Background
Larviciding is the application of biological or chemical insecticides that kill the immature stages of mosquitoes, and is one of the approaches of larval source management (LSM), along with habitat modification, habitat manipulation and biological control [
1]. The World Health Organization (WHO) recommends larviciding as a supplementary intervention against malaria in addition to the core vector control interventions of insecticide-treated bed nets (ITNs) and indoor residual spraying (IRS) [
2]. Larviciding is recommended in areas where the intervention is feasible and cost-effective, mostly in urban areas and during the dry season where breeding sites are ‘fixed, few and findable’ [
1,
2]. Biolarvicides,
Bacillus thuringiensis israelensis and
Bacillus sphaericus, are currently the most prominent larvicides as they are environmentally safe [
3]. However, under most environmental conditions, they have short residual effectiveness (
B. thuringiensis israelensis lasts for only 1–2 weeks [
7‐
9] and
B. sphaericus for 2–3 weeks [
6,
7]). Frequent applications have been widely recognized as a challenge for effective large-scale implementation [
11‐
15].
Larviciding was widely used in the first half of the 20th Century, most successfully outside sub-Saharan Africa, but fell out of favour after the introduction of IRS with DDT [
1,
13‐
16]. In the last decade, LSM, especially larviciding, has been reconsidered within an integrated vector management approach, especially as longer-lasting agents [
17‐
19], or novel deployment strategies and breeding site identification, i.e., using drones [
20,
21] might become increasingly available [
22]. Post-2000, pilot programmes of larviciding have been conducted in urban and in rural areas in multiple countries of Africa [
3,
9,
15,
21‐
39]. For example, the Urban Malaria Programme in Dar es Salaam, Tanzania, demonstrated operational feasibility and effectiveness of larviciding on larvae reduction and epidemiological outcomes in urban areas [
11,
26]. In Burkina Faso, a trial in 84 rural villages with
B. sphaericus applications during the main transmission season showed larviciding to be feasible and cost-effective when targeted to the most productive breeding sites [
25,
43]. Pilot implementations have previously been included in national malaria strategic plans in Eritrea [
7,
38,
44,
45], Zambia [
31,
46,
47] and Nigeria [
48]. Despite the long history of larviciding, its impact on malaria prevalence in humans [
9,
49] in different settings, and the influence of variations in its application, particularly frequency and timing of the year as well as duration [
6,
17,
40], remain insufficiently understood. For example, the application during the rainy season was described as impractical and less effective in study sites in Tanzania and The Gambia [
26,
40], but as feasible in the study in Burkina Faso [
24,
42], whereas its effectiveness during the dry season, as currently recommended, is still being debated [
1,
16].
Mathematical models have been used to simulate mosquito population dynamics and the relationship between larval stages and adult mosquitoes [
50‐
60]. However, most models consider only a small sub-set of the highly variable larviciding deployment scenarios. The models also include implicit assumptions about optimal deployment in relation to seasonality, (i.e., deployments either throughout the year, during rainy season or during dry seasons) and duration of larviciding effectiveness (i.e., constant or interrupted) without regard to re-treatment intervals and duration of product efficacy. While models have been used to simulate variations in the deployment strategies for other malaria control interventions, such as IRS [
61‐
64] and drugs [
64‐
67], larviciding strategies have not been investigated as much. In this study the impact of larviciding applications was simulated to assess the influence of different deployment strategies on expected entomological outcomes and malaria infections in humans for different seasonality and transmission settings.
