Study area
The study was carried out in Vallée du Kou, Burkina Faso (11˚23' N, 4˚24' W, Fig. 1), in an irrigated rice field area. The site is characterized by wooded savannah across 1,200 ha and contains seven discrete villages. The mosquito population in the village is resistant to insecticides and so a solution is required. Relatively high mosquito densities are observed annually during August and September, corresponding to the peak of the rainy season. Anopheles coluzzii is predominant throughout the year and An. gambiae is observed toward the end of the rainy season (frequencies fluctuating between five and 20%). Both species are highly resistant to pyrethroids and DDT (kdr frequency (0.8–0.95) [9, 37–39].
Study Design And Period
The study was carried out in three sequential phases that encapsulated product optimisation through to large-scale field deployment as follows:
-
Selection of the most effective PPF dose (May 2015, in the laboratory)
-
Trap manufacturing and impregnation (May – June 2015)
-
Large-scale field evaluation of the LFET (July to October 2015, in the field site.) (Table 1: Gantt chart of large-scale field evaluation of LFET)
Two villages, Vallée du Kou 3 (VK3) and Vallée du Kou 5 (VK5), were selected as intervention village (IV) and control village (CV), respectively. The control site was selected as it presents the same ecological characteristics in terms of mosquito densities and species as VK3, and was located 1 km away, which helped minimize any potential PPF contamination.
Prior to trap fabrication, a general survey of the houses and windows was conducted in VK3 using a Global Positioning System (GPS) [40] to count all the inhabited houses. Traps were then produced and deployed across the entire VK3 village.
Phase I: Selection Of The Most Effective Pyriproxyfen-dose
Laboratory mosquitoes
Mosquitoes were maintained at the Institut de Recherche en Sciences de la Santé (IRSS) insectary under controlled conditions (Temperature 27 ± 2 o C and relative humidity 80 ± 10%). For testing purposes, female and male mosquito cages were set up with pupae on the same day, maintained together until the day of testing (three days later at least), when females were removed from the cage. This time allowed them to mate before testing. Previous work done on the same laboratory colonies from the IRSS insectary showed an insemination rate of > 90%. Prior to cone testing, some females were checked to calculate the insemination rate [29]. Spermathecae were dissected under a magnifying glass 24 hours before bioassay to check this. Females were then blood fed when it was shown that the insemination rate was up to 90%. Three- to five-day-old female mosquitoes were starved for six hours by removing the 5% glucose (weight/volume, w/v) solution prior to blood feeding. The blood feeding took place at 18:00 for 45 minutes using direct rabbit feeding in the laboratory. The cage was blood-fed with males still inside. The non-blood-fed females were sorted and discarded, and only blood-fed mosquitoes were kept overnight and provided with glucose solution in cotton balls for the experiments. The laboratory mosquitoes were made up of susceptible An. gambiae-Kisumu, resistant laboratory strain An. coluzzii, and field An. gambiae s.l. collected at larval stage from breeding sites in both villages.
Effects of selected pyriproxyfen doses on mosquito fecundity and fertility
Prior to impregnating the trap net, preliminary laboratory testing with two doses of PPF liquid (10EC) was conducted to assess PPF efficacy on susceptible An. gambiae-Kisumu. After this, the most effective sterilizing dose was selected. A piece of net was impregnated at 20 mg/m2 or 30 mg/m2 of active ingredient (ai) selected according to previous studies [41, 42]. Three- to five-day-old blood-fed female mosquitoes were put in contact with impregnated pieces of net in WHO cones for three minutes. Then, these mosquitoes were transferred individually into cups (200 mL) containing a filter paper for oviposition 24 hours post exposure. A daily manual count of the number of eggs laid was performed to evaluate the fecundity. In cups where eggs were observed, water was added in order to hatch the eggs. The number of larvae were also counted. After the subsequent three days following oviposition, the number of females that laid eggs, and those that died before laying were counted. The fecundity and fertility of treated mosquitoes were assessed as compared to a control under laboratory conditions. It was shown that the PPF dose with the most desired effect on the tested mosquitoes was 30 mg/m2 ai. This was also the case in previous studies [29, 41, 42].
Phase Ii: Trap Manufacturing And Impregnation With Pyriproxyfen
Trap manufacturing
The traps were made from a metal frame (69 × 51 × 82.5 cm) and were fitted from the bottom to the top with a regular mosquito net to prevent any mosquitoes or other insects from escaping the trap once inside (see Sanou et al., 2021, [31] for more details) (Fig. 2). All of the windows of the inhabited houses (non-inhabited house windows were secured simply with a net, to reduce the number of mosquito-resting sites) were counted and measured to manufacture the traps accordingly. Each manufactured net covers the trap entirely, which itself fits into the window perfectly. Each trap’s net also has a sleeve for easy access, to open and/or close the window.
