This study aimed at developing a mosquito killing bait station that could be made at home using household materials. The ideal prototype was designed after investigating the optimal structure, substrate and deployment location of the bait. A stepwise approach to each design feature was performed using recycled materials, simple household utensils and ingredients that can easily be found in rural sub-Saharan Africa. First, the minimum effective concentration of ivermectin was determined, then different attractive fruit juices were tested, followed by different designs and last the final bait prototype was tested inside a semi-field system (SFS) to determine how effective the baits were when placed in different peri-domestic locations.
Determining the ivermectin LD90 against Anopheles arabiensis
Ivermectin was selected as the mosquitocidal ingredient, given its proven safety record in humans [20]. The endectocide’s mode of action makes it safer to vertebrates including humans as it targets glutamate-gate chloride channel present in invertebrates [20]. The channel does_not exist in vertebrates [21] and the drug has a low affinity for other mammalian ligand-gated chloride channels. Furthermore the drug is unable to cross the blood–brain barrier [21]. In addition, ivermectin is easily available and affordable in rural Eastern Africa, as communities commonly use it to deworm their cattle under the form of injectable Ivomec®. The drug is stable at room temperature and may be stored at 30 °C without losing its efficacy. Reports on ivermectin indicate there is gradual photo-degradation of the chemical when is in animal’s faeces [22]; however little information on whether ivermectin solution undergoes photo-degradation exists. Moreover the drug has a different mode of action to current insecticides highlighting the possibility of circumventing the issue of emerging insecticide resistance. There may be potential for cross-resistance between ivermectin and pyrethroid insecticides though little evidence of this currently exists and further investigation is needed [23, 24].
The minimum dose of Ivermec® needed to kill at least 90% of An. arabiensis was determined in the laboratory using dose response experiments inside standard insectary cages (30 × 30 × 30 cm). The mosquitoes used were An. arabiensis Ifakara strain, reared at 28 °C, 80% humidity and natural light conditions at the Ifakara Health Institute (IHI) insectaries in Bagamoyo, Tanzania. Larvae were reared on Tetramin® fish flakes, adults were maintained on 10% w/v glucose and blood feed on human blood for colony maintenance. The mosquitoes used in the dose response experiments were blood-naive, 3–6 days old and starved for 6–8 h before experiments. Serial dilutions of ivermectin in 10% w/v sucrose solution were prepared with reference to data collected by Allan [25], who reported that 0.014% of ivermectin in 10% w/v sucrose solution was sufficient to kill 90% of Anopheles quadrimaculatus. A food colouring agent (Carmoisin) was added to each dilution at concentration 0.5% v/v in order to easily identify if the mosquito had taken a sugar meal. Injectable Ivomec® was purchased at a “duka la mifugo”, local village shop selling veterinary and agricultural products, and used to create serial dilutions of ivermectin. A total of four replicates were performed for each ivermectin concentration after dilution: 0.2, 0.15, 0.1, 0.05, 0.025, 0.01, 0.005, 0.0025, and 0.001%. Forty female An. arabiensis were introduced into each cage. Containers with approximately 30 mL capacity were filled up to 2/3 with the test solutions and standard filter paper was rolled up like a tube and fit into the cup in a way that only the bottom part of the paper was in contact with the solution. The test solution then progressed up the filter paper through osmotic pressure thus allowing the mosquito to obtain a sugar meal from its surface. The test solutions were kept in the cages for around 12 h from 8 p.m. to 8 a.m. Mosquito knock down rate was observed at 3, 6, 24 and 48 h. All the mosquitoes that were no longer flying in the cage were removed with a syphon; their abdomens were squeezed onto white filter paper allowing the investigators to determine if the mosquito had sugar fed by visualizing the food dye that had been ingested. After each replicate the cage positions were changed to avoid any bias introduced due to cage positions.
