Multi-centre laboratory study to determine discriminating concentrations for broflanilide and isocycloseram resistance monitoring in mosquitoes

1 Introduction

The World Health Organisation recommends that programmes routinely monitor target mosquito populations for susceptibility to insecticides to guide decisions about which insecticide-based vector control interventions to deploy (WHO, 2022a). Field populations are regularly, typically annually, tested using bioassays measuring mortality on exposure to a known standard concentration of a given insecticide – known as the discriminating concentration (DC, sometimes referred to as the diagnostic concentration). Some insecticides are tested using treated filter papers in a WHO tube assay (Global Malaria Programme, 2022), and others in a treated Wheaton bottle via the WHO bottle bioassay (WHO, 2022b). The WHO has traditionally defined its discriminating concentrations (DCs) as twice the lowest concentration that systematically resulted in 100% mortality after 60 min of exposure and a holding period of 24 h using an insecticide-susceptible strain. The DC is calculated as twice the LC99 value (the concentration at which 99% of individuals are killed), as determined by a log-probit statistical model against a susceptible strain (World Health Organization (WHO), 2026). The concept of DCs has clear advantages in terms of the cost and efficiency of testing and has been widely adopted for monitoring insecticide resistance in mosquitoes and other disease vectors.

In 2017–2021, the WHO conducted a multi-centre study to establish and validate DCs of insecticides to enable the testing of resistance to insecticides in the main Anopheles and Aedes spp. mosquito vectors of human diseases (WHO, 2021). In 2022, the WHO published a methodology and a list of 12 new DCs to monitor insecticide resistance in Culex quinquefasciatus and Cx. tarsalis (WHO, 2023). Since these multi-centre studies, two new insecticides, both in class 30 of the Insecticide Resistance Action Committee (IRAC) classification, have been included in products that have been listed as prequalified by the WHO: broflanilide and isocycloseram (WHO, 2026). These are both IRAC class 30 (meta-diamide/GABA-gated chloride channel allosteric modulators), with a mode of action distinct from those of pyrethroids, carbamates, organophosphates, and organochlorines (IRAC, 2026).

For the purpose of establishing and validating discriminating concentrations for these new insecticides, a multi-centre study was therefore conducted in 2024–2025 to confirm and optimise, where necessary, the testing methods for performing dose–response experiments with broflanilide and isocycloseram. DCs of each compound were established using the data analysis method developed by Mara Kont (Corbel et al., 2023) and validated against well-characterised susceptible colonies of four mosquito species (An. gambiae, An. funestus, An. stephensi, and Ae. aegypti) with results generated from seven testing centres. Based on the experiences gained at the testing centres and the analysis of the data used to propose DCs, the WHO generic protocol for establishing DCs was further validated, and recommendations on best practice were formulated.

2 Materials and methods

The methods used in this study were pre-specified in the WHO generic protocol for determining discriminating concentrations, which is the same protocol applied in previous WHO-coordinated multi-centre studies (World Health Organization (WHO), 2026). The study design, species–compound allocation, replication targets, test conditions, and analytical approach were therefore fixed in advance and applied uniformly across all participating centres, with the aim of validating the protocol and generating publicly citable discriminating concentrations rather than comparing or optimising methodological choices.

2.1 Participating institutions

Seven internationally recognised laboratories with extensive entomological capacity participated in the study, which was coordinated by the Liverpool School of Tropical Medicine (LSTM) in collaboration with the WHO. These were: LSTM, Institut de Recherche pour le Développement (IRD), London School of Hygiene & Tropical Medicine (LSHTM), Centre for Research in Infectious Diseases (CRID), ICMR–National Institute of Malaria Research (NIMR), National Institute for Communicable Diseases (NICD), and Ifakara Health Institute (IHI). All participating institutions maintained well-characterised susceptible mosquito colonies and had adequate facilities and technical capacity to conduct laboratory testing. These institutions had previously participated in WHO multi-centre trials or had demonstrated strong capacity for insecticide testing.

2.2 Study design

The study followed a step-by-step approach (Figure 1) that had been previously used (WHO, 2021) and has now been formalised by the WHO as a generic protocol for studies designed to determine discriminating concentrations (World Health Organization (WHO), 2026), with the aim of establishing broflanilide and isocycloseram DCs for inclusion in the WHO manual for monitoring insecticide resistance in mosquito vectors and selecting appropriate interventions (WHO, 2022a).

