Disease control strategies that aim to reduce human-vector contact must consider the risk factors that influence where and when this contact occurs. The house is a common site of human-vector contact and this is reflected in the high degree of endophily and endophagy exhibited by important vector species, such as the malaria-transmitting mosquitoes Anopheles gambiae sensu stricto Giles and An. funestus Giles over much of Africa1, Culex pipiensquinquefasciatus Say, a domestic biting nuisance worldwide and a vector of lymphatic filariasis in tropical Africa, Southeast Asia, Western Pacific and the Americas23, and Aedes aegypti L. in West Africa, where it is a vector of yellow fever, dengue and Chikungunya viruses4. House-entering behaviour has evolved in some mosquito species and, more specifically, within some species populations but not others 45. This behaviour may partly be a consequence of their anthropophily; these species have evolved with their human hosts and in human-made environments6. For example, 'domestic' Ae. aegypti breed year round in clay jars maintained indoors for water storage 7. The attractiveness and ease of entry into a house by mosquitoes is affected by both structural factors and social practises. A recent study in a semi-rural area of The Gambia showed that An. gambiae sensu lato are more likely to enter houses with open eaves, mud brick walls and many occupants, whilst burning churai, a local incense, reduced house entry 8.

Studies that use entomological techniques such as light traps or knockdown catch, to monitor anopheline populations as part of malaria surveillance and risk assessment, will usually also trap a large number of culicine mosquitoes. Often the data relating to the culicine portion of the catch is not analysed or not published, yet some of the culicines trapped serve as disease vectors and major nuisance biters. Here we have assessed putative risk factors that determine the level of human exposure indoors to potential vectors of pathogens other than malaria. A multi-factorial risk factor analysis study was designed to highlight important spatial, compound-, house- and mosquito control-related parameters that affect house entry of culicine vectors and species that are a biting nuisance in The Gambia.

In this study culicines represented a substantial fraction of the average light trap contents: though there were more anophelines than culicines in all months of the trapping period, the culicine proportion was never smaller than 25% of the total monthly catch. The density of culicines is important as a contributor to the overall biting nuisance and its impact on malaria control. For example, it has been shown that a biting nuisance threshold of 10–15 mosquitoes/person/night exists, irrespective of mosquito species, below which bed nets are not generally used in Gambian villages 11. Culicines are also important as transmitters of disease. The dominant culicine species were Cx. thalassius, Cx. pipiens s.l., M. africanus, M. uniformis and Ae. aegypti. With the exception of Cx. thalassius, all of these species are potential vectors in this setting. Cx. pipiens quinquefasciatus is probably the most common man-biting culicine in Africa1213 and has been shown to be a vector of Wuchereria bancrofti 14 and Chikungunya 15 across much of the continent. Ae. aegypti is a vector of dengue, urban yellow fever and chikungunya in Cote d'Ivoire4 whilst M. africana and M. uniformis have been shown to be vectors of lymphatic filariasis, Rift Valley Fever and West Nile Virus in nearby Senegal16.

The results presented here identify some of the many factors that govern the risk of exposure to culicine vectors, particularly at the compound and household level. The urban-rural difference in catch numbers and proportion of Cx. pipiens s.l. may be explained by the availability of breeding sites, if one accepts that Cx. pipiens quinquefasciatus is the dominant member of the complex here. Cx. pipiens quinquefasciatus prefer breeding in household sewage systems, soakage pits, septic tanks and storm drains1718, structures more common to urban areas. This is also reflected in the relationship between Cx. pipiens quinquefasciatus and the distance to the nearest pit latrine in Farafenni. Cx. pipiens quinquefasciatus has a very short flight range19 and is likely to be caught not far from its breeding sites. 'Domestic' Ae. aegypti can breed year round in clay jars maintained indoors for water storage7 which may account for the greater prevalence of this species in town (22% v 18%). By contrast Cx. thalassius breeds in brackish water and is common to pools in the mangrove swamps in The Gambia20, while M. africana and M. uniformis breed in natural swamp areas and rice fields with emergent vegetation such as water lilies and water lettuce2122. The majority of these types of breeding site are found on the flooded alluvial plains bordering the River Gambia2324. Though the distance of each study house to the nearest larval habitat was not measured, it is the case that the rural settlements are closer to the alluvial plainsand therefore are exposed to more mosquitoes than those situated on the edge and in the middle of Farafenni town, further away from these sites825.

