Background
The latest World Malaria Report indicates that global efforts are dangerously off-track [
1] and will not meet the important targets of the Global Technical Strategy 2016–2030 to reduce mortality and case incidence by at least 40% by 2020 relative to the 2015 levels [
2]. Achieving the overall goals of elimination and eventual eradication will require major revitalization of proven strategies, but also introduction of new tools capable of complementing LLINs and IRS, and addressing gaps associated with challenges such as insecticide resistance [
3,
4], increased outdoor biting [
5,
6], sub-optimal user compliance [
7,
8] and high costs. Expanding the vector control toolbox is an important component of this new agenda, and various new options have been proposed in recent years [
9].
The use of spatial repellents [
10,
11] or odour-baited mosquito traps [
12,
13] have been proposed for consideration either singly, or in combination in the form of push–pull strategies [
14,
15]. The underlying assumption of push–pull is that the stimulo-diversionary effects on mosquitoes will ensure that host-seeking vectors repelled from their human targets can be trapped and killed, thereby preventing diversion to unprotected persons, and potentially improving communal protection by removing large densities of mosquitoes from circulation. Indeed, research on personal protection with topical repellents, such as DEET, has demonstrated that mosquitoes can move from protected to unprotected individuals [
16]. Even where repellents offer effective protection, poor compliance among users can significantly reduce this protection [
7,
8]. This is particularly a challenge with topical repellents, such as picaridin, for which despite high reported acceptance, actual daily user compliance was as low as 8% in a trial in Cambodia [
8,
17]. Because of the sub-optimal use even in areas with high access rates, the repellents in this study did not lead to any further reduction in malaria burden [
18].
Using spatial repellents may address these compliance challenges, more so with improved delivery formats such as transfluthrin-treated eave ribbons [
11,
19], which do not require retreatment for months and can be kept at safe distances from infants. However, even these do not fully address possible diversion to non-users under conditions of incomplete coverage [
20]. On the contrary, push–pull strategies, where host-seeking mosquitoes are repelled from their intended hosts and lured towards traps or other lethal sites [
21], could potentially address the compliance issues while also targeting outdoor-biting and insecticide-resistance mosquitoes. Indeed, push–pull has been successfully implemented against agricultural insect pests [
21,
22]and has previously demonstrated 95% efficacy against malaria vectors under controlled conditions [
15]. If baited with effective lures, traps function as pseudo-hosts and can attract and kill large densities of potentially infectious mosquitoes from an ecosystem [
23,
24].
Recently, a small-scale field study conducted in Kilombero valley implementing a non-optimized push–pull system, offered a marginal protection of 30% against wild populations of malaria vectors, with most of the benefits accrued in early evening hours when people are usually outdoors [
14]. The low efficacy was attributed to sub-optimal efficiency of the odour-baited device used in the study, and inconsistent levels of the lure, i.e. CO
2 gas from yeast-molasses fermentation. However, most importantly, that push–pull system was not optimized; type and number of traps, distance of the traps from huts, dose of the spatial repellent treatments, or directional orientation of the subunits were all assigned without any prior optimization, thereby not guaranteeing maximum protection against mosquito bites [
14].
This current study was, therefore, designed with two objectives, essential for eventual application of push–pull as a complementary tool for malaria prevention. First was to test different configurations of the push–pull system sub-units, i.e. repellents (push sub-unit) and traps (pull sub-units). The model system consisted of long-lasting spatial repellents (the recently developed transfluthrin-treated eave-ribbons technology [
19] and odour-baited traps (inverted version of the commonly used BG-Sentinel trap, recently evaluated for trapping malaria mosquitoes [
24,
25]. The study used a fixed dose of transfluthrin treatment on the eave ribbons, i.e. 0.25 g/m
2 ai, previously demonstrated to offer ~ 75% protection [
19], and also a fixed trap type, i.e. BG-Malaria [
26,
27], but the number of traps per hut, distance between the traps and huts, and use of the CO
2 gas as lure were varied to determine the optimum.