Step 1: Preliminary bottle bioassays for screening a range of serial insecticide concentrations to obtain 0–100% mortality in mosquitoes Approximately 10–12 serial concentrations were required to provide a range of mortality rates (from 0% to 100%) for each mosquito species and compound. Susceptible mosquitoes were exposed to serial concentrations of each insecticide for 1 h, with one replicate each of 50 mosquitoes per concentration and one replicate of 50 mosquitoes as an acetone-only control.

Step 2. Bottle bioassays to establish LC99 and LC100 values and propose tentative discriminating concentrations

For Step 2 testing, each centre was expected to complete three validated replicates for each concentration of each compound against each of its assigned representative species (Table 1). The overall aim was to achieve a complete dataset of three replicates from each of three centres for every compound–species pair. Testing centres tested a range of serial concentrations pre-selected by LSTM based on the results of Step 1. At least six test concentrations and a control were tested to generate “concentration-response” curves. For a Step 2 test to be considered valid, it was necessary to include at least two concentrations resulting in <50% mosquito mortality, one concentration resulting in approximately 50% mortality, two concentrations resulting in >50% mortality, and one concentration resulting in approximately 100% mortality. Each bioassay was performed in triplicate for a given mosquito species and strain (using at least 100 mosquitoes per concentration), wih different batches of mosquitoes tested on different days. The impregnated bottles were used only once.

Table 1. Susceptible mosquito species and strains used in the study, and the testing centres at which they were tested.

*LSHTM mosquitoes were tested at LSTM. This arrangement was necessary because the initial third testing centre for An. stephensi was forced to withdraw from the study, and a third centre could not be identified with both a susceptible colony and the capacity to test in the required timeframe. Mosquitoes were therefore from independent colonies of SDA500, but testing was performed in the same laboratory by the same operator as the LSTM SDA500 colony.

Step 3. Multi-centre validation of tentative discriminating concentrations against all mosquito species

Tentative DCs (TDCs) were calculated based on the data from Step 1, following discussions among the testing centres and WHO representatives. The TDC of a given insecticide was tested by three different laboratories against the same well-characterised susceptible strains in WHO bottle bioassays, with mortality recorded after 24 h. Bioassays included one replicate for each compound and mosquito species, with at least 100 mosquitoes tested per TDC to validate whether 100% of susceptible strains tested were killed. If this criterion was met, the concentration was proposed as the DC.

2.3 Mosquitoes

Four mosquito vector species (An. gambiae, An. funestus, An. stephensi, and Ae. aegypti) were included in the study based on three selection criteria: (i) species of major epidemiological importance in malaria or dengue transmission, (ii) representation of different geographical regions (Africa, Asia, and other relevant endemic areas), and (iii) availability of well-characterised, fully susceptible colonies maintained in at least three participating laboratories, ensuring cross-validation of results across multiple sites (Table 1). The mosquito species and susceptible strains tested at each centre were selected according to local colony availability and laboratory capacity.

2.4 WHO bottle bioassays

Given the chemical properties and modes of action of broflanilide and isocycloseram, and based on the experience of their manufacturers, who were consulted ahead of the study, the WHO bottle bioassay (WHO, 2022b; Corbel et al., 2023), with the addition of a surfactant (rapeseed oil methyl ester, MERO), was selected as the most appropriate testing methodology. Broflanilide was provided to LSTM for the study by Mitsui Chemicals Crop & Life Solutions, Inc., and isocycloseram was provided by Syngenta Crop Protection AG. Aliquots of each insecticide and of MERO were sent to each testing centre.

Bottles were coated with the required doses of insecticide and MERO surfactant at 1,500 ppm for Aedes aegypti and 800 ppm for Anopheles spp., according to the WHO bottle bioassay SOP (WHO, 2022b). The bottle drying time (i.e., the interval of time between coating the bottles and testing) was fixed at 24 h, and bottles were dried horizontally with their lids left open until the test started.

MERO was included following the manufacturers’ recommendations and in accordance with the WHO bottle bioassay SOP because both compounds are highly lipophilic and adhere to, and are poorly redissolved from, the bottle surface without a surfactant. Acetone/MERO-only controls were run alongside every assay, and the validity rule requiring control mortality below 20% would have identified and excluded any independent toxic effect of the surfactant. Because all centres used identical MERO concentrations (1,500 ppm for Aedes and 800 ppm for Anopheles), the surfactant was held constant across the comparisons made here and therefore does not affect the establishment of the discriminating concentrations determined (WHO, 2021). Whether the inclusion of MERO for some insecticides and not others confounds the conclusions drawn from susceptibility assays is an open question beyond the scope of this study.