Selecting the most attractive sugar concoction
In order to determine the most attractive sugar source to An. arabiensis to be used in ATSB; experiments were conducted in semi-field conditions during the months of March and April 2015 in Bagamoyo Tanzania. Average temperature was 28 °C, with minimum temperatures of 23 °C and maximum of 31 °C. A total of six cages (120 × 120 × 120 cm) were locally made using a metal frame, and screened with polyethylene net on all panels except for the bottom panel which was made of wood and lined with a light shade of vinyl floor sheet for easy visualization of knocked down mosquitoes. Each cage had a sleeve made of cloth on one of its net panels allowing easy access to the inside of the cage. Sugary concoctions were prepared using 10% sugar solution added to the pulp of the following locally bought fruits: papaya (Carica papaya), banana (Musa), tomato (Solanum lycopersicum), mango (Mangifera indica), orange (Citrus sinensis) and watermelon (Citrullus lanatus). Guava juice from Azam® was also purchased and tested based on previous studies conducted in Mali that had shown it to be attractive to sugar-seeking An. gambiae s.l. [14]. A control solution was also tested using only 10% sugar solution. Sugar baits were made using a used 0.5 L plastic water bottle, cut in half and lined with cotton cloth folded over its outer surface. When the concoctions were added to the bottle the liquid moved to the outer cloth through osmotic pressure creating a surface where mosquitoes could easily sugar feed. The baits were hung in the centre of each experimental cage. For each experimental round different fruit concoctions were compared to 10% w/v sucrose solution. A total of 40 starved female An. arabiensis were introduced into each cage. After 24 h mosquitoes were removed and feeding success was recorded by observing the food dye in the mosquitoes’ abdomens. Four replicates were done for each concoction type. This experiment was used to identify the sugary concoction that was most attractive to An. arabiensis and so most appropriate in a sugar bait.
Selecting the prototype design
Different bait/trap prototypes were designed using items commonly found in rural Tanzanian households. The aim was to create a prototype that effectively attracted mosquitoes searching for a sugar meal using basic domestic materials, easily available and with a minimal associated cost. Therefore, preference was given to materials that were commonly reused or thrown in the waste. Discussions were commonly held with locals regarding different materials that could be considered for this purpose. Three different prototypes, denominated A, B and C, were designed using materials such as cloth (“kanga”, “kitenge” and “tetroli”), different sized plastic bottles, string and pieces of sponge from old mattresses. Prototype A was made by cutting a 0.5 L plastic water bottle in half then lining it with cotton cloth folded over its outer surface, prototype B was made by cutting a 12 L bottle (‘maji ya uhai’) placing on its bottom a fitting piece of sponge and lined with black cloth; and prototype C was made by cutting 1.5 L bottle into half then the upper part of the cut bottle was seeped into the bottom part, the two parts were fixed with masking tape. The prototype C resembled the “honey trap” with the exception that it did not contain yeast. In order to maintain the CO2 production, yeast would need to be constantly added to the trap thus increasing the cost of the prototype making it unattractive to local communities.
Each type of prototype was assigned to one pair of experimental cage and tested using the best performing concoction determined in the previous objective. One prototype with ivermectin and another without ivermectin was placed into each cage in order to determine if mosquito feeding was deterred by ivermectin when using different prototype designs. A total of 40 starved female An. arabiensis were introduced into each cage. After 24 h mosquitoes were removed and feeding success was recorded by observing the food dye in the mosquitoes’ abdomens. Four replicates were done for each prototype.
Selecting the potential deployment location within the peri-domestic area
Experiments to determine the best deployment location of the prototype were conducted in a large screen house (22 × 29 m), also known as a biodome or semi-field system, during the months of May and June 2015, in Bagamoyo, Tanzania. The walls of the biodome are made of netting, which allows airflow, thus maintaining similar climatic conditions as outdoors but in a controlled environment (semi-field conditions). The biodome rests on a concrete slab surrounded by a narrow water trench that prevents ants and other animals from invading it and predating on the mosquitoes released during experiments. The biodome is roofed with polyvinyl sheets and divided into two compartments separated by a 29 × 4 m corridor. Each compartment contains two experimental huts (6.5 × 3.5 × 5.1 m). The experimental huts mimic traditional Tanzanian rural households in terms of size, structure and mosquito exit/entry points (eaves, windows and doors). Mosquito exit traps were fitted to the windows of the experimental huts and netting flaps were attached to the inner side of the eaves in order to funnel mosquitoes into the hut when entering it, but not allowing them to exit through the eaves. The exit traps worked in a similar way where mosquitoes are funnelled into the trap when attempting to exit the hut through the window but cannot return back into the experimental hut. Once a mosquito enters the hut the only way it can exit it is through an exit trap.