The second objective of the study was to compare efficacies of complete push–pull system versus either the traps alone or spatial repellents alone, for personal and household protection against indoor-biting and outdoor-biting malaria vectors.
Discussion
As the race to expand the malaria vector control tool box accelerates, additional evidence is required on new alternative tools to ascertain their suitability for use in various settings. This study evaluated the benefits of using push–pull as opposed to either the push component (spatial repellents) alone or the pull component (odour-baited mosquito traps) alone. The components were selected based on previous data, which had demonstrated high efficacies of both the transfluthrin-treated eave ribbons [
19] and the BG-Malaria trap [
23] against malaria vectors in Tanzania.
Spatial repellents and odour baited traps have both separately been proposed as potential new tools for expanding the malaria control tool box, though there is not yet adequate data to justify their large-scale application [
9]. For spatial repellents, current focus areas include developing strategies for delivery of active ingredients in ways that ensure long-term efficacy and safety [
34] and also improving user compliance, which has previously impaired field effectiveness [
17]. On the other hand, for odour-baited traps, focus areas include developing highly effective attractants [
23] and improved trapping devices [
23], but also on how the technologies can best be combined with existing interventions [
28,
31]. Though these two techniques on their own may complement LLINs and IRS [
28,
35], they both have unique challenges that have until now limited their individual appeal, especially under sub-optimal coverage [
20]. Indeed, a major reason for proposing push–pull instead of either traps only or repellents only is to counter the potential negative effects such as diversion of mosquitoes to non-users and to prevent excessive risk when traps are baited with attractants that increase mosquito densities in an environment. This study therefore also examined a range of configurations under which push–pull would be most effective in communities.
The dose of transfluthrin used here, i.e. 0.25 g/m2 was selected because it had yielded protection below 100% in earlier studies and, therefore, allowed further evaluation of push–pull instead of push alone. A major finding was that push–pull achieved greater protective efficacy than either component alone, with clear superiority superior over traps alone, but its advantage over spatial repellents alone was only marginal. Indeed, most of the gains obtained from push–pull in these experiments can be attributable to the push-component alone.
It had been expected that adding traps would help trap out mosquitoes repelled by the eave ribbons and therefore increase overall protection. However, this was not directly observable here, most likely because the current study assessed only personal and household-level protection, but not community-level protection. It is likely that given the high protective efficacies of the spatial repellents in this system, any additional benefits would only become apparent in community-level trials where both users and non-users are observed. Indeed, in studies my Menger et al. [
38], the value of push–pull was much more apparent when the personal protection data was incorporated into mathematical simulations of community-level impact. Future studies should therefore consider such assessments in both user and non-user households, and should also include tests on multiple
Anopheles species which may have different behaviours. Another possible reason for the marginal additional value of push–pull over spatial repellents alone could be that the mosquitoes were not only repelled but also killed by the transfluthrin. Indeed, in earlier trials where pyrethroid-susceptible
An. arabiensis were exposed inside the huts with transfluthrin-treated eave ribbons, these mosquitoes were consistently killed, suggesting multiple modes of action of transfluthrin [
19]. Such substantial killing-effect of transfluthrin would limit the diversionary effects without the need for trapping.
Also, push–pull systems involving traps and repellents can be also be cumbersome and expensive, especially where the traps are battery-powered and require regular replacement of lures as well as regular repairs. Spatial repellent products such as transfluthrin-treated eave ribbons can therefore offer a ready alternative, which if scaled up could match the overall efficacy of push–pull but at lower costs. Theoretically, an added advantage of high coverage with spatial repellents is that it could minimize the known diversion effects [
20], where mosquitoes bite non-users more than users of the spatial repellent products, especially if the repellent active ingredients also have a killing-effect on the mosquitoes.
