Study sites
Overall, 21 internationally recognized laboratories from five WHO regions participated in the study, which was jointly coordinated by Institut de Recherche pour le Développement (IRD) and the WHO (Fig. 1). The laboratories were selected because they were either formally designated WHO Collaborating Centres or had adequate facilities, capacity, and well-maintained mosquito colonies that were reported to be susceptible to the insecticides under consideration for evaluation.
Mosquito species and strains
Two major arbovirus vector species, Aedes aegypti and Ae. albopictus, and five major malaria vector species, Anopheles albimanus, An. gambiae ss, An. funestus s.s, An. minimus s.s. and An. stephensi, were included in the study. These species were selected because they play a primary role in the transmission of malaria or dengue, chikungunya and Zika viruses in different geographical regions of Africa, Asia-Pacific, Central and South America, and the Middle East, and each was held in colony by at least 3 different participating laboratories to enable cross-validation tests to be conducted at multiple sites. Different colonies of each species were used by the participating laboratories for testing insecticides (Table 1). The mosquitoes had all been in colony for several generations and each colony had previously been shown to be susceptible to the tested insecticides using biological and/or molecular assays.
Table 1
Mosquito species and strains selected for the WHO multi-centre laboratory study.
Species
|
Strains
|
Aedes aegypti (Linnaeus, 1762)
|
Bora Bora, New Orleans, Rockefeller
|
Aedes albopictus (Skuse, 1895)
|
Chengdu, EHI, NEA, Perols, VCRU
|
Anopheles gambiae s.s. (Giles, 1902)
|
Kisumu
|
Anopheles funestus s.s. (Giles, 1900)
|
Fang
|
Anopheles stephensi (Liston, 1901)
|
NDD, Puducherry
|
Anopheles minimus s.s (Theobald, 1901)
|
TM
|
Anopheles albimanus (Wiedemann, 1820)
|
Sanarate, Buenaventura
|
Insecticides
Following a WHO consultation in 2017 with experts and researchers from academia and industry, 18 insecticides were determined to be of high priority for testing against Anopheles and/or Aedes species [22]. Seven of these insecticides (see Table 2), belonging to 5 different mode of action classes were tested in glass bottles because they were not suitable for impregnation of filter papers due to the instability of those treated papers. These compounds are used in various formulations of insecticidal products that are currently in use or under WHO evaluation for vector control, including indoor residual sprays, insecticide-treated nets, space sprays, household pesticide products, and spatial repellents.
Table 2
Insecticides successfully tested with WHO bottle bioassays.
Class
|
Insecticide
|
CAS RN a
|
Test mosquitoes
|
Product type
|
|
|
|
Anopheles spp.
|
Aedes spp.
|
|
Pyrroles
|
Chlorfenapyr
|
122453-73-0
|
√
|
|
ITN
|
Neonicotinoids
|
Clothianidin
|
210880-92-5
|
√
|
√
|
IRS
|
Juvenile hormone mimics
|
Pyriproxyfen
|
95737-68-1
|
√
|
|
ITN / Larvicide
|
Butenolides
|
Flupyradifurone
|
951659-40-8
|
√
|
√
|
SP
|
Pyrethroids
|
Transfluthrin
|
118712-89-3
|
√
|
√
|
SS
|
Prallethrin
|
23031-36-9
|
|
√
|
SS
|
Metofluthrin
|
240494-71-7
|
|
√
|
SR
|
a CAS RN: Chemical Abstract Service Registry Number; b WHO Pre-qualification; HPP: Household pesticide products; IRS: Indoor residual spraying; ITN: Insecticide-treated nets; ATSB, Attractive toxic sugar baits; SS: Space spray; SR: Spatial repellents |
Study design
The first objective of the study was to develop and cross-validate protocols for testing the selected insecticides in a glass bottle bioassay. Six laboratories with strong expertise in testing and evaluation of public health pesticides were selected by the WHO to determine the most suitable technical parameters for applying the WHO bottle bioassay using each insecticide including the coating of bottles, the bottle drying time (i.e. 1 h, 2 h or 24 h), the mosquito exposure time (1 h or 2 h), the post-exposure holding period (24 h, 48 h or 72h) and appropriate test conditions, including ambient temperature, relative humidity and optimum concentrations of MERO®, a surfactant made of 81% rapeseed oil methyl ester that was supplied by Bayer CropScience (Mohnheim, Germany). The test protocols were validated in subsequent WHO consultations once results were consistent and reproducible across three independent laboratories for a given insecticide (see details in [22]).