Discussion
This modelling study investigated the impact of larviciding deployment strategies varying by coverage, duration, application frequency, and seasonal timing, for three transmission intensities (3, 10 or 90 ibpa), and five seasonality patterns, assuming homogeneous vector population similar to
An. gambiae. Overall, larviciding impacted the prevalence at a slower rate than the number of host-seeking mosquitoes and transmission intensity, while reduction in prevalence remained beyond the intervention period. The effective coverage during the intervention period, as a result of the efficacy duration of the larvicide, frequency and emergence reduction (effective coverage per single application), highly influenced the impact of larviciding. To ensure high impact, the product of the deployment factors need to be high, with regular deployments, tailored at the efficacy duration of the larvicide being more important than the effective coverage (number of emerging mosquitoes killed at each single round of larviciding) even at high coverage, when assuming short-lived larvicides. In highly seasonal settings, the deployment during the rainy season was predicted to have the highest impact on EIR and prevalence even at much lower coverage than during the dry season, and dry season larviciding had negligible impact. Larviciding at lower compared to high transmission intensity was further predicted to have a higher epidemiological impact with greater and longer lasting RR in the prevalence. This difference could be attributable to the differences in mosquito densities and faster rate of re-establishment at high transmission after lavicide decay to be effective (Additional file
2: Fig. S2.5).
Field observations [
4,
9] and simulations agree that larviciding reduces the number of emerging mosquitoes for the duration of the killing effect and that the vector population re-establishes immediately afterwards [
4,
37,
80]. In field studies, the time to reduce numbers of host-seeking mosquitoes varies between immediate impact and to lag times of two to three weeks [
4,
37,
80], with reductions in host-seeking mosquitoes ranging from very low to almost as high as the reduction in observed larval density [
6,
17,
40]. The simulations showed that prevalence is affected at a slower rate than the mosquitoes and transmission intensity and did not reach an equilibrium after one year of constant larviciding. This relates to the important role of duration of infection and parasite reservoirs in humans. It requires more time to clear infections in the human population by only reducing the mosquito population, whereas the reduction will also depend on malaria case management, which was not included in the simulations. However, this finding indicates that longer follow-up times would be required in field studies to capture impact of larviciding on prevalence with follow-up times varying depending on the seasonality.
Shorter intervals between deployments to reduce gaps in effective coverage over time were predicted to increase the average impact and reduce fluctuations in outcomes, as observed in two studies in Kenya [
6,
81]. In practice, the required deployment frequency depends on the emergence rate of new breeding sites and the persistence of the specific active agent [
5,
7,
17,
42,
82]. Notably, some programmes focus on treating only productive breeding sites [
10,
59], a strategy considered cost-effective in a rural district in Burkina Faso [
25]. Concentrating efforts on peri-domiciliary breeding sites has also been advocated [
83]. The appropriate deployment strategy to achieve high coverage of larviciding, or LSM in general, will further depend on dispersal of breeding sites and total land area to cover, surface area and quantity of breeding sites as well as their proximity to houses.
Larviciding is currently recommended by the WHO to be deployed in areas or seasons where breeding sites are fixed, few and findable, commonly associated with the dry season or urban areas [
1,
2], however, the simulation results suggest that larviciding in the dry season would have limited impact in seasonal settings. The results further suggest that the additional benefit of larviciding throughout the year would be marginal in highly seasonal settings. The greatest impact on prevalence was predicted when implementation preceded the peak in transmission, hence averting seasonal increases in host-seeking mosquito density. However, rainy season larviciding is more challenging, in particular because of proliferation of breeding sites and dilution of larvicide [
17,
29,
33,
81], while on the other hand emergence rates might be reduced when larvae are flushed away by very high rainfall [
18,
81,
84]. The trade-off between achieving high coverage (often described as more feasible in the dry season or arid areas [
1,
85]) and the epidemiological impact associated with a given coverage (in simulation estimated higher in the rainy season) must play out differently in diverse environments and might well account for some of the variation in seasonal patterns of impact observed in the field. In this modelling study, the operational challenge was attempted to reflect lowering the coverage during the rainy season while keeping the coverage during the dry season high, which did not change the recommended timing for larviciding unless coverage dropped to less than 20% of emerging mosquitoes killed. Alternative approaches to adjust for operational challenges would include simulating shorter effectiveness [
29] or more frequent deployments [
81] in the rainy season, presumably with similar implications. The results apply for settings with low vector densities and little to no transmission during the dry season, and where peak in transmission follows with one month lag after peak in rainfall.
