Insecticide treated nets (ITNs) and indoor residual spraying (IRS) are recognized as the most cost-effective methods for preventing malaria transmission by indoor-biting mosquitoes.^1,2 The success of these interventions relies on their ability to repel or kill endophagic (indoor feeding) mosquitoes, thus providing direct personal protection against exposure to bites, as well as reducing adult mosquito survival and human-feeding frequency.^3–10 The major malaria vectors of sub-Saharan Africa are Anopheles gambiae Giles and An. arabiensis Patton, which are members of the Anopheles gambiae sensu lato species complex, and An. funestus Giles, of the An. funestus s.l. species group.¹¹ These highly efficient vector species are generally considered to predominantly prefer feeding indoors (endophagy) at night (nocturnally), with peak biting activity typically occurring between midnight and the early hours of the morning, when most people in sub-Saharan Africa are asleep indoors.¹²

However, high coverage rates of ITNs or IRS can dramatically alter the composition of vector populations.^13–20 Consequently, biting activity of the persisting residual populations tends to be more evenly distributed across the night because mosquitoes feeding indoors in the middle of the night are selectively suppressed.^18–20 Together with emerging resistance to pyrethroids,²¹ the only class of existing insecticides suitable for use on ITNs, it is the host-seeking patterns of such residual vector populations that define the limit of how much malaria control can be attained with ITNs or IRS or both, as well as the ideal properties of complementary vector-control measures.^22–25

With so few reports with which to compare contemporary observations of changing vector behaviour^18–20 and those that may occur in the near future, it is important to examine existing data to get a broader appreciation of the range of values for the proportion of human exposure to insect vectors of disease that has occurred indoors in sub-Saharan Africa. Conventional indices of behavioural patterns of malaria vectors can substantively underestimate the potential protection provided by ITNs against exposure because they do not consider human indoor and outdoor movements.²⁶ In the study described here, records of indoor and outdoor mosquito biting distributions from 10 Anopheles populations at six rural sites in Africa have been combined with surveys of the times at which humans enter and leave their houses each night, to understand how much can reasonably be expected from the ongoing scale-up of ITNs and IRS.^22–26

Vector behavioural patterns differed considerably among the study locations and vector taxa, with peaks of biting activity occurring at various times from just after dusk to just before dawn (Figure 2). Biting rates that were clearly higher indoors than outdoors were not as ubiquitous as expected, and were found for only four of the An. gambiae s.l. populations and one of the An. funestus s.l. populations. Figure 2 illustrates a substantial degree of diversity in human and mosquito behaviour across Africa. The amount of time that residents spent indoors during the night varied from 8 hours in Ulanga to 12 hours in Rarieda. Figure 2

Hourly biting pattern of Anopheles gambiae sensu lato (panels on left) and Anopheles funestus sensu lato (panels on right) occurring both indoors (solid lines) and outdoors (dashed lines) at different study sites. The grey area represents the proportion of the human population predominantly spending time indoors during the times shown on the abscissa of each graph

Apart from a scope that spans only six sites in four countries, this study has a number of limitations relating to its technical methodology. Previous comparisons of questionnaires with direct observations for surveying the human behaviours described here suggest that these are approximately but not entirely consistent with each other.³⁵ In terms of mosquito behaviour, several of the sites in the study may not have captured some low levels of outdoor human exposure that occurred before HLC surveys that were begun in the evening and after they ended in the morning. The proportion of human exposure occurring indoors (π_i) may therefore have been slightly overestimated. However, examining the trends on either end of the activity profiles in Figure 3 reveals that in no case is this likely to result in overestimation by more than 10%. Additionally, the accuracy of these mosquito surveys is limited to some extent by the practical challenge of maintaining consistently sensitive human landing catches throughout the night.²⁸ Lack of explicit molecular data for distinguishing sibling species and molecular forms within the major vector taxa occurring in both Tanzania and Burkina Faso also introduces ambiguity into the interpretation of the results of the study. This limitation if of greatest significance for the Burkina Faso sites, where both populations of An. funestus s.l. exhibited early peaks of outdoor biting activity (Figure 2H and J) that contrast clearly with historical observations of peaks in feeding activity that occurred indoors during sleeping hours for both An. gambiae sensu stricto and An. funestus sensu stricto in other areas of Burkina Faso.^4,36 In the absence of molecular data with which to distinguish exactly which members of the An. funestus s.l. group contributed to these observations, we can only conclude that these distinct, early peaks of outdoor exposure may well be accounted for by secondary vectors, such as An. rivulorum or An. parensis,^11,37–39 that can replace An. funestus sensu stricto when selective pressure is applied by vector control but are of negligible relevance to malaria transmission.^5,15,40 This phenomenon also helps explain the discrepancy between the behaviour-weighted (Figure 3) and simpler binomial estimate (πiB, Table 2), and suggests that the latter may be more representative of exposure to important primary vectors: The more subtle weighted estimate captures the brief but intense period of largely outdoor exposure of a minority of residents to these peaks of biting activity by presumably secondary vectors that occurs between 20:00 and 21:00 hours (Figure 2).