Three- to five-day-old, non-blood-fed female mosquitoes were exposed to the treated bottles for 1 h according to the WHO bottle bioassay SOP (WHO, 2022b). Testing was performed, and mosquitoes were held until the 24-h mortality assessment at 27 °C ± 2 °C and 75 ± 10% relative humidity. All testing centres reported testing conditions within this range.

Several measures supported comparability across centres: a single source of each insecticide and of MERO was aliquoted centrally at LSTM and distributed to all testing centres, all of which followed the same WHO bottle bioassay SOP and analytical validity criteria.

2.5 Data analysis

To ensure consistency and robustness in defining the DCs before progressing to the validation step, bioassay data for Step 2 were included for analysis only if the control mortality was below 20%, a minimum of six concentrations was tested to generate concentration–response curves and estimate LC values, the goodness of fit met the threshold (P-value > 0.05), and the dataset included a concentration that resulted in 100% mortality (LC100), where available.

For each testing centre, data from three replicates were analysed separately to assess intra-laboratory variation. If no major differences were observed between replicates, the data were pooled to estimate LC values for that centre. Datasets from three independent laboratories were used to cross-validate the results and determine a tentative DC. Concentration–response curves were plotted for visualisation of the data. Concentration–response data for each insecticide–species combination were analysed using the Bayesian binomial modelling framework developed for the WHO bottle bioassay multi-centre study (Corbel et al., 2023; Kont et al., 2023). For each bioassay, the number of mosquitoes dead out of those tested was modelled using a five-parameter logistic regression on square-root-transformed concentration scales. Models were run on four Markov chains for up to 10,000 iterations, and convergence was assessed using the Gelman–Rubin statistic (R̂ < 1.01), with zero divergent iterations required. All models presented in Table 2 successfully met these convergence criteria.

Table 2. Summary of LC99 values, discriminating concentrations (DC = 2 × LC99), and observed LC100 values for broflanilide and isocycloseram across all participating centres, mosquito species, and strains in WHO bottle bioassays.

Values are presented to one decimal place as the median LC99 with lower and upper 95% confidence intervals, the corresponding DC, and the lowest tested concentration consistently resulting in 100% mortality (observed LC100). Model fit was assessed following Kont et al. (2023) (Kont et al., 2023) using the coefficient of determination (R²) and root mean square error (RMSE), calculated from a linear regression of observed against posterior median predicted mortality at each tested concentration; R² values approaching 1.0 and RMSE values approaching 0% indicate a good model fit. N/A indicates data were not available or were not determined. Results that did not meet the convergence criteria (Gelman–Rubin statistic R&#x0302; < 1.01 with zero divergent iterations) are flagged; see footnote† for details.

†Results marked as non-convergent should be interpreted with caution. For CRID Anopheles spp. (Broflanilide), the data comprised only one to two replicates with mortality at or near 100% across all tested concentrations, meaning that the dose–response curve could not be adequately resolved, and LC₉₉ estimates were extrapolated below the lowest tested dose. For CRID Anopheles spp. (Isocycloseram) and LSHTM An. stephensi SDA500 (isocycloseram), the tested concentration range was insufficient to capture the full dose–response relationship, precluding reliable model convergence.

The fitted curves and confidence intervals were used to assess consistency across testing centres and species and to derive LC values and tentative discriminating concentration estimates (Corbel et al., 2023). The basis for each DC was selected according to the following rules: (i) where the LC₉₉ was reliably estimated, the DC was set at twice the highest LC₉₉ across the susceptible strains tested; (ii) where an LC₉₉ could not be reliably estimated (because of insufficient data points or replicates), or where the highest calculated LC₉₉ was unrealistically high relative to the other sites, an observed LC₁₀₀ was used instead, defined as the lowest tested concentration that consistently produced 100% mortality at the centre with the highest LC estimate, provided this value did not exceed twice the highest reliable 2×LC₉₉ value from the remaining sites. For Step 3, the concentration–response bioassay data were analysed, and the concentration–response parameters were used to estimate the goodness of fit (with the P-value) and lethal concentrations (i.e., LC100 and LC99 with 95% confidence intervals) for each insecticide and mosquito species.

