filtered. For HPLC fluorescence detection, ivermectin was derivatized with N-methylimidazole/acetonitrile (1:1 v/v), triethylamine, trifluoroacetic anhydride/acetonitrile (1:1 v/v) and trifluoroacetic acid according to reported procedure46. Samples were quantified on an Agilent 1200 HPLC system with gradient elution (acetonitrile and purified water) on a reverse-phase, C18 column. Fluorescence detector wavelengths were 364 nm for excitation and 463 nm for emission. Extraction recoveries (mean ± s.d.) of the surrogate doramectin were 106.5 ± 21% in plasma and 103.2 ± 19.7% in dung samples. Each plasma or dung sample was divided into two technical replicates that were extracted and measured individually. Analytical limits of detection (LOD) and quantification (LOQ) were calculated as described in ref. 45: LOD = 1.47 ng ml−1 in plasma, 5.05 ng g−1 dw in dung; LOQ = 4.47 ng ml−1 in plasma and 15.31 ng g−1 dw in dung. Measured concentrations <LOQ were replaced with half the LOQ for further calculations. Data analysis was performed with ChemStation for LC 3D systems Rev. B.04.01 [481] and Microsoft Excel for Office 365.

Statistical analysis, mosquito survival and modelling

All statistical analyses and modelling were performed using R v.4.0.1 (18 May 2021, platform x86_64-apple-darwin17.0 (64 bit), ref. 63). The data and the R code are available in the referred repository, and the general methodology is described in the Supplementary Information Modelling chapter.

Pharmacokinetic parameters for plasma and dung concentration (maximum concentration (Cmax; peak plasma level), time to reach maximum concentration (Tmax), apparent elimination half-life (half-life; time for plasma concentration to be reduced by 50%), and area under the curve (AUC; total drug exposure over time)) were estimated with noncompartmental analysis using the package PKNCA (v.0.9.5, ref. 64). For the commercial formulation, parameters were estimated using data from the first 28 days, whereas for the long-acting formulation, data until 211 days after injection were used.

Our study was designed to detect hazard ratios from 1.5 and upwards, with a power of 0.9 and a type I error rate of 0.05. The sample size needed to detect this size effect was 171 individuals, which we obtained by exposing at least 40 mosquitoes on each of the 7 cattle of each arm (at least 280 mosquitoes per arm). In comparison, a sample size of 16 mosquitoes is needed to detect a hazard ratio of 4, which meets WHO criteria35 (hazard ratio ≥ 4) of endectocide products for malaria transmission control.

Kaplan–Meier survival curves and a Cox proportional hazards mixed model with cattle as a random effect were used to investigate mosquito survival differences between cattle clustered by experimental arm. Median survival times with 95% confidence intervals (CI) were plotted. Hazard ratios with associated P values were estimated before treatment and at each time point where mosquitoes were exposed for direct skin-feeding assays during the treatment efficacy follow-up (Supplementary Table 4).

We further explored the efficacy of the two formulations by considering the probability that a mosquito ingesting a blood meal containing ivermectin dies before becoming infected with P. falciparum sporozoites. The dose effect of ivermectin concentration on mosquito survival probability until a given day post ingestion of blood from treated animals (7, 10 or 13 days) was assessed using a multivariate log-logistic regression model with four input parameters. The lethal ivermectin concentrations inducing 7, 10 and 13-day cumulative LC50 and LC90 mosquito mortalities were estimated for each scenario and each treatment (Supplementary Table 5).

A modelling approach was developed to predict, in the field, the efficiency of ivermectin formulations in decreasing vector populations that transmit P. falciparum, with emphasis on ivermectin efficiency against infectious vectors. Efficiency was predicted under different field contexts relative to the proportion of cattle in a human–cattle host population, the long-lasting insecticide-impregnated net usage by human hosts (LLIN use proportion) and the intrinsic vector feeding preference (as measured experimentally using a dual-choice olfactometer). Four mathematical models were combined: (1) a generalized additive model and a one-compartment pharmacokinetic model describing the ivermectin cattle plasma concentration dynamics; (2) a Cox proportional hazards model to describe how this dynamic impacts mosquito survival; (3) a deterministic mosquito P. falciparum transmission model (the susceptible-exposed-infectious malaria transmission model from ref. 65), modified to take into account 2 host species (cattle and human) and 2 vector control interventions (the use of LLINs to protect humans and the injection of ivermectin formulations into cattle hosts); and (4) a vector behaviour model to take into account the mosquito intrinsic feeding preference and the probability of dying when encountering an LLIN. These models were all fed using parameters from the present study and parameters taken from the literature. The model descriptions, equations and parameters with the corresponding values used to feed the models are given and detailed in the Supplementary Information Modelling chapter.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.