Existing malaria vector control measures such as Long Lasting Insecticidal Nets (LLIN) and Indoor Residual Spraying (IRS) combined with Artemisinin Based Combination Therapy (ACT) drugs have significantly reduced the malaria burden in many parts of Africa (Battarai et al., 2007; Sharp et al., 2007a; Ceesay et al., 2008; O’Meara et al., 2008; WHO, 2009; Chizema-Kawesha et al., 2010; Ngomane and de Jager, 2012). The benefit of these interventions extends beyond the personal protection of the households that use them to protect entire communities by reducing either the infectiousness of blood stage parasites in human populations (Killeen et al., 2006a), or the abundance and survival of mosquito vectors (Killeen et al., 2007). It has been shown both theoretically (Killeen and Smith, 2007; Killeen et al., 2007) and in the course of operational control (Hawley et al., 2003; Klinkenberg et al., 2010) that significant community-wide reductions in transmission can be obtained even when intervention coverage levels are modest (35–75%). However, it is unlikely that these packages of interventions of their own will be sufficient to achieve malaria elimination in the most endemic settings where transmission rates are extremely high (Gillies and Smith, 1960; White, 1969; Oyewole and Awolola, 2006; Bayoh et al., 2010; Van Bortel et al., 2010; Bugoro et al., 2011a; Reddy et al., 2011; Russell et al., 2011). Even should the considerable financial, logistic, and behavioral obstacles that currently limit attainment of 100% coverage be overcome (Vanden et al., 2010; Larson et al., 2012), the combined use of effective anti-malarial drugs and these vector control interventions are not predicted to be sufficient for elimination in these settings (Killeen et al., 2000; Griffin et al., 2010) Because they do not cover the full spectrum of all locations where mosquito exposure occurs, and even if only a small percentage of mosquitoes remain and bite outside, their existence could be enough to prevent the transition from very low to zero transmission.

The World Health Organization (WHO) defines malaria elimination as meaning the permanent reduction “to zero incidence of locally contracted cases, although imported cases will continue to occur and continued interventions measures are required” (WHO, 2008b). Achieving this goal will require full understanding of where and when persons are most exposed to the bites of mosquito vectors in order to target interventions where they can achieve maximum impact. While elimination is possible in some settings with low malaria transmission intensity (WHO, 2009; Griffin et al., 2010), and where the dominant vectors exhibit the stereotypical behaviors of biting indoors and late at night where they can be targeted by LLIN and/or IRS (Mabaso et al., 2004; Sharp et al., 2007b; John et al., 2009; WHO, 2009), it is unlikely that these methods will be sufficient to push prevalence below the WHO-defined pre-elimination threshold (<1 case/1000 population/year) in areas of high transmission (Molineaux and Gramiccia, 1980; Kleinschmeidt et al., 2009; Russell et al., 2010) and where the majority of human exposure to transmitting mosquitoes occurs outside human dwellings to which most current interventions are restricted (Taylor, 1975; Pates and Curtis, 2005; Tirados et al., 2005; Oyewole and Awolola, 2006; Geissbühler et al., 2007; Griffin et al., 2010; Van Bortel et al., 2010; Bugoro et al., 2011a; Yohannes and Boelee, 2012). To date, probably the most comprehensive attempt to achieve local elimination within an endemic region of Africa was made in the Garki region of northern Nigeria in the 1970’s (Molineaux and Gramiccia, 1980; WHO, 2008b). Before initiating this campaign, malaria transmission within this region was extremely high (example as indexed by an estimated annual entomologic inoculation rate, EIR of 20–120 infectious mosquito bites per person per year). During the elimination program which ran from 1972 to 1973, near complete coverage of all households (97–99%) with propoxur-based IRS combined with mass drug administration of the anti-malarials sulfalene and pyrimethamine (73–92% coverage) was attained throughout the study area. Although these intense efforts led to a drastic reduction in malaria prevalence within the region from 70 to 1%, the threshold for local elimination was not even approached (Molineaux and Gramiccia, 1980). A critical factor in the failure of elimination was incomplete suppression of vector populations due to the existence of low level outdoor-feeding (exophagy) and resting (exophily) behaviors within a small proportion of locally important vector populations (Molineaux and Gramiccia, 1980). This small proportion of mosquitoes with atypical behaviors was sufficient to prevent elimination even across a short period of time in which the potential for insecticide resistance was not recorded. Furthermore, by preventing the rapid achievement of elimination, the existence of these vectors may enhance the likelihood and spread of Insecticide Resistance by necessitating the continued application of high dose formulations over longer periods of time.

