Modelling the impact of ivermectin-based optimal strategies on malaria control: the role of formulation, coverage, and mosquitocidal efficacy timing

Summary

Background

Despite wide use of insecticidal nets and indoor spraying, residual malaria transmission persists due to insecticide resistance and mosquito behaviours like early and outdoor biting. Ivermectin (IVM) can kill mosquitoes after blood feeding, regardless of the timing or location of the bite. However, oral IVM's short efficacy duration and required frequent dosing limits epidemiological impact. Using mosquitocidal efficacy data from a long-acting injectable IVM formulation (BEPO® technology), we modelled its effect on malaria transmission, comparing it with oral regimens evaluated in the fields.

Methods

We developed a mathematical model incorporating IVM's pharmacodynamic persistence and stratifying humans by age and sex. Three regimens were evaluated: (i) monthly oral IVM at 3 × 300 μg/kg for 4 months, (ii) monthly oral IVM at 1 × 400 μg/kg for 3 months, and (iii) long-acting BEPO® injection at 0.6 or 1.0 mg/kg. The model includes post-exposure effects on mosquitoes and the target population excludes women of childbearing age or pregnant women, and children under 15 kg. Mosquitocidal efficacy data were derived from published human oral studies and from preclinical studies in livestock (cattle) for injectable formulations.

Findings

Campaigns delivering ivermectin are projected to lower clinical prevalence, with slightly greater benefits in Burkina Faso–like seasonal settings when aligned with the transmission peak. Under perennial transmission and 50–90% target population coverage, oral regimens reduced total cases (all ages) by 7–11%, 9–14%, and 24–38% over one year, depending on the regimen and the data source. A single long-acting injectable IVM campaign achieved 16–29%, rising to 33–57%, 48–74%, and 59–78% with two, three, and four campaigns, respectively.

Interpretation

Long-acting injectable ivermectin offers increased and sustained efficacy and simpler delivery to tackle residual malaria transmission. Phase 1 trials are timely and warranted.

Funding

The African Consortium in Modelling for Effective Vector Control (ACoMVeC) project INV-047049, IMPACT (Unitaid) and Institut de Recherche pour le Développement (IRD).

Research in context.

Evidence before this study

We conducted a systematic literature review using PubMed and Google Scholar up to January 26, 2026, to identify studies employing mathematical modelling to assess the impact of ivermectin (IVM) administered to humans on malaria prevalence and incidence. The search strategy combined the keywords (“ivermectin” OR “endectocide”) AND “malaria” AND (“modeling” OR “modelling”), with no restrictions on language or publication date. Additional sources, including unpublished literature, were also consulted. Four published modelling studies were identified. Selvaraj et al. (2019) showed that substantial reductions in malaria burden (30%) were achievable only with IVM efficacy durations exceeding 60 days, especially when combined with seasonal malaria chemoprevention (SMC). Slater et al. (2020) predicted that oral IVM (1 × 400 μg/kg or 3 × 300 μg/kg administered monthly for three consecutive months) could reduce malaria incidence by up to 70% in settings with highly seasonal, moderate transmission. Yeo et al. (2022) developed a physiologically based pharmacokinetic (PBPK) model to reproduce plasma exposure profiles and evaluate alternative regimens, including complex formulations. Dongmo et al. (2024) introduced a model structured by IVM exposure time and used mathematical analysis to identify optimal deployment strategies for reducing prevalence, supported by a detailed mathematical analysis.

Added value of this study

We present a mathematical model that jointly incorporates human sex, chronological age, and time since IVM exposure as continuous variables in both humans and mosquitoes. This approach enables a detailed and realistic assessment of IVM intervention effects by capturing heterogeneity in drug efficacy across the population. To our knowledge, no prior model has evaluated a long-acting injectable IVM formulation based on experimental pharmacodynamic data from livestock studies. The framework is flexible, allowing comparison between oral and injectable formulations and doses, and assessing epidemiological scenarios involving either i) standard or accelerated loss of ivermectin efficacy, ii) exclusion of women of childbearing age or only pregnant women from treatment, and iii) examining outcomes in young children (5–15 years) relative to the entire population, highlighting the added benefit or reduced impact of treatment in this subgroup.

Implications of all the available evidence

Our analysis offers a robust basis for comparing IVM formulations across malaria transmission settings, explicitly accounting for demographic heterogeneity and IVM effects on mosquito mortality. It highlights the greater potential of the long-acting injectable formulation over oral regimens by identifying optimal combinations of coverage and efficacy duration. As such formulation already exists-with promising safety profiles and simplified logistics–our findings support its integration into malaria control efforts. Evaluating pharmacokinetics and efficacy in humans is a critical next step. Overlooking this opportunity could delay progress on residual malaria control. Together with IVM's known benefits for other parasitic diseases, our results strengthen the case for advancing to a phase 1 clinical trial.