3 Results

3.1 Step 1. Preliminary bottle bioassays for screening a range of serial concentrations of insecticides to obtain 0%–100% mortality in mosquitoes

Preliminary data generated at LSTM and IRD confirmed the suitability of a 24-h bottle drying time before testing, a 1-hour exposure period, and 24-h post-exposure mortality as the suitable endpoint. The initial concentrations used by each centre for testing in Step 2 were established based on preliminary testing at LSTM.

3.2 Step 2. Bottle bioassays to establish LC₉₉ and LC₁₀₀ and propose tentative discriminating concentrations

In total, 62,673 Aedes and Anopheles mosquitoes were tested in WHO bottle bioassays treated with broflanilide and isocycloseram (Table 2A).

Table 2A. Total number of mosquitoes tested in WHO bottle bioassays against broflanilide and isocycloseram, by species.

Where datasets were suitable, concentration–response curves were plotted for each insecticide–species combination by testing centre; where data could not be fitted by the model, no curve was generated. The fitted curves and confidence intervals were used to assess consistency across testing centres and species and to derive LC values and tentative discriminating concentration estimates.

3.2.1 Broflanilide

Fittable concentration–response data were generated for An. gambiae by LSTM, IRD, and IHI (Figure 2), for An. funestus by LSTM and NICD (Figure 3), for An. stephensi by LSTM (LSTM and LSHTM SDA500 colonies) and NIMR (Figure 4), and for Ae. aegypti by IHI, NIMR, LSTM, and IRD (Figure 5). CRID data for An. gambiae and An. funestus could not be fitted to the model, so no curves were generated for those sites. Replicate counts, the numbers of concentrations, and concentration ranges for each centre are given in Table 2B.

Table 2B. Total number of replicates, concentrations, and the concentration range tested for each.

3.2.2 Isocycloseram

Fittable data were generated for An. gambiae by IHI, IRD, and LSTM (Figure 6), for An. funestus by NICD, CRID, and LSTM (Figure 7), for An. stephensi by NIMR and LSTM (LSTM SDA500 colony) (Figure 8), and for Ae. aegypti by IHI, IRD, LSTM, and NIMR (Figure 9). CRID data for An. gambiae and the LSTM data for the LSHTM An. stephensi colony could not be fitted to the model.

For both broflanilide and isocycloseram, the fitted models produced steep slopes with narrow confidence intervals at higher mortality levels, indicating consistent concentration–response relationships. LC values were generally lower for An. gambiae and Ae. aegypti than for An. funestus and An. stephensi, reflecting differences in baseline sensitivity. The study was not designed to attribute these differences in baseline sensitivity to specific mechanisms, but several factors may plausibly contribute and merit investigation in future work, including differences in body size and cuticular surface area between species, variation in cuticular penetration and metabolic detoxification capacity, and possible differences in target-site sensitivity. Resolving their relative contributions would require dedicated biochemical and genetic characterisation beyond the scope of this study.

While some inter-laboratory variation was observed, particularly in LC₉₉ estimates, the overall trends were coherent and suitable for establishing tentative discriminating concentrations based on pooled or representative site data according to the study protocol (Table 2).

Step 3. Multi-centre validation of tentative discriminating concentrations against all mosquito species

Tentative discriminating concentrations (TDCs) were proposed for each testing centre for each compound against An. gambiae, Ae. aegypti, An. funestus, and An. stephensi based on data from that centre (Table 4). These values were determined using either a calculated 2×LC₉₉ or an “observed LC₁₀₀.” The observed LC₁₀₀ was used in cases where LC estimates differed substantially between testing centres and the highest calculated LC₉₉ was unrealistically high compared with the data generated elsewhere. In those situations, the observed LC₁₀₀ was defined as the lowest tested concentration that consistently produced 100% mortality at the centre with the highest LC estimate, provided this value was not more than twofold higher than the highest confidence 2×LC₉₉ from the remaining sites. Each proposed TDC was tested by at least three independent laboratories to validate its suitability as a proposed DC. To address the large discrepancies observed in DC estimates from Step 2, IRD and IHI conducted additional Step 3 tests using additional doses. IRD tested Ae. aegypti with broflanilide at 3 µg and isocycloseram at 6 µg, and An. gambiae with broflanilide at 3 µg and isocycloseram at 2 µg. IHI also tested Ae. aegypti with isocycloseram at 6 µg. These additional data were generated to support the selection of the final DC values.

