This study was conducted under ambient conditions in a semi-field system measuring 29 x 21 meters built from a fabricated greenhouse frame modified to make two compartments with a central corridor and an opaque polyethylene roof for rain protection (Fig. 1)  . The semi-field system (SFS) is located at 6° 8’ S, 30° 37’ E at the Ifakara Health Institute in Bagamoyo district in Tanzania. Bagamoyo district experiences annual rainfall between 800 mm and 1000 mm, temperatures between 22ºC and 33ºC, and mean relative humidity of 73%. This evaluation followed the WHO Guidelines for efficacy testing of topical repellents .
This study was divided into two parts to accommodate the circadian rhythm of mosquitoes investigated. An. gambiae s.s. and An. arabiensis were tested at night and Ae. aegypti in the early morning. A pilot study for the first six hours after repellent application was conducted and found that a longer testing period was required. The final study tested the repellent from six to and twelve hours post-application.
A study for Anopheles
In the pilot study of Anopheles strains, repellents were applied at 17:45 hours and mosquito collection was conducted between 18:00 hours to 00:00 hours (Fig. 2A). In the final study, repellents were applied at 17:45 hours and mosquito collection was conducted between 00:00 hours to 06:00 hours (i.e. 6-12 hours after the application of repellents) (Fig. 2B).
Both the pilot and final study consisted of two fully balanced (3x3) Latin squares (LS) design conducted in two chambers of the SFS simultaneously over nine nights using six volunteers. In each LS, three volunteers rotated sequentially between the three mosquito collection positions each day inside the chamber and swapped repellents after every 3-days. After nine days of the study period, each volunteer had tested each of the repellents at each of the three mosquito collection positions inside the SFS chamber three times. The study flow plan for An. gambiae s.s. and An. arabiensis is shown in (Fig. 2).
A study for Aedes
For Ae. aegypti repellents were applied at 05:50 hours in the pilot study and at 23:45 hours in the final study. Mosquito collection was conducted from 06:00 hours to 10:00 hours in both the pilot and final study. The pilot study consisted of two fully balanced (3x3) LS conducted in two chambers of the semi-field system simultaneously over nine nights using six volunteers. In the final study, one fully balanced (3x3) LS was conducted in one chamber over nine nights using three additional volunteers. In both studies, volunteers rotated sequentially between three collection positions each day inside the chamber of the SFS and switched repellents after every 3-days. After the nine-days study, each volunteer had tested each of the repellents at each of the three mosquito collection positions inside the SFS chamber three times. The study flow plan for Ae. aegypti is shown in (Fig. 3).
This study used nulliparous female laboratory-reared mosquitoes, aged 5-7 days old, sugar starved and reactive to human odour. Mosquito strains used were pyrethroid-resistant (20% mortality with pyrethroid) An.arabiensis (Kingani, colonized Tanzania 2006), fully pyrethroid susceptible An. gambiae (Kisumu, colonized Kenya 1975) and fully pyrethroid susceptible Ae. aegypti (Bagamoyo, colonized Tanzania 2018). Mosquitoes were reared following the MR4 guideline . Before the experiment, An. arabiensis were lightly marked by placing them in a cup coated with a fluorescent dye to make them distinguishable from the morphologically identical An. gambiae. By very lightly marking the mosquitoes there is no significant effect on their fitness nor host preference . Mosquitoes were then sugar-starved for 8 hours. About 30 minutes before the experiment, 100 female mosquitoes that were responsive to human odour were selected and transported in boxes to the SFS chambers to acclimatize with the ambient environmental conditions.
MAÏA® ointment and 20% DEET were shipped to Ifakara Health Institute (IHI) Vector Control Product Testing Unit (VCPTU) in plastic jars by MAÏA Africa SAS . The amount of MAÏA® repellent ointment and DEET 97% (reference number 26028, lot number 2436308) received at IHI was 600 and 200 ml respectively. After the products were received, they were stored the same day at room temperature between 25º and 29ºC until used in the experiment. The 20% DEET in ethanol (V/V) was prepared in-house before the experiment.
Nine male volunteers between the age of 24 to 30 years were recruited after signing informed consent forms written in Swahili. All volunteers were tested for malaria parasite infection using SD BIOLINE Malaria Ag P.f  rapid malaria diagnostic kits before participating in the study and once per week during the study period as part of IHI health and safety procedures. Only male volunteers were recruited for cultural reasons as women would be expected to work overnight with men.
Allocation of volunteers
At the beginning of the study (pilot study), six male volunteers were assigned into two groups of 3 volunteers. One group was assigned to chamber ‘1’ and the other in chamber ‘2’ of the SFS (Fig. 1). Inside each chamber, three mosquito recapture positions were marked each 9 meters from the mosquito releasing point (Fig. 4). Each volunteer was randomly assigned to one of the three mosquito recapture positions inside one chamber and rotated between positions nightly. After nine days of experiments, it was discovered that the complete protection time of both repellents was above six hours. Therefore, the final study of 12hours recapture period was set up with an additional three volunteers recruited to run the Ae aegypti which was conducted immediately after the Anopheles experiment.
