Study design
The study had limited funding, and as such, it was important to consider various factors in developing the sampling. The development of the sampling strategy took place during a two-day meeting with the LF Programme officers and entomologists. Following extensive discussions, the least expensive option of using community volunteers as against trained entomologists was adopted for the study.
Factors considered in designing the sampling strategy
The factors we considered in designing the sampling strategy include financial and human resources, workload and the local epidemiological context.
Financial resource factors affecting the sampling strategy are: (i) available budget; (ii) distances to be covered and associated transportation costs; (iii) daily allowances for entomologists, community volunteers and supervisors; and (iv) sourcing and purchasing of consumables and supplies.
Human resource factors considered are: (i) availability and skills of experienced entomologists versus community volunteers; (ii) need for supervision of volunteers during collections; (iii) training requirements (best practices for mosquito collection and storage) and logistics (training venues, meals and transportation allowances); (iv) need to develop human capacity for future public health needs; and (v) security and safety of trained entomologists.
Sampling decisions were also based on workload: (i) one-time versus repeated mosquito collections (i.e. probability of catching an infected mosquito in a one-time collection as compared to repeated collections); (ii) number of households that can be sampled per day; and (iii) total number of mosquitoes required.
Finally, the regional, national, and local epidemiological context affecting sampling are related to: (i) vector transmission dynamics (impact of seasonality on mosquito species and LF transmission); (ii) number of transmission assessment survey (TAS) evaluation units (EU) to investigate (weighing relative benefits of sparse coverage of all EUs against comprehensive coverage of one EU); and (iii) relative emphasis on high-risk versus low-risk areas in the sample.
Study area
The study was conducted in the Savanes Region in the northern part of Togo (Fig. 1). Three of the five districts in the region, Kpendjal, Cinkassé and Tone, were previously endemic for LF. The survey was conducted in the three districts grouped into one evaluation unit (EU) because of their proximity to Ghana in the West, Burkina Faso to the North and Benin to the East, where transmission of LF is reported to be ongoing.
Selection of sampling sites
A two-stage sampling method was used to select the sampling sites. In the first stage, villages were chosen, and in the second, households (HHs) were selected within each village. [13]. All communities with population greater than 5000 were excluded from the sampling because the potential for transmission in urban areas is low [14, 15]. Thirty villages were selected in the EU with probability proportional to size.
In addition, eight additional villages previously known to have reported a microfilaremia positive case, either through monitoring and evaluation, TAS or passive surveillance, were also assessed. The last microfilaremia positive case was identified in 2015. One of the villages purposefully selected was also selected by probability proportionate to size; therefore, 37 villages were surveyed in total. The geolocation of each surveyed village was recorded.
Households were sampled to cover the entire village as much as possible, by the end of the sampling period. Each village was divided into four approximately equal sections. Households in each section were numbered consecutively and selected randomly (using a dice). The household head was approached, and consent sought. If a household refused to participate, a different household was selected. New households were selected for each sampling day, and households from which mosquitoes were previously collected were excluded from the selection unless the number of households in the section was exhausted.
Mosquito collection
This entomological study was undertaken over five months during the peak dry season (October 2016–February 2017), with the aim of collecting as many mosquitoes as possible. In each study community, mosquito collection was done twice every month. The estimated sample size was 2000 vector mosquitoes per IU, required to estimate an infection rate of 1% with a power of 0.80 [16].
In each village, community volunteers were identified and trained for mosquito collection and storage. Mosquito collections were primarily done using the pyrethrum spray catch (PSC) method. On each mosquito collection day, households were randomly selected from each section and mosquitoes collected, using the PSC method. The day before the collection, consent was obtained from occupants of the households, they were asked to keep bedroom doors and windows closed the following morning. Mosquitoes were collected early in the morning between 05:00 h and 08:00 h by two trained collectors. The occupants were asked to remove or cover all food items in the room. Potential mosquito hiding places (under the bed, tables) were searched and disturbed to displace any resting mosquito White sheets were laid on the floor and other surfaces in the rooms. The room was then sprayed with pyrethrum insecticide and left for about 15 min, after which the white sheets were inspected for any dead or knocked down mosquitoes.
In the eight purposefully selected communities, mosquito collection was also done using human landing collection (HLC) method and exit trap collection (ETC). Both HLC and ETC were undertaken in randomly selected households, different from the households where PSC was undertaken.
All collected mosquitoes were placed in a Petri dish labelled with the village code. Each Petri dish contained silica gel in a ball of cotton wool, to keep the mosquitoes dry. The Petri dishes from all villages were sent to the district on a specified date. Once a month, a central team visited all districts to collect the mosquitoes and deliver them to the entomology laboratory of the department of “Unité de Recherche en Ecotoxicologie (URET)” of Sciences faculty of the University of Lomé (Togo).
Sample processing
Mosquito genera and species identification were conducted at the entomology laboratory of the University of Lomé using morphological identification keys [17, 18]. All mosquitoes collected by the PSC and ETC were grouped in pools of 25 or less, according to the village and method of collection. The pooled mosquitoes were then sent to the NTD reference laboratory of the Parasitology Department, Noguchi Memorial Institute for Medical Research (University of Ghana) for molecular identification of W. bancrofti infection in the vector species. DNA was extracted from the mosquito pools using the DNeasy Tissue Kit (Qiagen, Valencia, California, USA), and molecular identification of W. bancrofti was done using the LAMP method [19,20,21]. All reactions included a positive (W. bancrofti DNA) and negative (water) control (Fig. 2) The positive control is W. bancrofti DNA extracted from microfilariae positive mosquitoes, in previous studies from Sierra Leone [20]. All positives were confirmed using the conventional PCR method for the determination of W. bancrofti infection [22].
Data analysis
The number of mosquitoes collected in each month was evaluated regarding the rainfall data for each collection month. The Poolscreen 2.0 software was used to analyse the pool screening results [23]. The survey costs were grouped into categories and presented in a table. Survey costs were divided into the following categories: personal allowance (research team and vector collectors), supplies, transportation, communication, and others. The unit cost per sample was estimated as the sum of the unit cost for collecting each sample and the unit cost for laboratory analysis.
All levels of statistical significance were determined at the 95% confidence limit. Graphs were drawn using GraphPad Prism 7 (GraphPad Software, Inc., La Jolla, California, USA) and Microsoft Excel (Microsoft Corporation, Redmond, Washington, USA). The geolocation data were imported into QGIS 2.18 (QGIS Geographic Information System, QGIS Development Team, Open Source Geospatial Foundation Project) for mapping.