Evaluation of mosquito electrocuting traps as a safe alternative to the human landing catch for measuring human exposure to malaria vectors in Burkina Faso

Measurement of malaria transmission and evaluation of vector control requires estimation of human exposure to malaria-infected mosquitoes [1]. This exposure is often estimated in terms of the Entomological Inoculation Rate (EIR [2]) defined as the mean number of malaria-infected mosquito bites a person would be expected to receive in a given setting [1, 3]. Accurate estimation of exposure to mosquito bites is crucial for evaluating interventions, thus there is an urgent need for reliable and robust methods to give unbiased estimates of exposure in a range of settings [3]. Several methods have been used to measure mosquito host-seeking behaviour and human exposure to mosquitoes. Historically, the human landing catch (HLC) has been the most commonly used method for African malaria vectors and is considered a gold standard approach for direct measurement of human-mosquito contact in both indoors and outdoors settings [4]. In this method, human volunteers expose part of their body, usually the lower legs, to lure host-seeking mosquitoes that are then collected upon landing [4].

Although HLC provides a direct measurement of human exposure to bites, its estimates can be biased due to variation in the skill of mosquito collectors and their attractiveness to mosquitoes [5‐8]. HLC also raise ethical concerns as collectors are exposed to potentially infectious mosquito bites [9]. While this risk can be minimized by providing malaria prophylaxis to collectors, protection cannot be guaranteed in areas of drug resistance or where mosquitoes are carrying other pathogens, such as arboviruses [10, 11]. One African study indicated that HLC participants had no increased risk of malaria [12], but there remains a concerns about disease exposure in areas where other mosquito-borne pathogens are circulating.

Due to these limitations of the HLC, a range of alternative “exposure-free” methods have been developed. Most common is the CDC light trap [4, 13‐15], a trap that can be placed next to a person sleeping under a bed-net and used to collect mosquitoes that would have otherwise have fed on them [14]. Although effective and easy to use in indoor environments [16], this method is harder to implement outdoors and may not accurately reflect human exposure in this setting [16‐18]. Furthermore, CDC light catches can be affected by variation in the trap-light intensity [19, 20] and colour [16]. Other “exposure-free” methods include the human-baited double net trap (HDN) [18], Suna Trap [21], Host Decoy Trap (HDT; [22]), Ifakara tent trap design C (ITT-C) [23] and the Mbita trap [11]. Of these the last two have the same limitation as the CDC light trap of not being suitable or representative for measuring exposure in outdoor environments. For example, the tent trap only samples mosquitoes that are capable of entering a small enclosed structure, therefore, disproportionately catches indoor biting mosquito species [24]. The HDN was as efficient as the HLC in collecting outdoor anthropophilic mosquito. However, like the Tent Trap, it may also be selectively biased towards indoor biting mosquitoes, or sample vectors that enter the net to rest instead of biting [18, 25]. Similarly the Mbita trap had poor performance relative to the HLC in a setting where most vectors were exophilic and zoophilic [26]. Both the SUNA and Host Decoy Trap have shown promise for sampling outdoor biting malaria vectors [21, 22]; although may under [27] or overestimate [22] human exposure relative to the HLC. Given the growing recognition of outdoor biting as a major source of residual transmission in Africa [28‐30] there is a clear need for improved methods that can reliably and safely measure exposure outside of homes.

The mosquito electrocuting trap (MET) has been developed as a representative and safer alternative method to the HLC for measuring human exposure to mosquito vectors both indoors and outdoors [17, 31, 32]. As previously described [31], the MET builds on previous work using electrified nets and grids to trap flies [33, 34] and mosquitoes [35‐39] attracted to hosts or their odours. This trap consists of four panels that can be assembled into a box around the lower legs of seated human [17, 31] (Additional file 1: Figure S1), or an entire host (human or cow) [32]. Each panel consists of an electrified surface that allows free air movement and is safe to use in close proximity to a human volunteer, and intercepts and kills mosquitoes just before they land on hosts. An advantage of this method is that in addition to protecting participants from mosquito bites, it can be used in a standardized way in both indoor and outdoor environments. This method has shown promise as alternative to the HLC for sampling malaria vectors in Tanzania [17, 31, 32]. For instance, the first prototype achieved a sampling efficiency of ~ 60% relative to the HLC for sampling Anopheles arabiensis outdoors in rural Tanzania, falling to 20% when used indoors [31]. Further study on an improved prototype carried out in an urban area indicated the MET had a similar performance to the HLC [17]. A recent study evaluated a further prototype of the MET in which the electrified trapping panels were expanded to encompass the whole body of a human volunteer or calf [32], with the performance of the MET exceeding that of the HLC. The MET has not been tested yet outside Tanzania thus its effectiveness in different ecological settings is unknown. There is a need to evaluate the MET in west African settings where vector species composition, ecology and biting behaviour is often markedly different from East Africa and to see how its performance varies between sites and seasons.