Introduction

Substantial progress in malaria control over the past two decades–largely driven by global eradication efforts, the scale-up of insecticide-treated nets (ITNs), and widespread use of artemisinin-based combination therapies (ACTs)–has led to marked reductions in disease burden.^1,2 However, these gains are increasingly threatened by physiological resistance to insecticides and antimalarial drugs, alongside behavioural adaptations in mosquito vectors.³ While next generation long-lasting insecticidal nets (LLINs) incorporating synergistic compounds or new insecticides offer enhanced protection,⁴ they remain ineffective against outdoor- or early-biting mosquitoes, thereby contributing to residual transmission.⁵ The recent resurgence in malaria cases underscores the urgent need for complementary vector control tools that address the full spectrum of Anopheles biting behaviours, whether innate or shaped by control-driven selection pressures.⁶

Systemic endectocides offer a promising complementary strategy by targeting mosquitoes regardless of biting time or location.⁷ Among these, ivermectin (IVM) is the most widely studied candidate. It is already deployed in mass drug administration (MDA) campaigns against onchocerciasis and lymphatic filariasis,⁸ and is lethal to mosquitoes that ingest it via blood meals from treated humans or animals.^9,10 Unlike conventional insecticides, IVM is effective against resistant mosquito populations and demonstrates activity across all known Anopheles species.¹¹ Standard oral doses (150–200 μg/kg) result in plasma concentrations that kill Anopheles mosquitoes for up to one week.¹² A pioneering trial administering this dose every three weeks led to a significant reduction in clinical malaria incidence in children under five.¹³ Phase 1 clinical studies have shown that higher dosing regimens, such as 1 × 400 μg/kg or 3 × 300 μg/kg, are well tolerated and have a favourable safety profile with mosquitocidal efficacy in laboratory settings extended approximately to 7–10 days¹⁴ or 14–21 days.¹⁵ Modelling studies have estimated that such regimens could reduce malaria incidence by up to 70% in regions with moderate to highly seasonal transmission.¹⁶

Despite these promising data, phase 3 trials have failed to demonstrate a significant benefit of monthly repeated IVM MDA (3 × 300 μg/kg) for three or 4 consecutive months when combined with, respectively, dihydroartemisinin–piperaquine (DHAP)¹⁷ or seasonal malaria chemoprevention (SMC).¹⁸ Possible explanations to discrepancies between model projection and field outcomes include vector bionomics,¹⁹ laboratory overestimation of IVM's mosquitocidal effect, and unexpectedly rapid clearance from the host. Indeed, RIMDAMAL II reported that nearly 50% of participants had undetectable plasma levels by day 14 post-treatment,¹⁸ highlighting the need for formulations that sustain mosquitocidal levels longer. Two phase 3 randomised controlled trials conducted by the BOHEMIA consortium in Kenya and Mozambique evaluated monthly administration of oral ivermectin only at a dose of 1 × 400 μg/kg over three consecutive months during the rainy season.²⁰ In Kenya, this regimen resulted in a 26% reduction in malaria incidence among children aged 5–15 years compared with the control arm.²¹ In contrast, no significant reduction in malaria incidence was observed in Mozambique under the same intervention strategy.²² Challenges in delivery logistics, resulting in suboptimal population coverage, have been cited as an explanation for the lack of effectiveness.

Veterinary injectable formulations of IVM have already demonstrated extended efficacy (1–2 months) against Anopheles species in Asia,²³ and more recent formulations have shown mosquitocidal activity lasting over six months.²⁴ Although previous individual-based models have explored the impact of long-acting endectocides by assuming efficacy durations of 14–90 days,²⁵ these models lacked integration of empirical pharmacokinetic and entomological data-limiting their relevance to real-world applications.

In this study, we present a structured modelling framework to assess the impact of both long-acting injectable administered to cattle only and oral formulations of IVM administered to humans on malaria transmission. The model integrates (i) the chronological age and sex of the human host population, (ii) the post IVM exposure time structure in humans, capturing pharmacodynamic effects and governing its impact on vector mortality dynamics, and (iii) the time since parasite infection in mosquitoes—accounting for transmission potential during parasite development. Key innovations of the proposed framework include: (i) the explicit incorporation of the duration of IVM efficacy, termed the formulation's “pharmacodynamic persistence”, in both mosquitoes and humans exposed to IVM; and (ii) its flexibility to evaluate a range of epidemiologically relevant scenarios, taking into account variations in efficacy loss, differential treatment of specific subpopulations, and added benefit or reduced impact across transmission settings and formulation types.

The long-acting IVM formulation (LAIF) is the lead formulation mdc-STM-001 from the BEPO® technology selected during preclinical cattle studies.²⁶ Mosquito survival data were obtained through experimental designs in which cattle received doses of 0.6–1.0 mg/kg and mosquitoes originated from field populations characterised by widespread insecticide resistance.²⁶ The LAIF demonstrated mosquitocidal efficacy lasting at least two months. Survival data for oral regimens were drawn from published sources.^14,15 We used mosquito survival data, which constitute the only empirical evidence available across published experiments assessing the mosquitocidal activity of IVM over relevant time windows. The analysis focuses on 4-day cumulative mosquito mortality to capture the acute effect of ivermectin, as previously considered by.²⁵ While their study used a 3-day window, our assessment²⁶ indicated that ivermectin induced mortality occurs predominantly within 2- to 3-days, justifying the use of a 4-day cumulative measure. Instead of relying on a fixed crude hazard ratio (CHR) threshold to define IVM efficacy, we optimised this parameter to estimate the duration beyond which the drug no longer provides a meaningful effect, thereby guiding the scheduling of IVM campaigns. This optimised duration is not unique and may vary slightly between formulations due to differences in pharmacokinetics and routes of administration.

