Study areas
The study was carried out in two ecological settings, Vallée du Kou and Soumousso (Fig. 1), located near Bobo-Dioulasso, Burkina Faso. Vallée du Kou (11˚23' N, 4˚24' W) is a village located to the north-west of Bobo-Dioulasso, characterized by over 1,200 hectares of wooded savannah. It is a rice growing area with high mosquito densities throughout the year. The trial was conducted in one village of the seven neighbourhoods separated from one another by rice fields, Vallée du Kou 3 (‘VK3’). This site was chosen due to its proximity to the main tarred road. Given the presence of surface water all year round, mosquitoes are found with ease, with a peak density observed in August-September, during the rainy season. Anopheles gambiae and Anopheles coluzzii are present with a predominance of An. coluzzii throughout the year. Both species in this area are highly resistant to pyrethroids and DDT (kdr based mechanism, 0.8-0.95), and with an increase in ace-1 resistance frequency has also been reported [20,21].
Soumousso (11˚04' N, 4˚03' W), located in the south of Bobo-Dioulasso, in contrast with VK3, it is a drier setting where the dominant species are An. gambiae, and a mixture of mostly Anopheles funestus, Anopheles arabiensis, Anopheles coluzzii [22]. In this area, the mosquito density is lower compared to that of Vallée du Kou, and the dynamics of the mosquito population follow the two main seasons, with fewer mosquitoes in the dry season compared to the rainy season.
Description of the traps
The original LFET (prototype 1, P1) was designed by Diabaté and collaborators in 2013 [19] and was made from a metal frame (length = 69 cm, width = 51 cm, height = 165 cm) fitted with a regular mosquito net to prevent mosquitoes and other insects from escaping the trap once they enter it (Fig. 2). A funnel made from a metal frame was inserted at the top of the trap in such a way that mosquitoes approaching the window go first through the larger opening of the funnel and enter the trap through the small and rectangular opening. The first (large) opening of the funnel is 70 cm long and 54 cm diagonally, while the second (small) opening in bottom 13.3 cm long and large of 11.2 cm. The small opening of the funnel is 10 cm away from the backside of the trap. The funnel is inserted in the frame in a way that allows the mosquitoes to enter the trap easily but prevents them from escaping. Once the mosquitoes enter the trap, they have a large space beneath the funnel where they disperse. For a mosquito to escape, it would have to fly up towards the small opening of the funnel and navigate through the 10 cm space separating the small opening of the funnel and the back of the frame. Ultimately, mosquitoes continue to fly to exhaustion before finding a way out. The principle of the LFET is to confine the mosquito inside the trap until dehydration and death. For the purpose of this experiment, traps were fitted with three sleeves on the side (one below, one in the middle and one on the top) through which mosquitoes were aspirated. The trap was secured to the windows using nails.
The Medium (prototype 2, P2) is a smaller version of the original LFET. The funnel dimensions are similar but its height (82.5 cm) is half that of the original LFET (165 cm) (Fig. 3). Similarly, the whole trap was also covered with a net, fitted with three sleeves for mosquito collection.
The Prototype 3 (P3) was made using a metal funnel frame with the following measurements: length = 81 cm, width = 16 cm, height = 80 cm (Fig. 4). Here, the small funnel opening used as entry is a circular metal funnel, instead of rectangular as in the previous traps, with an opening size of 16.5 cm in diameter. Two circular openings on the right and on the left (the distance between both circular openings is 12.5 cm) were made to fit the window size. Depending on the space available around the window of the house, the trap position may allow the use of only one of the openings (either right or left). When one opening is used as entry, the second would be closed and covered with netting. The volume of the trap was made of a horizontal and rectangular metal frame (length = 110 cm, width = 16 cm, height = 56 cm). A net covered the trap on the backside of the funnel and was fitted with five sleeves allowing mosquito collection.