An important aspect for consideration here is that the spatial repellent used here was a pyrethroid, transfluthrin. The study therefore does not recommend that the transfluthrin-treated eave ribbons are used for resistance management but rather to offer additional protection against mosquitoes that bite outdoors or indoors at times when people are not using their LLINs. This is a key limitation of this approach and suggests that the search for new active ingredients, particularly those that are non-pyrethroids should continue. The concept of spatial repellents for resistance management could then be applicable if other active ingredients are used. Nonetheless, it was interesting that in a previous study [
19], pyrethroid-resistant
An. arabiensis were sufficiently repelled by transfluthrin-treated eave ribbons. Moreover, when inside huts with 0.02% transfluthrin-treated ribbons, the mosquitoes also died in very high proportions (mortality of 99.5%). This indicates providing both repellent and killing activity could be highly effective against resistant
An. arabiensis. However, the same study also found that pyrethroid-resistant
An. funestus, which dominate malaria transmission in rural south-eastern Tanzania were only modestly affected by the transfluthrin-treated eave ribbons [
19]. This suggests indeed that the efficacy of transfluthrin-based products would be limited in certain settings. Therefore, to sustain effectiveness and enable resistance management efforts, it is recommended that transfluthrin-treated eave ribbons should be combined with non-pyrethroid interventions, for example organophosphate-based or carbamate-based IRS indoors.
It was hypothesized at the start of the study that increasing trap densities could improve protection, by mass-trapping host-seeking mosquitoes. However, this study determined that while doubling trap densities from 0.5 to 1 trap/hut was beneficial, further increase to 2 traps/hut was detrimental. Probably, this is because the traps increased the concentration of the odour lures (CO
2) in the environment and kept the mosquitoes active enough to increase rather than decrease biting. Obviously even if the study had determined that increasing trap densities would be beneficial, the economic cost would likely be exorbitant. Future developments in trapping technology could potentially lead to higher trapping efficiencies and exclusion of the need for industrial CO
2 gas as used in these experiments, a development that could greatly improve the appeal of host-seeking mosquito traps for control. In this study, presence of CO
2—baited traps at the peri-domestic areas seemed to increase proportions of mosquitoes biting outside and slightly reduced indoor mosquito-biting risks. The pull-subunits may indeed increase mosquito biting risks to people engaged in various outdoor activities such as story-telling, cooking, dish-washing and drinking [
29,
36]. Overall, these
findings should however not be interpreted to mean that traps have no value, as assessments here were mostly of personal protection. It is possible that adding traps into the systems would indeed address potential diversion problems while also trapping and killing large number of mosquitoes, thereby contributing to mass community-level benefits for users and non-users. Thus, additional studies are required in field settings to more accurately measure such community-level outcomes including any potential diversion from users to non-users, and distances over which such diversion can occur. One concern with the
trapping systems used here was the difficulty in standardization, possibly as a result of differential airflow in the systems, which resulted in discordant trap efficacies between studies in the first tests versus those done in the tunnel. For example, in the first test, one trap for two houses had no significant impact, whereas two traps for two houses had a significant effect, (Fig.
4). However, in the tunnel tests, one trap had no significant impact with just one hut in the system, whatever the distance. In future tests, these differences could be avoided by conducting all studies in similar setting and by improving standardization and airflow in the experimental systems.
The transfluthrin treated hessian eave-ribbons used as an intervention offered significant protection against indoor and outdoor malaria vectors. These results corroborate findings from previous studies, which showed that transfluthrin-treated hessian ribbons can offer more than 75% protection against mosquito bites [
10]. In this study, the ribbons were wrapped along the eave spaces of the experimental huts, without blocking the entire eave space. The eave-ribbons with higher concentrations above 0.02% transfluthrin were tested in the same chamber and found to offer 99%–100% protection against indoor and outdoor mosquito biting risks [
19]. The mosquito biting protection offered by higher concentrations of transfluthrin treated eave-ribbons described by [
19] affected mainly personal and household protection levels and not communal protection level. Though the diversion problem was not explicitly tested in this study, it is one of the aspects that could potentially be addressed by addition of trapping in the field settings, as the traps would take out the mosquitoes before they bite unprotected persons. There is currently an ongoing study by Ifakara Health Institute, which assesses the diversion effects of transfluthrin-treated eave ribbons (Mwanga et al. pers. commun.). The number of mosquitoes trapped by the baited-traps was also slightly lower when both treated eave-ribbons and traps were tested together. This was probably due to repellent effect or the feeding-inhibition effect of transfluthrin [
37]. Such effects probably also reduced the number of mosquitoes, which might be trapped by the baited-trap at the peri-domestic areas.
