Our results show that at the end of rainy season inhabitants of Goden are potentially exposed to at least 10 infective bites/person/night, despite 5 years since the beginning of LLIN mass distribution campaign in Plateau Central region of Burkina Faso . Indeed, the fraction of bednet-protected people (about two third of the whole population ) are likely exposed to mosquito-bites in early evening and early morning, when not sleeping under the LLIN, and receive much less infective bites than estimated in the present study. However, the high biting rates observed between 9 and 10 PM (when most people are inside houses and possibly not yet protected by LLINs; CNRFP, unpublished data) suggest actual high levels of exposure of Goden population. This levels of exposures are unusually high compared to EIRs reported (per night) before LLIN distribution in the same village (mean 7.4, min 2.5, max 17.0 ) as well as in the Plateau Central region of Burkina Faso (mean 2.3, min 0.2, max 8.3;[44,45]), as well as those reported in other regions of Burkina Faso (2.4; ) and in other sub-Saharan African countries where LLINs are in use (ranging from 0.04 to 3.4 [47–63]).
The estimated EIR value is due to both high malaria vector densities (HBR=23.5 females/hour/person) and very high levels of infectivity in the vector population (SR=5.8%). Both biting rate and infectivity rate are in line with the few literature data available for An. gambiae s.l. in Goden (1.8<SR<12.1) and in the same geographical area before LLIN implementation (0.5<HBR<26.3 [64,65]; 3<SR<10 [64–66] in 6 villages in a radius of 40 km from Ouagadougou) and comparable to SR in Goden in 2011 (6.9%, ) and 2012 (6.6%, ). Notably, even though SRs were assessed by different approaches in different studies, results by recently developed rDNA-based TaqMan assays [33,67] do not significantly differ to those obtained by traditionally used CSP-ELISA .
Data also shed some light on other entomological factors, which could have reduced LLIN-effectiveness after 5-year implementation (i.e. endophagy vs exophagy, time of biting and genetic resistance to insecticides) in the two main malaria vector species in the village, i.e. An. coluzzii and An. arabiensis.
First, our results does not suggest endophagic preferences in either vector species, in agreement with what observed in An. coluzzii in other settings characterized by massive LLIN coverage (North-West Burkina Faso , Benin  and Bioko island ). Although most studies focusing on resting mosquitoes indirectly suggest high endophagy for An. coluzzii/An. gambiae, the few studies carried out by HLC both indoors and outdoors show lack of preference for the biting location even before LLIN implementation [70–77]. Overall, this highlights that anthropophily is the main driver of the endophagic behaviour in An. coluzzii/An. gambiae. Consequently, the supposed higher exophagy of An. arabiensis is a consequence of its generalistic host preference [78,79]. Thus, in cases of reduced human-host availability indoors due to LLIN, both An. coluzzii and An. arabiensis do not require a secondary adaptation to bite outdoors, as they are already adapted to do it. Notably, mathematical models suggest that even relatively modest changes in outdoor biting can have a substantial public health impact (e.g. a 10% increased outdoor biting activity could result in 10.6 million additional malaria cases in whole Africa, even assuming a 100% LLINs coverage) .
Second, no differences in biting rates were observed between 21:00 and 4:00 in vector species. Indeed, a peak of activity during night hours is typical for An. gambiae s.l. in the absence of bednet protection [71,72,74,75,80–85]. To our knowledge, a lack of peak of activity was observed in An. coluzzii only in Burkina Faso  and in Bioko Island  where, two years after the introduction of LLINs, the proportion of host-seeking events changed towards a lack of significant differences throughout the night. On the other hand, a wide range of peak biting times (i.e. early, late or “central” night activity) has been reported in An. arabiensis after control interventions [52,86–89]. According to theoretical predictions, lack of biting time peak is indicative of a situation in which the selective pressure exerted by LLIN has altered the typical biting pattern, but has not been yet sufficient to trigger a strong shift towards earlier and later biting times (Figure 2; ) Expanding the duration of HLC before dusk and after dawn would allow to analyse more in detail a possible peak of biting activities to earliest and latest hours to access hosts unprotected by LLINs, as reported by . In particular, measuring the numbers of host-seeking mosquitoes at times when people are still engaged in working activities outdoors would give a better estimate of actual risk of malaria transmission in the area.
Third, while we did not carry out insecticide resistance bioassays, we genotyped a subsample of collected specimens for the locus L1014 of the sodium-gated voltage-channel gene - known to be one of the markers associated to pyrethroid resistance - to speculate on the possible role of insecticide resistance in contributing to high levels of transmission in the village by increasing survival of endophagic vectors entering in contact with the bednet. We found frequencies of the L1014F allele of 67% and 38% in An. coluzzii and An. arabiensis, respectively (without significant differences between indoor and outdoor collections), suggesting pyrethroids are selecting target site resistance in the study site. Notably, the deterrent effect of LLINs on partially resistant (either behaviourally or genetically) mosquitoes is known to diverge biting activity to unprotected hosts both indoors and outdoors [92–95].
Effectiveness of ITN/LLINs relies on anthropophilic - and therefore endophagic - behaviour of vector populations and susceptibility to insecticides used to impregnate the bednets. A limited number of studies showed that changes in vector biting behaviour, as well as shift in species dominance, following LLIN implementation can undermine the efficacy of such control measures (Additional File 2). A species shift was repeatedly shown in East Africa where, after LLIN introduction, the former most abundant highly anthropophilic vector, An. gambiae, has been outnumbered by the more generalist An. arabiensis, leading to an overall decrease in sporozoite rates [51–53,88,96–100]. To our knowledge, shifts in vector dominance after LLIN introduction has never been investigated in West and Central Africa (where An. coluzzii is also present), with the exception of Bioko island, where a progressive replacement of An. gambiae by An. coluzzii was observed as indoor control measures were applied . Our results do not show a shift in species composition following 5-year LLIN implementation in the study area in Burkina Faso where, even before bednet introduction, An. coluzzii and An. arabiensis were the main circulating vectors and An. gambiae was uncommon . Nevertheless, results suggest that LLINs have elicited a “behavioural resilience” (sensu Govella et al., ) in An. coluzzii, leading to the biting rhythms observed in the present study [72,75,84], as well as to a higher zoophagy, as detected in 2011 . Indeed, An. coluzzii is known to be characterized by high ecological plasticity in the exploitation of different habitats [103–109], as well as by an opportunistic host-seeking behaviour [16,69,110]. Overall, the ecological plasticity of the two main vectors maximises their capacity to reach the human host when not protected by bednets, consistent with the high biting and sporozoite rates observed in the study site.