After validating the test protocols, we conducted a multi-centre study with serial concentrations of the insecticides to generate concentration-response data for each insecticide-species combination. For insecticides with a killing action, the first step of the study (Step 1) comprised of preliminary bioassays to scope the range of insecticide concentrations that would kill between 0 and 100% of the tested mosquito colonies using a small number of mosquitoes (n = 50 per concentration). Based on Step 1, a range of 6 insecticide concentrations were chosen to produce a complete set of bioassays in triplicate (Step 2) (n = 300 per concentration). Based on the results, we estimated the lethal concentrations that would kill 50% and 99% (LC50 and LC99) of the tested mosquito colonies using a statistical model specifically developed for analysing complex toxicological datasets (see details in Statistical analysis section). For pyriproxyfen, that inhibits or reduces the fertility and fecundity of adult female mosquitoes, we estimated the concentrations that inhibit oviposition by 50% and 99% (OI50 and OI99) by the end of the observation period.
WHO bottle bioassay procedure
Preparation of stock solutions
The following insecticides were tested using high purity technical-grade active ingredients: transfluthrin (99.2%), flupyradifurone (98.4%) and clothianidin (99.4%) from Bayer CropScience (Mohnheim, Germany); metofluthrin (96%), prallethrin (93.3%) and pyriproxyfen (99.6%) from Sumitomo Chemical Co. Ltd. (Tokyo, Japan); and chlorfenapyr (100%) from BASF (Ludwigshafen, Germany). The initial stock solutions of each insecticide were prepared by diluting them in analytical grade acetone, while for clothianidin and flupyradifurone the stock solutions were prepared by dissolving them in a mixture of acetone and MERO® (0.903 density) according to the manufacturer’s instructions to prevent crystallization (Table 3). After a preliminary experiment to determine sublethal concentrations, MERO® was used at 1500 ppm (equivalent to 170 µl MERO® mixed in 100 ml of acetone) or 800 ppm (equivalent to 89 µl MERO® mixed in 100 ml of acetone) concentration to prepare the stock solutions for coating bottles for testing Aedes spp. and Anopheles spp., respectively. For testing with An. albimanus, the concentration of MERO® was reduced to 200 ppm (equivalent to 22 µl MERO® mixed in 100 ml of acetone) to avoid high mortality in control mosquitoes. The glass bottles with stock solutions were then wrapped in aluminium foil to avoid exposure to UV rays in sunlight and closed with tightly fitting caps to prevent evaporation of acetone before being stored at 4–6°C until use. From the stock solution ten-fold serial dilutions were then prepared.
Table 3
Optimised test conditions and specific endpoints for each insecticide and mosquito species in the WHO bottle assay.
Insecticide class
|
Insecticide
|
Mosquito species
|
Bottle drying time (h)
|
Exposure time (h)
|
Recording time (h)
|
Surfactant and solvent control
|
Endpoint
|
Pyrroles
|
Chlorfenapyr
|
All Anopheles species
|
24 h
|
1 h
|
72 h
|
Acetone
|
Mortality
|
Neonicotinoids
|
Clothianidin
|
An. gambiae, An. funestus, An. stephensi, An. minimus
|
24 h
|
1 h
|
24 h
|
Acetone + MEROa 800 ppm
|
An. albimanus
|
24 h
|
1 h
|
24 h
|
Acetone + MEROa 200 ppm
|
Ae. aegypti, Ae. albopictus
|
24 h
|
1 h
|
24 h
|
Acetone + MEROa 1500 ppm
|
Butenolides
|
Flupyradifurone
|
An. gambiae, An. funestus, An. stephensi, An. minimus
|
24 h
|
1 h
|
24 h
|
Acetone + MEROa 800 ppm
|
An. albimanus
|
24 h
|
1 h
|
24 h
|
Acetone + MEROa 200 ppm
|
Ae. aegypti, Ae. albopictus
|
24 h
|
1 h
|
24 h
|
Acetone + MEROa 1500 ppm
|
Pyrethroids
|
Transfluthrin
|
All Anopheles and Aedes species
|
24 h
|
1 h
|
24 h
|
Acetone
|
Prallethrin
|
Ae. aegypti, Ae. albopictus
|
24 h
|
1 h
|
24 h
|
Acetone
|
Metofluthrin
|
Ae. aegypti, Ae. albopictus
|
24 h
|
1 h
|
24 h
|
Acetone
|
Juvenile hormone mimics
|
Pyriproxyfen
|
An. gambiae, An. funestus, An. stephensi
|
2 h
|
1 h
|
72 h for mortality; 7 days for oviposition b
|
Acetone
|
Oviposition inhibition
|
a MERO: 81% rapeseed oil methyl ester (Bayer CropScience)
|
b The 7-day period includes a 72-h holding period in which mosquitoes are kept in paper cups to record mortality, followed by an additional 96 h of individual chambering of surviving females to record oviposition.
