The study area consisted of all 127 rural villages and the semi-urban town that were part of the extended health district of the Kossi region in Northwestern Burkina Faso, close to the Mali border. It stretched over a total surface of about 4,770 km2 and covered some 156,000 inhabitants. The study area is heterogeneous regarding its ecology. While the Northern parts towards the Sahel often feature sandy soils with high infiltration rates and lower numbers of environmental mosquito breeding sites, the picture is different in the South where there are more stagnant water bodies and wet rice growing areas. The eastern border of the district is characterized by wetlands around the Sourou valley.
The study was designed as a cluster randomized trial, administering different larviciding choices to mosquito breeding sites. Reporting followed the CONSORT guidelines for randomized trials where applicable. Three larviciding choices (i: untreated control, ii: treatment of all breeding sites and iii: risk map based larvicide application to only the most productive breeding sites) were performed within a total of 9 village clusters (Figure 1). Villages were clustered to avoid spill-over effects caused by the flight range of mosquitoes (19,20). Three clusters always represented areas that were similar in surface water availability, soil type, vegetation and other geographical factors (ecozone). Larviciding choices were randomly assigned to the predefined clusters, with the criterion to have each larviciding choice represented in each geographical ecozone (figure 1). The study comprised three years, a baseline year without intervention (2013) and two intervention years (2014 + 2015). Here we present results from the baseline and the first intervention year, in which the abundance of non-anophelines was seized. Larviciding with Bti VectoBac® WG, AM65–52 strain (Valent BioSciences Corporation, IL, USA) was performed during and after the rainy season in the public space in and around villages but not in private compounds. VectoBac® was diluted in pond water, that was filtered through cotton cloth and brought out onto the water surface using inox steel knapsack sprayers (Mesto®, Freiberg, Germany). Prior to the intervention we identified the optimum dosages for field application (21). Maps with all publicly accessible water bodies were generated during field visits for villages with exhaustive larviciding. For villages that received guided treatment, remote sensing derived risk maps for larval productivity were used (9).
- Adult mosquito monitoring
The primary outcome to assess larviciding efficacy was the abundance of different mosquito species. For the collection of adult mosquitoes Center for Disease Control light traps (Model 512, John W. Hock Company, Gainesville, Florida) were used. Indoor and outdoor captures were performed in 27 villages in 2013, and in 36 villages in 2014; additionally, the seven town quarters of Nouna were included. Light trap captures were performed every two weeks, following a rotating system with two independent fieldwork teams, covering 4 villages per night, resulting in a total of at least 10 sample rounds per village per rainy season.
Per village, three places were chosen regarding their central position in the village and in agreement with the household head, where a light trap each was positioned about one meter above the ground. Light traps were installed in a distance of approximately 100 to 150 meters from each other to detect possible local differences in vector abundance between different places within one village. The traps inside houses were installed near the sleeping places equipped with untreated bed nets, the traps outside were put beside the house within the common courtyard, where people sat in the evenings. Mosquitoes were collected between 18:00 and 06:00 hours to fully cover the peak biting period. Species determination was performed using microscopes, following the WRBU (Walter Reed Biosystematics Unit) identification keys (22).
Statistical analysis was performed using Stata/IC 14.2 for Windows (StataCorp LLC, 4905 Lakeway Drive, College Station, TX 77845, USA). Count of female mosquitoes collected per night per trap was used as the outcome variable. A negative binomial regression (Stata function “nbreg”), corresponding to a generalization of a Poisson distribution, was performed to model the count data, which showed over-dispersion. The random effect was integrated at village level.