The metal manufacturer first produced a sample of each trap size. Then, the remainder of the traps were produced, painted with neutral oil (Fig. 3a), labelled according to dimensions, before transportation to VK3. The trap samples were also sent to a tailor in order for the nets to be made to size (Fig. 3b).
Trap impregnation with pyriproxyfen
After the traps were manufactured, they were impregnated with PPF. Through weighing the equipment, the amount of water and PPF solution needed for each net size was determined (Supplementary Table 1) according to the WHO insecticide impregnation process with pyriproxyfen 30 mg ai /m2 (PPF). For each set of dimensions, calculations were made to ensure the right amount of insecticide and water necessary to entirely coat the net (Supplementary Table 2). After the nets were coated, all trap nets were dried overnight indoors and wrapped into labelled sachets for easy identification prior to their transfer into the field for installation by the team and local workers (Fig. 3c).
Phase Iii: Large-scale Field Evaluation Of The Lehmann Funnel Entry Trap (Dup: Abstract ?)
Experimental design
The large-scale field trial was designed with one intervention area and one control area, and the entomological endpoints were assessed simultaneously in both villages after the traps were deployed. The villages selected were: Vallée du Kou 3 (VK3) for the intervention village (IV) and Vallée du Kou 5 (VK5) for the control village (CV). These two villages were selected because they had similar ecological properties in terms of mosquito densities and species [43, 44] and were situated 1 km apart, which minimizes any potential PPF contamination between the two sites. A general survey of the houses and windows was conducted in VK3 using a Global Positioning System (GPS) [40] to count all the inhabited houses. Each household was georeferenced (Fig. 4a, b, c).
Deployment and installation of the traps in VK3
For trap installation, the village was divided into six line-bands from the north to the south separated by green areas used for circulation in the village (Fig. 4d). All eaves and holes in the houses were blocked using cloth or sponge and a curtain was placed at each door by a large team of local and technical workers. In total, 1,313 traps were placed within the windows of houses to intercept incoming mosquitoes, and a new curtain made from regular cloth was provided to each house in VK3 to block mosquitoes from entering through the door. No constraints were required on the use of the doors or windows, and occupants were free to go to bed at any time.
Mosquito collection
To assess the trap performance, a monthly mosquito collection was performed from 12 selected traps (trapped mosquitoes) and matching houses (indoor resting mosquitoes) in VK3, while only indoor resting mosquitoes from eight houses were collected in VK5. Single-room houses (a single house with one window and one door) were randomly selected according to their geographic location (central, north, east, west, and south) in both villages. Houses were located far from each other, spaced at least 10 m apart to avoid human attractivity bias, and were monitored over nine days per month in both villages from July to October 2015. Mosquitoes were manually collected (on 36 collection days over the four months of the trial) with mouth aspirators in the traps and matching houses (for two hours) by three experienced collectors. The mosquito collection was simultaneously carried out in VK3 and VK5.
To provide evidence of the impact of the traps being deployed on mosquito density reduction at village level, a pyrethrum spray catch (PSC) [45] was carried out simultaneously one day per month over the four month-trial, in 10 randomly selected houses in each village. These houses were different ones to the regular study houses in the villages.
Assessment of mosquito species identification and allelic frequencies of kdr mutation in trap and house collection from Vallée du Kou over the study period
Collected mosquitoes (traps and houses) were morphologically identified to genus, species, and physiological status [46], and then counted. A sub-sample was then preserved in 80% ethanol vials for subsequent genotyping to species level and to check on the frequency of the knock down resistance (kdr) mutation [47, 48].
Assessing wild female mosquito parity during trap deployment
In addition, to assess mosquito population age structure [41] from July to October 2015, around 55 unfed female An. gambiae collected from the traps and houses were dissected and classified into parous and nulliparous mosquitoes, according to Detinova’s protocol [49].
Efficacy bioassays
The efficacy of PPF-treated traps was assessed on two mosquito strains including susceptible An. gambiae-Kisumu and wild adult An. coluzzii collected at larval stage.