Table 4. Summary of mortality results from bottle bioassays conducted across multiple testing centres using broflanilide- and isocycloseram-coated bottles.

Each centre tested relevant mosquito species at the control (0 µg/bottle) and selected validation concentrations, recording 24-h post-exposure mortality. All data represent a single biological replicate, except for the NICD data, where an additional two biological replicates were completed.

The majority of Step 3 tests resulted in 100% mortality at all doses tested across all testing centres, and the TDCs were confirmed. The only dataset where this was not the case was the NICD dataset, where 14 µg broflanilide resulted in 98% mortality and 57 µg isocycloseram resulted in 93% mortality for An. funestus. A second experiment conducted done at NICD with two technical replicates using 14 µg broflanilide, which resulted in 99% mortality. Finally, a set of three replicates was conducted using 15 µg broflanilide and 60 µg isocycloseram, yielding 100% mortality for both compounds across all three replicates. The DCs were therefore deemed to be appropriate, albeit with the caveat that variable results were observed at this one testing site.

The DCs based on the complete dataset from this study are shown in Table 5. Some values were rounded for simplicity to generate the final DCs recommended to the WHO for adoption for each compound–species pair.

Table 5. Proposed tentative discriminating concentrations (TDCs) for broflanilide and isocycloseram in WHO bottle bioassays against Aedes and Anopheles species based on the analysed Step 2 data.

4 Discussion

The WHO provides guidance on how to assess insecticide resistance in mosquito vectors in a standardised way using a discriminating concentration (DC) assay (World Health Organization (WHO), 2022) designed to detect the early signs of emerging resistance to an insecticide. The data generated by monitoring mosquito populations for their susceptibility to a range of insecticides can inform decisions by those designing and implementing vector control strategies.

A WHO-sponsored multi-centre study was conducted in 2024–2025 to propose DCs for two new insecticides, broflanilide and isocycloseram, for resistance monitoring in Ae. aegypti, An. gambiae s.s., An. funestus, and An. stephensi. Both insecticides belong to IRAC class 30 and are incorporated into vector control products that are prequalified by the WHO for public health use. Seven laboratories from Africa, Asia, and Europe participated, with the study coordinated by the Liverpool School of Tropical Medicine (LSTM) in collaboration with the World Health Organisation (WHO).

The study followed a standardised, three-step approach: preliminary testing to establish suitable concentration ranges (Step 1), multi-centre concentration–response testing to estimate LC99/LC100 and propose tentative discriminating concentrations (TDCs) (Step 2), and multi-centre validation of the TDCs (Step 3). Well-characterised, fully susceptible laboratory colonies were used for all assays. WHO bottle bioassays, with 1 h exposure, 24 h drying, and 24 h holding time, were confirmed as the most appropriate method for both compounds, with MERO surfactant at 1,500 ppm for Aedes spp. and 800 ppm for Anopheles spp.

In total, 62,673 mosquitoes were tested in WHO bottle assays across all centres. Validated TDCs were established for each species–compound combination, generally supported by three fully replicated datasets from partner sites, along with separate confirmation that the selected TDC resulted in 100% mortality of all susceptible strains tested. Although generally successful in generating the desired dataset, the study encountered some logistical challenges that impacted the data that could be collected. Limited capacity for An. stephensi testing, following the unforeseen withdrawal of one testing centre from the study, left only LSTM able to test this species. The London School of Hygiene and Tropical Medicine (LSHTM) held an independent colony of the same strain, SDA500, and although they were unable to join the study, they provided mosquitoes to LSTM. Although this ensured An. stephensi testing could proceed in three sites, the lack of multi-centre data restricted inter-laboratory comparison and reduced the strength of evidence for validation. The delays caused by this change to the study plan meant that seven rather than the nine planned replicate tests were possible. The completion rate for all other insecticide–species combinations was 100%.

Limited studies are available on isocycloseram, and therefore comparison of these DC values with existing datasets is not possible; however, the values generated were broadly in line with the manufacturer’s expectations. Our broflanilide discriminating concentrations are broadly consistent with the preliminary values reported by Ando et al. (2026), including the wide between-site variation among susceptible strains, reinforcing that this variability is a general feature rather than a study-specific artefact. As both compounds act on a target site distinct from those of the established insecticide classes (IRAC class 30), limited cross-resistance with pyrethroids and other older classes is expected, which is the basis for their deployment against resistant populations; monitoring with the DCs established here will allow this expectation to be tested as field use expands.