We use this model to quantify both the population-level and age-specific impacts of IVM, calibrated with epidemiological data from Burkina Faso.²⁷ In addition, injectable and oral formulations are compared in terms of their effectiveness in reducing clinical malaria cases over a one-year period. The analysis further examines targeted scenarios, such as the exclusion of women of childbearing age or only pregnant women from treatments, and compares outcomes in young children relative to the entire population. We also assess formulation performance in different transmission settings (perennial versus seasonal), and evaluate the effect of treatment timing relative to the transmission peak. Overall, our results provide actionable insights to guide the optimisation of IVM-based malaria control under realistic operational scenarios.

Methods

Malaria transmission mathematical model

We developed a mathematical model of malaria transmission that incorporates IVM-based interventions, structuring the human population by infection status, age, sex, and time since IVM exposure, and mosquitoes by infection status, time since parasite infection, and the timing of their contact with treated humans. This structure captures key biological and behavioural differences that influence infection risk and mosquito–host contact. We use sex (male/female) as the model stratum. References to ‘women’ (e.g., pregnant women; women of childbearing age) follow programmatic eligibility language and correspond to female sex individuals. The model reflects parasite transmission via the mosquito gonotrophic cycle and infection from both symptomatic and asymptomatic humans. A key feature of the model is the inclusion of interactions between uninfected mosquitoes and IVM-treated but uninfected humans, allowing for potential drug uptake during blood feeding. This process depends on the time elapsed since human drug exposure, an often overlooked pathway in previous modelling studies. By distinguishing between transmission and pharmacological exposure pathways, the model enables evaluation of targeted IVM deployment strategies and estimation of coverage thresholds needed to reduce transmission. Refer to Fig. 1 for a graphical outline of the model. Full mathematical details are provided in Supplementary A.

Fig. 1.

Compartmental representation of the IVM-based mathematical model. Ivermectin is administered to a target human population, who upon treatment enters a state of exposure to the drug. As time progresses (τ), the effectiveness of the drug declines, gradually reducing its physiological impact. On the vector side, the model also accounts for the time since infection (σ), which influences the progression within the infected mosquito class. In parallel, mosquitoes become exposed to ivermectin through blood meals taken from treated individuals. This contact results in a transition to an ivermectin-exposed class of mosquitoes, characterised by an increased mortality rate that depends on the time elapsed since the human host had received ivermectin at the moment of the blood meal (τ). Other transmission processes, including the well-established progression of humans from susceptibility to infection, recovery, and gradual loss of immunity, as well as the susceptible–infected–infectious dynamics in mosquitoes shaped by the extrinsic incubation period, together with background and disease-induced mortality and population turnover, are incorporated into the model, although they are not explicitly described here. For a more detailed compartmental diagram, see Figure S1.

Long-acting injectable formulation

The long-acting IVM formulation (LAIF named mdc-STM-001) used in this study was formulated using the patented BEPO® injectable platform,²⁸ combining IVM with biocompatible polymers and a solvent to ensure injectability. Using cattle model, it was selected during preclinical studies as the optimal candidate formulation for future human use, based on combined pharmacokinetic and mosquitocidal properties.²⁶ The model includes datasets related to three case scenarios: (i) mdc-STM-001-0.6-vk5: cattle injected at the dose 0.6 mg/kg, direct skin feeding experiments with Anopheles coluzzii mosquitoes derived from VK5, a village near Bobo-Dioulasso, in the Kou Valley, a region with intense insecticide use and documented vectors resistance²⁹; (ii) mdc-STM-001-0.6-kis identical to (i) with mosquitoes from the Anopheles gambiae s.s. Kisumu strain, fully susceptible to insecticides^30,31; and (iii) mdc-STM-001-1.0: cattle injected at the 1.0 mg/kg dose with wild Anopheles coluzzii mosquitoes from VK5 village.

Oral formulation

Data were retrieved from two distinct phase 1 clinical trials where adult men were administered IVM tablets from commercial formulations (Table 1). Participants received weight-adjusted IVM doses under three regimens: a 400 μg/kg single dose administered once (1 × 400 μg/kg,¹⁴) and a 300 μg/kg dose, administered daily for three consecutive days (3 × 300 μg/kg), either alone¹⁴ or with adjunct DHAP the first day.¹⁵ All oral regimens were tested using Anopheles gambiae s.s. strain mosquitoes, fully susceptible to insecticides. After feeding on treated hosts, mosquito survival was monitored from days 2–28. Injectable formulation assessments occurred from 2 to up to 126 days post-treatment. Mosquito mortality was analysed by Kaplan–Meier curves, providing discrete timepoint estimates but lacking continuous resolution.²⁶ Our model predicts daily IVM-induced mosquitocidal efficacy, incorporating continuous variables for time since exposure in humans, which shape its downstream effect on mosquitoes and enable detailed, time-dependent evaluation of intervention effectiveness.

Table 1.

Main information of the control strategies.

