Malaria vectors and their blood-meal sources in an area of high bed net ownership in the western Kenya highlands

Female mosquitoes take blood meals to obtain nutrients needed for egg development. In the process of obtaining blood meals from humans they can unwitting gain the potential to vector malaria parasites between hosts. The blood-meal source of malaria vectors affects the mosquito population’s capacity to transmit malaria, as illustrated by Garret-Jones’ model on vectorial capacity [1]. One of the factors included in the model is the frequency with which each mosquito bites a person, i.e., the probability that a particular mosquito will bite a human being rather than an animal on a given day.

Africa has the most efficient human malaria vectors, Anopheles gambiae sensu stricto and An. funestus s.s. and the most virulent malaria parasite, Plasmodium falciparum [2‐5]. Their efficiency as malaria vectors is mainly due to their highly anthropophilic biting behaviour [2, 6‐8]. These mosquitoes are the key malaria vectors in the western Kenya highlands [9, 10] where they have been shown in the past to prefer human blood meals and to feed indoors [6‐8]. Similar feeding behaviours of malaria vectors were reported from the Kenyan coast about two decades ago [11]. These studies were done long before the introduction and scaling-up of insecticide-treated nets (ITNs)/long-lasting insecticidal nets (LLINs) and indoor residual spraying (IRS) to control malaria vectors.

Current malaria vector control methods are designed to prevent human-vector-contact and distribution of LLINs is aimed at universal coverage of all households in sub-Saharan Africa [12]. Insecticides on the net repels, disables and/kills mosquitoes that come into contact with them [13, 14]. The net also provides a physical barrier for mosquitoes to bite well-covered people. In areas where bed net coverage is high, malaria vectors might find it increasingly difficult to find a successful blood meal from their favourite host [10, 14, 15]. This study investigated malaria vector abundances and their blood-meal sources in an area of high bed net ownership in the western Kenya highlands.

A total of 193 residential houses were mapped within the boundaries of the three study areas but household data on mosquito control methods could only be obtained from 153 houses and data on domesticated animals from 152 houses (Table 1). This was due to absence of the residents of the house on the survey dates or refusal to consent to be enumerated. A total of 568 persons were counted in the 153 houses, with an average of 3.71 (3.41–4.02) persons/household (Table 1). Little less than a fifth (18.3 %) of the population was under the age of 5 years. The three study sites were similar in their mean number of household members and in the mean number of domesticated animals kept per household. The only difference between sites was bed net ownership which was highest in Ebulako with a mean of 2.13 nets per household and lowest in Inavi with a mean of 1.43 nets per household (Table 1).

Out of the 153 households that were sampled, 132 (86.3 %) had bed nets with a total of 511 residents. Twenty-one (13.7 %) households were without bed nets and had a total of 57 residents. Fifty (32.7 %) households had one bed net each and with a total of 132 residents. Forty-four (28.8 %) households had two bed nets each and a total of 186 residents. Twenty-nine (19.0 %) households had three bed nets each and a total of 140 residents. Six (3.9 %) households had four bed nets each and a total of 32 residents. Only one (0.7 %) household had five bed nets with four residents. Two (1.3 %) households had seven bed nets each and with a total of 17 residents. On average, one bed net was used by 2.1 (568/268) people. Bed nets were the main malaria prevention method used in the communities. These were supplemented by anti-mosquito sprays in 5.2 % (8/153) of the households and mosquito coils in 2.0 % (3/153). No other mosquito control measures were used within the three sites.

Over a half (57.2 %, 87/152) of the interviewed households had domesticated animals. A total of 785 animals were kept, most of them being chickens (Table 2). Most of the domesticated animals slept in the main house (86.0 %), few in the sheds (5.7 %), in the kitchen (5.2 %), and outside (3.1 %) in the open (Table 2). The majority of households (83.6 %, 158/189) had people and domesticated animals sharing the same houses, including kitchens, at night. It is especially notable that the majority of cattle, goats and chickens were kept in the main house during the night (Table 2).