The analytical validity criteria were applied uniformly regardless of the number of replicates available; therefore datasets retained for LC estimation met the WHO standards; datasets that did not meet these criteria, including some with fewer replicates or an insufficient concentration range, were not fitted and are identified as such. The datasets that could not be fitted to the model (CRID An. gambiae and An. funestus) did not weaken the resulting recommendations. In each case, the discriminating concentration was anchored on the fitted dose–response curves from the other participating centres, and the same compound–species combinations were subsequently confirmed in Step 3 validation, where the proposed values produced 100% mortality. These fitting failures therefore reflect a range-finding and replication limitation at a single site rather than a contradiction of the proposed discriminating concentrations.

Some inter-laboratory variation in LC estimates was observed, most visibly in the upper part of the concentration–response curve. Much of this reflects estimation uncertainty rather than genuine disagreement between sites. The largest apparent discrepancy, the An. funestus isocycloseram LC₉₉ of 290.3 µg/bottle at LSTM against 26.1 µg/bottle at NICD, illustrates this: the LSTM estimate carries a very wide confidence interval (164.6–48,940 µg/bottle) and a low R² (0.6), marking it as a poorly constrained fit rather than a reliable value in conflict with the other centre. Some between-site variation in LC values among susceptible strains is expected nonetheless, as also seen in the WHO 2017–2023 multi-centre study on which the present protocol is based (WHO, 2021); plausible contributors include differences in bottle coating and insecticide handling, colony history, environmental conditions, and operator technique. Differences in results between centres could be minimised in the future by standardising mosquito-breeding protocols, equipment, and ensuring strong and consistent training for testing across centres. Taking advantage of opportunities for transfer of skills and techniques between centres would benefit future studies.

Within-bioassay variability, measured as the mean absolute deviation of observed mortality from the model’s best-fit line, was low across the great majority of curves: the median was approximately 4%–5% mortality, and 28 of 30 site–species–compound combinations fell below 10%. The two exceptions were both isocycloseram curves: An. funestus at LSTM (the same fit responsible for the wide LC₉₉ interval noted above) and An. stephensi at NIMR. This degree of agreement matches that reported for the WHO bottle bioassay in the predecessor multi-centre study (WHO, 2021; Corbel et al., 2023), confirming that the assay is reproducible within sites and that the greater uncertainty in some upper-tail (LC₉₉) estimates stems from a few poorly constrained fits rather than from general assay noise.

Consistent with previous applications of this analytical framework (Corbel et al., 2023; Kont et al., 2023), the limited number of replicates per insecticide–species combination at each centre meant that between-site and between-strain variability could not be formally determined; instead, within-bioassay variability was characterised. The study design (adopting the highest LC across susceptible strains and confirming each tentative DC independently) ensures that the residual between-site variation does not compromise the proposed discriminating concentrations.

Some suggestions on the methods used to generate and analyse data for the purpose of establishing DCs are proposed below, based on the experience gained during this multi-centre study under four headings.

4.1 Selecting concentrations for Step 2

An LC99 is an inherently challenging target to estimate due to its dependence on observations very close to, but not reaching, 100% mortality. Mortality observations of 100% are far less useful for this purpose, as they only identify that the LC99 threshold has been exceeded, but not by how much. As a result, LC99 estimates carry substantial error unless the experimental design is explicitly targeted at observing the 95%–99% range.

A common misconception is that a full dose–response curve with a wide spread of mortality values is required for the purpose of obtaining the target LC value. For robust LC99 estimates, rich information near the target value matters far more than accurately defining the shape of the lower parts of the curve. Thus, to estimate a target LC value, it is ideal to conduct a preliminary range-finding step that approximates the concentration needed, followed by a targeted step that aims above and below that dose range. As highlighted previously (Mechan et al., 2024; Harvey et al., 2025), more robust bioassay data do not necessarily require larger sample sizes; instead, mosquitoes can be better targeted towards relevant concentrations. In the final analysis, data generated during the range-finding step can be included in the final modelling analysis to strengthen the estimates.