The susceptible An. gambiae-Kisumu strain mosquitoes were released into traps to assess how long a PPF-treated net effect can last, and to follow the degradation dynamic over time. The morning following the blood meal, between 06:00 and 07:00 am, mosquitoes were transported by car while covered with a wet cloth (to maintain humidity) to the field. Once in the field, one hour rest time was given to the mosquitoes prior to the release. This rest time coincided with the monthly mosquito collection, where traps were emptied. Around 150 blood-fed female An. gambiae-Kisumu were then released 12 h after a blood meal once per month in VK3 (~ 50 females in each of two out of 12 monthly selected monitored traps) and in VK5 (50 females into one house). Mosquitoes were allowed to be in contact with the trap net for 30 mins, after which all of these mosquitoes were recaptured and brought to the laboratory for oviposition (fecundity) and egg hatchability (fertility) assessment. Only valid mosquitoes were oviposited. Two releases (July and August 2015) of the An. gambiae-Kisumu susceptible strain were performed over the study period. After each release-recapture process in VK5, an indoor spray with insecticide (Kaltox Paalga, SAPHYTO, Burkina) was performed to kill the non-recaptured mosquitoes.
Anopheles gambiae s.l. mosquito larvae were randomly collected from breeding sites in VK3 and VK5 to evaluate the effect of PPF over time on the wild mosquito population. These larvae were reared at the IRSS insectary until adulthood. Three to-five-day-old female mosquitoes were processed in the same way as described in the above section prior to their release. A total of 100 blood-fed female An. gambiae were simultaneously released in VK3 (in two selected traps in the third and fourth regular monthly collection traps) and in VK5 (50 blood-fed females into one house), once a month over the two-month trial.
Effect of pyriproxyfen-treated traps on wild blood-fed mosquitoes
Blood-fed mosquitoes were collected from traps and houses in VK3 and VK5 to assess whether females that encountered PPF-treated traps overnight had reduced fecundity and fertility. These blood-fed mosquitoes were collected from each of the 12 randomly selected traps once per month over three months in VK3. Similar mosquito sampling was performed inside eight houses in VK5 and considered as controls. About 25 blood-fed female mosquitoes per trap per day in VK3 and the same per house in VK5 were morphologically identified to species. They were then brought back to the IRSS laboratory on the same day and were allowed to lay eggs into a single cup containing Whatman filter paper and 5 mL of water for a week. Cups were checked daily, and females that laid eggs were subsequently removed, killed, dried, and preserved into silica 1.5 mL cryotubes labelled for subsequent analysis. Eggs were counted under stereomicroscope and then hatched into rearing trays (43 × 26 × 15 cm) filled with 1 liter of tap water with TetraMin baby fish food (TetraMin®, Germany).
Furthermore, to evaluate the impact of PPF on egg development as compared to a control as previously described [29, 30], about 100 field blood-fed An. gambiae s.l. were randomly collected from the monthly collection traps in VK3 and compared to about 100 blood-fed females collected from VK5 testing houses. These mosquitoes were kept individually in 20 mL cups with Whatman paper and 5 mL of water, and around 30 of them were dissected per day at 24 h, 48 h and 72 h post-collection over three months.
Physical conditions and cleanliness of the traps during the trial
Immediately after trap installation, 50 nets were sampled and were checked for their physical integrity in 50 selected households over two months (one and two months after installation). The assessment of the integrity of the fabric was performed by visual examination without removing the nets. Any holes observed were assigned to one of four size categories according to WHO guidelines [50]: a hole size of 0.5-2.0 cm or ‘< a thumb-sized opening’; a hole size of 2.0–10.0 cm or ‘> a thumb but < a fist’; a hole size of 10–25 cm or ‘> a fist but < a head’; and a hole size of > 2.5 cm or ‘> a head’.
General trap integrity was assessed based on two measurements:
-
The proportion of nets with any observed hole(s). The integrity of the nets was determined by counting the number of tears and holes as described: total number of coded nets with at least one hole of size (1-4) ×100 / total number of nets assessed in surveyed households.
-
The proportionate holes index (pHI) for each net, calculated as the sum of the holes weighted by size for each net. For this group, the weights used to calculate the pHI were 1, 23, 196 and 576 as described below: pHI = 1 × number of size − 1 holes + 23 × number of size − 2 holes + 196 × number of size − 3 holes + 576 × number of size − 4 holes. To better correlate the holes index to an integrity status (net condition) for each sampled net, the pHI was categorized into ‘good’ (pHI ≤ 64), ‘serviceable’ (pHI ≤ 768) and ‘replace’ (pHI > 768).
The trap net dirtiness (/cleanness) was also evaluated, and nets were classified and categorized into ‘clean’, ‘a bit dirty’, ‘dirty’, and ‘very dirty’. When the net was deemed irreparably damaged by dirtiness, it was replaced by another net of the same size.
Socio-anthropological investigation on the use of the traps
Qualitative and quantitative surveys were conducted from March to August 2015 from the beneficiaries of the LFET traps in VK3. The qualitative survey consisted of individual interviews with members of the community about their perception of the traps. There was direct observation of trap being used in the village. The usage of traps was witnessed elsewhere. The quantitative survey covered 276 inhabitants and was based on the level of acceptance of the trap by its users, the trap’s perceived effectiveness, and the limits of the trap.