However, the most important limitations of this study are fundamental in nature and relate to the relevance of the π_i parameter itself. Estimates of the proportion of mosquitoes that make contact with treatable surfaces while resting within houses (π_r) would be far more directly relevant to community-level transmission control with IRS rather than ITNs, but field methods for measuring such a quantity have yet to be developed. Also, the proportion of exposure occurring during sleep (π_s) is a more directly relevant determinant of protection with ITNs than is π_i, but could only be estimated for the Kenyan and Zambian sites. The assumption that the latter only modestly overestimates the former obviously introduces some degree of systematic inaccuracy. Examining the two sites for which both quantities are estimable suggests quite modest differences between these alternative estimates of individual protective coverage. However, a very different picture emerges when the same estimates are considered in terms of the gaps in protective coverage that allow malaria transmission and vector populations to persist, highlighting the crucial importance of high biological coverage against the transmission of malaria and accurate ways of measuring such cverage.³³ In Rarieda, biological coverage gaps of only 5% for An. gambiae s.l. and 2% for An. funestus s.l. are apparent when calculated as the complement of π_i, but this contrasts dramatically with values of 18% and 8%, respectively, for the complement of π_s. In Luangwa, corresponding coverage gaps of 11% and 8% when estimated on the basis of π_i are approximately doubled to 23% and 14% when based on the π_s measurement, which more accurately reflects protective coverage with nets.

Despite these limitations, a number of clear, useful, and broadly applicable conclusions can nevertheless be drawn from the present study. It appears that the mosquito taxa that are responsible for most of the malaria transmission in Africa have only mild and inconsistent preferences for feeding indoors. However, biting contact with humans overwhelmingly occurs indoors simply because this is where people spend most of the hours of darkness, which is when these vectors are active. These findings are consistent with the long-standing rationale for prioritization of ITNs and IRS for malaria prevention in Africa, and support their continued upscaling across the continent.^41,42 However, some human exposure to vector mosquitoes occurred outdoors in all sites in the present study (Figure 3), suggesting that additional vector control measures⁴³ that complement ITNs and IRS by targeting this gap in de facto protective coverage may well be required if malaria transmission is to be eliminated in such settings.²⁴

It is particularly encouraging that most of the human–vector interaction found in the present study occurred indoors in the most recently surveyed Rarieda and Luangwa sites, both of which had high ITN coverage at the time. In the Rarieda site, ITNs have had a clear^44,45 and sustained^46–48 impact on malaria transmission and malaria-related morbidity and mortality. Furthermore, substantive changes in vector population composition have occurred, with An. gambiae s.s. all but disappearing, leaving An. arabiensis, which is known to be capable of feeding extensively on humans early in the evenings, before humans go indoors,^13,26,49 as the only remaining vector species of the An. gambiae s.l. complex.¹⁴ The continued high proportions of human exposure to transmission occurring indoors in the absence of personal use of an ITN (π_i) up to at least 2009 may well help explain why supplementing ITNs with IRS confers additional incremental protection in a nearby district,²⁷ despite dramatic changes in vector population composition, and may underpin similar observations elsewhere.⁵⁰

However, these continued high proportions of human exposure to bites by extensively modified residual vector populations¹⁴ in Rarieda contrast strongly with recent observations of dramatic declines in this proportion following ITN and IRS scale-up in Equatorial Guinea²⁰ and in the Ulanga site itself,¹⁹ as well outside of Africa in the Solomon Islands.¹⁸ It is therefore clear that summary estimates of relevant mosquito–human interaction quantities, such as P_i, P_fl, and π_i, should be regularly monitored by national malaria control programmes and carefully considered by policy makers, product manufacturers, and public health funding bodies.^23–25 Care should be taken not to misinterpret such reports of declining proportions of human exposure occurring indoors: These measurements do not necessarily reflect a failure of ITNs or IRS. Instead, they often represent the characteristics of persisting populations of zoophagic and exophagic mosquitoes following the successful control^{5,13–15,19,20,40,51} and even elimination¹⁸ of anthropophagic and endophagic vector populations by ITNs or IRS. By definition, less anthropophagic mosquitoes are less efficient vectors of these malaria parasite species because Plasmodium falciparum and P. vivax are strict anthroponoses that only infect human hosts. Indeed many of these anthrophophagic mosquitoes, such as the An. quadriannulatus, An. rivulorum, An. parensis, An. vaneedeni, and An. leesoni found in Luangwa, are considered to play a negligible role in sustaining the transmission of malaria.^11,37 It may therefore be inappropriate to judge the ongoing effectiveness of commonly used vector control measures such as ITNs on the basis of contemporary measures of mosquito–human interactions because these measures reflect the characteristics of the surviving mosquito populations only. Quantitative estimates of mosquito behavioural parameters, such as those presented here (Figure 3, Table 2), collected before the scale up of ITNs or IRS (Ulanga, Kourowego, Oubritenga), or at least before these interventions had substantially reduced π_i values (Rarieda, Luangwa, Rufiji), may therefore be more representative than contemporary measurements for evaluating the ongoing impact of ITNs on vectors of historical importance. Such historical reference values are therefore crucial to balanced interpretation of contemporary estimates and observations of longitudinal trends. The consistency of the summary values presented in Table 2 suggests that it may be reasonable to extrapolate this range of values beyond the 6 study sites named above so that they may even constitute useful historical reference values for rural African vector populations generally.