The model results suggest that the relative impact would be greater at low than at high transmission in which high coverage would be needed. Nevertheless, larviciding has been successfully deployed in moderate to high transmission areas in several studies [
27,
37,
86]. One study in particular showed that larviciding could be implemented at high transmission in highly seasonal areas with findable breeding sites [
86]. Reduction in prevalence has rarely been studied in larviciding field studies [
9] although one study reported a reduction of more than 70% [
26]. Based on the simulations, such high reductions would only be achievable at very high coverage and long duration of effective larviciding, either with more frequent deployments or longer residual activity (e.g., longer than 120 days), and seems unlikely to be achieved with larviciding alone as reported in the study.
In most instances, larviciding, recommended as a supplementary intervention [
1], will be deployed alongside ITNs or IRS, to reduce transmission and create a context where larviciding is more effective. While interactions were not explicitly modelled, implicitly synergistic effects with these interventions were assumed by simulating low pre-larviciding transmission intensity. This assumption is supported by the higher impact at low transmission seen in the predictions. For the same reason, synergies with chemotherapeutic interventions can also be anticipated [
87]. Further synergies are likely where there is insecticide or drug resistance because larvicides have different biochemistry and act independently of host-seeking and resting behaviour of adult mosquitoes [
14], and they can address transmission that is refractory to the core interventions. In practice, larviciding might also be combined with other LSM approaches [
1,
88,
89] that together reduce the adult mosquito emergence in an area.
Although the measurement of coverage is critical for predicting the impact of a larviciding programme, there is no standardization of operational coverage measures. These have been variously defined as the number of treated breeding sites out of the total identified, the proportion of the surface area of water bodies that are treated, or even the proportion of larvae covered by larvicide out of all larvae within a breeding site (Additional file
3). Targeting specific areas (or selection of sub-sets of breeding sites or other criteria) reduce the denominators in such calculations. All these measures of coverage are challenging to estimate [
8,
12,
15,
90], especially since the proportion of breeding sites identified varies in each setting and over time. Regardless of the suitability of the local settings, the effectiveness also depends on the performance of field staff, community engagement and supervision [
91,
92].
The simulations of larviciding in Mbita, western Kenya, attempted to calibrate the model to allow for these factors. The results emphasize the difficulty of correctly reproducing the impact of larviciding, and on estimating coverage levels that would be feasible, despite accounting for details of deployment. In the simulations, the vector population immediately increased between the larvicide applications, whereas in the field measurements adult densities remained relatively low [
6]. Hence, the low levels in host-seeking mosquito density maintained throughout the intervention period of two years could only be reproduced with constant high effective coverage. It could be that the sampling under-represented the true adult mosquito density in the community or that the simulated re-treatment intervals underestimated the effectiveness in practice. Another reason could be additional use of ITNs or other factors not accounted for in the simulations that lowered the transmission throughout the study period.
In contrast to the homogeneous vector populations in the simulations, multiple vector species are usually present in the field, and some of this variation in outcomes result from environmental and ecological factors that cannot be captured in the model. For instance, larviciding of rice fields has been found to be impractical in The Gambia, due to low accessibility [
40], but was feasible in Tanzania and Rwanda [
36,
93]. One study in Kenya found positive effects of dry season implementation on mosquito density and clinical malaria using long-lasting larvicides [
17]. Another study in western Kenya reported higher effectiveness during the rainy season, using short-lived larvicides [
94]. For instance, while a high number of breeding sites existed throughout the dry season in an urban setting (Dar es Salaam) [
11], they substantially varied by season in the rural village setting in Mbita, western Kenya [
6]. Hence, field operations should always consider local climate, breeding site permanence based on water sources and characteristics, dominant vector species, available resources, and engagement of the community [
95]. The diversity of operational implementation and outcomes highlights the need for more setting-specific guidelines for larviciding to differentiate between strategies for different localities. For instance, in Tanzania, the national malaria strategic plan includes larviciding the whole country [
96], but heterogeneities in malaria epidemiology and environmental factors represent a huge challenge for planning appropriate large-scale strategies [
97,
98] and implementation will require a thorough assessment of the context at local level.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.