At the compound level, in the villages only, it was found that the presence of horses tethered near the house at night was associated with lower numbers of mosquitoes in the trapping room, but that cows increased the likelihood of culicinesother than Cx. pipiens s.l., entering the house. The presence of smaller domestic animals was not recorded as these were usually kept further from the houses or outside the compound at night. The presence of livestock has previously been shown to attract culicines into a compound26. These findings are of significance to mosquito control since zooprophylaxis, the use of large domesticated animals to reduce biting frequencies, has been successful against culicines in some parts of the world2728. Nevertheless, caution should be exercised in recommending zooprophylaxis against culicines, particularly in areas where they are vectors of zoonotic arboviruses that require amplification in domestic animalsOur findings suggest horses may be an important bloodmeal for many culicines here but that cattle, probably through production of CO₂, act as attractants but not bloodmeals for the culicines found, with the exception of Cx. pipiens s.l.. However, blood meal analysis or host choice studies are needed to confirm this. Horses have been shown to reduce culicine house entry in another study in this area29. The reasons for the attractiveness of horses may be partly due to the high bed net usage in The Gambia30 making it difficult for a blood-questing mosquito to locate a human host, as well as the unusually high density of equines in the Senegambia region31.

Both cattle and humans are primary blood meal sources for Cx. pipiens quinquefasciatus in western Kenya12, which feed opportunistically from humans when they enter houses but largely from cattle outside 32. It seems plausible that the Cx. pipiens quinquefasciatus in this study behaved similarly. That is, they are more catholic in their feeding behavior than reported from other parts of the world e.g. in south India33.

Within the house the important variables were the status of the eaves, the height of the walls, the roofing material and the number of occupants. Open eaves are the main route by which many mosquitoes, including Mansonia spp, enter homes, and closing eaves has been shown to reduce the number of mosquitoes entering a house3435. Mansonia spp fly at a very low (<0.15 m) height above the ground36 and so it seems that they must fly upwards when encountering a vertical wall surface, following the cues emanating from the eaves and becoming channeled indoors through the open eaves by the overhanging roof. This behavior is a common trait of endophagic mosquitoes and probably hasa genetic basis. For example, house entering by Ae. aegypti in Kenya was shown to be a characteristic of an indoor form37 and genetically determined5. The importance of eaves to Cx. pipiens quinquefasciatus house entry is less clear. For example indoor-resting density was suppressed by 61–96% by deltamethrin-impregnated curtains hung in the eaves and doorways of huts in Southern India38, but in Trinidad door and window curtains and eave treatment were not related to the number of Cx. pipiens quinquefasciatus caught indoors39.

Wall height had a weak affect on mosquito numbers caught indoors and was only associated significantly with the other culicines group caught in town. Cx. thalassius was the dominant mosquito in that group and has been shown to be prevented from entering houses by increasing wall height. For example, a 62% decrease in Cx. thalassius house entry was observed between a hut with a 60 cm wall compared to one with a 172 cm wall35. The average wall height in Farafenni houses was significantly higher and the minimum-maximum range greater than in village houses which may partly explain why the trend was not observed in the villages.

Thatched roofs are typically associated with open eaves and poorer quality houses in this setting so it is surprising that they are linked with a reduction in Cx. pipiens s.l., albeit in town houses only. Metal roofed houses are more permanent structures, and it has been reported elsewhere that Culex spp are more abundant in permanent houses (typically houses with cement walls and metal roofs) compared to traditional thatched homes 394041. Why this should be the case is unclear.

Routes of entry into houses are recognized by olfactory cues i.e. the carbon dioxide and body odors emanating from the hosts inside42. That more culicines are found in rooms with higher numbers of people is unsurprising since it has been shown that odours emanating from individuals attract mosquitoes towards a human host4043. An increased number of occupants are likely to increase the attractiveness of a house to culicines. Occupant number will increase at night as people go to bed, and this coincides with periods of high activity in two of the most common species caught in this study. Cx. pipiens quinquefasciatus is active throughout the night in many locations 3, with peak activity from 22–23:00 h44, and M. uniformis tends to feed indoors post-dusk and leave the subsequent night for outdoor resting sites45. Furthermore, while Ae. aegypti is typically considered a diurnal species, in Côte d'Ivoire it has shown persistent indoor biting activity during the night up to 06:00 h with peak activity at 00:00 h both inside and outside 4. The age, sex and ethnicity of the occupants had no obvious effects in this study. Pregnancy has been identified as a risk factor for attractiveness to Mansonia spp46 but was not recorded here.

Because there may be unmeasured biological or social confounders operating at a scale not explored here, caution is needed before interpreting these results as indicating causal associations, and therefore it is important that these associations be tested by rigorous experimental trials in which each variable can be isolated. Equally it is important to note that much of the culicine-borne disease transmission and biting nuisance will occur away from the home owing to the opportunistic feeding behavior that is characteristic of some of the species captured here. As a result, culicine control campaigns should not ignore biological or chemical approaches to reduce breeding sites. Nevertheless, it is likely that the household factors shown here to be statistically associated with risk of culicine house-entry are relevant to disease transmission. Therefore, these results suggest that closing eaves, covering pit latrines or treating them with polystyrene beads, and zooprophylaxis using horses in rural areas, are simple household control measures that could reduce the number of culicines found indoors.

The authors declare that they have no competing interests.

MK coordinated the fieldwork, data collection and sample identification, and drafted the manuscript. PW assisted the data collection and sample identification. CG carried out the PCR. MJ participated in the design of the study and provided access to the demographic database. SWL conceived of the study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.