The prototype 4 (P4) is similar to P3 in terms of size and funnel type but has an additional circular funnel frame. This small (9 cm long) circular funnel gives access to the space beneath.
The distance between the end of the circular pipe and the back of the trap is 10 cm. In addition, P4 was equipped with a mirror for the inhabitants’ personal use, which would make it valuable for other functions beyond controlling mosquito populations. This may help increase the acceptability and sustained use of the trap (Fig. 5). As with P3, this trap was also outfitted with five sleeves enabling mosquito collection.
Study design and mosquito collection
Three of each of the new LFET prototypes (Medium/P2), P3 and P4) were tested in the two selected ecological settings along with three of the original LFET design (P1) for comparison purposes. The performance of the traps was assessed in terms of the number of malaria mosquitoes trapped as well as other mosquitoes entering the house through the window. A total of 12 houses, corresponding to 12 traps, were chosen in each site (a total number of 24 traps produced for both sites). Only houses with a single room, single window (similar size with a metal frame) and single door were selected for the study. Each of the houses were at least 10 m apart from each other. All of the traps were installed on the same day between 15:00 and 17:00 with a two-day rotation between houses according to a Latin square plan, to reduce biases linked to house inhabitants’ attraction. After installation, all the traps were simultaneously used, and checked every morning for eight consecutive days per month, for collection of all mosquitoes dead or alive in the traps for morphological identification. To ensure that mosquitoes had no other alternative except the windows to enter the house, small holes in ceilings and walls were blocked using sponges or cloth, and a curtain was placed at the entrance of each house (Fig. 6). The inhabitants were informed of the aim of the study and, therefore, free to use their doors as they wished. The window where the LFET was fitted was left open throughout to allow mosquitoes to enter the house. Traps were installed the day before prior to mosquito sampling over eight consecutive days per month, from September to November in VK3 and Soumousso.
Traps were emptied from 07:00 to 09:00, using a mouth aspirator in addition to pyrethrum spray catches [23] performed in the corresponding house. All the mosquitoes caught were kept in a single cup, then killed with chloroform and morphologically identified on site. The daily work on site consisted of mosquito collection, sorting, identification, sexing according to Gillies and De Meillon [24], counting and scoring per genus, species and physiological status (unfed, blood fed, gravid), and the numbers were recorded on a spreadsheet. All traps set up each month were removed at the end of the eight-day period.
Statistical analysis
Microsoft Excel 2007 (Microsoft®, New York, USA) was used to record the data, and R-3.6.2 (package dplyr, questionr, and coin) was used for statistical analyses and to produce the graphs.
In the study, three main variables were analysed: (i) T = Tt + Th, T = Total number of An. gambiae sensu lato (s.l.) collected in the trap and in the house, Tt = total number of mosquitoes collected in trap, Th = total number of mosquitoes collected in the house; (ii) P (%) = (Tt/T)*100, P = Proportion (%) of mosquito entry reduction in the house and (iii) Dr = T/24, Dr = daily removal of mosquitoes per site over 24 days.
The number of mosquitoes in the traps and the matching houses did not follow a normal distribution. Therefore, a one-way non-parametric analysis of variance (Kruskal Wallis test) was used to determine whether there was difference between the traps in terms of numbers of mosquitoes collected and to assess their overall performance.
The Post-hoc test was used for multiple comparisons of mean numbers of mosquitoes between traps (Bonferroni). In order to meet this Post-hoc test assumption, a Tukey test for multiple comparisons was used to confirm the results.
The effects of monthly collection in mosquito density reduction were evaluated using a Levene’s test for homogeneity of variance (centre = median) with a Wilcoxon rank sum test and a p-value bonferroni adjustment method. A Tukey multiple comparison of means with 95% family-wise confidence level test was used for pair-wise comparison between LFET prototypes.
To assess whether the traps caught more mosquitoes than those that entered in the matching houses, a comparison using Wilcoxon rank sum test with a holm p-value adjustment method was used.