Case scenario	IVM dose regimen (mg/kg)	IVM formulation	Target human population	Number of cycles: nc	Mosquito species
BOHEMIA	1 × 0.4	Oral	Males older than 5 years.	3	An. gambiae s.s (S), Kilifi
RIMDAMAL IIS	3 × 0.3			4	An. gambiae s.s (S)
RIMDAMAL IIK	3 × 0.3			4	An. gambiae s.s (S), Kilifi
mdc-STM-001-0.6-vk5	1 × 0.6	LAIF	Females aged 5–15 or over 45 years.	{1,2,3,4}	An. coluzzii (R), VK5
mdc-STM-001-0.6-kis	1 × 0.6			{1,2,3,4}	An. gambiae s.s (S), Kis
mdc-STM-001-1.0	1 × 1.0			{1,2,3,4}	An. coluzzii VK5 (R), field

Target populations correspond to those excluding all women of childbearing age, serving as the reference scenario; outcomes are primarily evaluated for the entire population. Scenarios excluding only pregnant women and outcomes in young children (5–15 years) are also addressed. Hereafter, LAIF denotes long-acting injectable formulation. RIMDAMAL II_S refers to the administration schedule used in the RIMDAMAL II trial, and the mosquito survival data for the 3 × 300 μg/kg regimen extracted from the work by Smit and colleagues,¹⁵ while RIMDAMAL II_K refers to data from Kamau and colleagues.¹⁴

Model parameterisation and statistical analysis

To model mosquito mortality, we used experimental survival data from mosquitoes exposed to various IVM formulations. These data were fitted using the death probability function P_m,IVM(τ) (Supplementary Eq. (B.6)) where τ denotes the time since exposure. This function depends on three parameters: τ₅₀ represents the time at which the death probability is 50%, α controls the steepness of the death probability curve (slope) and D is the upper asymptote of the death probability. Parameters were estimated with 95% binomial confidence intervals (CI) for each formulation using nonlinear least squares in R with the dose–response curve (drc) package.³²

The estimates were as follows, mdc-STM-001-0.6-vk5: α = 1.48 (95% CI, 1.06–1.90), τ₅₀ = 64.42 (52.65–76.19), D = 1; mdc-STM-001-0.6-kis: α = 1.23 (0.69–1.77), τ₅₀ = 147.98 (90.94–205.03), D = 1; mdc-STM-001-1.0: α = 1.18 (0.58–1.79), τ₅₀ = 168.10 (98.17–238.03), D = 1; BOHEMIA: α = 2.31 (1.54–3.09), τ₅₀ = 5.0 (4.09–5.90), D = 0.8626; RIMDAMAL II_K: α = 1.63 (0.80–2.47), τ₅₀ = 2.77 (1.14–4.40), D = ; RIMDAMAL II_S: α = 1.54 (1.30–1.78), τ₅₀ = 9.09 (8.26–9.92), D = 0.8632. Based on this death probability function, we derived a time-dependent adjusted mortality rate μ_m,IVM(τ) (Supplementary Eq. (B.7)), reflecting the additional mortality over baseline daily mortality attributable to IVM exposure. We also computed the CHR (Supplementary Eq. (B.8)) to quantify the time-varying relative risk of mortality due to IVM. The function outputs including estimates of P_m,IVM(τ), μ_m,IVM(τ), and the corresponding dynamics are presented in Figure S2.

Model diagnostics were assessed using residual plots (Figure S3), which showed residuals randomly scattered around zero with no discernible systematic patterns, indicating that the log–logistic model adequately captured the dose–response relationship. Prediction accuracy was further quantified using the root mean squared error (RMSE), computed in the same units as the dose–response. Full model parameter details are in Supplementary B.

Initial conditions and IVM campaigns

To assess IVM's impact on malaria incidence, we simulated an endemic equilibrium under perennial transmission with a total human infection prevalence of 31%, including 9% clinical cases, values typically observed in West African field trials^16,33 and which is also consistent with annual prevalence reported in the Kenya field trial.²¹ The simulation assumed a mosquito-to-human density ratio of 11.66, consistent with estimates for this region³⁴ (Figures S4 and S5A–B). We additionally explored seasonal transmission scenarios, where total infection prevalence ranged from 24% to 45% and clinical cases from 6% to 14% depending on the season, consistent with field trial data from the same region.^13,35 These field trials relied on RDT-based malaria prevalence estimates, which were used to validate the calibration of our model. The prevalence of infectious mosquitoes was also calibrated to range between 5% and 10%, depending on the transmission setting and the time of year. The modelled population comprised 813,610 individuals, reflecting demographic characteristics typical of a representative setting in Burkina Faso, including age structure, sex distribution (52% female), and malaria-specific parameters such as transmission intensity and immunity profiles. The reference target population representing the subgroups considered operationally feasible for IVM-based interventions corresponded to 60.20% of the total population, excluding all women of childbearing age. The full age and sex distribution, including the assumptions leading to this target, is provided in Table S1. The model also accounts for the greater likelihood of pregnant women to progress to symptomatic infection. Injectable administrations varied in the number of treatment cycles (n_c), while oral regimens were fixed, with schedules defined following BOHEMIA²⁰ and RIMDAMAL II³⁶ trials (Table 1). In addition, we examined a range of epidemiologically relevant scenarios capturing variations in IVM efficacy loss and population subgroups.

