In Zambia, both LLINs and IRS have been dramatically scaled up in recent years [14], and the official national target is to achieve 100% coverage of all populations at risk with locally appropriate interventions by 2012 [15]. The country was an early recipient of donor support to scale-up malaria control efforts through funding from the Global Fund, the President Malaria Initiative, World Bank Booster Program and Bill & Melinda Gates Foundation [16].

In Zambia, IRS was introduced in 2003, and as of the 2011–2012 transmission season, IRS services are offered at varying levels of coverage in all 72 districts. Many areas of endemic transmission in Africa, including Zambia, have conducted extensive LLIN distribution campaigns which have dramatically increased bed net coverage in previously unprotected populations. Zambia has distributed nets both through mass campaigns and through antenatal care (ANC) clinics with the aim to achieve high coverage [17]. However, even given high levels of net ownership, usage levels of nets in many of these areas is unsatisfactory [18,19]. Furthermore, recent studies have shown that the current ITN distribution strategies, that miss households occupied by the elderly and those without children or ANC access, cannot reach 100% coverage, even if the number of nets should be sufficient to reach this level [17].

Both the personal and community-level impact of IRS and ITNs are entirely dependent upon mosquitoes entering houses, [9,20,21] and thus understanding the behaviour of malaria vectors in relation to vector control interventions is fundamental; importantly, behaviour is highly variable across the spectrum of vector species [7]. Despite the massive investment in the use of LLINs and IRS as front-line vector control measures across the tropics, there are only a few estimates of the proportion of malaria transmission occuring indoors and can be directly prevented by these measures [9,11,13,22-24]. Zambia is primarily reliant upon these measures for national-scale malaria vector control, and estimates of the proportion of human exposure occurring indoors are important to understand the success of malaria control efforts.

Anopheles quadriannulatus is readily susceptible to malaria infection [33] and does occasionally feed upon humans [34], as documented here in south-east Zambia (Figure ​(Figure2),2), but it is widely regarded as being as preferentially zoophagic and therefore of negligible importance to malaria transmission [35-37]. The large number of An. quadriannulatus collected by human landing catches in this setting supports the view that host choice is very plastic in this species and that they may well feed on humans in some settings. Torr and colleagues [34] suggest that An. quadriannulatus has no specific preference for animals and also feeds on humans, but simply responds in proportion to overall body mass, which is consistent with how anthropophagic species respond to individual humans [38]. While Anopheles quadriannulatus readily feed on humans in Luangwa, no sporozoite-infected specimen was identified in this setting and we could find no report of such an occurrence in the literature, so it is most likely of negligible importance to malaria transmission.

The predominant malaria vector species caught in this study area was An. funestus s.s., which is commonly regarded as one of the most efficient malaria vector species in the world because of its very high degree of anthropophagy and endophagy [25,32,39-42]. Anopheles funestus s.l. in the study area showed no clear preference for feeding indoors or outdoors (Figure ​(Figure22 and Table1). Nevertheless, estimates of the proportion of human exposure which is preventable through the personal protection arising from LLINs use are as high as any published estimate [9,11] in the current study, and indicate that both IRS and LLINs remain excellent options for malaria vector control and should achieve their full potential personal and communal protective effects in this setting. This is contrary to the recent evidence from other parts of the world that suggests that behavioural changes of malaria vectors toward feeding predominantly in the early part of the evenings and outdoors render LLINs less protective [7,10-13,43,44].

The estimate that more than half of all transmission, presumably mostly by An. funestus s.s., accrued by LLIN users occurs indoors (π_i,n>0.5) suggests that supplementing LLINs with IRS might well achieve incremental impact upon malaria transmission in this setting. For effective combined implementation of IRS and LLINs, the insecticide of choice for IRS should ideally have a different mode of action from those used on LLINs to minimize risk of rapid selection of resistance traits in vector populations [45]. As slightly less than half of residual transmission for LLIN users occurs outdoors, there is a limit to how much incremental control can be achieved by complementing LLINs with further indoor vector control using IRS, regardless of how efficacious the product used for the latter is. Therefore, measures which protect against outdoor exposure or which suppress mosquito breeding [46] may well be required to supplement even a combination of LLINs and IRS targetted at endophagic vectors to go beyond malaria control and achieve local elimination in this setting.

Both recent [12,13] and historical [47] reports indicate that as LLINS and IRS are scaled up, previously ignored outdoor-biting mosquitoes become culpable for a greater proportion of malaria transmission. In most cases, this increased proportion of transmission exposure that occurs outdoors appears to be arising from selective suppression of indoor-biting vectors by ITNs [48-50] or IRS [47,51]. Evidence of heritable changes in behavioural traits within single species also seem to be associated with selective suppression of specific molecular [12] and chromosomal [52,53] forms. Perhaps the main limitation of this study is that, while most residents slept under an LLIN, coverage remains incomplete, falls short of the Roll Back Malaria targets [54] and was only achieved relatively recently. It is therefore possible that further selective suppression of the endophagic vectors may occur as LLIN usage rates are increased and sustained. If this plausible scenario were to occur, the majority of human exposure might well occur outdoors so options such as repellents [55], insecticide-treated clothes [56] and insecticide-treated cattle [57] might be higher priority, possibly rendering supplementary IRS redundant. It is therefore important to distinguish between two quite different scenarios: 1) Full coverage of LLINs has achieved maximal suppression of endophagic and endophilic vectors and could be genuinely supplemented with complementary measures, and 2) Partial LLIN coverage with incompletely suppressed indoor transmission so that IRS essentially fills gaps in coverage and therefore partially substitutes for LLINs as a means to tackle persistent indoor transmission. We therefore suggest that national systems for monitoring intervention coverage, malaria risk, and insecticide resistance should now be supplemented by field surveys of vector population composition and the relevant behaviours they exhibit.