While historically transmission in much of Africa has been dominated by vector species that primarily feed and rest indoors where they can be efficiently targeted with domestic insecticides (Gillies and DeMeillon, 1968; White, 1974; Gillies and Coetzee, 1987), there is growing evidence from across the continent that the widespread use of LLINs and IRS is driving vector species composition toward those with more flexible behaviors (Braimah et al., 2005; Pates and Curtis, 2005; Tirados et al., 2005; Antonio-Nkondjio et al., 2006; Oyewole and Awolola, 2006; Geissbühler et al., 2007; Bayoh et al., 2010; Reddy et al., 2011; Russell et al., 2011). For instance Anopheles gambiense sensu stricto has been historically viewed as the most significant vector of malaria in Africa (Gillies and DeMeillon, 1968; White, 1974; Gillies and Coetzee, 1987; Kiszewski et al., 2004). However in the wake of the widespread deployment of domestic insecticidal interventions, this species is in significant decline in many areas; with the majority of remaining transmission now being dominated by An. arabiensis (Braimah et al., 2005; Tirados et al., 2005; Bayoh et al., 2010; Russell et al., 2011); a closely related sibling species that can exhibit much more flexible behaviors including biting people and resting outdoors, and switching their biting between humans and common domestic animals such as cattle (Gillies and DeMeillon, 1968; White, 1974; Gillies and Coetzee, 1987; Tirados et al., 2005).

Even low levels of exophagy, exophily, or zoophagy (Table 1) may substantially attenuate the impact of LLIN and IRS because this allows mosquitoes to obtain blood while avoiding fatal contact with insecticides (Smith and Gillies, 1960; Taylor, 1975; Molineaux and Gramiccia, 1980; Pates and Curtis, 2005; Geissbühler et al., 2007; Killeen and Smith, 2007; Govella et al., 2010; Bugoro et al., 2011a). Furthermore, the deployment of insecticides inside houses is likely to place strong selection on even highly endophilic species to switch their behaviors, which could not only prevent the achievement of elimination but undermine the continued effectiveness of these interventions (White, 1974; Pates and Curtis, 2005; Van Bortel et al., 2010; Bugoro et al., 2011a). For example in Bioko Island, Equatorial Guinea, An. gambiae s.s. were historically documented to feed almost entirely indoors (Molina et al., 1996). However, following initiation of mass coverage campaigns with indoor-insecticide based interventions, the behavior of this vector species has switched to an almost 50:50 split between indoor and outdoor biting (Reddy et al., 2011).

With the shift in policy from targeted to universal distribution of LLINs to each sleeping space in many parts of Africa (Teklehaimanot et al., 2007; WHO, 2008a; Kilian et al., 2010), it has been increasingly observed that mosquitoes can re-adjust their biting activity so that much of it occurs outdoors and before bed time across different settings (Rubio-Palis and Curtis, 1992; Trung et al., 2005; Bayoh et al., 2010; Van Bortel et al., 2010; Bugoro et al., 2011a,b; Reddy et al., 2011; Russell et al., 2011). This is due to the fact that indoor biting species (Smith and Gillies, 1960; Gillies and Furlong, 1964; Gillies and DeMeillon, 1968; Pates and Curtis, 2005; Killeen et al., 2006b) and subpopulations (Coluzzi et al., 1979; Molineaux and Gramiccia, 1980; Bendesky and Bargmann, 2011) will be more affected, leaving predominantly the outdoor-feeding, “early biting” residual populations and individuals (Gillies and Smith, 1960; Taylor, 1975; Bayoh et al., 2010; Bugoro et al., 2011a,b; Russell et al., 2011) to maintain residual transmission. Evidence from the previous Global Malaria Elimination campaign (1950s) suggests this is not only a recent phenomenon. Following the widespread implementation of IRS in the South Pare Region of Tanzania (1955–1959), the highly endophilic vector An. funestus disappeared leaving only an An. gambiae s.l. population which exhibited extremely exophilic behavior (Smith and Gillies, 1960). Similarly, An. funestus was also replaced by the highly zoophagic and exophilic species An. rivulorum and/or An. parensis on at least three distinct occasions following the implementation of IRS in South Africa, Kenya, and Tanzania (Gillies and Smith, 1960; Gillies and Furlong, 1964). Note that not all of the mosquito vector behavioral shifts mentioned above may necessarily hinder malaria control. For example, while a shift in vector behavior to biting people outside or before bedtime would clearly allow them to evade control, the impacts of others such as increased zoophagy are not clear cut. For example, while shifting their behavior from biting people to livestock would allow vectors to avoid insecticides, but it would concurrently prevent them from becoming infected or transmitting. Thus if interventions prompted increased zoophily in vector populations, it could actually help enhance control and prospects for elimination if it resulted in a long term, stable shift away from biting humans, and did not just provide them with a short term strategy to maintain their populations in the face of insecticide use, before later returning to feeding on humans when insecticide coverage dropped and/or resistance emerged (see reference Ferguson et al., 2010 for more review).

More recently, the long term use (∼10 years) of ITNs at high coverage in western Kenya and south eastern Tanzania has been accompanied by a substantial shift in malaria vector species composition as manifested by the progressive diminishment of the importance of the highly endophagic/endophilic An. gambiae s.s. (Bayoh et al., 2010; Russell et al., 2011). Examples of similar phenomenon have also been observed outside Africa. In the Solomon islands, for example, intense IRS campaigns conducted in the 1960s appeared to have eliminated the major vectors An. punctulatus and An. koliensis which predominantly rest indoors, but recent programs combining ITNs and IRS have had negligible impact upon human biting rates because the remaining, current primary vector species An. farauti feeds and rests predominantly outdoors (Bugoro et al., 2011a). Similar phenomena may be in Asia where human exposure to mosquito bites predominantly occurs outside houses and before bed time (Van Bortel et al., 2010).