|
Process of coating and drying of bottles
Wheaton® bottles with a volume of 250 ml were coated in the testing laboratories according to the CDC guidelines [23]. Each bottle and its cap were coated with 1 ml of insecticide solution by rolling and inverting the bottle until all visible signs of liquid had disappeared. In parallel, a control bottle was coated with either 1 ml acetone alone or with 1 ml mixture of acetone and MERO® according to the solvent used for the compound. After coating the bottles were opened and left horizontally in the dark to dry for 24 h, except for pyriproxyfen coated bottles that were dried for only 2 h before being used for testing.
Test conditions
To avoid any influence of the environmental conditions on the test results, mosquitoes were maintained at 27° ± 2°C temperature and 80% ± 10% relative humidity during the exposure and holding periods.
Test procedures for insecticides with a killing action
The test conditions for bottle bioassays of each insecticide are summarized in Table 3 and the detailed Standard Operating Procedure (SOP) is available from the WHO website [24]. Briefly, 100 non-blood-fed females, aged 3–5 days (4 replicates of 25 mosquitoes each) were exposed to a range of serial concentrations with at least 5 concentrations for 1 h and two replicates of 25 mosquitoes were included as a control. After the exposure, mosquitoes were gently removed from the bottles using a mechanical aspirator and transferred into paper cups covered with netting, and provided with cotton pads soaked in 10% sucrose solution. Knockdown was recorded at the end of the 1 h exposure and mortality was recorded at 24 h post exposure, except for chlorfenapyr for which a 72h holding period post exposure was found necessary for recording mortality in mosquitoes.
Test procedures for insecticides with sterilizing properties
Test conditions for pyriproxyfen, a juvenile hormone mimic with sterilizing properties, are summarized in Table 3, and the SOP is available from [25]. For pyriproxyfen, for which the outcome is oviposition inhibition, only 5 to 7-day-old, blood-fed female mosquitoes that were allowed to mate with healthy males for 2–3 days in colony cages prior to blood-feeding were used. The females were allowed to blood feed for 1 h prior to the test. Briefly, 100 female mosquitoes were exposed in four batches of 25 for 1 h to each of the serial insecticide concentrations or a control. After exposure, mosquitoes were gently removed from the bottles using a mechanical aspirator and transferred into paper cups and provided access to cotton wool pads soaked in a 10% sucrose solution. Mortality was recorded up to 72 h after the initial 1 h exposure. After 72 h, surviving females were individually kept in paper cups for another 96 h and the proportion of females laying and the number of eggs laid by each female were recorded for both the control and the treatments.
Endpoints for susceptibility testing
For insecticides with a killing action, the 24 h mortality, or 72 h mortality for chlorfenapyr, were used as the final endpoint for insecticide susceptibility. The mortality of the test sample or control was calculated from the sum of dead mosquitoes across replicates expressed as a percentage of the total number of mosquitoes exposed. If the mortality in the control was ≥ 20%, the test was discarded and repeated. When control mortality was > 5% but < 20%, the test mortality was corrected with control mortality as part of the model fitting process, using methods analogous to the Abbott’s formula as per WHO guidelines [13]. For pyriproxyfen, the oviposition inhibition rate (OI%) was calculated as the proportion of egg-laying females exposed to pyriproxyfen against the those in the control, assessed at 7 days after the 1 h exposure. The total reduction in oviposition rate was obtained by calculating the percentage reduction in the number of females that laid eggs in treatments versus the number of females laying eggs in the control for each pyriproxyfen concentration. If the oviposition rate at the end of 7 days after exposure was < 30% in the control mosquitoes, the test was discarded and repeated.