The addition of odour-baited traps to the eave-ribbons, to form push pull showed modest improvements on personal protection, unlike in previous studies, where presence of baited-traps outdoor undermined the efficacy of push–pull system [
38]. The study by Menger et al. indicated no additional effect of having push and pull subunits at the peri-domestic areas. They however also concluded that mosquito biting protection was mainly offered by the push-subunits, which is similar to the findings of this current study [
38]. Additionally, during a recent small-scale field evaluations of push–pull, it was determined that presence of the system at the peri-domestic areas undermined the effects of the odor-baited mosquito landing box (MLB) [
14]. However, testing both push-and-pull subunits in this current study did not affect the indoor mosquito biting protection, as this was mainly offered by the push-subunits alone. The study suggested the necessity of optimization studies on the number and the distance of pull subunits needed to offer maximal protection against mosquito bites. The distance between the pull and the human volunteer needed to be optimized to prevent the mosquito attraction competition as reported previously [
12,
19]. These aforementioned challenges have now been tackled in the series of push–pull optimization experiments reported in this current study.
During evaluation of the optimal distances between the push and pull subunits, when the baited trap was
situated at either 5 m or 15 m away from the treated ribbons, the trap caught a higher average number of
mosquitoes compared to other distances tested. Although, the lower biting protection conferred by the pull-subunit situated at either 5 m or 15 m away from the push-subunit was best for offering communal
protection, the best configuration of push–pull was that with traps located at least 15 m away from the huts.
Since outdoor biting was greater than indoor-biting, and because traps placed at 30 m away from huts
resulted in the greatest reduction in outdoor bites, it can be argued that where outdoor-biting
An. arabiensis
are the main vectors, then the traps should be placed at least 30 m from the huts, (Fig.
5).
Mosquito trapping in these experiments was done by human volunteers outdoors and CDC light traps indoors. Overall, as shown in Tables
1,
2 and
3 and Fig.
3, more mosquitoes were caught outdoors than indoors, which may create an impression that there was lower biting risk indoors. This is mostly because in these experiments, a fixed number of mosquitoes were released each night and outdoor trapping preceded indoor trapping. Though these findings do not invalidate the percentage protection values calculated, there are still many settings across Africa where substantial proportions of malaria transmission events actually occur indoors as opposed to outdoors, in which cases prioritization of indoor protection remains a key. Interpretation of the data should therefore consider the fact that different trapping methods were used indoors and outdoors.
One limitation of this study is that it lacks the field data assess other factors which might influence the efficacy of the push–pull system. These factors may include airflow (wind), which was limited in the semi-field system compared to field settings. Another is the use of pyrethroid susceptible An. arabiensis as the only test organism. Further studies are, therefore, needed to assess impact against resistant mosquitoes, and also against other species such as An. funestus, where it is the dominant vector. Lastly, these current studies also used untreated nets inside the study chambers, which may not represent actual situation in field settings. Future field tests should, therefore, consider using LLINs as primary intervention.
Authors’ contributions
ASM and FOO conceived the study and developed the study protocol. ASM, MK, RM, EM and HK conducted the experiments, logistic support and supervised data collection. ASM, HSN and FOO analysed the data. ASM and FOO wrote the manuscript. ASM, HSN, EPAB, EWK, AEE and FOO reviewed the manuscript. All authors read and approved the final manuscript.