Over the last two decades, malaria control has been scaled up throughout sub-Saharan Africa with an emphasis on the distribution of long-lasting insecticidal nets (LLINs), targeted application of indoor residual spraying (IRS), and improved diagnostics and case management. As a result, the burden of malaria has declined substantially with a 40% reduction in incidence and a 50% reduction in prevalence between 2000 and 2015. While LLINs contributed an estimated 68% of the decline in malaria prevalence, IRS was responsible for 13%¹.

The efficacy of insecticide-treated nets was demonstrated in a series of cluster randomized, controlled trials². Formal randomized controlled trials of IRS have also demonstrated the efficacy of IRS^3,4. Furthermore, there is a long history of programmatic implementation of IRS in many settings of the world which resulted in reduced malaria burden and even elimination in some settings². The use of both LLINs and IRS for malaria control has a direct impact on mosquito bionomics. LLINs and IRS have multiple effects on mosquito populations which may result in reduced malaria transmission including: reduced indoor Anopheles densities^5–7, shifts in vector species composition⁸, changes in the time and location of mosquito biting^9–11, and changes in host selection¹², and increases in early exophily¹³.

In western Kenya, vector control included universal coverage of LLINs through periodic mass campaigns and routine distribution to high-risk groups as well as IRS in specifically targeted areas. The first mass LLIN distribution occurred in 2006 and targeted children <5 years of age. Additional distributions aiming for universal coverage occurred in 2011 and 2014 leading to 54% of households in the lake endemic zone having one LLIN for every two residents¹⁴. The region also bears the highest malaria burden nationally^14–16. Implementation challenges facing LLINs include incomplete coverage^14,17–19, widespread pyrethroid resistance^20–23 and possible changes in vector behaviour⁹.

IRS in western Kenya was based exclusively on pyrethroids until 2012^24–26. However, spraying was interrupted between 2013 and 2017 due to widespread pyrethroid resistance in local malaria vector populations and the lack of a registered, non-pyrethroid insecticide in the country. In response to widespread pyrethroid resistance, the Kenyan National Malaria Control Programme (NMCP) developed an insecticide resistance management strategy involving the rotation of different non-pyrethroid classes of insecticides used in IRS every two years in endemic and epidemic-prone areas where 80% or more households own one or more LLIN²⁷. This is in accordance with global insecticide resistance management strategy aimed at delaying the rise and spread of insecticide resistance to new classes of insecticide while preserving pyrethroids for use in bednets²⁸. In 2017, IRS was re-introduced using a microencapsulated formulation of pirimiphos-methyl (Actellic 300CS (capsule suspension)). The insecticide has been reported to be effective against pyrethroid-resistant Anopheles mosquitoes^29–31 and has a relatively long residual effect on sprayed wall surfaces, of up to twelve months^29,30,32.

Between 2008 and 2012, IRS campaigns in western Kenya were conducted using pyrethroid based insecticides. These campaigns resulted in significantly lower parasitaemia and cases of clinical malaria in IRS districts compared to non-IRS districts²⁶. The protective efficacy of ITN plus IRS compared with ITN only in the same area was measured at 62%, whereas, 60% fewer anophelines were observed in the sprayed districts compared to unsprayed districts post-IRS²⁵. Anopheles funestus accounted for 1.7% of total anophelines collected in the sprayed areas before IRS and 2.7% after IRS²⁵, an indication that the species was not affected by spraying. Following reports of increasing insecticide resistance, the malaria vector control policy changed to the use of non-pyrethroid insecticides for IRS. IRS was reintroduced in 2017 with an organophosphate, Actellic 300CS. We sought to determine efficacy of the Actellic CS based IRS campaign on entomological and epidemiological indices against a background of moderate to high coverage of pyrethroid LLINs¹⁴ and extensive pyrethroid resistance.

Our results demonstrate a significant reduction in An. funestus indoor resting densities, biting rates and sporozoite infections, as well as a decline in malaria test positivity rates and case-count at health facilities after one round of Actellic 300CS IRS in Migori County, western Kenya. Human biting rates and sporozoite infections in Anopheles mosquitoes are the most direct entomological measures of malaria infection risk. We observed moderate biting and sporozoite rates in both intervention and non-intervention sites before IRS and the unsprayed sites after IRS. However, after IRS, An. funestus biting rates were nearly zero, and no sporozoite infections were detected post-IRS. Susceptibility tests confirmed that the major vector species, An. arabiensis and An. funestus, were both resistant to pyrethroid insecticides but were susceptible to pirimiphos-methyl, the active ingredient in Actellic 300CS.