Relative gain

The relative gain of implementing an IVM strategy ϕ, computed over a one-year period and denoted as r_gain(ϕ), is defined as the ratio between the total number of clinical cases observed during the annual IVM campaign and total baseline cases in the absence of intervention (Supplementary Eq. (B9)). For example, a relative gain r_gain(ϕ) of 20% means that, at its optimal effectiveness, the strategy ϕ reduces the total number of clinical cases by 20% over a one-year period. In addition, we defined the difference in performance between two strategies, ϕ₁ and ϕ₂, as Δ(ϕ₁,ϕ₂) = r_gain(ϕ₁)-r_gain(ϕ₂), which corresponds to the percentage difference in impact between the two strategies. For example, a positive value of 5% indicates that strategy ϕ₁ is 5% more beneficial than strategy ϕ₂, while a negative value implies that ϕ₂ outperforms ϕ₁.

Ethics

Injectable formulations was imported into Burkina Faso with clearance from the Direction Générale des Services Vétérinaires of Burkina Faso (Permit No 2020/199) and was stored at room temperature, protected from light, until administration to cattle. The study protocol was approved by the Institutional Animal Ethics Committee of CIRDES (document reference number: 004-10/2020 CEB-CIRDES).

Role of funders

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

The clinical outbreak patterns across all age groups under different IVM formulations in a perennial transmission setting are shown in Fig. 2, for target population coverages of 50%, 70%, and 90%, corresponding to total population coverages of 30.1%, 42.1%, and 54.1%, respectively. The BOHEMIA and RIMDAMAL II_K scenarios exhibit a modest impact on clinical cases (Fig. 2A, E and I), with their relative gain between 7 and 14% (Fig. 3) across a target population coverage range of 50–90%. By comparison, RIMDAMAL II_S induces a more pronounced reduction in clinical cases (Fig. 2A, E and I). The corresponding relative gain ranges from 24 to 38% (Fig. 3) across target population coverage levels between 50% and 90%. By contrast, with a single IVM campaign, the injectable formulation approaches the impact of RIMDAMAL II_S, though remaining lower (Fig. 2A, E and I), with relative gains ranging from 16 to 29% (Fig. 3A). However, when two or more campaigns are implemented, the injectable formulation consistently outperforms the oral RIMDAMAL II_S, particularly as population coverage increases (Fig. 3B–D). Among injectable treatments and in the context of insecticide-resistant mosquito populations, mdc-STM-001 administered at 1.0 mg/kg outperforms the 0.6 mg/kg dose, particularly when at least two treatment campaigns are implemented (Figs. 2 and 3). At 0.6 mg/kg, malaria case reductions are slightly greater in susceptible than resistant mosquito settings, although species effects cannot be separated. For reference, the corresponding dynamics, expressed as incidence rate per 1000 population, are presented in Figure S6, while the corresponding dynamics of the infectious mosquito population under the same configuration as in Fig. 2 are shown in Figure S7.

Fig. 2.

Clinical cases under the effect of IVM across different levels of target population coverage (TP). The oral formulation dynamics are consistent across columns, with 3 cycles for BOHEMIA and 4 cycles for RIMDAMAL II_S and RIMDAMAL II_K scenarios (see Table 1). Only the number of cycles for the injectable formulation varies across columns. In this case, target population coverages of 50%, 70%, and 90% correspond to total population coverages of 30.1%, 42.1%, and 54%, respectively. (A–D) Effect of the number of cycles n_c (A: n_c = 1, B: n_c = 2, C: n_c = 3, D: n_c = 4) on clinical case dynamics at 50% coverage. (E–H) Same as (A–D) at 70% coverage. (I–L) Same as (A–D) at 90% coverage.

Fig. 3.

The relative gain from the introduction of IVM. The corresponding dynamics are illustrated in Fig. 2. Detailed relative gain values with 95% uncertainty intervals are provided in Table S2. Relative gain r_gain(ϕ) represents the percentage reduction in total clinical cases over a one-year period achieved by strategy ϕ relative to baseline without intervention. In this case, target population coverages of 50%, 70%, and 90% correspond to total population coverages of 30.1%, 42.1%, and 54%, respectively. Black arrows indicate the 95% uncertainty intervals. (A) Effect of the number of cycles (n_c = 1) and target population coverage on the relative gain. (B–D) Same as (A) for n_c = 2, 3, 4, respectively.

In the previous scenario (Fig. 2), the next IVM campaign is introduced once the protective effect of the prior campaign has waned and is no longer sufficient to control clinical cases. However, alternative scenarios can also be considered. For instance, in the basic scenario reflecting field practice, oral regimens are administered every 30 days, whereas the interval is assumed to be extended to 60 days for the injectable formulation (Figure S8). Overall, while the epidemiological dynamics of both scenarios differ markedly (Fig. 2 vs. Figure S8), their relative gains are approximately the same (Figure S9).

A global sensitivity analysis was conducted to identify the parameters most responsible for uncertainty in projected reductions in clinical malaria cases (Figure S10). The IVM regimen (comprising oral and long-acting formulations), through its associated duration of mosquitocidal efficacy emerged as the dominant driver of uncertainty, followed by the number of administration cycles. Secondary contributions arose from population coverage, while exclusion criteria related to sex and pregnancy and ivermectin clearance play comparatively minor roles. These rankings indicate that uncertainty in model projections is primarily driven by assumptions related to formulation/regimen mosquitocidal properties and deployment intensity. Additional methodological details and quantitative results are provided in Supplementary B.