A total of 41,800 mosquitoes were collected over 324 trapping days, of which 41,395 were females (Additional file 5: Table S2). Most of the female mosquitoes were anophelines (86.4%), with the remainder being culicines (Additional file 5: Table S2). Anopheles gambiae s.l. represented 97.7% of all anophelines, (Additional file 5: Table S2). Within the subset of An. gambiae s.l. individually analysed to species level (n = 3199, 36.3% of total), An. gambiae constituted 41.58%, An. coluzzi 58.17% and An. arabiensis 0.25%. No molecular identification of species within the Anopheles funestus group was performed because of the small number collected indicated this is not a major vector in the area (n = 35). There was seasonal variation in vector species composition, with the proportion of An. coluzzi within the An. gambiae s.l. complex varying from ~ 75% to ~ 44% between the dry and wet season (Additional file 6: Table S3).

Here the performance of the METs was evaluated as an alternative to the gold standard “HLC” for estimating human exposure to malaria vectors. This was the first time that the trap was evaluated outside Tanzania and in a West African setting. In general, the MET caught fewer An. gambiae s.l. than HLC with relative performance being higher in outdoor (52%) than indoor environments (43%). The overall efficiency (combining in and outdoors) of the MET (~ 46%) was similar to that described for first prototype trialled in rural Tanzania by [31], but below the near 100% relative performance reported with further prototypes tested in Tanzania [17, 32]. However, estimates of vector species composition, seasonal dynamics, biting behaviour (indoor vs outdoor) and malaria infections rates were generally similar between MET and HLC collections. This strengthens evidence that METs can provide a safe alternative to the HLC for characterizing attributes of malaria vector populations; even though they may require location-specific calibration for prediction of vector density.

It is unclear why MET performance was relatively lower in this study. However, several factors may account for this. One possibility is that the current study incorporated more intra-site variability. All previous work in Tanzania has involved evaluation at a limited number of fixed sampling points in a few sites. Here the METs were tested at multiple households across 12 different villages and noted considerable variation in MET relative performance between sites (17–100%). Thus, local characteristics of the study site may have a significant impact on trap performance. The relatively lower sampling efficiency of the MET here compared to Tanzania could also be due to operational problems that arose after the first batch of METs had been in continuous use for several months, exacerbated by wear and tear during the regular transport between villages (up to 100 km apart, on poor roads). These operational problems included short-circuits, and power supplier failure in addition to dipping in current/voltage, some of which may not have been noticed until traps failed. Although only data from days in which both MET and HLC collections were conducted was used for analysis, these faults indicate that the MET prototype may need further improvement for stable use over long periods of time. Additionally, there were small differences in trap design between the prototype used here and in Tanzania, which may have contributed to the reduced performance. For example in contrast to previous studies in Tanzania [17, 31], the MET prototype here used white non-treated net to protect the part of participant’s bodies that were not in the trap. It has been shown that An. gambiae s.l. are more attracted to traps with high visual contrast [22], and the use of white netting to protect participants here may have diminished the contrast between the trap and host bait compared to previous versions. Another factor that can make difference is the vector ecology and species composition. The major vectors in areas where the MET has been used in Tanzania is An. arabiensis [17, 31] whereas An. gambiae and An. coluzzi were the main vectors in our study area in Burkina Faso [42, 43]. Cuticular hydrocarbon composition (CHC) varies between Anopheles species [59‐61], and it is known that the electrical conductivity of insects can vary with their CHC, water content and body size [62]. Therefore, the variation in the MET performance between the current study and those carried out in Tanzania could also be due to local variation in vector species composition.

The results from the present study suggested METs performed better in outdoor ~ 52% relative sensitivity compared to the HLC) than indoor (~ 43%) settings. Earlier trials in Tanzania also found MET performance to be higher outdoors than inside houses [31]. It is unclear why MET sampling efficiency tends to be higher outdoors, with further work required to address this bias. Given the growing recognition of the importance of outdoor biting in maintaining residual malaria transmission [28‐30] and current lack of satisfactory alternatives to the HLC for measuring this, the MET can serve a useful purpose even if only suitable for use outdoors. The relatively good performance of the MET relative to the HLC for sampling malaria vectors outdoors reported here and elsewhere [17, 32] indicate that it is suitable for monitoring exophagic and zoophilic vector [32] populations.

The relative efficiency of the MET for collection of An. gambiae s.l. across dry and wet seasons was evaluated, and its ability to reflect seasonality in vector abundance relative to the HLC standard. Both trapping methods confirm strong temporal variability in vector abundance, likely due to seasonality and meteorological conditions as has been widely documented in Burkina Faso and other parts of West Africa [63, 64]. The current results indicate that the relative performance of the MET compared to the HLC stays constants across seasons, and that both methods predict similar seasonal trend in vector abundance. Additionally, there was no evidence of density dependence in the sampling efficiency of METs over a wide range of An. gambiae s.l. density. This contrasts with results from an earlier prototype where MET performance showed signs of density dependence indoors but not outside [17], but another study also found no density dependence [31]. However, this previous study was conducted over a relatively short period (21 nights) and did not encapsulate the seasonal extremes in vector density incorporated here. Based on the current and previous studies, it can be concluded that the MET can provide relatively accurate estimates of vector population dynamics that are unbiased by season or underlying density. An investigated was also undertaken to assess whether the performance of the MET relative to the HLC decreased over the course of a sampling night as could be indicative of battery drain. Consistent with previous studies [17, 31], there was no detection of any difference in MET sampling efficiency throughout the night when it was used indoors. However, there was a reduction in relative MET performance throughout the night when used outdoors. Such a decrease in MET sampling efficiency outdoors was reported with an early MET prototype in Tanzania [31], but not in a follow up with a new version [17]. It is unclear why MET sampling efficiency falls during the night in outdoor but not indoor settings. One possibility is variation in microclimatic conditions like humidity, which is generally higher outdoors than indoors. Humidity can trigger more rapid discharge of batteries [65]. To maintain consistent MET performance when used outdoors, batteries could be changed during the sampling night.