Two mattresses were placed inside each hut and volunteers were asked to sleep in the huts from dusk until dawn. In order to determine the effect of treated and untreated bed nets on mosquito sugar feeding; two huts in one biome’s compartment were given Olyset® LLINs and the other two in another biome’s compartment were given non-treated nets. Four potted Ricinus communis plants were placed at the midpoint of each compartment in between both experimental huts. The best performing prototype and sugary concoction determined in previous experiments was used for this experiment. A total of 24 sugar baits were made and deployed at different locations of the biodome. Each location type was assigned a different food dye in order to be able to recognize where the mosquito had fed: (a) eight sugar baits were placed indoors (2 per hut) containing red food colouring; (b) eight sugar baits were placed outdoors directly outside the huts containing blue food dye; and (c) eight sugar baits were placed outside amidst the R. communis vegetation containing green food dye (Fig. 1a, b). To maximize outdoor mosquito recapture three artificial resting boxes were randomly placed in outdoor locations. A total of 150 female mosquitoes were released each night in each of the compartments. All dead and alive mosquitoes in each collection site (exit trap, ceilings, resting indoors in baits, resting outdoors in boxes, on plants and outdoors in baits) were separately collected at 07:00. Alive mosquitoes were then knocked down in a freezer and inspected for presence or absence of food colouring in their midguts by squeezing their abdomens onto white filter paper. While inspecting the presence or absence of food dye in the mosquito midgut; observation on whether mosquito had half or fully sugar fed was investigated. A total of 16 nights were conducted, nets and baits were maintained in fixed locations and volunteers were rotated in order to control any bias caused by differences in individual attractiveness to mosquitoes.
Consideration on the ATSB safety
Into each prototype, one litre of ASB solution was sufficient to make the sponge mattress wet enough for mosquito to feed. In total 10 mL of 1% ivermectin was used to make 0.01% ivermectin concentration needed per prototype. Since this amount of ivermectin contained in one prototype is more than the ivermectin recommended dose for children with 15 kg, it is important to grill the prototype when is applied in the field in order to assure the ATSB trap’s safety. Currently there is inadequate information on the possibility of non-targeted organisms feeding on these traps; however more studies should be conducted to investigate the possible adverse effects of these traps to non-targeted organisms.
Data analysis
Data analysis was done with STATA version 13 (Stata corp, College Station, TX).
Ivermectin LD90 against Anopheles arabiensis
A binomial generalized linear model with a logit link was used to compare the cumulative proportion of knocked down mosquitoes for ivermectin concentration to control (10% sucrose solution) at 6, 24, and 48 h. Other sources of variation, such as experimental day, replicate, mosquito age and cage position in the insectary were included also in the model. Odds ratio and 95% confidence intervals were derived from the model.
Sugar concoction and prototype design
A binomial generalized linear model with a logit link was used to compare the proportion of sugar fed mosquitoes on each different sugar concoction containing ivermectin compared to control (10% sugar solution). Other sources of variation such as experimental day, replicate, mosquito age and cage position in the insectary were included also in the model. Odds ratio and 95% confidence intervals were derived from the model. The same analysis was done with the data from the different prototype designs using prototype A without the addition of ivermectin as reference.
Selecting the deployment location within the peri-domestic area
A logistic regression was done to compare the number of mosquitoes that took a full sugar meal or a half sugar meal between the three possible locations: indoors, outdoors close to the huts and outdoors close to vegetation. The model took into account the day as a random effect; the experimental hut and the net type as fixed effects. Sample size calculations for the ivermectin dose response experiments, selection of the most attractive concoction and selection of the best prototype experiments were carried out in a similar fashion. For each experiment, to be able to detect a minimum of 20% difference in outcome with an alfa of 0.05 and 95% confidence interval; it was calculated that a minimum of forty mosquitoes were needed per arm.