Similar reductions in An. funestus populations to near elimination were observed in the Asembo Bay area of western Kenya following the scale-up of pyrethroid-treated nets⁵ although An. funestus later returned as the primary malaria vector in the region presumably due to the development of pyrethroid resistance³⁴. The complete elimination of An. funestus following effective IRS campaigns have been reported in South Africa, Mauritius and the Pare/Taveta area of Tanzania and Kenya^49,50. An. funestus is particularly sensitive to effective insecticides used in IRS, probably due to the highly endophilic and anthropophilic nature of the species^51–53, leading to high levels of exposure to treated surfaces.

In contrast, An. arabiensis indoor resting densities, human-biting rates, and sporozoite infection rates all reduced only marginally in sprayed areas post-IRS. With the decline in An. funestus, An. arabiensis became the predominant vector species in the sprayed areas (although the densities of An. arabiensis did not increase). IRS has a limited impact on the population of An. arabiensis, despite full susceptibility to pirimiphos-methyl in WHO susceptibility tests. This lesser impact of IRS on An. arabiensis is therefore unlikely to be due to insecticide resistance but may be attributable to the behaviour of this species. Blood-meal host analysis showed that An. arabiensis fed more frequently on cattle than humans, unlike An. funestus that fed more frequently on humans. This finding is supported by a previous study in western Kenya that reported An. arabiensis fed predominantly on cattle (65% of blood meals on cattle; 22% mixed bovine/human; 13% human)⁸. Furthermore, results from a deterministic model, developed using data from Kilombero, Tanzania, suggested that An. arabiensis fed outdoors on both humans and cattle and rapidly exited houses without fatal exposure to insecticidal nets or IRS⁵⁴. A recent trap evaluation in western Kenya observed over sevenfold more An. arabiensis collected by cow odour compared to human odour outdoor⁵⁵. Therefore, it is likely that a significant population of An. arabiensis rests predominantly outdoors and feeds primarily on cattle, but occasionally bites humans and transmits malaria, albeit less efficiently than the more anthropophilic vector An. funestus. Therefore, it is possible that the population of An. arabiensis collected indoors by light traps and PSC represents only a proportion of a larger outdoor population. These factors may explain the lesser impact of IRS on An. arabiensis.

IRS with Actellic 300CS had a prolonged residual activity of at least ten months post-IRS, as measured by wall bioassays. As spraying was conducted in February, the insecticide provided protection throughout the periods of highest malaria transmission during the long (April–June) and short (October–November) rainy seasons. This long residual efficacy of Actellic 300CS makes the insecticide particularly useful in providing all-year-round protection with just one spray round each year. Similar prolonged residual activity of Actellic 300CS and control of pyrethroid-resistant mosquitoes have been reported in other countries³². As this is the first report of the efficacy of IRS with a non-pyrethroid insecticide in Kenya in the wake of pyrethroid resistance, strict adherence to the insecticide resistance management strategy^27,28 is needed. The Kenya National Malaria Control Program (NMCP) has a responsibility to fast track registration of additional non-pyrethroid classes of insecticides to enable rotation of insecticide in IRS for effective resistance management.

This study presents three novel findings that are crucial for malaria control and elimination in Kenya. Our study demonstrated that a single spray-round with Actellic 300CS is effective for year-round control of malaria in the region. Well-timed IRS, just before the peak malaria transmission, suppressed the local malaria vector populations through both the long and short rain periods, hence reducing disease transmission even during the peak transmission seasons that coincide with the long and short rain seasons. Second, the impact of IRS on An. arabiensis was less than that on other vectors, making it the dominant malaria vector sustaining residual transmission. While IRS was effective in controlling An. funestus, An. arabiensis would require alternative control approaches for malaria elimination. Finally, the study demonstrates late morning biting by An. funestus, a phenomenon not before investigated in the region. This presents a new challenge facing vector control given that the tools currently deployed for vector control are limited to the household and mostly only to the times when people are asleep (LLINs).