In addition, all inconveniences reported by users of the traps were recorded by a social worker and reported for subsequent remedial measures. A follow-up survey according to the WHO indices [50] was conducted once the traps were installed, any mishandling was also reported, and action was taken to resolve any problems raised either by a social worker or the users themselves.
Parameters Measured And Statistical Analysis
The outcomes of the large-scale field trial were:
-
Mosquito density – calculated as the number of mosquitoes caught in the trap out of the total number of mosquitoes collected in the trap in the matching house (house where a LFET was placed and where the daily mosquito collection was performed)
-
Direct mortality – calculated as the number of dead mosquitoes caught in the traps out of collected mosquitoes from the trap.
-
Allele frequency – calculated as the number of resistant mosquito homozygote resistant (2nRR) and hybrid resistant (nRS) out of the 2x total of homozygote resistant (RR), and susceptible (SS), and RS the hybrid Resistant-susceptible.
-
Age structure (parity) – defined as the number of female mosquitoes having laid eggs (parous) and nulliparous out of a total of dissected mosquitoes.
-
Physical integrity of the traps – defined as a check of how a trap’s net was maintained in the intervention village. This physical integrity data was analysed using descriptive statistics (mean, median, interquartile range) to compare pHI values.
-
Acceptability – defined as a socio-anthropological investigation on the use of traps by end-users and calculated by counting the “Yes” or “No” answers to the relevant questions, relating to the total number of interviewees.
Descriptive data was summarized, inputted, and cross-checked in Microsoft Excel 2007 (Microsoft®, New York, USA), and R-4.0.4 was used to produce tables, graphs, counts, means and standard errors.
A Generalized Linear Mixed Models (GLMM) with a Poisson or negative binomial distribution was used to choose the suitable distribution of the mosquitoes collected in the traps and houses from VK3 and VK5 respectively. A zero inflated Poisson mixed regression modelling tool was used to estimate the intervention (trap) effect on daily numbers of mosquitoes collected while accounting for a possible spatial variation in terms of total mosquitoes collected between VK5 and VK3. This was to check for a possible difference in terms of physiological status (gravid, blood fed, unfed), or for a possible temporal variation induced by monthly weather conditions (rainfall, humidity, or temperature). Therefore, two models were built including two random-intercepts, one random coefficient and zero inflation terms. The model structure is defined with trap, gonotrophic _status, status, rainfall, and humidity considered as the fixed effects (fixed effects = trap + gonotrophic _status + status + rainfall + humidity) and the total number of collected female mosquitoes as the response variable.
Model 1: The probability of inflation terms was constant
Y ∼ fixed-effects + random¬ (month) + random (village) + zero-inflation (∼1) + residual error
Model 2: The probability of inflation terms was dynamic depending on the trap variable
Y ∼ fixed-effects + random¬ (village /month) + zero-inflation (∼trap) + residual error,
where Y is the total number of collected female mosquitoes.
The sub-model of model 2 was built on basic count, zero inflated, and altered models to account for zero values in the data. In these models, the rainfall and humidity fixed effects were removed to compare their Root Mean-Square Error (RMSE) and the Median Absolute Error (MAE) used to choose the best model. The lower the RMSE, the better the model. The model with the lowest AIC (Akaike information criterion) was considered the best model to fit the data [51]. Akaike information criterion and recent developments in information complexity were used as these methods consider both suitability and complexity of the model to check the performance of zero-inflated models.
To prevent the model from overfitting, trap data was split into a training set (0.8) and a test set (0.2). The different models were then tested on the training set and confirmed with the test set. For all analyses, the level of significance chosen was 5% (Supplementary Fig. 1).
To assess the dynamic of mosquitoes between VK3 and VK5, pyrethrum spray catch data were analysed using a non-parametric pairwise test, Anova-glmmTMB.
A one-way Anova was used to compare the age distribution of mosquitoes between VK3 and VK5.
The PPF effect on female mosquito fecundity was calculated as the mean number of eggs per female that contributed to the oviposition. The fertility was measured as the mean number of larvae per female that contributed to oviposition. Following this, a non-parametric pairwise test, Anova-glmmTMB, was used to compare the reduction of fecundity and fertility between the group exposed to PPF-treated net and the control group. As the number of eggs and larvae from the different treatments did not follow a normal distribution, a one-way non-parametric analysis of variance (Kruskal Wallis test) was used to determine whether there was any difference between 30mg/m2 and 20 mg/m2 ai doses in terms of fecundity and fertility reduction.
Weather related temporal measures such as the mean temperature, the mean relative humidity and rainfall, were collected from Burkina Faso’s meteorological station (http://www.meteoburkina.bf ).