Application of repellents
Volunteers wore shorts and washed their lower limbs using water without soap before starting the experiment and they also wore closed shoes and a mesh bug jacket to ensure mosquitoes have access to the lower limbs only. Volunteers were non-smokers, and were requested to not drink alcohol or use perfumed soaps or ointment during the entire study period. We calculated a lower limb-skin surface area for each volunteer using the following formula at the beginning of the study; Area = ½ (K+A) x L
Where “L” represents the leg length between the knee and the ankle and “K” represents the circumference at the knee and “A” represents the circumference of the ankle area. The average lower limb skin-surface area of volunteers was 1259.2 cm2. The amount of repellents used on the day of experiment was measured using Ohaus CS200 weighing scale (Ohaus Corporation, USA). The average amount of MAÏA® repellent ointment applied per limb was 2.52 grams corresponding to a target dose of 2 mg/cm2 (approximately 0.3 mg of DEET per cm2). The average amount of 20% DEET in ethanol applied per limb was 2.4g corresponding to 1.9mg/cm2 (approximately 0.38 mg of DEET per cm2). All volunteers applied repellents using latex-gloved hands to minimize absorption onto the hands. Empty plastic cups that contained MAÏA® ointment were weighed to determine the amount left in the cups after application. Repellents for the Anopheles study were applied at 17:45 hours and for the Aedes study at 05:50 hours during the pilot study, while during the final study repellents were applied at 17:45 hours for the Anopheles study and at 23:45 hours for the Aedes study. In the final study, all participants rested with their trousers rolled up to prevent abrasion of the repellents after repellent application.
Duration of the study
Six-hour pilot test
During the first 9 days, 50 An. arabiensis and 50 An. gambiae were released in each of the SFS chambers and testing was conducted for 6 hours. During the 6 hours recapture period, no mosquitoes were recaptured by volunteers who applied MAÏA® and 20% DEET. Therefore, we extended the recapture period to 12-hours in order to confidently determine the duration of complete protection of MAÏA® repellent ointment and 20% DEET.
During the LS with 12-hours of recapture period, the same volunteers, chambers and rotation schedule was used for Anopheles experiment: 100 An. arabiensis and 100 An. gambiae were released (25 per release) starting at 00:00 hours. Three additional volunteers were recruited following study procedure and were assigned to the Ae. aegypti experiments in which 100 Ae. aegypti were released with 25 released every hour between 06:00 and 10:00hours.
Volunteers recorded the time of a first mosquito recapture of each experiment (Anopheles or Aedes) and placed in a separate cup labelled with the time of recapture, volunteer’s code, position and repellent (treatment) code. Volunteers collected subsequent mosquitoes that landed for a 6-hour collection period up to 12-hours post-application of repellent with cups labelled with repellent code, position and hour of collection. Cups were changed after every hour. At the end of recapture time, mosquitoes were killed by refrigeration at -4ºC for about 40 minutes and then sorted to species level. If fewer than 50% of mosquitoes were recaptured by a negative control volunteer, the data were discarded and the replicate was repeated. Temperature, relative humidity and wind speed were recorded on the day at the beginning of the experiment.
Data management and statistical analysis
Data were recorded in paper forms and then double entered and cleaned in Microsoft Excel 2016. Data analyses were performed in Stata 15.1 (Stata Corp, USA). Descriptive analyies of mosquitoes recaptured by repellents was performed. The mean complete protection time (CPT) of each repellent for each mosquito strain was estimated using the Kaplan-Meier survival analysis. The protective efficacy (PE) was established for the data collected up to 12-hours and calculated using this formula; P = ((C – T)/C)*100; where ‘P’ represents the percentage protection, ‘C’ represents the number of mosquitoes recaptured on the negative control (ethanol) and ‘T’ represents the number of mosquitoes recaptured on volunteer’s lower limbs treated with either MAÏA® or 20% DEET.
Statistical analysis was performed using a mixed-effects binary logistic regression to compare the protective efficacy between MAÏA® and 20% DEET (as the reference in the statistical model). Several models were tested using the proportion of recaptured mosquitoes as the outcome variable and repellent type (treatments), volunteer, position of the volunteer and time of recapture as fixed effects and day of test as random effects. The best-fit model was determined using the Aikaike’s Information Criterion (AIC) and the model with the smallest AIC value was selected. In the Anopheles data, repellent and time after application were fixed effects and the day of test was a random effect. In the Aedes data, the type of repellent, time of recapture and position of the volunteer were fixed effects and the day of test was a random effect.