Biting by An. funestus in the intervention and non-intervention areas before IRS and non-intervention areas after IRS occurred mostly indoors late at night corresponding to the period when most people were indoors and in bed. Late night, indoor biting by An. funestus has been previously reported, dating back to the pre-bed net era. For instance, in 1975, 94% of An. funestus were observed to bite after midnight, with another a peak in the hours before dawn in Kano plain of western Kenya⁵⁶. Similar late night, indoor biting was observed in the same study area in 1996⁵³ and more recently the vector species have been reported to persistently bite indoors, late in the night despite high coverage in insecticide-treated nets⁵². While all these studies observed a biting peak at dawn, the collections ceased at 07:00 hours. With extended collections to monitor An. funestus biting in the morning, we observed biting until 11:00 hours. Similar findings of day-biting An. funestus in the presence of LLINs have been recently reported in Senegal⁹ and Benin⁵⁷. This observed extended biting behaviour in An. funestus not previously investigated may potentially undermine the effectiveness of LLINs as people may be exposed to mosquito bites while away from the protection of their bed nets. However, one round of IRS with Actellic 300CS substantially reduced the number of An. funestus collected and it was not possible to detect biting either indoors or outdoors in the sprayed areas post-IRS.

In an analysis of risk factors associated with the indoor occurrence of mosquitoes, fewer An. funestus and An. arabiensis were collected by light traps in houses with closed eaves. Similar results were observed for An. arabiensis collected by PSC. Open eaves are known to be the main route for indoor entry of Anopheles mosquitoes^11,58 and blocking them has been demonstrated to be effective in preventing Anopheles house entry⁵⁹. Closing eave spaces or deploying vector control tools in these spaces may present an additional intervention to the current vector control tool kit for reducing the indoor occurrence of mosquitoes in addition to IRS and LLINs. However, new tools targeting outdoor Anopheles populations, in particular, An. arabiensis, are needed to control vector species that are less well controlled by interventions that target the indoor environment such as IRS and LLINs.

Reductions in malaria cases at the health facilities within sprayed areas post-IRS provided further evidence of the impact of a single round of IRS on malaria transmission. Health facility-based surveys of malaria cases in febrile patients have been useful as part of rapid analysis of changes in local malaria epidemiology^60–63. Malaria infection is highly correlated with febrile cases reported at the health facilities⁶². Furthermore, a systematic review of febrile illness over 20 years in sub-Saharan Africa reported a dramatic reduction in the proportion of fevers associated with Plasmodium falciparum malaria⁶⁴. Consequently, reductions in malaria cases likely contribute considerably to the reductions in febrile illnesses presenting at health facilities. The use of routine Health Management Information System (HMIS) data to evaluate malaria control interventions⁶³, however, suffers from incompleteness in reporting and variation in the utilization of the health system⁶⁵. We extracted data from the primary records, health facility laboratory registers and observed a reduction in confirmed malaria cases in health facilities within sprayed areas post-IRS, with no change in the number of cases detected at the facilities in control regions. Reduction in An. funestus densities, sporozoite rates and man-biting rates coupled with reduced malaria cases following one round IRS with pirimiphos-methyl provide compelling evidence of the effectiveness of IRS in malaria transmission reduction when implemented with an effective insecticide to which mosquito populations are susceptible.

This study compared entomological indices and malaria test positivity rates between one district receiving IRS in addition to LLINs and the neighboring districts receiving only LLINs. This approach was driven by the fact that it was an effectiveness study rather than a highly controlled study to assess efficacy. However, the study does raise questions about the optimal approach to evaluate vector control interventions such as IRS or larval source management which target population-wide suppression of malaria vectors and where the scale of the study may affect the magnitude with which the results can be generalized. While cluster randomized trials might be ideal for the evaluation of these tools, the flight range of mosquitoes should be accounted for as migration from outside clusters and/or spill-over effects from intervention clusters into control clusters may bias results towards the null⁶⁶. However, larger clusters, in addition to being logistically unfeasible, may result in clusters that are spatially distant and less similar which may also bias the results. In this study, the non-random selection of areas for IRS, the low number of clusters and their spatial distance are a potential limitation of the study. However, the before and after design coupled with multiple endpoints indicating similar trends in mosquito densities and malaria-related outcomes strongly suggest that the results are robust.

IRS with pirimiphos-methyl was highly effective for the control of indoor biting and indoor resting, pyrethroid-resistant An. funestus and resulted in substantially reduced numbers of this primary vector species coupled with reduced malaria cases. Due to the long residual effect of pirimiphos-methyl, it was possible to achieve year-round protection with a single round of IRS. Sustaining these gains is a priority for the Kenya NMCP and development partners and IRS should continue to be implemented to sustain the impact on An. funestus. However, there was less of an impact of spraying on An. arabiensis populations, likely due to their exophilic nature. Additional control measures are needed to control outdoor biting and resting An. arabiensis.