The differences in performance across IVM administration regimen, evaluated over all age groups, are illustrated in Fig. 4. The results show that RIMDAMAL II_K and BOHEMIA exhibit similar performance, with BOHEMIA performing marginally better from 1 to 3% (Fig. 4A) across target population coverage levels between 50% and 90%, while RIMDAMAL II_S outperforms BOHEMIA from 15 to 26% (Fig. 4A). With a single campaign, the injectable formulation outperforms RIMDAMAL II_K and BOHEMIA, with the difference in performance increasing as target population coverage rises, but about 5–11% lower than RIMDAMAL II_S (Fig. 4A). With two campaigns, the injectable formulation surpasses RIMDAMAL II_S by 9–19% and exceeds RIMDAMAL II_K and BOHEMIA by at least 24% (Fig. 4B). The advantage becomes even more pronounced with three and four administration campaigns, achieving an added benefit of up to 62% and 66%, respectively –further underscoring the strong potential of injectable long-acting formulation (Fig. 4C and D). Detailed differences in performance are provided in Table S3.

Fig. 4.

Performance between regimens: oral vs oral, injectable vs oral, and injectable vs injectable. Subscripts on the x-axis indicate specific scenarios (e.g., BOH for BOHEMIA; RII_S and RII_K for RIMDAMAL II_S and RIMDAMAL II_K; 0.6-vk5 for mdc-STM-001-0.6-vk5; 0.6-kis for mdc-STM-001-0.6-kis; 1.0 for mdc-STM-001-1.0). Performance corresponds to the difference in relative gain between two strategies, Δ(ϕ₁,ϕ₂), indicating the percentage by which one strategy outperforms another. (A) Effect of the number of cycles (n_c = 1) on the performances between regimens. (B–D) Same as (A) for n_c = 2, 3, 4, respectively.

Discussion

Parasite transmission–blocking compounds, such as systemic endectocides, have the potential to complement existing disease control efforts to tackle residual malaria transmission and accelerate the decline in global malaria burden. In this study, we developed a mathematical model that incorporates both oral ivermectin formulations that have been evaluated in randomised mass drug administration trials and long-acting injectable formulations currently under development and evaluated in cattle species. This injectable formulation is issued from unique long lasting delivery technology and is designed to meet, with a single campaign, the minimum WHO requirements for endectocides against malaria.²⁶

A key strength of our proposed model lies in its structured framework, which incorporates (i) the chronological age and sex of the human host population, (ii) the time since IVM exposure in humans, shaping its downstream effect on mosquitoes, and (iii) the time since parasite infection in mosquitoes. This level of detail enables a more realistic representation of intervention dynamics and provides important insights into heterogeneous drug effects. Additionally, by explicitly modelling the persistence of IVM's effect –referred to here as the formulation's “pharmacodynamic persistence”—in both vectors and treated hosts, the framework captures the progressive decline of IVM's mosquitocidal effect. This innovation enhances our ability to compare oral and injectable formulations under more biologically realistic conditions.

The primary aim of this study was to assess the potential impact of a long-acting injectable formulation of IVM on malaria incidence. This formulation, currently undergoing preclinical laboratory evaluation, was compared to existing oral regimens with a view to informing the design of future phase III clinical trials. The comparison was carried out across various epidemiological settings, including areas of perennial transmission as well as seasonal contexts characterised by alternating periods of high and low transmission.

In perennial transmission settings, the limited effectiveness of BOHEMIA and RIMDAMAL II_K regimens, reinforces the need for more robust alternatives. The long-acting injectable formulation represents a promising option in this regard. A single long-acting injectable formulation campaign already achieves substantial prevalence reductions, approaching those observed with the oral RIMDAMAL II_S regimen. When two long-acting injectable formulation cycles are administered, the intervention outperforms RIMDAMAL II_S, and with three cycles, it achieves roughly twice the reduction (Fig. 4). Alternative relative gain values for oral regimens under three or four treatment cycles are reported in Table S5. To assess the role of transmission context, we examined the added benefit or reduced impact of IVM formulations under seasonal compared with perennial settings, with particular emphasis on the timing of campaign initiation relative to the seasonal peak. For both oral regimens and two cycle long-acting injectable formulation, the predicted impact is slightly higher in seasonal settings than in perennial transmission settings (Fig. 5). By contrast, late initiation or suboptimal timing reduces the effectiveness of long-acting injectable formulation, particularly when more than two cycles are deployed. The corresponding epidemiological dynamics for these seasonal scenarios are illustrated in Figures S11 and S12. These results highlight the strong potential of the long-acting injectable formulation in seasonal settings, where, compared with oral regimens, it more effectively curbs epidemic peaks and sustain reduced clinical incidence thereafter.

Fig. 5.

Difference in performance, Δ(ϕ_Seasonal,ϕ_Perennial), between seasonal vs. perennial transmission contexts. The oral formulation dynamics are consistent across panels, with 3 cycles for BOHEMIA and 4 cycles for RIMDAMAL II_S and RIMDAMAL II_K. (A) Effect of the number of cycles (n_c = 1) on performance in seasonal and perennial transmission settings. (B–D) Same as (A) for n_c = 2, 3, 4, respectively.