A total of 80,331 mosquitoes were collected: 76.5 % (61,459/80,331) of them were collected exiting through windows at night; 9.9 % (7943/80,331) were collected resting indoors; 7.4 % (5919/80,331) resting outdoors in AgREPOTs and 6.2 % (5010/80,331) resting outdoors in SRB (Table 3). A total of 1911 malaria vectors were unfed and most of them (94.9 %, 1814/1911) were caught exiting through the window traps. The majority (94.6 %) (76,033/80,331) of the mosquitoes were Culex species, 4.3 % (3493/80,331) were An. gambiae s.l. and 1.0 % (805/80,331) was An. funestus s.l.. Most of the females caught were unblood-fed: 80.6 % (1602/1987) of the collected An. gambiae s.l.; 72.8 % (398/546) of the An. funestus s.l. and 47.2 % (23,689/50,205) of the Culex species. The overall mean density of An. gambiae s.l. in PSC was 0.17 (95 % confidence interval (CI) 0.13–0.22) and 0.04 (95 % CI 0.03–0.06) in WET. The An. funestus s.l. overall mean density was 0.02 (95 % CI 0.01–0.04) in PSC and 0.01 (95 % CI 0.01–0.02) in WET. Means of An. gambiae s.l. and An. funestus s.l. in AgREPOTs and SRB could not be modelled due to the low numbers that were collected outdoors. The mean density of Culex spp. was 3.06 (95 % CI 2.75–3.39) in PSC, 2.79 (95 % CI 1.93–4.03) in WET, 0.85 (95 % CI 0.74–0.97) in AgREPOTs, and 3.14 (95 % CI 3.14–3.14) in SRB. Average monthly numbers of adult mosquitoes collected in WET, AgREPOT and SRB were significantly lower than those collected by PSC, except for An. funestus s.l. in WET which was not significantly different (Table 4).

The mean density of blood-fed malaria vectors per house was low in the study areas in general, however, significant differences were observed between the study areas (Table 5). It was 8.2–9.4 times more likely to collect a blood-fed An. gambiae s.l. in Muluhoro than in the two other study sites and 9.6 times more likely to collect a blood-fed An. funestus s.l. (Table 5). Out of the 208 blood-fed adult An. gambiae s.l. mosquitoes collected (Table 3), 23.6 % (49) were analysed by ELISA to determine the sources of their blood meals and also by PCR to determine their species.

Blood-meals sources identified from An. gambiae s.s. were 53.1 % from humans, 26.5 % from goats and 18.4 % from bovines (Table 6). Some An. gambiae s.s. mosquitoes obtained blood-meals from single sources, that is, 26.5 % from humans, 8.2 % from bovines, 2.0 % from goats. One (2.0 %) An. arabiensis obtained blood from humans only (Table 6). No blood-meal from chickens was identified from both An. gambiae s.s. and An. arabiensis. Mixed blood-meal sources identified from An. gambiae s.s. included: 24.5 % (12/49) human/goat, 10.2 % (five) human/bovine, 8.2 % (four) human/bovine/goat and 8.2 % (four) bovine/goat (Table 6). A total of 564 An. gambiae s.l. were further identified by PCR: 383 (67.9 %) were An. gambiae s.s., four (0.7 %) were An. arabiensis and 177 (31.4 %) were not identified due to amplification failure.

There has been a dramatic increase in the proportion of households owning at least one bed net, from 3 % in 2004 to estimated 49 % (range 44–54 %) in 2013 in sub-Saharan Africa (WHO 2014). In Kenya, LLIN ownership was estimated in 2008 to be 56 % nationwide [29] and more recent reports from western Kenya show ownership and usage to exceed the World Health Organization (WHO) target of 80 % coverage [10, 30, 31]. The latter can be confirmed by the here presented data. In the study area in the western Kenya highlands bed net coverage was 86 % and the recommended minimum bed net coverage ratio of one LLIN per two persons at risk of malaria [32] has been achieved. However, this achievement may be distorted by the existence of a gap between bed net coverage and usage, which was not further explored in this study but which has been shown to vary among sites in the area and seasons with greater usage during high malaria transmission seasons [20].

The physical barrier preventing a blood meal combined with a repellent effect of the LLINs may have led fewer malaria vectors to rest, feed and remain indoors until they became gravid. This is in agreement with other studies that reported a reduced likelihood of malaria vectors to obtain blood meals indoors but an increased likelihood to exit from houses with bed nets [33, 34]. Outdoor mosquito collection tools used in this study, namely AgREPOTs and SRBs however, failed to sample the mosquitoes outdoors. Hence, the blood-meal sources results presented in this study has the bias of only using mosquitoes collected indoors. Therefore, there is an urgent need to investigate where malaria vectors rest outdoors after leaving the houses in this study area in order to develop more efficient sampling tools.