4.2 Ensuring power for robust LC calculations

Sample size calculations should focus on the precision required for the LC estimate. In principle, the first decision is the level of error that can be tolerated, for example, ± 10%. In practice, the process becomes a comparison of alternative sample size scenarios and the resource implications of each. Precision depends on the number of concentrations tested and the number of replicates at each concentration. For any fixed total number of mosquitoes, there will be an optimal balance between these two elements. Identifying the minimum total number of mosquitoes for a given level of precision is an optimisation problem that usually requires simulation. Such approaches are well established but rarely straightforward for end users because they rely on realistic assumptions about variance that must be informed by existing data or set conservatively (Mechan et al., 2024; Harvey et al., 2025).

Although the LC99 is conceptually the most relevant value for setting a discriminating concentration, its practical limitations outlined above and in previous studies (Corbel et al., 2023), together with resource limitations, mean that a trade-off exists between the usefulness of the LC selected and the robustness of its estimate. An LC50 is typically estimated with the greatest precision because data are generally more balanced above and below that point (unless explicitly designed otherwise). Intermediate values such as LC80 or LC90 may offer a more realistic compromise between biological relevance and statistical confidence.

4.3 Defining a DC

While the statistical principles underlying estimation of LC values are well defined, the step of converting an LC to a discriminating concentration is open to improvement. Notably, the confidence interval around the LC estimate is not currently considered in the calculation of the final DC (World Health Organization (WHO), 2026). Even when using a Bayesian modelling approach that better captures heterogeneity, the uncertainty interval is not considered in this final step (Kont et al., 2023). If the true LC99 lies at the lower bound of the confidence interval, a discriminating concentration based on the mean could be set too high, reducing the sensitivity of the assay, and early signs of resistance may be missed. Thus, a conservative approach would be to use the lower bound of the 95% confidence interval when defining the discriminating concentration.

The current practice of doubling the LC99 assumes that unobserved variation in susceptible populations needs to be accommodated, i.e., the range of susceptible strains assessed does not capture the range of all possible susceptible populations. However, the choice of a multiplier of two appears arbitrary. If the multiplier is set too high, there is a risk that early signs of resistance will be missed.

Increasing the robustness of LC estimates could reduce or remove the need for such a multiplier if the susceptible populations used were reasonably representative. In recognition of this, the study design applied here and proposed by the WHO (World Health Organization (WHO), 2026) tests three susceptible strains of each species and adopts the highest LC99 or LC100 as the basis for the DC. The more strains that are included, the greater the confidence that the full range of susceptible phenotypes is represented; however, in practice, a multiplier may still be required, depending on the selected LC value.

4.4 Setting DC values

Deciding on a DC is a challenge in situations in which LC estimates vary substantially between testing centres or susceptible strains. Selecting the highest value is sensible, provided there is no evidence of quality control issues affecting one or more strains, e.g., genetic contamination causing resistance or poor rearing leading to smaller adults. All strains should be profiled (Williams et al., 2019) for phenotypic susceptibility before testing, but additional checks may be needed when results appear anomalous. These could include genotyping for known resistance markers or morphometric comparisons such as wing measurements. If no explanation for an outlier can be found, all data should remain in scope, and the highest LC should be used nonetheless to avoid setting a DC too low and thus causing false-positive reports of resistance. If resources allow, further strains or repeated testing could help determine whether the outlier represents true biological variation or an anomaly. Even so, a decision needs to be made whether to retain the highest LC or exclude problematic data, and this decision should be guided by clear criteria.

5 Conclusion

This multi-centre study validates the generic protocol for establishing DCs published by the WHO (World Health Organization (WHO), 2026), having successfully generated and validated discriminating concentrations for broflanilide and isocycloseram against An. gambiae, An. funestus, An. stephensi, and Ae. aegypti. This is the first time a DC has been established for isocycloseram. For broflanilide, the generated DCs reflect those previously proposed, including the large variation between different sites with susceptible mosquitoes of the same species (Lees et al., 2020; Ando et al., 2026). These values will enable national programmes and research institutions to integrate these insecticides into their resistance monitoring activities.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

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

Publisher’s note

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

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmala.2026.1874255/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Yousif Elsafi Himeidan, National Center for the Prevention & Control of Plant Pest & Animal Diseases (Weqaa Center), Saudi Arabia

Reviewed by: Hien Sawdetuo Aristide, Research Institute for Health Sciences (IRSS), Burkina Faso

Genevieve Tchigossou, International Institute of Tropical Agriculture (Benin), Benin