Statistical analysis
Fitting concentration–response relationships
A binomial model using a 5-parameter logistic function was developed to analyse intensity bioassay data from the WHO multi-centre study. The same framework was used to analyse data from the killing bioassays and those which affected the fecundity of mosquitoes. In both cases, a binomial sampling distribution was used to describe the outcome following exposure to control or insecticide treatment:
where \({y}_{i}\) is the number of mosquitoes that died or were inhibited from ovipositing in bioassay \(i\); \({n}_{i}\) is the number of mosquitoes tested in the bioassay; and \(0\le {p}_{i}\le 1\) is the mean proportion of mosquitoes that died or were inhibited from ovipositing, which is assumed to follow a logistic curve:
$${p}_{i}=D+ \frac{A-(D- Z)}{{[1+{e}^{B\bullet (\text{ln}\left({x}_{i}\right)-C)}]}^{E}}.$$
Here, \({x}_{i}\ge 0\) is the insecticide concentration and \(A\), \(B\), \(C\), \(D\) and \(E\) are all parameters that influence the shape and position of the logistic curve, while A, B, D and E are being strictly non-negative. For each unique intensity bioassay, or set of bioassay runs in each laboratory, for each species and insecticide, the parameters were estimated by fitting the model to each data point within that group. In all runs, parameters A and D were set to 0 and 1, respectively, so that the resulting dose–response curve represents the estimated mortality without the background mortality, Z.
The model was fitted using a Bayesian framework. Priors were defined as B ~ N(3,1), C ~ N(3,5), E ~ N(3,5), \(Z\) ~ N(0,5). Individual parameter values for each curve can be found in Additional file 1: Table S1. The model was run using the probabilistic programming language Stan [26] in R v4.0.2 [27]. The model was run on 4 Markov chains for 5000 Markov Chain Monte Carlo (MCMC) iterations (or 10 000 iterations if the model did not converge at 5000) with 50% of iterations discarded as a warm-up. Non-convergence indicates that the MCMC sampler has not managed to sample from the true population distribution and may, therefore, bias the resulting estimates. Increasing the number of iterations while running the sampler enabled for reaching convergence more easily. However, this is computationally more expensive and so was only performed on the runs where convergence was not initially reached. The model was fitted to all bioassays from each laboratory for each unique combination of insecticide, and species tested, generating one curve for each laboratory, insecticide and mosquito species. Uncertainty around this estimate was generated from the range of concentration–response curves provided by each bioassay. Within each laboratory, a curve was fitted to each individual bioassay replicate and the minimum and maximum estimates around the point estimate were used to capture the uncertainty. A single curve indicates that there was only a single replicate for that combination of insecticide and mosquito species. Curves were plotted by extrapolating across the observed range of insecticide concentrations that were tested across all laboratories.
If multiple institutions provided data for specific insecticide and species combinations, the mean of the different institutions’ fitted estimates was computed when generating statistics at the insecticide and species level.
Model variability
Model variability was estimated by computing the absolute difference in mortality or oviposition inhibition of each individual data point from the best fit line for each MCMC iteration at the insecticide, species, and location level. The median of all iterations for each insecticide, species and location was computed to generate a measure of variability, which is interpreted as the percent variability in mortality or oviposition inhibition from the best fit line. This measure allows the amount of within-bioassay variability to be quantified in terms of mortality or oviposition inhibition. Estimates were generated for each mosquito species-insecticide combination with the variability reflecting the average within-bioassay variability between different laboratories with its 95% confidence intervals.
Where multiple institutions provided data for specific insecticide and species combinations, the mean of the different institutions’ fitted estimates was computed when generating estimates for the insecticide and species levels.