Animal husbandry is common and a key economic activity within the western Kenya highlands [22]. Livestock are mainly kept for food, sale of their products and as a financial security in times of need including the payment of pride prices. Dogs are kept to offer security at night and cats to keep rats and snakes away from the house. Due to increased human population density [21] and livestock theft cases [22]; the practice of most households sharing the same houses at night with livestock is common in these areas. This is done to enhance livestock’s security at night. This puts both people and domesticated animals in close vicinity to malaria vectors at night.

An interaction between increased bed net coverage of humans and the presence of domesticated animals in the same houses at night likely plays a key role in this increased observation of animal and mixed human-animal blood meals taken by local malaria vectors. This is likely enhanced by the physical, chemical (excito-repellence, especially pyrethroid insecticides) and community-wide barriers that LLINs offer humans against malaria vectors [14]. Malaria vectors that fail to obtain blood meals from humans are compelled to seek and obtain from other readily available animal sources. A similar diversion of An. gambiae s.s. and An. arabiensis mosquitoes from feeding on humans to cattle was reported about three and half decades ago and was attributed to insecticidal spraying using fenitrothion [35].

Blood-meal sources were tested only for origin from humans, goats, bovines, and chicken, but there is a possibility that malaria vectors obtain blood meals from all available domesticated animals, as shown by a large number of previous and recent studies indicating that the major malaria vectors in sub-Saharan Africa readily adapt to available blood-meal hosts even if they have a preference for human hosts [6, 7, 11, 36‐45]. Approximately two decades ago, the human blood index (HBI) for An. gambiae s.s. collected indoors within the region north of the Lake Victoria (Kisian, Saradidi and Mumias) was 88–97 % an indication that they had fed exclusively on humans [6‐8]. However, in this study, the HBI was 53 % and, may be more importantly, the majority of those that had fed on humans had also fed on animals. This strongly suggests a shift in blood-meal sources as a result of the interaction between increased bed net coverage and close proximity of domesticated animals. Comparing the data from this study with the data from Shililu et al. collected in this region two decades ago before bed nets were used in large numbers and treated with insecticides [8, 46], blood-meal sources from humans have dropped by 43 % (96 % minus 53 %), and from bovines has increased by 13 % (5 % minus 18 %). Mutuku et al. [39] reported a similar reduction in HBI and increased blood-meal sources from cattle and goats following increased ITN use on Kenya’s south coast. The key implication of this shift of malaria vector blood-meal sources from humans to domesticated animals is a reduction in malaria transmission as the Plasmodium parasites that cause human malaria do not develop fully in the domesticated animals. Infective stages of the malaria parasite (sporozoites) injected in animals by malaria vectors, in the process of taking blood meals, reach a dead end in their development cycle. This is enhanced by the fact that malaria vectors obtain blood meals from multiple hosts, including different types of domesticated animals. This may have contributed to the decline in malaria epidemics, prevalence, incidence, and distribution that is being reported in Kenya [47].

A shift of malaria vector blood-meal sources from humans to animals presents an opportunity for their control, especially in areas where livestock is kept in close vicinity to people. For example, insecticides can be applied on livestock in order to kill the malaria vectors when biting [48‐50]. However, the few relevant studies conducted to date have focused on pyrethroid-treated cattle [48, 49]. The insecticides tested are identical to those used for mosquito control indoors and subject to increasing levels of resistance [51]. Accordingly, treating cattle with pyrethroids is unlikely to be sustainable and would contribute to even more rapid development of resistance. Furthermore, it has been shown, with few exceptions, that the impact of pyrethroids on cattle is effective for less than 1 week [52‐54] which would require weekly re-application. In addition, some pyrethroids have shown repellent effects on mosquitoes [55] which can be highly counterproductive, as vectors could be diverted to feed on people, thereby increasing transmission [56]. Consequently, there is a need to develop cattle-targeted interventions based on insecticides with a completely different mode of action to insecticides currently used (i.e., pyrethroids, organophosphates) or proposed (i.e., chlorfenapyr) to control mosquitoes indoors. Promising novel control agents might be cattle endectocides or insect growth regulators [57]. It has been suggested that treatment of livestock in the entire community at the same time can maximize protection from malaria vectors and other livestock ecto-parasites and biting flies [58]. Such an approach is likely to get community support due to its added advantage of protecting both humans and livestock. Zooprophylaxis and the use of insecticide-treated cattle have been thought to work well for An. arabiensis, which shows naturally a more zoophilic behaviour [59, 60] but has rarely been suggested for other malaria vectors. However, the findings of this study suggest that it can also work for An. gambiae s.s., in areas where humans are well covered by bed nets and where malaria vectors, humans and livestock are at close vicinity at night.