In addition to assessing overall relative gains across all age groups, we specifically examined outcomes in children aged 5–15 years, consistent with²¹ (Figure S9). Across all strategies, this subgroup benefits from additional reductions compared with population-wide averages. Oral regimens provide an added benefit of 1–3% over their baseline gains reported in Table S2. The injectable formulation yields a marginal added benefit of about 1–2% with a single cycle, which increases with repeated administrations and can reach up to 7% in this age group. Importantly, while children 5–15 years were a particular focus, other age groups were not neglected: although some subgroups are not directly eligible for IVM administration, both oral and injectable campaigns nonetheless induce notable reductions in clinical prevalence within these populations as well (Figure S13).

In addition to our first scenario, which excluded all women of childbearing age, we also examined the impact of excluding only pregnant women from IVM administration on the relative gains observed across the entire population. For comparability, the reference results in Table S2 were obtained under the latter assumption. When only pregnant women are excluded, the operational target population rises to 83.03%, with target population coverages of 50%, 70%, and 90%, corresponding to total population coverages of 41.5%, 58.1%, and 74.7%, respectively; the resulting additional benefits are shown in Figure S9. In this case, BOHEMIA and RIMDAMAL II_K show an added benefit of 2–3%, RIMDAMAL II_S 6–8%, and a single LAIF campaign 5–8%, with values increasing to 10–13% for two or more LAIF cycles over their baseline gains reported in Table S2.

A potential faster loss of IVM efficacy, as suggested by field observations on oral regimens indicating that treated individuals may lose drug effect up to twice as fast,^14,18 was evaluated by comparing scenarios of accelerated versus standard efficacy decline. The differences in performance under these scenarios are shown in Figure S14. Specifically, BOHEMIA and RIMDAMAL II_K exhibit a reduction of 1–4% in relative gains, RIMDAMAL II_S decreases by 7–10%, and LAIF loses 1–3% with a single cycle, reaching a maximum of 9% with more than two campaigns. Given that LAIF already achieve substantial relative gains (Table S2), this potential faster loss of efficacy has a limited effect on its overall performance. These findings highlight the strong potential of LAIF, whose benefits are expected to be more substantial and sustained when applied under field conditions.

The limitations associated with the use of derived hazard ratios from the IVERMAL trial¹⁵ to model IVM impact,¹⁶ including their tendency to overestimate mosquitocidal effects, are explicitly discussed to contextualise the results for the RIMDAMAL II_S formulation and avoid overinterpretation. Moreover, in the IVERMAL trial, adjunction of DHAP to ivermectin treatment may have increased IVM bioavailability,³⁷ potentially through reduced metabolisation,³⁸ an effect not accounted for in previous models, which attributed the mosquitocidal effects observed solely to ivermectin. These combined regimens could therefore be more effective than ivermectin alone if implemented in suitable contexts with sufficient coverage and adherence, but this remains to be confirmed. Note that our estimated reductions in clinical cases for oral regimens are more moderate to those reported in¹⁶ and are closer to field observations. An additional potential explanation for the overestimation reported in earlier models may reflect structural limitations in capturing the rapid loss of ivermectin efficacy in treated hosts, particularly when relying on discrete hazard ratio assumptions. This comparison is facilitated by the fact that both models rely on closely aligned epidemiological assumptions and parameterisation strategies. Moreover, compared with the recent BOHEMIA trial,²¹ which evaluated ivermectin alone and reported a 26% reduction in incidence among children aged 5–15 years over six months, our model predicts a 9–14% reduction over one year under comparable conditions. As a 26% six-month impact is compatible with a 9–14% one-year impact, this supports the realism of our projections.

Given that high population coverage is often targeted in mass drug administration strategies, achieving such levels in real world settings remains challenging. Adolescents and adult males are frequently underrepresented due to mobility related to education and work, making sustained coverage, particularly with repeated rounds, difficult to achieve. As a result, coverage levels of 70–90% of the eligible population, although desirable, may be unrealistic in routine implementation.²² To reflect these operational constraints, we therefore consider lower and more plausible coverage assumptions. In particular, a coverage level of 50% of the eligible population is explored as a conservative worst-case scenario, providing valuable insights for programmatic planning and future implementation trials.

While this study focused on evaluating the standalone impact of IVM, particularly in relation to formulation properties and deployment schedules, long-acting injectable endectocides also offer promising opportunities when combined with other interventions that share similar delivery timelines, such as mass drug administration (MDA) with antimalarials or seasonal malaria chemoprevention (SMC), although potentially requiring fewer repeat doses than oral treatments while maintaining program compatibility. A natural extension of this work will involve integrating long-acting IVM into multi-intervention models that better reflect realistic programmatic contexts, to inform policy-relevant deployment strategies. Such an integrated modelling framework will incorporate structural heterogeneities, pharmacokinetic–pharmacodynamic variability, and recent field evidence,^18,21,36 thereby providing a more comprehensive basis to inform policy decisions and optimise intervention packages across diverse transmission settings.

While a log–logistic function is used here to model mosquito mortality, alternative functional forms such as Weibull or Gompertz distributions could also be considered reasonable. However, the primary objective here is not to identify a statistically optimal mortality model per se, but rather to adopt a biologically plausible and commonly used frame that captures the gradual decay of ivermectin-induced mosquitocidal effects over time, a feature effectively represented by the log–logistic function.