Despite the limitations described in the two opening paragraphs of this discussion, measurements of π_i are very useful for approximately assessing de facto protective coverage of humans with ITNs and IRS.^30,31,33 The proportion of human exposure to bites that occurs indoors can be most directly applied to estimating the maximum level of personal protection that can be realistically expected with indoor vector control measures, or combinations thereof.³⁰ However, the relevance of this behavioural parameter extends far beyond personal protection because it is critically important as a determinant of the greater community-level effects that ITNs and IRS can have when used by the majority of the population.³¹ Even though π_i does not directly reflect the probability of insecticide contact while resting, the high estimates for Rarieda help rationalise evidence for an incremental impact of IRS as a supplement to ITNs in a neighbouring district.²⁷ In Luangwa, similar estimates have been used to infer that IRS may also be a useful supplement to ITNs in that setting,³⁰ and the consistently high values presented here are in accord with recent reviews suggesting that this combination of malaria-preventive measures may have broad potential in Africa.^50,52

Beyond IRS and ITNs, π_i is also informative as a primary determinant of target product profiles for complementary measures designed to fill the coverage gaps created when mosquitoes feed outdoors.^24,33 It has long been recognized that pre-existing behavioural resistance traits, specifically preferences for feeding outdoors, usually limit the impact of vector control far more than does physiological resistance to the relevant active ingredients of insecticides.^53–55 In fact, many of the diverse primary vectors distributed across tropical America and Asia are predominantly exophagic.^53–57 Furthermore, residual mosquito populations that have persisted after the scale-up of ITN and IRS in Africa and the Pacific are often perfectly capable of mediating stable, endemic transmission of malaria because they include primary vectors that are behaviourally resistant to these measures.^{14,18–20,26,49,58,59} The primary parameter that determines the comparative merits of vapour-phase insecticides that can be used in outdoor spaces, as opposed to contact insecticides, which by definition require a treatable surface to which they can be applied, is the proportion of human exposure occurring indoors.^24,33 The consistently high values for this quantity reported in Figure 3 and Table 2 confirm that ITNs and IRS using contact insecticides are indeed the logical first choices of intervention, while the intermediate values reported recently from residual populations across the tropics^18,19,26 suggest that supplementary use of vapour-phase repellents may well effectively complement these traditional approaches in such situations.²⁴ In addition to the usual assays of physiological susceptibility to insecticides that are already integral to choosing vector control measures,^21,60 up-to-date surveys of vector behavioural characteristics will also be essential to underpinning the selection of alternative or additional vector control technologies.

A recent modelling analysis cautions that supplementing existing ITNs or IRS with indoor use of spatial repellents may undermine and reverse the impact of the former upon historically important anthropophagic and endophagic vectors that have been suppressed but persist and can therefore recover if they are deterred from houses where they would otherwise be killed.²⁴ When deciding about whether to supplement ITNs with IRS, it is essential to consider not only the contemporary values of such behavioural quantities for surviving residual vector populations, but also the normal range of values for historically important vectors that need to be suppressed indefinitely.³³ To conclude, we believe that historical values for such behavioural parameters, recorded before the occurrence of wholesale changes in vector population composition, are likely to be more useful for rationalizing the impact of ongoing interventions, while equivalent, contemporary surveys of surviving residual populations are more appropriate for informing strategies to augment existing means of vector control and for ultimately eliminating the transmission of malaria.³³

The MTIMBA project was initiated by the International Network for the Demographic Evaluation of Populations and Their Health (INDEPTH) and was financed by the Multinational Initiative for Malaria/The Special Programme for Research and Training in Tropical Diseases Research and the Roll Back Malaria Programme. The MTC project was funded by the Bill & Melinda Gates Foundation (Award 45114, coordinated by Professor Frank Collins and Dr. Neil Lobo at Notre Dame University, South Bend, IN, USA). Studentship for B.H. was financed by the Stipendienkommission für Nachwuchs-Kräfte aus Entwicklungsländern of Basel Kantonal, Basel, Switzerland.