The malaria vector species composition in this study area varied notably compared to that of previous MET trials in Tanzania. Specifically An. coluzzi and An. gambiae were the dominant vector species here compared to An. arabiensis and An. funestus in Tanzania [17, 32, 66, 67]. Previous work in Tanzania indicated MET capture efficiency varied between malaria vector species (e.g. An. arabiensis and An. funestus [31]). However, vector species composition was similar in collections made by HLC and MET here; indicating no differential sampling performance between An. coluzzi and An. gambiae. Further calibration may be required to ensure the MET gives unbiased estimates of composition of malaria vector species in new settings. Similar to previous studies [17, 32], we found no difference in malaria sporozoite rates between vectors in HLC and MET collections. Thus, the MET also appears to yield unbiased estimates of appropriate for estimating of An. gambiae s.l. infection rates and transmission potential.

Finally, Three key human-mosquito exposure metrics were evaluated to assess whether they were reliably predicted by the MET: the proportion of (i) indoor biting (P_i), (ii) An. gambiae s.l. bites occurring during times when most people are indoors (P_fƖ,) and (iii) human exposure to An. gambiae s.l. bites that would occur indoors in the absence of personal or household physical protection (π_i) [50]. A higher proportion of outdoor biting by An. gambiae s.l. was found than previously reportedly in Burkina Faso [68‐70]. In general, estimates of these three exposure-metrics were similar between HLC and MET collections. However, the MET tended to slightly underestimate all three metrics likely because of its slightly lower sampling performance in indoor versus outdoor settings. However even this with bias estimates of exposure as calculated by the different trapping methods were generally within a few percentage points of one another. For operational use, estimates of exposure derived from MET collections could be adjusted to compensate for this bias.

The multi-site nature of this study allowed assessment of wider aspects of MET feasibility for programmatic sampling. In contrast to previous trials in Tanzania where the MET was used in fixed, single locations [17, 31]; here was carried out in 12 villages requiring the MET to be moved every few days and sometimes as far as 100 km. The integrity of electrified surfaces on the METS were checked before and after transport in the field. The output voltage was also regularly checked during collections to ensure it was meeting the necessary target. On occasions where voltage output was suboptimal (~ 0.4% of days), MET operation was stopped and the problem reported to technical support team. Overall, MET collections were performed on ~ 17% fewer sampling hours than the HLC. However, this does not represent the proportion of times that the MET failed. Most of these MET hours (~ 9%) were lost while waiting for a replacement unit to be made and delivered (~ 4-week period). The most frequent problem encountered with MET use was power failure due to short-circuiting (~ 6% of time) with occasional sparking on the frame. Therefore, further improvements in MET design are needed to resolve this issue. In addition, it was noted that short-circuiting was more likely to occur when there was high level of moisture in the environment (e.g. rainy season, times of high humidity). This was probably due to small water droplets condensing on the frame and occasionally running down the wires. Regular wiping of the MET surface (e.g. during 15 min break periods from sampling) could help avoid a build-up moisture of trap surface. Alternately, redesigning the trap with wires running horizontally instead of vertically will prevent droplets from running down into the frame. METs were subjected to heavy use in this study, under challenging field conditions. It is perhaps not surprising that traps exhibited some degree of physical damage and breakage under these intense circumstances. These issues could be resolved by making future prototypes more robust, and/or keeping METs in fixed locations rather than in constant transport. In addition, on some other nights, MET sampling was intentionally stopped (~ 1% of the sampling hours) due to high wind and rainfall that was anticipated to drive water onto the MET surface and cause short-circuiting. Even with these difficulties, the METs still performed relatively well and consistently with the HLC in this study. To increase the protection of volunteers from bites of very small biting insects (those with wingspan less than 5 mm) that may be present at some study sites, we recommend fitting fine-mesh insecticide-free netting on the inner panel of MET surfaces with very small holes.

This is the first-time that the MET was evaluated outside of East Africa. Overall, the MET collected proportionately fewer malaria vectors than the HLC, and slightly overestimated the proportion of outdoor biting. However, the performance of METs relative to the HLC was consistent over time, and provided similar estimates of seasonal dynamics, biting behaviour, species composition and infection rates in malaria vector populations. Thus, despite some technical problems arising after prolonged MET usage under field conditions, we conclude it presents a promising and safer alternative for monitoring human exposure to malaria vectors in outdoor environments.