In the approach proposed here, we assume that ivermectin (IVM) primarily exerts a lethal or mosquitocidal effect on vector populations. This constitutes a potential limitation, as IVM may also induce non-lethal effects on mosquitoes, such as reduced fecundity or altered biting behaviour (e.g., Refs.^9,12,39), as well as effects on intrinsic parasite development.^40,41 However, given the limited and inconclusive nature of available data on these secondary mechanisms, we assume – consistent with previous studies⁴² –that such effects are likely to be of lesser importance compared to IVM's primary lethal action. Furthermore, a future extension of the model could incorporate human genetic factors, which have a non-negligible impact on human-to-mosquito transmission parameters.⁴³ Age-dependent clearance of ivermectin was not explicitly modelled due to the current lack of robust, age-stratified pharmacokinetic data, particularly for long-acting injectable formulations; however, the model structure developed in this manuscript is fully compatible with the incorporation of age-stratified pharmacokinetic processes once such data become available.

The projected impact of LAIF is based on mosquito mortality data from preclinical livestock studies, because human data for long-acting injectable ivermectin were not available at the time of analysis. This approach is methodologically justified because ivermectin's mosquitocidal effect is driven by the drug concentration ingested during the blood meal, which reflects a direct mosquito–drug interaction independent of host species. Livestock studies provide well-characterised PK/PD curves linking systemic exposure to mosquito mortality, enabling a biologically grounded and experimentally robust first approximation of impact.²⁶ We acknowledge that PK differences between cattle and humans such as body fat composition, metabolic clearance, and volume of distribution may alter systemic exposure, and that venous plasma concentrations may not fully represent the drug levels ingested by mosquitoes.⁴⁴ Consequently, our projections should be interpreted as first-order estimates. Planned human PK bridging studies of long-acting injectable formulations will be essential to refine and validate these projections, and the model can be updated as soon as human data become available.

Prolonged exposure to long-acting IVM formulations may exert selective pressure on targeted parasite populations, raising concerns about the potential emergence of resistance. However, the routine exclusion of vulnerable subgroups, such as children under five and women of childbearing age, from IVM-based interventions may provide a natural safeguard, as these unexposed populations could serve as reservoirs of genetic diversity for both ectoparasites and endoparasites, thereby helping to mitigate resistance evolution.⁴⁵ Exclusion of these subgroups, while necessary for safety reasons, may influence population-level impact estimates by leaving higher-risk subgroups untreated; statistical analyses were therefore needed to assess the sensitivity of model outcomes to alternative eligibility assumptions. Beyond its impact on model outcomes, the exclusion of women of childbearing age raises important ethical and equity considerations. While such exclusions may be justified at early stages for safety reasons, they are unlikely to be appropriate for later-stage development. Progression to Phase 1 and subsequent trials should prioritise the inclusion of women, with adequate contraceptive coverage and safety monitoring to ensure equitable and generalisable outcomes.

From an implementation perspective, long-acting injectable ivermectin may reduce logistical complexity by limiting the number of campaigns needed to sustain protection. Fewer rounds could ease delivery, reduce personnel requirements, and improve adherence. However, deployment would still require trained staff, appropriate storage, monitoring of injection-site adverse events, and cost–benefit evaluation relative to oral regimens, all of which are critical for large-scale feasibility. Should these formulations prove highly effective in reducing malaria morbidity or accelerating elimination, their widespread deployment will also require careful monitoring, including pharmacological surveillance, entomological assessments, and predictive modelling to optimise formulation characteristics and delivery schedules while minimising resistance risk. In parallel, alternative endectocides such as nitisinone⁴⁶ and strategies targeting outdoor-biting vectors are in development.⁴⁷ Comparative evaluations of the cost-effectiveness of IVM-based tools versus these emerging interventions will be crucial to inform integrated, sustainable malaria control programmes.

Contributors

FJD: Software, Formal analysis, Investigation, Visualisation, Writing–original draft, Validation. RDD: Conceptualisation, Methodology, Investigation, Visualisation, Supervision, Writing–original draft, Validation. KM: Resources, Supervision, Data curation, Writing–original draft, Validation. AP: Data curation, Software, Writing–review & editing, Validation. NM: Resources, Writing–review & editing, Validation. SB: Data curation, data management, Writing–review & editing, Validation. SLD, CSW, HSP, RKD: Writing–review & editing, Validation. KM & SB have accessed and verified the underlying data.

All authors have read and approved the final version of the manuscript.

Data sharing statement

This study was sponsored by UNITAID and IRD, all investigators had access to the raw data and were free to participate in data analysis and result interpretation. All data relevant to this article will be made publicly available in the Dataverse of the Research Institute for Development within three weeks of the article's publication (https://dataverse.ird.fr/Id=doi:10.23708/T8TWVE). All analyses were conducted using R version 4.5.1, with dose–response modelling implemented using the drc package (version 3.0-1); reproducible scripts are available at https://doi.org/10.5281/zenodo.19594552. Further code and information will be made available upon reasonable request to the corresponding authors.

Declaration of interests

SLD is currently an employee and shareholder of Medincell, and did not influence or contribute to sample collection, data collection, sample analysis, data analysis, evaluation, model design, decision to publish, or conclusions presented in this paper. The remaining authors have nothing to declare.